Monitoring of Industrial Sounds, Seals, and Bowhead Whales Near BP's

MONITORING OF INDUSTRIAL SOUNDS, SEALS, AND BOWHEAD WHALES NEAR BP’S NORTHSTAR OIL DEVELOPMENT, ALASKAN BEAUFORT SEA, 1999–2002

and

Greeneridge Sciences Inc.

for

BP Exploration (Alaska) Inc. Dept of Health, Safety & Environment 900 East Benson Blvd. Anchorage, AK 99519-6612

and

National Marine Fisheries Service Anchorage, AK, and Silver Spring, MD

December 2003 MONITORING OF INDUSTRIAL SOUNDS, SEALS, AND BOWHEAD WHALES NEAR BP’S NORTHSTAR OIL DEVELOPMENT, ALASKAN BEAUFORT SEA, 1999–2002

edited by

W. John Richardson 1 and Michael T. Williams 2

1 LGL Ltd., environmental research associates 22 Fisher St., POB 280, King City, Ont. L7B 1A6 (905) 833-1244; [email protected]

2 LGL Alaska Research Associates Inc. 1101 E. 76th Ave, Suite B, Anchorage, AK 99518 (907) 562-3339; [email protected]

for

BP Exploration (Alaska) Inc. Dept of Health, Safety & Environment 900 East Benson Blvd. Anchorage, AK 99519-6612

and

National Marine Fisheries Service Anchorage, AK, and Silver Spring, MD

December 2003 Suggested format for citations: Richardson, W.J. and M.T. Williams (eds.) 2002. Monitoring of industrial sounds, seals, and bowhead whales near BP’s Northstar Oil Development, Alaskan Beaufort Sea, 1999–2002. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 343 p.

Individual chapters can also be cited, as follows: Williams, M.T., W.J. Richardson and R. Rodrigues. 2003. Introduction. p. 1-1 to 1-12 In: … LGL Rep. TA2707-4. Williams, M.T. and R. Rodrigues. 2003. BP’s activities at Northstar, 1999–2002. p. 2-1 to 2-29 In: … LGL Rep. TA2707-5.

Ice-Covered Season Blackwell, S.B., C.R. Greene Jr. and W.J. Richardson. 2003. Drilling and operational sounds from an oil production island in the ice-covered Beaufort Sea. p. 3-1 to 3-26 In: … LGL Rep. TA2704-2.*

Moulton, V.D., R.E. Elliott and M.T. Williams. 2003. Fixed-wing aerial surveys of seals near BP’s Northstar and Liberty sites, 2002. p. 4-1 to 4-35 In: … LGL Rep. TA2702-2.

Moulton, V.D., W.J. Richardson, T.L. McDonald, R.E. Elliott, M.T. Williams and C. Nations. 2003. Effects of Northstar on local abundance and distribution of ringed seals (Phoca hispida) of the Alaskan Beaufort Sea. p. 5-1 to 5-24 In: … LGL Rep. TA2702-4.*

Break-Up and Open Water Seasons Blackwell, S.B. 2003. Sound measurements, 2002 open-water season. p. 6-1 to 6-49 In: … LGL Rep. 2705-2.*

Greene, C.R. Jr., M.W. McLennan, S.B. Blackwell, T.L. McDonald and W.J. Richardson. 2003. Acoustic monitoring of bowhead whale migration, autumn 2002. p. 7-1 to 7-31 In: … LGL Rep. TA2706-2.*

Greene, C.R. Jr., M.W. McLennan, R.G. Norman, T.L. McDonald and W.J. Richardson. 2003. Using DIFAR sensors to locate calling bowhead whales and monitor their migration. p. 8-1 to 8-38 In: … LGL Rep. TA2706-3.*

McDonald, T.L. and W.J. Richardson. 2003. A statistical approach for assessing displacement of bowhead whale locations associated with elevated sound levels. p. 9-1 to 9-26 In: … LGL Rep. TA2706-4.*

Richardson, W.J., C.R. Greene Jr. and T.L. McDonald. 2003. Acoustic localization of bowhead whales near a Beaufort Sea oil development, 2001-2002: Evidence of deflection at high-noise times? p. 10-1 to 10-42 In: … LGL Rep. TA2706-5.*

Numbers of Seals and Whales Potentially Affected, Nov. 2001 – Oct. 2002 Moulton, V.D., M.T. Williams, W.J. Richardson and T.L. McDonald. 2003. Estimated numbers of seals and whales potentially affected by Northstar activities, Nov. 2001 – Oct. 2002. p. 11-1 to 11-31 In: … LGL Rep. TA2707-6.*

* Chapters 3 and 5–11 have been revised relative to those in a draft of this report that was dated May 2003.

ii Table of Contents

TABLE OF CONTENTS

ACRONYMS AND ABBREVIATIONS...... xi EXECUTIVE SUMMARY...... xiii Introduction...... xiii BP’s Activities at Northstar, 1999–2002 ...... xv Drilling and Operational Sounds, Winter 2001 & 2002 ...... xv Fixed-Wing Aerial Surveys of Seals Near BP’s Northstar & Liberty Sites, Spring 2002...... xvi Northstar Effect on Seal Densities...... xvi Sound Measurements, 2002 Open-Water Season ...... xvii Acoustic Monitoring of Bowhead Whale Migration, Autumn 2002 ...... xviii DIFAR Sensors for Monitoring Bowhead Whale Migration...... xix Statistical Approach for Assessing Displacement...... xx Deflection of Bowhead Migration near Northstar?...... xx Estimated Numbers of Seals and Whales Potentially Affected, Nov. 2001 – Oct. 2002...... xxi

CHAPTER 1: INTRODUCTION ...... 1-1 Background...... 1-2 BP Business Rationale for the Northstar Monitoring Program...... 1-3 Chronology of Marine Mammal Authorizations ...... 1-3 Monitoring Tasks...... 1-4 Monitoring During Ice-Covered Seasons...... 1-5 1. Fixed-wing Aerial Surveys of Seals ...... 1-5 2. Monitoring of Seal Structures in the Sea Ice Adjacent to Northstar...... 1-6 3. Helicopter Survey of Seal Breathing Holes...... 1-7 4. Sound Measurements during Winter ...... 1-7 Monitoring During Open-Water Seasons...... 1-7 5. Sound Measurements during Summer and Autumn ...... 1-7 6. Island-based Visual Monitoring during Spring Break-up, 2000...... 1-8 7. Acoustic Monitoring of Migrating Bowhead Whales...... 1-8 Other Components of This Report...... 1-9 Acknowledgements...... 1-9 Literature Cited ...... 1-10

CHAPTER 2: BP’S ACTIVITIES AT NORTHSTAR, 1999–2002 ...... 2-1 Abstract...... 2-2 Introduction...... 2-2 Overall Sequence of Northstar Development...... 2-3 Seasonal Transportation Needs ...... 2-3 Ice Roads ...... 2-4 Transition and Open-Water Seasons...... 2-4 Initial Construction ...... 2-8 Ice Road Construction, Early 1999...... 2-8 Ice Road Construction, Winter 1999-2000...... 2-9 Island and Pipeline Construction, Early-Mid 2000 ...... 2-11

iii Table of Contents

Later Construction Phases and Initial Drilling...... 2-14 2000 Break-up and Open-Water Season ...... 2-14 2000–2001 Ice-covered Period...... 2-16 Ice-Roads and Transportation...... 2-16 Activities On and Near Northstar Island...... 2-18 Backfill Along Pipelines...... 2-19 2001 Break-up and Open-water Season ...... 2-20 Transportation and Sealift...... 2-20 Island Activities...... 2-21 Initial Production and Continued Drilling ...... 2-22 2001–2002 Ice-Covered Season ...... 2-22 Ice Roads and Transportation ...... 2-22 Activities On and Near Northstar Island...... 2-23 2002 Break-up and Open-Water Season ...... 2-25 Transportation...... 2-25 Island Activities...... 2-25 Other Major Offshore Activities...... 2-26 Conclusions...... 2-28 Acknowledgements...... 2-28 Literature Cited ...... 2-29

ICE-COVERED SEASON

CHAPTER 3: DRILLING AND OPERATIONAL SOUNDS FROM AN OIL PRODUCTION ISLAND IN THE ICE-COVERED BEAUFORT SEA ...... 3-1 Abstract...... 3-2 Introduction...... 3-2 Background ...... 3-2 Methods ...... 3-5 Equipment...... 3-5 Field Procedures ...... 3-5 Island Activities During Recordings ...... 3-7 Signal Analysis...... 3-7 Results...... 3-8 Underwater Sound...... 3-8 Broadband Levels ...... 3-8 Spectral Characteristics ...... 3-10 In-Air Sound...... 3-10 Broadband Levels ...... 3-10 Spectral Characteristics ...... 3-12 Iceborne Vibrations ...... 3-13 Broadband Levels ...... 3-13 Spectral Characteristics ...... 3-16 Summary of Results ...... 3-17 Discussion...... 3-18

iv Table of Contents

Sounds during Drilling ...... 3-18 Sounds during Production ...... 3-20 Sounds from Power Generation...... 3-21 Background Levels...... 3-21 Audibility Ranges and Conclusions...... 3-22 Acknowledgements...... 3-24 Literature Cited ...... 3-24

CHAPTER 4: FIXED-WING AERIAL SURVEYS OF SEALS NEAR BP’S NORTHSTAR AND LIBERTY SITES, 2002...... 4-1 Abstract...... 4-2 Introduction...... 4-2 Observed Ringed Seal Densities in the Area...... 4-4 Factors Influencing Numbers of Seals Counted ...... 4-6 Factors Affecting Seal Abundance and Distribution ...... 4-6 Factors Affecting the Proportion of Seals Hauled Out ...... 4-7 Effects of Industrial Activity on Ringed Seal Distribution ...... 4-7 Objectives of BP Aerial Survey ...... 4-9 Methods ...... 4-10 Survey Design ...... 4-10 Survey Procedures...... 4-12 Data Recording Procedures ...... 4-13 Analysis Procedures ...... 4-13 Statistical Tests...... 4-14 Description of the Study Area...... 4-15 Water Depth...... 4-15 Ice Conditions...... 4-15 Results...... 4-18 Ringed Seal Abundance and Distribution ...... 4-18 Observed Ringed Seal Behavior...... 4-18 Bearded Seal Sightings...... 4-18 Factors Affecting Ringed Seal Abundance and Distribution ...... 4-24 Water Depth...... 4-26 Distance from Ice Edge ...... 4-26 Observed Ringed Seal Densities Near Development Sites...... 4-27 Northstar ...... 4-27 Liberty ...... 4-29 Discussion...... 4-29 Observed Ringed Seal Densities, Abundance and Distribution ...... 4-29 Factors Affecting Observed Ringed Seal Density...... 4-31 Water Depth Effects ...... 4-31 Distance from Ice Edge ...... 4-31 Observed Seal Densities Near Northstar ...... 4-31 Acknowledgements...... 4-32 Literature Cited ...... 4-32

v Table of Contents

CHAPTER 5: EFFECTS OF NORTHSTAR ON LOCAL ABUNDANCE AND DISTRIBUTION OF RINGED SEALS (PHOCA HISPIDA) OF THE ALASKAN BEAUFORT SEA ...... 5-1 Abstract...... 5-2 Introduction...... 5-2 Methods ...... 5-3 Survey Design and Procedures...... 5-3 Analysis Procedures ...... 5-4 Statistical Tests ...... 5-7 Power Analysis...... 5-8 Results...... 5-9 Observed Seal Densities and Distribution...... 5-9 Environmental Factors...... 5-10 Factors Affecting Local Abundance...... 5-10 Factors Affecting Proportion of Seals Hauled Out...... 5-10 Northstar Effects...... 5-10 Seal Densities Observed near Northstar ...... 5-10 Seal Densities with Allowance for Other Factors...... 5-13 Power of Poisson Regression Analysis...... 5-13 Other Anthropogenic Effects...... 5-16 Discussion...... 5-16 Observed Seal Densities ...... 5-16 Northstar Effects...... 5-16 Other Anthropogenic Effects...... 5-18 Environmental Factors...... 5-19 Effectiveness of Intensive Aerial Surveys...... 5-20 Survey Design...... 5-20 Analysis Approach ...... 5-20 Statistical Power ...... 5-20 Conclusions...... 5-21 Acknowledgements...... 5-21 Literature Cited ...... 5-21 In Memory of Tom Blaesing...... 5-24

OPEN-WATER SEASON

CHAPTER 6: SOUND MEASUREMENTS, 2002 OPEN-WATER SEASON ...... 6-1 Abstract...... 6-2 Introduction...... 6-2 Background ...... 6-2 Objectives and Approach ...... 6-3 Methods ...... 6-4 Boat-based Recordings in Open Water ...... 6-4 Equipment...... 6-4 Field Procedures ...... 6-6 Cabled Hydrophones and ASARs ...... 6-7

vi Table of Contents

Equipment...... 6-7 Field Procedures ...... 6-8 Signal Analysis...... 6-10 Boat-based Recordings ...... 6-10 Cabled Hydrophone Signals ...... 6-11 Results...... 6-13 Boat-based Recordings...... 6-13 Underwater Sounds...... 6-13 Airborne Sounds...... 6-18 Summary of Boat-Based Recordings in Open Water ...... 6-20 Near-Island Cabled Hydrophone Recordings...... 6-23 Vessel Sounds...... 6-25 Overall Underwater Sound ...... 6-27 Calculation of the ISI in 2002...... 6-31 Summary of Cabled Hydrophone Recordings...... 6-33 Discussion...... 6-35 Boat-based Recordings...... 6-35 Underwater Sounds...... 6-35 Airborne Sounds...... 6-37 Near-island Cabled Hydrophone Recordings...... 6-38 Vessel Sounds...... 6-39 Overall Underwater Sound ...... 6-39 Sound Field Around Northstar ...... 6-40 Summary...... 6-41 Boat-based Recordings in Open Water ...... 6-42 Near-island Cabled Hydrophone Recordings...... 6-42 Acknowledgements...... 6-43 Literature Cited ...... 6-43 Appendix 6.1: Definition of Industrial Sound Index (ISI) in 2001...... 6-45

CHAPTER 7: ACOUSTIC MONITORING OF BOWHEAD WHALE MIGRATION, AUTUMN 2002...... 7-1 Abstract...... 7-2 Introduction...... 7-3 Background ...... 7-3 Specific Objectives...... 7-4 General Approach...... 7-5 Methods ...... 7-6 Overview of Methods...... 7-6 Refinements in Methods during 2002 ...... 7-6 1. Time and Direction Calibration ...... 7-9 2. Gain Imbalance Between the Directional Sensors...... 7-10 3. Industrial Sound Index (ISI) ...... 7-12 4. High Background Noise Cutoff Distance ...... 7-13 Offshore Distances ...... 7-15 Results...... 7-16

vii Table of Contents

DASAR Installation Data ...... 7-16 Ping Calibration Measurements...... 7-16 Time Calibration...... 7-16 Reference Axis Calibration...... 7-16 Underwater Sounds near Northstar...... 7-18 Temporal Variation...... 7-18 Spectral Characteristics ...... 7-18 Industrial Sound Index (ISI) ...... 7-18 Whale Call Analysis...... 7-21 Number and Locations of Whale Calls Detected...... 7-21 Accuracy of Whale Call Localization...... 7-23 Discussion...... 7-24 Ping Calibrations ...... 7-24 Underwater Sounds near Northstar...... 7-28 Whale Calls and Locations...... 7-29 Acknowledgements...... 7-30 Literature Cited ...... 7-30

CHAPTER 8: USING DIFAR SENSORS TO LOCATE CALLING BOWHEAD WHALES AND MONITOR THEIR MIGRATION...... 8-1 Abstract...... 8-2 Introduction...... 8-3 Background ...... 8-3 Methods ...... 8-4 DIFAR Principles ...... 8-5 Directional Autonomous Seafloor Acoustic Recorder (DASAR) ...... 8-7 Digital Recording ...... 8-8 Transponder for Health Checks...... 8-8 Physical Construction and Batteries ...... 8-8 The DASAR Array...... 8-11 Time and Reference Axis Calibration ...... 8-12 Gain Imbalance Between the Directional Sensors...... 8-14 Localization Process and Its Accuracy...... 8-14 High Background Noise Cutoff Distance...... 8-14 Results...... 8-14 Calibration of DASAR Clock Times and Reference Axis Bearings ...... 8-14 Whale Call Localization ...... 8-16 Localization Accuracy...... 8-20 High Background Noise Cutoff Distance...... 8-20 Discussion and Conclusions ...... 8-24 Potential Improvements to DASAR Performance...... 8-24 Utility of DASARs in Localizing Sound Sources ...... 8-24 Acknowledgements...... 8-25 Literature Cited ...... 8-25 Appendix 8.1: Gain Imbalance Between Directional Sensors and Its Effect On Bearing Accuracy...... 8-27 The Effects of a Gain Imbalance ...... 8-27

viii Table of Contents

Measuring the Actual Error and Gain Factor ...... 8-31 Results ...... 8-33 Appendix 8.2: Localization Accuracy ...... 8-36

CHAPTER 9: A STATISTICAL APPROACH FOR ASSESSING DISPLACEMENT OF BOWHEAD WHALE LOCATIONS ASSOCIATED WITH ELEVATED SOUND LEVELS ...... 9-1 Abstract...... 9-2 Introduction...... 9-2 Methods ...... 9-4 Response Distribution ...... 9-4 Quantile Regression...... 9-5 Form of Relationship...... 9-7 Expected Quantiles Under No Effect ...... 9-9 Assuming Responses are Independent...... 9-9 Interdependence of Responses...... 9-10 Example: Northstar and Bowhead Whales, 2001 ...... 9-12 Methods...... 9-13 Results ...... 9-15 Discussion...... 9-23 Acknowledgements...... 9-24 Literature Cited ...... 9-25

CHAPTER 10: ACOUSTIC LOCALIZATION OF BOWHEAD WHALES NEAR A BEAUFORT SEA OIL DEVELOPMENT, 2001–2002: EVIDENCE OF DEFLECTION AT HIGH-NOISE TIMES? ...... 10-1 Abstract...... 10-2 Introduction...... 10-2 Methods ...... 10-4 Whale Call Localization ...... 10-5 Sounds from Northstar ...... 10-6 Statistical Analyses...... 10-8 Exclusion of Non-Northstar Vessels in 2002 ...... 10-8 Quantile Regression of Offshore Distances vs. ISI ...... 10-9 Averaging Time for ISI ...... 10-9 Expected Quantiles if No Effect ...... 10-9 Estimating Offshore Displacement vs. ISI ...... 10-10 Results...... 10-10 Whale Call Locations ...... 10-10 Sounds 400 m from Northstar ...... 10-12 Call Locations vs. Sound Level near Northstar...... 10-14 2001 Results ...... 10-14 2002 Results ...... 10-15 Discussion...... 10-18 Effect of Northstar-Related Sound on Call Locations...... 10-19 Statistical Analysis Issues...... 10-24 Possible Attraction to Weak Industrial Sounds ...... 10-26 Advantages and Limitations of the Acoustical Monitoring Approach...... 10-26

ix Table of Contents

Ability of Approach to Meet Objectives ...... 10-30 Objective 1...... 10-30 Objective 2...... 10-30 Objective 3...... 10-31 Objective 4...... 10-32 Objective 5...... 10-33 Objective 6...... 10-34 Acknowledgements...... 10-34 Literature Cited ...... 10-35 Appendix 10.1: 2002 Results with Different ISI Averaging Times...... 10-38 Appendix 10.2: Estimated Displacement in 2001...... 10-42

CHAPTER 11: ESTIMATED NUMBERS OF SEALS AND WHALES POTENTIALLY AFFECTED BY NORTHSTAR ACTIVITIES, NOV. 2001 – OCT. 2002 ...... 11-1 Abstract...... 11-2 Introduction...... 11-2 2001–2002 Ice-Covered Season...... 11-3 Possible Displacement from Northstar Development Area...... 11-4 Acoustic Monitoring...... 11-8 Fixed-wing Aerial Surveys...... 11-9 Combined Impact of Northstar Activities during Ice-Covered Season ...... 11-10 2002 Break-Up and Open-Water Seasons ...... 11-11 Estimated Number of Seals Present...... 11-12 Break-Up Period ...... 11-12 Open-Water Period ...... 11-12 Combined Break-Up and Open-Water Periods ...... 11-13 Estimated Number of Whales Potentially Affected...... 11-13 Bowhead Whales ...... 11-13 Gray Whales ...... 11-22 Beluga Whales...... 11-22 Effect on Accessibility to Hunters ...... 11-23 Seals...... 11-23 Bowhead Whales...... 11-24 Summary...... 11-25 Seals...... 11-26 Whales ...... 11-27 Availability for Subsistence...... 11-27 Acknowledgements...... 11-28 Literature Cited ...... 11-28

x Acronyms & Abbreviations

ACRONYMS AND ABBREVIATIONS

~ approximately ACS Alaska Clean Seas AEWC Alaska Eskimo Whaling Commission ASAR Autonomous Seafloor Acoustic Recorder ASL Above Sea Level BPXA BP Exploration (Alaska) Inc. BACI Before-After/Control-Impact, a type of study design BIC Bayesian Information Criterion C.F.R. Code of Federal Regulations C.H. Cabled hydrophone CI Confidence Interval CIDS Concrete Island Drilling System DASAR Directional Autonomous Seafloor Acoustic Recorder DAT Digital Audio Tape dB decibel, a logarithmic measure of sound strength dBA “A-weighted” decibel scale, for in-air sounds DIFAR Directional Frequency and Recording – a directional sonobuoy EIS Environmental Impact Statement FAA Federal Aviation Administration FLIR Forward Looking Infra-Red ft foot or feet (1 foot = 0.305 m) FWS Fish & Wildlife Service GPS Global Positioning System Hz hertz, or “cycles per second”; standard measure of sound frequency IHA Incidental Harassment Authorization ISI Industrial Sound Index (refers to sound level in 5 specific one-third octave bands, centered at 31.5–80 Hz) ITC International Transducer Corp. LoA Letter of Authorization kg kilogram (= 2.20 lb) km kilometer (1 km = 3281 ft, 0.62 land miles, or 0.54 n.mi)

xi Acronyms & Abbreviations lb pound (= 0.454 kg) m meter (1 m = 1.09 yards or 3.28 ft) mi land or statute mile (1 mi = 1.61 km or 0.87 n.mi.) MLLW Mean Low Low Water MMPA Marine Mammal Protection Act MMS Minerals Management Service, U.S. Dept. of the Interior n.mi. nautical mile (1 n.mi. = 1.15 land miles or 1.853 km) NMFS National Marine Fisheries Service, U.S. Dept. of Commerce NSB North Slope Borough PLQ permanent living quarters PV particle velocity RL received level rms root mean square (a type of average) rpm revolutions per minute s.d. standard deviation SDC Steel Drilling Caisson SPDSL Sound Pressure Density Spectrum Level SPL Sound Pressure Level UIC Underground Injection Control [well] µPa micropascal, a measure of pressure USFWS U.S. Fish & Wildlife Service UTM Universal Transverse Mercator VFR Visual Flight Rules WDBO West Dock Base of Operations

xii Executive Summary

EXECUTIVE SUMMARY

Introduction

BP Exploration (Alaska) Inc. (BP) began constructing oil-production facilities for the Northstar Development during early 2000, and began producing crude oil from the Northstar Unit on 31 October 2001. The unit extends from 3.2 to 12.9 kilometers [km] (2 to 8 miles) offshore from Point Storkersen, northwest of the Prudhoe Bay industrial complex. Northstar is approximately 54 miles (87 km) northeast of Nuiqsut, the closest Native Alaskan (Inupiat) community (see map). The Northstar Development was built on the submerged remnants of Seal Island. Seal Island was an artificial island constructed in 1982, used for exploration drilling during the 1980s, and subsequently abandoned. The Northstar Development includes a gravel island for the main facilities and two pipelines connecting the island to the existing infrastructure in Prudhoe Bay. One pipeline transports crude oil to shore, and the other transports natural gas to the island for field injection. The facilities on the island include prefabricated modules for living quarters, utilities, and warehouse/shop. Also present are a drilling rig and facilities for waste grind and injection and for oil production. The production facilities include gas turbine engines to operate power generators and gas compressors. Transportation between the mainland and Northstar is primarily via ice roads in winter, vessels in summer, and helicopters during the transitional seasons. The Northstar Devel- opment is, to date, the only offshore oil production facility in the Beaufort Sea north of the barrier islands.

Location of the Northstar Development at Seal Island in the central Alaskan Beaufort Sea.

xiii Executive Summary

“Northstar Island” was constructed and the associated pipelines were installed during early 2000. Other construction activities continued through mid-2001 with the installation of facilities on the island necessary for processing and handling oil from the reservoir. The production modules were brought on line and oil production began on 31 October 2001. Oil production at Northstar has continued from then to the present. This report describes BP’s Northstar activities, and the associated marine mammal and acoustic monitoring projects, during the November 2001 through October 2002 period. To provide a more comprehensive summary, considerable additional information is included about BP’s Northstar activities and some aspects of the monitoring work in previous years. This report is not intended to be a final report on all aspects of the Northstar monitoring work, as some work is ongoing. However, it does begin to compile the results from multiple years as well as fulfilling BP’s reporting obligations for Nov. 2001–Oct. 2002. BP’s business rationale for the marine mammal and acoustical studies was driven both by corporate values and by regulatory requirements. BP supports studies that objectively assess environmental effects that may result from BP operations. BP’s company goal of “no harm to the environment” drives the design and implementation of studies intended to understand and minimize the effects of BP operations. In addition, monitoring and reporting were required to satisfy BP’s commitments under Letters of Author- ization (LoA) issued by the National Marine Fisheries Service (NMFS) to BP, and under a North Slope Borough (NSB) zoning ordinance. Seven different monitoring tasks were identified during BP’s initial planning and during discussions with NMFS and stakeholders before Northstar construction began. Various tasks have been implemented in different years depending on BP’s activities in those years and the outcome of monitoring in previous years. The “X” symbols in the following table summarize the monitoring undertaken each year since 1999.

Marine mammal and acoustical studies conducted by BP for Northstar during 1999–2002.

Monitoring Task 1999 2000 2001 2002 Ice-Covered Season 1.Fixed-wing aerial surveys of ringed seals (LGL) a XXXX 2.Monitoring of seal structures in the sea ice (LGL) X X X 3.Helicopter survey of seal breathing holes (LGL) X 4.Acoustic monitoring during winter (Greeneridge Sciences) X X X

Open Water Season 5.Sound measurements July, Aug., Sept. (Greener. Sciences) X X X 6.Island-based visual monitoring (LGL) X 7.Bowhead migration monitoring (Greener. Sciences & LGL) X b XX

a Aerial surveys also done in 1997–98. b Data on bowheads in autumn 2000 were incomplete and of limited use.

This report describes the results from 2002 for the four monitoring studies active in 2002. How- ever, this report also includes some of the results obtained during pre-2002 phases of those four studies. The report consists of an Introduction (summarized here), a chapter describing BP’s Northstar activities from 1999 through October 2002, three chapters describing monitoring tasks during the winter and spring (ice-covered season), five chapters describing monitoring tasks during the open-water season, and a concluding chapter estimating the numbers of seals and whales potentially affected. The following sections are summaries of Chapters 2 to 11 of the report.

xiv Executive Summary

BP’s Activities at Northstar, 1999–2002

BP’s Northstar Oil Development is centered at Northstar Island, located in the Beaufort Sea ~6 miles (10 km) offshore of Point Storkerson along the north coast of Alaska, and about 3 miles (5 km) offshore of the barrier islands. As one condition of the authorizations received from the National Marine Fisheries Service concerning potential “taking” of whales and seals, BP must report on the activities conducted in support of Northstar development. Chapter 2, by M.T. Williams and R. Rodrigues (of LGL Alaska), describes BP’s Northstar activities from 1999 through October 2002. For this purpose, the operational year is divided into two periods: the ice-covered season, nominally from 1 Nov. through 15 June, and the open-water season from 16 June through 31 Oct. Ice roads are constructed each year during the ice-covered period to facilitate transportation of personnel and equipment to and from the island. Equipment used during construction of ice roads includes powered augers and pumpers mounted on rolligon-type off-road vehicles, snow blowers, front-end loaders, bulldozers, dump trucks, and tanker trucks. After ice roads are completed, standard pick-ups, SUVs, and buses are used to transport personnel, and tractor trailers are used for equipment and supplies. During the open-water period crew vessels and barges are used to transport personnel and re-supply the island. Helicopters are used during transition periods (freeze-up and break-up), and intermittently at other times when ice and water conditions do not permit use of land-based vehicles or vessels. An ice road was constructed during the ice-covered period in 1998–99 but in April 1999 construction was postponed due to regulatory problems, and the ice road was abandoned. Ice roads were re-built during the 1999–2000 ice-covered period, and used to haul gravel for island construction (Feb.–Apr. 2000) and to install subsea pipelines (Mar.–May 2000). Island construction continued during the 2000 open-water period; equipment used included cranes, pile drivers, and various types of heavy equipment. During the 2000 open- water period, three sealift barges delivered modules for living quarters, utilities, pipe racks, and a drilling rig. During the 2000–01 ice-covered period, construction and equipment-installation continued, and the first wells were drilled. During August 2001, three sealift barges delivered processor modules, gas-turbine generators and compressors, and other equipment to the island. This equipment was installed by late autumn of 2001, and the gas-turbine powered generators replaced diesel generators as the primary source of power. Oil production began on 31 Oct. 2001 and has continued to the present. Production operations involve use of up to three generators and two high-pressure compressors powered by gas-turbine engines, along with various other types of equipment. Drilling operations continued for most of the November 2001 through October 2002 period, although drilling during the open water season was limited to depths above the oil reservoir. Various testing, training, monitoring and inspection activities, including spill response activities, occurred on an intermittent basis. Drilling and Operational Sounds, Winter 2001 & 2002

Recordings of sounds underwater and in air, and of iceborne vibrations, were obtained in March 2001 and February–March 2002. (Similar measurements during initial construction of Northstar in early 2000 are not covered by this report.) The purpose of the recordings was to document the levels, characteristics, and range-dependence of sounds and vibrations produced by Northstar-related industrial activities, such as drilling and oil production, when Northstar is surrounded by shore-fast ice during winter. The measurements were needed to help interpret data on occurrence of ringed seals near Northstar during winter and spring. Chapter 3, by S.B. Blackwell (Greeneridge Sciences) et al., describes this work, and has been submitted (Aug. 2003) for journal publication.

xv Executive Summary

Drilling sounds were readily identifiable underwater and produced the highest broadband (10– 10,000 Hz) levels, with a maximum of 124 dB re 1 µPa 1 km (0.6 mi) from the drill rig. Drilling led to increased underwater levels, mainly at 700–1400 Hz. In contrast, drilling did not increase broadband levels in air or in the ice over those during oil production. The contribution of production operations to overall sound levels at Northstar was not very prominent, but during production levels underwater increased somewhat in the 125–160 Hz band and at certain lower frequencies associated with power generation. Repeated measurements at increasing distances from the sound source allowed computing propagation loss. In all three media, broadband sound levels decreased by ~20 dB / tenfold change in distance. Background levels underwater were reached by 9.4 km (5.8 mi) during drilling and 3–4 km (1.9–2.5 mi) without. In air, background levels were reached 5–10 km (3.1–6.2 mi) from Northstar, depending on the wind but irrespective of drilling. In the ice, background levels were reached 2–10 km (1.2–6.2 mi) from the island depending on the geophone channel and the wind, but irrespective of drilling. A comparison of the recorded sounds with harbor and ringed seal audiograms showed that sounds were probably audible to seals, at least intermittently, out to ~1.5 km in water and ~5 km in air. Fixed-Wing Aerial Surveys of Seals Near BP’s Northstar & Liberty Sites, Spring 2002

Chapter 4, by V.D. Moulton et al. (LGL), documents ringed seal (Phoca hispida) densities and local distribution in the area of landfast ice that surrounded BP’s Northstar oil development in spring 2002. Aerial surveys for seals were conducted during 30 May–7 June 2002 in a manner consistent with surveys done during each spring since 1997. There had been construction activities at Northstar from early 2000 through late 2001, and oil production and drilling at Northstar from then through the time of the aerial surveys in the spring of 2002. These industrial activities had the potential to affect the local abundance and distribution of ringed seals. The survey design provided for intensive and repeated coverage of the study area within each spring season. In 2002, the surveys covered 3618 km2 or 1396 mi2 of landfast ice habitat. The overall observed density in 2002 was 0.72 seals/km2. Excluding areas where the summer water depth was <3 m, where seals were rarely seen, the overall observed density in 2002 was 0.83 seals/km2, as compared with 0.43 to 0.63 seals/km2 in 1997–2001. In 2002, sightings were most common in shallower portions (5–15 m or 16–49 ft) of the landfast ice zone, and in the zone 10–15 km or 6–9 mi shoreward of the fast ice edge. Seal sightings may have been most common at these intermediate distances from the ice edge because there were numerous cracks in the ice in this zone, and these provided suitable habitat for hauling out to molt. This may also explain, at least in part, why the overall ringed seal density for 2002 was higher than during previous survey years. Univariate analysis provided no conclusive evidence that Northstar industry activities significantly affected the observed ringed seal density in 2002. (The following chapter describes a more definitive multivariate analysis of the combined 1997–2002 data, taking account of the confounding effects of weather, ice, and temporal factors that are known or expected to influence seal haulout and distribution.) Northstar Effect on Seal Densities

Chapter 5, by V.D. Moulton (LGL) et al., investigates how densities of ringed seals were affected by construction and operational activities at Northstar Island. This chapter is being (Dec. 2003) submitted for journal publication. Intensive and replicated aerial surveys of seals on landfast ice were conducted

xvi Executive Summary during three spring seasons before island construction began (1997–99); after a winter of intensive island construction (2000); and after more limited construction plus drilling (2001) and drilling plus oil production (2002). A Poisson regression model examined seal densities relative to distance from Northstar after allowance for environmental covariates known or suspected to influence seal distribution and haulout behavior. A post hoc power analysis indicated that the study design and Poisson regression approach (with covariates) had high power to detect small-scale changes in seal densities near Northstar if such changes had occurred. Seal densities near Northstar during spring were not significantly affected by industrial activities in 2000, 2001 and 2002. Other human activities, including Vibroseis surveys, that occurred in other portions of the study area during one or more survey years apparently did have detectable negative effects on seal densities. Also, various habitat, temporal, and weather factors had significant and strong effects on seal densities. This study shows that effects of the Northstar offshore oil development on the local distribution of basking ringed seals are no more than slight, and are small in comparison to the effects of natural environmental factors. An understanding of the latter effects is essential when assessing potential impacts of industrial activity on ringed seals. Sound Measurements, 2002 Open-Water Season

The primary objective of the acoustic measurement program at Northstar during the 2002 open- water season was to document underwater and airborne sounds during the first open-water season with oil production operations. Greeneridge Sciences Inc. made sound measurements during three days in Sep- tember 2002 at seven distances (¼ to 20 n.mi., or ½–37 km) north of Northstar Island. In addition, two near-island cabled hydrophones and (for backup) autonomous seafloor recorders, all located about ¼ n.mi. (420 m) off the north shore of Northstar, recorded underwater island sounds continuously for 31 days from 31 Aug. to 1 Oct. in 2002. Chapter 6, by S.B. Blackwell (Greeneridge), describes this study. Measurements at the seven distances showed that, in the absence of vessels, both underwater and in-air broadband levels were the lowest recorded in the three years for which these types of measurements have been made (2000–02). Background levels (lowest recorded values) were 87–90 dB re 1 µPa underwater and 37–40 dBA re 20 µPa in air at the more distant stations. Maximum levels reached 116 dB re 1 µPa underwater and 56 dBA re 20 µPa in air. While both drilling and oil production were taking place, underwater and airborne sounds reached background levels ~3.5 km (2.2 mi) from Northstar. This is comparable to values obtained in similar conditions in other studies. When the crew boat or other operating vessels were present, sound levels were higher, up to 128 dB re 1 µPa at 3.7 km. The cabled hydrophone showed that broadband (10–1000 Hz) levels recorded in 2002 spanned a narrower range than in 2001. The 95th percentile was lower in 2002 (117.2 dB re 1 µPa) than in 2001 (122.8 dB). The 5th percentile was higher in 2002 (94.8 dB) than in 2001 (87.8 dB). Median values were comparable in the two years: 103.0 dB in 2002 vs. 102.6 dB in 2001. Drilling and oil production, present in September 2002 but absent in September 2001, may have been responsible for the higher 5th percentile values in 2002. Vessels were an important but intermittent sound source around Northstar during both years, and many “spikes” (peaks) in broadband levels could be attributed to the crew boat and barge traffic. The ISI (Industrial Sound Index) is an index used to quantify the amount of industrial noise emanating from Northstar over specified time intervals. The index is used in analyzing how Northstar sounds affect the migration path of bowhead whales (see Chapters 7–11). The index is introduced in this chapter because it is a measure of Northstar sound. Five different methods were tested to determine which freq-

xvii Executive Summary uency bands should be used in defining the ISI for the period 31 Aug.–23 Sept. 2002, when bowhead migration was studied. One method provided equivocal results, but four methods supported use of the five one-third octave bands centered at 31.5–80 Hz to define the ISI in 2002. The times with high ISI values were generally times when vessels were active near Northstar. Acoustic Monitoring of Bowhead Whale Migration, Autumn 2002

During the bowhead whale migration in September 2002, Greeneridge Sciences, on behalf of BP, implemented an acoustic monitoring program north-northeast of Northstar Island. Chapter 7, by C.R. Greene Jr. (Greeneridge Sciences) et al., describes this effort. The primary purpose was to assess the effects of Northstar production activities, especially their underwater sounds, on the migration corridor and calling behavior of bowhead whales. This was to be done by comparing the offshore distances of calling bowheads at times with varying levels of industrial sound. If the closest bowhead whales detected by their calls tend to be farther offshore at times when Northstar sounds are stronger than average, this would be an indication some whales are affected by Northstar activities. The geographic scale of any such effect would provide evidence of the magnitude of the effect, if it occurs. This project in 2002 was a follow up to similar efforts in 2000 and 2001. The 2000 work demonstrated that the acoustic localization approach had promise but equipment problems limited the results. The 2001 results provided some evidence that the southern edge of the bowhead migration corridor was deflected offshore at times with high levels of underwater sound near Northstar, but additional data were needed to better characterize the effect and to address remaining uncertainties. The study in September 2002 occurred in much the same way as in 2001. This project required continuous measurements of sound levels near the island over time, and of the locations of calling bowhead whales over the same time. (1) To monitor sounds near the island, two cabled hydrophones and (for backup) two omnidirectional ASARs were installed ~400 m (¼ mi) north of the north edge of the island. These instruments provided a continuous record of underwater sounds near Northstar. (2) To detect and determine the source locations for whale calls in the nearshore part of the bowhead migration corridor, 11 “Directional Autonomous Seafloor Acoustic Recorders (DASARs) were installed in a grid pattern 6.5 to 22 km (4 to 13.7 mi) offshore of Northstar Island. These two types of acoustic data permit analyzing whale call locations in relation to near-island sound levels, the key project objective. Both the near-island hydrophones and the DASARs were installed at the same locations as used in 2001. For 2002, simultaneous data of the two types were available from 31 Aug. to 23 Sept., and cabled hydrophone data were in fact available until 1 Oct. The sounds received by cabled hydrophone in 2001 have been analyzed as broadband signals (10–1000 Hz) and as one-third octave band levels for the 31 Aug.–1 Oct. period. Compared with 2001, the minimum broadband level ~400 m from the island was about 10 dB higher in 2002, while the maximum levels (generally due to vessel activity near the island) were about 6 dB lower in 2002. The higher minimum level in 2002 may be attributed to a difference in sounds from the drilling and production activities of 2002 as compared with the construction activities on the island during 2001. However, in both years, the times with highest sound levels near Northstar were those when vessels were operating nearby. A total of 10,587 bowhead whale calls were recorded in ~24 days. This is about the same number of calls that were detected during 36 days in 2001. However, in 2002 a higher proportion of the bowhead calls were detected by two or more DASARs, thus providing the intersecting bearings necessary for

xviii Executive Summary localization. Also, bearing accuracy was improved in 2002, a higher proportion of the calls were detected by >2 DASARs, and DASARs operated more reliably. (Ten of 11 operated nominally until overturned by a storm on 23 Sept.) Overall, data quality was improved in 2002 compared with 2001. Chapter 7 describes the 2002 procedures and results. Subsequent chapters provide more specifics about the acoustic localization method (Chapter 8), statistical approach (Chapter 9), and results from 2001 and 2002 with regard to the key question of Northstar effects on migrating whales (Chapters 10, 11). DIFAR Sensors for Monitoring Bowhead Whale Migration

Chapter 8, by C.R. Greene Jr. (Greeneridge Sciences) et al., describes the acoustic methods developed and used to determine locations of calling bowhead whales in the southern part of their migration corridor near Northstar, with emphasis on the latest (2002) procedures. Some 2002 results are presented for illustrative purposes. This chapter has been submitted (Nov. 2003) for journal publication. The southern part of the bowhead migration corridor is exposed to sounds from BP’s activities at and near Northstar. BP needed to determine whether these sounds deflected migrating whales farther offshore. Bowheads call during fall migration, and an acoustical approach to localizing calling whales was selected in preference to aerial surveys. The specific technique chosen was to incorporate DIFAR (Directional Frequency and Recording) sonobuoy techniques into autonomous seafloor recorders, designing and building an array of 11 DASARs (Directional Autonomous Seafloor Acoustic Recorders). When two or (preferably) more DASARs detected the same call, the location of the whale was determined from the intersection of the bearings. The chapter summarized here describes the specialized acoustic methods used to determine the locations of the calling bowhead whales, and to show the types and precision of the data acquired. DIFAR uses an omnidirectional pressure hydrophone and two horizontal, orthogonal particle- velocity hydrophones having dipole patterns. The relative amplitudes of a sound signal at the two dipoles provide a measure of the angle to the source, while the relative phases of the dipole signals with respect to the omni signal resolve the ambiguities. The digital recording system was based on a Persistor Instru- ments single-board computer and a 30-GByte disk drive. An acoustic transponder provided a way to check the operational health of each deployed DASAR. Batteries were sufficient to record for 45 days. Each unit was installed with a tag line on the bottom between the unit and an anchor, both of whose positions were measured by GPS during deployment. No surface buoys were used because of hazards from drifting ice floes. The array of 11 DASARs was configured as two overlapping hexagons in which the unit-to-unit separation was 5 km (2.7 n.mi.). Calibration transmissions from a J-9 sound projector at GPS-measured locations within and around the array provided measures of the individual DASAR clock drift (for timing whale call arrivals) and reference axis direction (for relating measured whale call bearings to grid north). Standard errors of the bearing measurements to the J-9 projector at distances of 3–4 km (~2 n.mi.) was 2.22º without correcting for gain imbalance in the two directional sensors, and 1.35º with corrections. During 23 days in 2002, 10,587 bowhead calls were detected and 8383 were localized. Of those, 7922 were deemed sufficiently accurate to use in subsequent analyses, with 5837 of those being detected by three or more DASARs. Uncertainty in locations varied with distance and other factors; 90% confidence ellipses were calculated to have an average major axis length of 500–1000 m at distances 10 km from the array center.

xix Executive Summary

Statistical Approach for Assessing Displacement

In many environmental monitoring and impact studies, the animals closest to the source of a stimulus are potentially affected by the stimulus, while animals far from the source are not. In such cases, the primary interest can be in detecting and estimating the magnitude of stimulus effects on the lower tail of a distance-to-source distribution, where maximal effects might occur. Statistical procedures appropriate for this situation are not well developed. In Chapter 9, T.L. McDonald (WEST Inc.) and W.J. Rich- ardson (LGL) propose a suitable statistical method based on quantile regression. Chapter 9 has been submitted (Dec. 2003) for publication as a journal article. The proposed method estimates deflection of animal locations (or alternative variables that are subject to disturbance effects) in relation to levels of a time-varying stimulus that diminishes in intensity through space (such as sound). The method focuses on univariate distance distributions and estimates the relationship between a distribution’s lower quantiles and levels of the stimulus. To do this, quantile regression is applied. Quantile regression requires that animal locations and levels of the stimulus be recorded over a period of time, and that stimulus levels vary over this interval. Complications include overcoming lack of independence in responses, non-linearity of quantile relationships if there is no effect, and the fact that researchers may not know (a priori) how long levels of the stimulus need to be elevated before a response occurs and becomes large enough to be detectable. Approaches that can deal with these complications are described. Chapter 9 illustrates the application of the statistical method by analyzing the acoustic monitoring data on the 2001 bowhead migration past Northstar, a source of time-varying levels of underwater sound. The 5th quantile (=5th percentile) of the offshore distances of bowhead calls is examined in relation to the variable levels of sound emanating from the Northstar area. Although the primary purpose of this chapter is to describe the development of the statistical technique, analysis of the 2001 data identified an apparent offshore deflection of the closest bowhead whales at times when underwater noise from the Northstar area was strongest. Deflection of Bowhead Migration near Northstar?

The main objective of BP’s acoustic monitoring of bowhead whales during early autumn was to determine whether underwater sound from Northstar and its support vessels deflected the southern part of the bowhead migration corridor farther offshore at times when these sound levels were above average. Chapter 10, by W.J. Richardson (LGL), C.R. Greene Jr. (Greeneridge), and T.L. McDonald (WEST), describes the results from 2001 and 2002. The analysis of the 2001 data that is summarized in this chapter is the same analysis as was described in the preceding chapter concerning development of the statistical technique. Calling whales were localized for 36 days in 2001 and 23 days in 2002. In those periods, totals of 1121 and 2057 calls (respectively) were localized within a 20 × 26 km (12 × 16 mi.) area seaward of Northstar where comparable data could be obtained at all times. For those calls, quantile regression was used to determine the relationship between the 5th quantile distance from shore and an index of the level of underwater sound coming from Northstar Island and its support vessels. In both 2001 and 2002, the apparent southern edge of the migration corridor was slightly farther offshore at the noisiest times as compared with typical times, with ≥95% confidence (one-sided). In both years, the maximum displacement effect was evident when an industrial sound index (ISI) was averaged over ~70 min. When the 70-min ISI was high, the 5th quantile of the offshore distances was displaced off-

xx Executive Summary shore by ~2.3–3.3 km (1.4–2.1 mi) in 2001, and by ~2.2–4.7 km (1.4–2.9 mi) in 2002, as compared with that expected with no displacement. However, there was considerable uncertainty in the estimated displacement distances. The high sound levels occurring at times with evidence of displacement were mainly from vessels, not Northstar island itself. Estimated Numbers of Seals and Whales Potentially Affected, Nov. 2001 – Oct. 2002

A Letter of Authorization (LoA) issued by NMFS to BP authorized the “taking” of small numbers of seals and whales incidental to Northstar production activities. The numbers potentially affected during the ice- covered, break-up, and open-water periods in November 2001 through October 2002 have been estimated based on monitoring results. An estimated 110 ringed seals, 1 bearded seal, and probably no spotted seals were present near BP’s Northstar activities during the combined ice-covered, break-up, and open-water seasons of 2001–02. These figures are less than the corresponding “takes” authorized by the LoA issued by NMFS to BP, and most of these seals do not seem to have been negatively affected by Northstar activities. These totals exclude ~143 seals that dived in response to aircraft overflights during the spring aerial surveys; those incidents occurred over a much larger area and were separately authorized via a Scientific Research Permit issued by NMFS to LGL. The overall results suggest that any effects of Northstar production activities on seals were minor, short-term, and localized, with no consequences for the seal populations. There was no indication of any effects of the Northstar development on availability of seals to subsistence hunters during the 2002 ice-covered, break-up, or open-water periods. Acoustic localization data indicated that, during the late summer and early autumn of 2001 and 2002, a small number of bowhead whales in the southern part of the migration corridor (closest to Northstar) were affected by vessel or Northstar operations. This occurred during 4–5% of the time when levels of underwater sound from these activities were highest. At these times, most “Northstar sound” was from maneuvering vessels, not the island itself. The best (though quite imprecise) estimates of the number of bowheads that were apparently deflected offshore by ≥2 km (1.2 mi) were 13 bowheads in 2001 and 19 bowheads in 2002, i.e., <0.2% of the population. Actual numbers would vary relative to these imprecise estimates depending on the accuracy of assumptions and sampling variability. It is possible that the apparent deflection effect was, at least in part, attributable to a change in calling behavior rather than actual deflection, but in either case there was a change in the behavior of a small number of whales. The occurrence of a small “Northstar effect” on a low number of bowheads in each of 2001 and 2002 would be well within the provisions of the LoA. Based on the acoustic monitoring results, it is unlikely that the apparent displacement extended far enough east and northeast to affect whales being hunted near Cross Island. There was no specific information on numbers of gray or beluga whales (if any) that may have been close enough to Northstar to be disturbed by Northstar production and drilling operations in 2002. Given that gray whale sightings have been infrequent this far to the east in previous years, it is most likely that no gray whales were affected by Northstar in 2002. For belugas, the estimated numbers that might approach within an assumed 1–2 km (0.6–1.2 mi) disturbance radius are 10–20 in an average year.

xxi CHAPTER 1:

INTRODUCTION 1

Michael T. Williams, a W. John Richardson, b and Robert Rodrigues a

a LGL Alaska Research Associates, Inc. 1101 E. 76th Ave., Suite B, Anchorage, AK 99518 (907) 562-3339; [email protected]

b LGL Ltd., environmental research associates 22 Fisher St., POB 280, King City, Ont. L7B 1A6 (905) 833-1244; [email protected]

for

BP Exploration (Alaska) Inc. Dept. of Health, Safety & Environment 900 East Benson Blvd. P.O. 196612 Anchorage, AK 99519-6612

LGL Report TA2707-4

May 2003

1 Chapter 1 In: W.J. Richardson and M.T. Williams (eds.). 2003. Monitoring of industrial sounds, seals, and bowhead whales near BP’s Northstar Oil Development, Alaskan Beaufort Sea, 1999-2002. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences, Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 1-2 Monitoring at Northstar, 1999–2002

BACKGROUND

BP Exploration (Alaska) Inc. (BP) began constructing oil-production facilities for the Northstar Development during early 2000, and began producing crude oil from the Northstar Unit on 31 October 2001. The unit extends from 3.2 to 12.9 kilometers [km] (2 to 8 miles) offshore from Point Storkersen, northwest of the Prudhoe Bay industrial complex. Northstar is approximately 54 miles (87 km) northeast of Nuiqsut, the closest Native Alaskan (Inupiat) community (Fig. 1.1). The Northstar Development was built on the submerged remnants of Seal Island. Seal Island was an artificial island constructed in 1982, used for exploration drilling during the 1980s, and subsequently abandoned. The Northstar Development includes a gravel island for the main facilities and two pipelines connecting the island to the existing infrastructure in Prudhoe Bay. One pipeline transports crude oil to shore, and the other transports natural gas to the island for field injection. The facilities on the island include prefabricated modules for living quarters, utilities, and warehouse/shop. Also present are a drilling rig and facilities for waste grind and injection and for oil production. The production facilities include gas turbine engines to operate power generators and gas compressors. Transportation between the mainland and Northstar is primarily via ice roads in winter, vessels in summer, and helicopters during the transitional seasons. The Northstar Devel- opment is, to date, the only offshore oil production facility in the Beaufort Sea north of the barrier islands.

In 1998 and early 1999, BP’s plans were to reconstruct the gravel island by hauling gravel from the Kuparuk River Delta over ice roads to the old Seal Island location. However, island construction was delayed until the following year by regulatory questions not associated with seals or whales. “Northstar Island” was constructed and the associated pipelines were installed during early 2000. Other construction activities continued through mid-2001 with the installation of facilities on the island necessary for processing and handling oil from the reservoir. The production modules were brought on line and oil production began on 31 October 2001. Oil production at Northstar has continued to the present.

FIGURE 1.1. Location of the Northstar Development at Seal Island in the central Alaskan Beaufort Sea. Chapter 1: Introduction 1-3

This report describes BP’s Northstar activities, and the associated marine mammal and acoustic monitoring projects, during the November 2001 through October 2002 period. In order to provide a more comprehensive summary, considerable additional information is included about BP’s Northstar activities and some aspects of the monitoring work in previous years. This report is not intended to be a final report on all aspects of the Northstar monitoring work, as some work is ongoing. However, it does begin to compile the results from multiple years as well as fulfilling BP’s reporting obligations for Nov. 2001–Oct. 2002: • Chapter 1 of the report is a general introduction. • Chapter 2 contains a detailed description of BP’s activities during the construction and production phases of the Northstar project (1999 through late 2002), including specific details for the Nov. 2001–Oct. 2002 period. • Chapters 3–5 describe some aspects of the physical acoustics and seal studies conducted during the ice-covered season, including only those tasks active in Nov. 2001–Oct. 2002. • Chapters 6–10 describe the physical acoustics and whale studies during the open water season. • Chapter 11 evaluates the number of seals and whales potentially affected by Northstar activities during Nov. 2001–Oct. 2002.

BP BUSINESS RATIONALE FOR THE NORTHSTAR MONITORING PROGRAM

BP’s business rationale for the marine mammal and acoustical studies was driven both by corporate values and by regulatory requirements. BP corporate values support studies that objectively assess environmental effects that may result from BP operations. BP’s company goal of “no harm to the environment” drives the design and implementation of studies intended to understand and minimize the effects of BP operations. In addition, monitoring and reporting were required to satisfy BP’s commitments under Letters of Authorization (LoA) issued by the National Marine Fisheries Service (NMFS) to BP, and under a North Slope Borough (NSB) zoning ordinance. Six studies were done to satisfy the requirements for marine mammal monitoring and acoustical measurements as defined by NMFS in incidental take authorizations issued by NMFS to BP. Those authorizations also require BP to provide reports describing BP’s activities under the authorizations. The general requirements flow from the provisions of the Marine Mammal Protection Act (MMPA), which prohibits “taking” of whales and seals, including disturbance, unless a specific authorization for incidental “taking” has been issued by NMFS. BP requested and received the appropriate authorizations from NMFS. Similarly, the zoning ordinance issued by the NSB to BP for construction of Northstar included requirements for some types of monitoring done in 2002 (and earlier years) and described in this report. The marine mammal and acoustical monitoring described in this and prior reports meets the requirements of the authorizations and zoning ordinance.

CHRONOLOGY OF MARINE MAMMAL AUTHORIZATIONS

Well before the start of Northstar construction, BP, NMFS, and various other stakeholders anticipated that, during construction and subsequent operation of Northstar, some seals and whales could be disturbed in a manner that might be considered “taking” under the MMPA. Disturbance of seals and whales by certain operations of the offshore oil industry has been documented previously (Richardson et al. 1995). Consequently, in November 1998, BP requested that NMFS promulgate regulations allowing for the issuance of incidental take authorizations under section 101 (a) (5) (A) of the MMPA to allow “taking” of small numbers of seals and whales. 1-4 Monitoring at Northstar, 1999–2002

Williams and Richardson (2000, 2001) described the early stages of the permitting process for the Northstar Development project. The Regulations that had been requested by BP were issued by NMFS on 25 May 2000 (NMFS 2000). The initial Letter of Authorization (LoA) under those Regulations was issued by NMFS on 18 Sept. 2000 to authorize potential “taking” of whales and seals incidental to construction. A second LoA was issued by NMFS on 14 Dec. 2001 to authorize potential “taking” of whales and seals incidental to production and maintenance operations at Northstar through Nov. 2002 (NMFS 2001). The latest LoA was issued on 9 Dec. 2002 and is valid through 30 Nov. 2003. The present report is primarily concerned with the Nov. 2001 through Oct. 2002 period, but also includes related information from previous reporting periods.

MONITORING TASKS

Incidental take authorizations normally require that monitoring studies be conducted to assess the nature, amount, and geographic extent of any “taking” that did occur, and associated effects on subsistence hunting (Swartz and Hofman 1991; NMFS 1996). BP’s petition for regulations, supplemented by subsequent detailed monitoring plans, proposed to conduct several types of monitoring studies during both the ice-covered season and the open-water season. The study design and results of some or all the projects were reviewed by a peer-review and stakeholder group convened by NMFS at least once each year from 1998 through 2002. The monitoring studies were designed to assess whether effects on seals and whales were limited to behavioral disturbance and/or localized displacement, the nature and extent of such effects, whether effects on these mammals and their populations were negligible, and whether there were adverse effects on availability of seals and whales for subsistence. The studies were also designed to provide the data needed to estimate the numbers of seals and whales that might be affected by BP’s Northstar-related activities. Some of the key aspects of the monitoring were use of innovative, adaptive techniques, and use of statistically powerful study designs. During the ice-covered season, ringed seals were studied using four different techniques. One of these methods—intensive, replicated, fixed-wing aerial surveys of basking seals—was applied in six consecutive spring seasons, including three years prior to Northstar construction (1997–99) as well as three subsequent years (2000–02). Another of the techniques used for studying seals was the intensive application of specially trained dogs to find seal holes and lairs, and monitor their fate over the course of the winter, in relation to distance from industrial activities. This approach was applied during the winters of 1999–2000 and 2000–01. For the open-water season, BP proposed a novel acoustic localization method to monitor the bowhead migration. This approach was approved by NMFS and the stakeholders, developed in 2000, and applied successfully in 2001 and 2002; it is planned to continue in 2003. In addition to these and other biological studies, extensive physical acoustic measurements have been included. These acoustic data were needed as a basis for evaluating the results of seal and whale studies in the context of their exposure to Northstar sounds. Monitoring of marine mammals under the provisions of the NMFS authorization process began in 1999, and has continued during each successive year. BP’s aerial surveys of seals hauled out on the landfast ice in the central Alaskan Beaufort Sea began in 1997 (Miller et al. 1998) and continued in 1998 (Link et al. 1999) in anticipation of the subsequent development of Northstar and possible future development of Liberty. Those surveys were incorporated into the overall Northstar monitoring plan in 1999. Physical acoustics measurements began in 2000, the year when specific Northstar construction activities began. Previous Northstar-specific annual reports described BP’s Northstar activities and the associated marine mammal and acoustical monitoring during • the construction of ice roads in early 1999 (Richardson and Williams [eds.] 2000), Chapter 1: Introduction 1-5

• the construction of Northstar Island and the associated pipelines during 2000 (Richardson and Williams [eds.] 2001), and • the final stages of construction (and initial drilling) during parts of the November 2000 through October 2001 period (Richardson and Williams [eds.] 2002a). In addition, nine manuscripts that describe the results of the monitoring studies completed at Northstar to date are published or in press (Moulton et al. 2002b, 2003), or are in active preparation. This last category includes chapters 3, 5, 8 and 9 from the present report, which have been prepared in a format close to that of the anticipated journal articles. Additional journal manuscripts are planned as part of the ongoing Northstar studies. The present report provides corresponding information for the initial year of oil production (and continued drilling) from November 2001 through the open water season 2002. This report incorporates and supersedes two preliminary (“90-day”) reports previously submitted concerning BP’s activities and monitoring work during the ice-covered and open-water portions of the Nov. 2001–Oct. 2002 period (Richardson and Williams [eds.] 2002b; Richardson [ed.] 2003). The following paragraphs outline the main monitoring tasks undertaken during the period of Northstar construction and (in some cases) initial operations. Seven monitoring tasks were identified early in the planning process for the Northstar monitoring. Table 1.1 shows which monitoring tasks were active during each year of Northstar monitoring to date. Monitoring During Ice-Covered Seasons

1. Fixed-wing Aerial Surveys of Seals Systematic, intensive and replicated aerial surveys were used to study the distribution and abundance of seals around the locations of the Northstar and potential Liberty Developments. In particular, a key objective was to determine whether, and to what extent, oil development affected the local distribution and abundance of ringed seals. The surveys began in 1997, to provide pre-development “baseline” data (Miller et al. 1998). At the time, it was expected that construction of Northstar might begin in 1998 or 1999. The surveys continued, in a consistent way, each year from 1997 through 2002, encompassing three years of pre-development data (1997–99) and three years with construction and then oil production (2000–02). The surveys were conducted in late May and early June, the season when many ringed seals

TABLE 1.1. Marine mammal and acoustical studies conducted by BP for Northstar during 1999–2002.

Open Water Season 5.Sound measurements July, Aug., Sept. (Greener. Sciences) X X X 6.Island-based visual monitoring (LGL) X 7.Bowhead migration monitoring (Greener. Sciences & LGL) X b XX

a Aerial surveys also done in 1997–98. b Data on bowheads in autumn 2000 were incomplete and of limited use. 1-6 Monitoring at Northstar, 1999–2002

haul out on the surface of the ice to bask and molt. A fixed-wing aircraft was used to conduct closely spaced, replicated surveys of a study area that was ~75 km (47 mi) wide and extended from the shoreline to ~40 km (25 mi) offshore. Each year’s results have been described in an annual report, and the more recent reports have included an analysis of the data from 1997 to date concerning effects of Northstar on numbers of seals seen on the ice adjacent to winter activities (e.g., Moulton et al. 2002a). A journal paper on the effects of natural factors on the numbers of seals seen on the ice (based on the 1997–99 data) has been published (Moulton et al. 2002b). A preliminary paper on the influence of Northstar on local seal distribution, based on the 1997–2001 data, is forthcoming (Moulton et al. in press). The 2002 results are described in Chapter 4 of this report. Chapter 5 is a detailed analysis of the effects of Northstar on local seal distribution, based on all the data (1997–2002), and incorporating a statistical power analysis. Chapter 5 has been written ready for submission as a journal paper. This aerial survey project was planned to end after the 2002 and combined 1997–2002 data are analyzed and published. No additional surveys are required under the LoAs (NMFS 2001), and none are planned. 2. Monitoring of Seal Structures in the Sea Ice Adjacent to Northstar On-ice monitoring of seals along the Northstar ice roads was required during the ice-covered period of 1998–99. The Northstar ice road was surveyed for the presence of marine mammals (ringed seals and polar bears) in order to avoid potential harassment of marine mammals during ice-road construction and to amend construction activities, if necessary, to avoid injury or mortality to ringed seals. The technique approved for use that year, avalanche probes, was shown to be unreliable as a method for locating seal holes and lairs (“structures”) beneath the snow (Williams and Perham 2000). Although not specifically relevant to the NMFS incidental-take authorization, the area was also searched for polar bear dens using an airborne Forward Looking Infra-Red (FLIR) system. On-ice monitoring of seals along the Northstar ice road was again initiated in late 1999 amid concerns that the aerial survey project might not detect effects of the small scale suspected to occur in response to industrial activities at Northstar. In December 1999, trained dogs were used to detect ringed seal structures. After the winter and spring construction of Northstar, the fates of the structures found in December were determined. In addition, a second search with dogs was completed over a subsection of the original study area (Williams et al. 2001). In late 2000 and early 2001, trained dogs were again used to study the creation and abandonment of seal structures in the sea ice surrounding BP’s Northstar construction and drilling activities. This project was an expanded and modified version of the one conducted in 1999–2000, taking account of a statistical power analysis and a BP initiative to resolve potential industrial effects on ringed seal use of structures in the sea ice. In addition to dog-assisted searches for structures during early, mid- and late winter, temperature sensors and data loggers were placed in ringed seal structures to determine the dates of abandonment. This project was reported in Williams et al. (2002). A journal manuscript on the 2000–01 study is in preparation, planned for submission in 2003. The LoA issued to BP by NMFS for the late 2001 and 2002 period required on-ice searches for seal lairs only if BP’s activities moved onto previously undisturbed parts of the landfast ice after 1 March, i.e., during the season when ringed seals give birth to pups in lairs under the snow. In recent years, BP’s activities have not moved into such areas in late winter. Consequently, no on-ice searches were done during the winter of 2001–02, and this topic is not further addressed in this report. This project is not Chapter 1: Introduction 1-7 expected to be implemented at Northstar in future, given that BP does not anticipate, as part of its regular Northstar activities, to move onto previously undisturbed parts of the landfast ice during late winter. 3. Helicopter Survey of Seal Breathing Holes A pilot study to test the feasibility of using a helicopter survey to enumerate the numbers of active vs. inactive seal holes as a function of distance from infrastructure at Northstar was conducted during spring 1999. Unfortunately, ice, snow, and meltwater conditions in different parts of the study area were judged to be too variable to allow reliable comparison of the abandonment rate in the ice-road vs. control areas. Surveyors concluded that the temporal window to perform this type of survey was quite narrow and unpredictable; the proposed survey methodology suffered from being extremely dependent on ice and meteorological conditions for its success. After consultation with the peer/stakeholder group, this survey method was discontinued after the 1999 season. The results of the pilot study in 1999 are described in Lawson and Williams (2000). This study is not part of the current monitoring project, and it is not further addressed in this report. No further work of this type is planned. 4. Sound Measurements during Winter One of the monitoring requirements has been to document the levels, characteristics and range- dependence of sounds and vibrations from the main industrial activities occurring during construction and operation of Northstar. This was to be done during both winter and summer. During February, March, and May 2000, digital recording equipment was used to record underwater and airborne sounds, and iceborne vibrations, at locations near Northstar Island while the island was being reconstructed. Activities recorded included construction of ice roads, truck traffic along those ice roads, general construction activities at the newly rebuilt island, trenching through the ice and blasting across a barrier island during installation of pipelines, and both vibratory and impact driving of sheet-piles (Greene and McClennan 2000). A journal manuscript on that work is nearing completion. In subsequent winters, the requirement has been to document the levels, characteristics, and range- dependence of sounds and vibrations produced by any Northstar-related industrial activities whose acoustic characteristics were not adequately studied during previous winters. Recordings were made again in early March 2001 and in late February and early March 2002 to document sounds and vibrations during drilling and production operations. The 2001 data were reported by Blackwell and Greene (2002a). Chapter 3 of this report describes the combined 2001 and 2002 winter acoustic measurements; that chapter is expected to be submitted in 2003 for journal publication. No major new sound sources became active at Northstar during the winter of 2002–03. Therefore, no additional winter acoustic measurements were needed or obtained during the winter of 2002–03. Monitoring During Open-Water Seasons

5. Sound Measurements during Summer and Autumn In parallel with the winter acoustic measurements, boat-based underwater and airborne recordings were obtained during the 2000, 2001 and 2002 open water seasons to examine the levels, characteristics, and range-dependence of sounds from the industrial activities taking place on and near Northstar Island during summer/early autumn. The boat-based methods provided sequences of recordings at varying distances, as needed to assess range dependence. In addition, at fixed locations a few hundred meters north of the island, cabled hydrophones and Autonomous Seafloor Acoustic Recorders (ASARs) were deployed on the bottom to document variability in the sounds propagating into the water from Northstar 1-8 Monitoring at Northstar, 1999–2002 and associated vessels. The latter measurements also were used to analyze the relationship between variability in the sounds emanating from Northstar vs. locations where calling bowhead whales were detected (see “7”, below). The 2000 and 2001 results were described by Blackwell and Greene (2001) and (2002b), respectively. Chapter 6 describes the corresponding results from the 2002 open-water season. A journal manuscript describing sounds from Northstar industrial activities during summer and early fall is planned. Under the LoA, additional boat-based acoustic measurements are required after 2002 only if some potentially significant source of previously unstudied sound is introduced into the Northstar operations. Additional continuous near-island measurements will be needed in 2003 in support of the planned continuation (in 2003) of the acoustic monitoring of bowhead calls (see “7”, below). 6. Island-based Visual Monitoring during Spring Break-up, 2000 Marine mammal observers were stationed at Northstar Island on selected dates during June and July 2000 to document reactions of seals near the island to impact driving of conductor and insulator pipes, and to implement mitigation measures if necessary. Impact pile driving was expected to produce stronger sounds than any other industrial activities occurring at and near the island. There was concern that some seals or whales might receive sound levels exceeding 190 or 180 dB re 1 µPa (rms), respectively (NMFS 2000). These are the levels above which NMFS is concerned that there might be some physical effects on the auditory or other systems. The results of the acoustic and marine-mammal observations were described by Blackwell and Greene (2001) and Lawson and Williams (2001), respectively. The sounds measured in the water adjacent to the island (and at greater distances) were well below the levels of concern established by NMFS. Thus, there was no requirement under the LoA to continue this type of monitoring after July 2000 (NMFS 2001). No other sound source producing levels as high as those of the impact pile driver has operated subsequently at Northstar, nor is any such activity planned for the future. Hence, this monitoring task was not continued after 2000, and it is not further described in this report. A journal manuscript concerning the pile-driving sounds and associated seal observations is nearing completion (Blackwell et al. in prep.). It is expected to be submitted in 2003 for journal publication. 7. Acoustic Monitoring of Migrating Bowhead Whales One of the key components of the monitoring required by both NMFS and the NSB has been monitoring of the autumn bowhead migration past Northstar. The main concern is the possibility that migrating bowheads might be deflected offshore when passing Northstar as a result of exposure to underwater sounds from Northstar activities. A novel acoustic localization approach for monitoring the bowhead migration was proposed by BP, approved by NMFS and the stakeholders, and first implemented in 2000. This approach is based on detecting bowhead calls, estimating the locations of the calling whales, and determining whether those locations vary as a function of the underwater sounds originating at and near Northstar. As compared with a more traditional aerial survey approach, the acoustic method was expected to be more sensitive (and more cost-effective) for detecting an effect of Northstar. In addition to using an innovative acoustic data-collection system, this monitoring approach required a specialized statistical analysis approach. The analysis had to be designed to identify and quantify any deflection of the whales traveling closest to shore (in contrast with the usual approach of testing for a change in “average” values). Statistical power analyses conducted in advance of the fieldwork suggested that the proposed acoustic and statistical methods could be effective, and that aerial surveys could not be effective (McDonald 2001; Nations and McDonald 2001). Chapter 1: Introduction 1-9

The acoustic technique was implemented by designing new sound recording devices termed Direc- tional Autonomous Seafloor Acoustic Recorders (DASARs). Each one of these can determine the direction from which a sound is received. The source of a sound that is received by two or more DASARs can then be estimated by triangulation. DASARs have been deployed at a grid of 10 locations northeast of Northstar during the autumns of 2000, 2001 and 2002 to detect calling bowhead whales in the southern part of the bowhead migration corridor, where any deflection was expected to occur. The DASARs have been improved each year since 2000, as have the analytical methods for extracting whale call data and for calculating more precise locations of calling whales. The 2000 deployment of the DASARs provided information of value for improving the equipment and procedures, but technical problems resulted in imprecise localizations and an inadequate record of sounds emanating from Northstar (Greene et al. 2001). These problems were largely solved in 2001 through partial redesign of some equipment, additional testing, and deployment of multiple backup instruments. The necessary types of call localization and Northstar sound data were collected from 29 Aug. to 3 Oct. 2001 (Greene et al. 2002). The study was repeated in 2002, with some further improvements, and the required types of data were collected from 31 Aug. to 23 Sept. 2002, when a storm disrupted the equipment. Despite the early termination of the field season in 2002, results from 2002 were enhanced by better equipment reliability and associated improvements in the precision of the estimated locations of calling whales. The sample size for 2002 was in fact higher than for 2001. The specific 2002 results are described in Chapter 7. Chapters 8–10, which are intended to form the basis for journal publications, draw together the combined 2001 and 2002 acoustic monitoring results concerning DASAR development and call localization methods (Chapter 8), statistical approach (Chapter 9), and Northstar effect (if any) on the migration corridor (Chapter 10). Collectively, Chapters 7–10 show that the acoustic localization method is suitable for the intended purposes. The same acoustic localization method is scheduled to be applied again in 2003. That will increase the available sample size and will allow for more precise quantification of the scale of any effects.

OTHER COMPONENTS OF THIS REPORT

The majority of this report (Chapters 3–10) is a description of the marine mammal and acoustic monitoring done during 2001–02 and (in part) during previous years. However, two chapters respond to other requirements of the Letter of Authorization issued by NMFS: Chapter 2 is a description of BP’s activities. Besides including the required description of activities during the Nov. 2001 to Oct. 2002 period, Chapter 2 also includes considerable information about the earlier stages of Northstar construction. This integrated account of BP’s activities at Northstar is intended to provide the background necessary for later chapters that describe monitoring work during some or all of these years. Chapter 11 is a description of the numbers of whales and seals that might have been affected by Northstar activities during the Nov. 2001 to Oct. 2002 period. This chapter is based on the monitoring results presented in most preceding chapters.

ACKNOWLEDGEMENTS

The studies described in this report were funded by BP Exploration (Alaska) Inc. We thank Dr. Bill Streever, Allison Erickson, and Wilson Cullor of BP’s Environmental Studies Program for their specific support of the 2002 monitoring work, and Dr. Ray Jakubczak for overall support of the Northstar 1-10 Monitoring at Northstar, 1999–2002 monitoring efforts. Dr. Bill Streever of BP and Gillian O’Doherty of LGL reviewed earlier versions of this chapter. We thank everyone who contributed each of the Northstar seal, whale, and acoustics studies, as listed in the Acknowledgements sections of subsequent chapters and of the previous earlier reports. Also, we thank participants in the peer/stakeholder groups convened each year by the National Marine Fisheries Service. Those groups discussed the plans for and results of the monitoring studies. Particip- ants in this group have varied from meeting to meeting, but included representatives of NMFS (R. Ang- liss, Dr. D. DeMaster, K. Hollingshead, and others), the Alaska Eskimo Whaling Commission (M. Ahmaogak, J. Lefevre, T. Napageak), the North Slope Borough (Mayor G. Ahmaogak, Dr. T. Albert, J.C. George, Dr. T. O’Hara), and the Minerals Management Service (S. Treacy, F. Wendling), plus a variety of individual researchers with expertise on topics related to some of the work described in this report.

LITERATURE CITED

Blackwell, S.B. and C.R. Greene Jr. 2001. Sound measurements, 2000 break-up and open-water seasons. p. 7-1 to 7-55 In: W.J. Richardson and M.T. Williams (eds., 2001, q.v.). LGL Rep. TA2429-2. Blackwell, S.B. and C.R. Greene Jr. 2002a. Sound and vibration measurements during winter drilling at Northstar in early 2001. p. 3-1 to 3-25 In: W.J. Richardson and M.T. Williams (eds., 2002, q.v.). LGL Rep. TA2571-2. Blackwell, S.B. and C.R. Greene Jr. 2002b. Sound measurements, 2001 open-water season. p. 7-1 to 7-39 In: W.J. Richardson and M.T. Williams (eds., 2002, q.v.). LGL Rep. TA2572-2. Blackwell, S.B., J.W. Lawson and M.T. Williams. In Prep. Noise from impact pipe driving and construction of an oil production island: tolerance by ringed seals, Phoca hispida. Greene, C.R., Jr., and M.W. McLennan. 2000. Sound and vibration measurements during Northstar construction in early 2000. p. 4-1 to 4-31 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of ringed seals and sounds during construction of BP's Northstar oil development, Alaskan Beaufort Sea, winter and spring 1999-2000: 90-day report. LGL Rep. TA2426-1. Rep. from LGL Ltd., King City, Ont., and LGL Alaska Res. Assoc. Inc., Anchorage, AK, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 107 p. Greene, C.R., Jr., M.W. McLennan, T.L. McDonald and W.J. Richardson. 2001. Acoustic monitoring of bowhead whale migration, autumn 2000. p. 9-1 to 9-57 In: W.J. Richardson and M.T. Williams (eds., 2001, q.v.). LGL Rep. TA2431-2. Greene, C.R., Jr., M.W. McLennan, T.L. McDonald and W.J. Richardson. 2002. Acoustic monitoring of bowhead whale migration, autumn 2001. p. 8-1 to 8-79 In: W.J. Richardson and M.T. Williams (eds., 2002, q.v.). LGL Rep. TA2573-2. Lawson, J.W. and M.T. Williams. 2000. Helicopter survey of seal breathing holes, 1999. p. 4-1 to 4-12 In: W.J. Richardson and M.T. Williams (eds., 2000, q.v.). LGL Rep. TA2350-2. Lawson, J.W. and M.T. Williams. 2001. Island-based visual monitoring during spring break-up, 2000. p. 8-1 to 8-14 In: W.J. Richardson and M.T. Williams (eds., 2001, q.v.). LGL Rep. TA2430-2. Link, M.R., T.L. Olson and M.T. Williams. 1999. Ringed seal distribution and abundance near potential oil development sites in the central Alaskan Beaufort Sea, spring 1998. LGL Rep. P-430. Rep. from LGL Alaska Res. Assoc. Inc., Anchorage, AK, for BP Explor. (Alaska) Inc., Anchorage, AK. 58 p. McDonald, T.L. 2001. Statistical power of aerial surveys of bowhead whales. p. C-1 to C-11 In: W.J. Richardson and M.T. Williams (eds., 2001, q.v.—Appendix C). Chapter 1: Introduction 1-11

Miller, G.W., R.E. Elliot and W.J Richardson. 1998. Ringed seal distribution and abundance near potential oil development sites in the central Alaskan Beaufort Sea, spring 1997. LGL Rep. TA2160-3. Rep. from LGL Ltd., King City, Ont., for BP Explor. (Alaska) Inc., Anchorage, AK. 43 p. Moulton, V.D., R.E. Elliott, W.J. Richardson and T.L. McDonald. 2002a. Fixed-wing aerial surveys of seals near BP's Northstar and Liberty sites in 2001 (and 1997-2001 combined). p. 5-1 to 5-60 In: W.J. Richardson and M.T. Williams (eds., 2002, q.v.). LGL Rep. TA2570-3. Moulton, V.D., W.J. Richardson, T.L. McDonald, R.E. Elliott and M.T. Williams. 2002b. Factors influencing local abundance and haulout behaviour of ringed seals (Phoca hispida) on landfast ice of the Alaskan Beaufort Sea. Can. J. Zool. 80(11):1900-1917. Moulton, V.D., W.J. Richardson, M.T. Williams and S.B. Blackwell. 2003. Ringed seal densities and noise near an icebound artificial island with construction and drilling. Acoust. Res. Let. Online. in press. Nations, C. and T.L. McDonald. 2001. Statistical power of acoustic surveys for calling bowhead whales. p. D-1 to D-9 In: W.J. Richardson and M.T. Williams (eds., 2001, q.v.—Appendix D). NMFS. 1996. Small takes of marine mammals; harassment takings incidental to specified activities in arctic waters; regulation consolidation. Fed. Regist. 61(70, 10 Apr.):15884-15891. NMFS. 2000. Taking marine mammals incidental to construction and operation of offshore oil and gas facilities in the Beaufort Sea/Final Rule. Fed. Regist. 65(102, 25 May):34014-34032. NMFS. 2001. Taking and importing marine mammals; taking marine mammals incidental to construction and operation of offshore oil and gas facilities in the Beaufort Sea. Fed. Regist. 66(246, 21 Dec.):65923-65935. Richardson, W.J. (ed.). 2003. Monitoring of industrial sounds and whale calls during operation of BP’s Northstar Oil Development, Alaskan Beaufort Sea, summer and autumn 2002: 90-day report. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Springs, MD. 66 p. Superseded by the present report. Richardson, W.J. and M.T. Williams (eds.). 2000. Monitoring of ringed seals during construction of ice roads for BP's Northstar oil development, Alaskan Beaufort Sea, 1999. [Final, Dec. 2000.] Rep. from LGL Ltd., King City, Ont., and LGL Alaska Res. Assoc., Anchorage, AK, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 153 p. Richardson, W.J. and M.T. Williams (eds.) 2001. Monitoring of industrial sounds, seals, and whale calls during construction of BP's Northstar oil development, Alaskan Beaufort Sea, 2000. [April 2001 ed.] Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 316 p. Richardson, W.J. and M.T. Williams (eds.). 2002a. Monitoring of industrial sounds, seals, and whale calls during construction of BP’s Northstar Oil Development, Alaskan Beaufort Sea, 2001. [Oct. 2002 ed.]. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 309 p. Richardson, W.J. and M.T. Williams (eds.). 2002b. Monitoring of ringed seals, sounds and vibrations during operation of BP’s Northstar Oil Development, Alaskan Beaufort Sea, winter and spring 2001-2002: 90-day report. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 86 p. Superseded by the present report. Richardson, W.J., C.R. Greene Jr., C.I. Malme and D.H. Thomson. 1995. Marine mammals and noise. Academic Press, San Diego, CA. 576 p. 1-12 Monitoring at Northstar, 1999–2002

Swartz, S.L. and R.J. Hofman. 1991. Marine mammal and habitat monitoring: requirements; principles; needs; and approaches. U.S. Mar. Mamm. Commis., Washington, DC. 16 p. NTIS PB91-215046. Williams, M.T. and C. Perham. 2000. On-ice monitoring of seals along the Northstar ice roads, 1999. p. 2-1 to 2-10 In: W.J. Richardson and M.T. Williams (eds., 2000, q.v.). LGL Rep. P446-3. Williams, M.T. and W.J. Richardson. 2000. Introduction [1999]. p. 1-1 to 1-9 In: W.J. Richardson and M.T. Williams (eds., 2000, q.v.) LGL Rep. TA2348-1. Williams, M.T. and W.J. Richardson. 2001. Introduction [2000]. p. 1-1 to 1-17 In: W.J. Richardson and M.T. Williams (eds., 2001a, q.v.) LGL Rep. TA2432-5. Williams, M.T., J.A. Coltrane and C.J. Perham. 2001. On-ice location of ringed seal structures near Northstar, December 1999 and May 2000. p. 4-1 to 4-22 In: W.J. Richardson and M.T. Williams (eds., 2001, q.v.). LGL Rep. P485-2. Williams, M.T., T.G. Smith and C.J. Perham. 2002. Ringed seal structures in sea ice near Northstar, winter and spring of 2000-2001. p. 4-1 to 4-33 In: W.J. Richardson and M.T. Williams (eds., 2002, q.v.). LGL Rep. P557-2. CHAPTER 2:

1 BP’S ACTIVITIES AT NORTHSTAR, 1999–2002

Michael T. Williams and Robert Rodrigues

LGL Alaska Research Associates Inc. 1101 E. 76th Ave, Suite B, Anchorage, AK 99518 (907) 562-3339; [email protected]

for

BP Exploration (Alaska) Inc. Dept of Health, Safety & Environment 900 East Benson Blvd., POB 196612 Anchorage, AK 99519-6612

LGL Report TA2707-5

May 2003

1 Chapter 2 In: W.J. Richardson and M.T. Williams (eds.). 2003. Monitoring of industrial sounds, seals, and bowhead whales near BP’s Northstar oil development, Alaskan Beaufort Sea, 1999–2002. Rep. from LGL Ltd., King City, Ont., for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 2-2 Monitoring at Northstar, 1999-2002

ABSTRACT

BP’s Northstar Oil Development is centered at Northstar Island, located in the Beaufort Sea ~6 miles (10 km) offshore of Point Storkerson along the north coast of Alaska. BP began to construct the gravel island in early 2000. Island construction and installation of facilities were completed by late 2001, when BP began producing crude oil from Northstar. As one condition of the authorizations received from the National Marine Fisheries Service concerning potential “taking” of whales and seals, BP must report on the activities conducted in support of Northstar development. For the purposes of describing BP’s Northstar activities, the operational year is divided into two periods: the ice-covered period from 1 Nov. through 15 June, and the open-water period from 16 June through 31 Oct. Ice roads are constructed each year during the ice-covered period to facilitate transportation of personnel and equipment to and from the island. Equipment used during construction of ice roads includes powered augers and pumpers mounted on rolligon-type off-road vehicles, snow blowers, front-end loaders, bulldozers, dump trucks, and tanker trucks. After ice roads are completed, standard pick-ups, SUVs, and buses are used to transport personnel, and tractor trailers are used for equipment and supplies. During the open-water period crew vessels and barges are used to transport personnel and re-supply the island. Helicopters are used during transition periods (freeze-up and break-up), and intermittently at other times when ice and water conditions do not permit use of land-based vehicles or vessels. An ice road was constructed during the ice-covered period in 1998–99 but in April 1999 construction was postponed due to regulatory problems, and the ice road was abandoned. Ice roads were re-built during the 1999–2000 ice-covered period, and used to haul gravel for island construction (Feb.–Apr. 2000) and to install subsea pipelines (Mar.–May 2000). Island construction continued during the 2000 open-water period; equipment used included cranes, pile drivers, and various types of heavy equipment. During the 2000 open- water period, three sealift barges delivered modules for living quarters, utilities, pipe racks, and a drilling rig. During the 2000–01 ice-covered period, construction and equipment-installation continued, and the first wells were drilled. During August 2001, three sealift barges delivered processor modules, gas-turbine generators and compressors, and other equipment to the island. This equipment was installed by late autumn of 2001, and the gas-turbine powered generators replaced diesel generators as the primary source of power. Oil production began on 31 Oct. 2001 and has continued to the present. Production operations involve use of up to three generators and two high-pressure compressors powered by gas-turbine engines, along with various other types of equipment. Drilling operations have continued for most of the November 2001 through October 2002 period, although drilling during the open water season was limited to depths above the oil reservoir. Various testing, training, monitoring and inspection activities, including spill response activities, occur on an intermittent basis.

INTRODUCTION

BP Exploration (Alaska) Inc. (BP) is producing crude oil from an artificial gravel island (Northstar Island) rebuilt in 2000 on the submerged remnants of Seal Island. Seal Island was constructed in 1982, used for exploration drilling during the 1980s, and subsequently abandoned. Northstar Island includes a gravel island work surface to support the main facilities required for drilling and oil production, and two pipelines connecting the island to the existing infrastructure in Prudhoe Bay. One pipeline transports crude oil to shore, and the other transports natural gas to the island for field injection and use in power generation. The facilities on the island include prefabricated modules for living quarters, oil production, utilities, warehouse/shop, waste grind and injection, and gas flaring. There is also a drilling rig. Chapter 2: Description of BPs Activities 2-3

Before construction and oil production began, BP requested authorizations from the National Marine Fisheries Service (NMFS) to allow incidental taking of small numbers of whales and seals. The authorizations issued by NMFS required BP to provide summaries of the dates and locations of BP’s activities associated with the Northstar project. This chapter summarizes BP’s activities during the ice- covered and open-water periods from 1999 through October 2002, including somewhat more details for the November 2001 though October 2002 period. All of this information has been described previously in a series of annual and 90-day reports (see Williams and Richardson 2000; Williams and Perham 2001a,b; Perham and Williams 2002a,b; Perham 2002; Williams and Rodrigues 2003). The present chapter brings this information together in one document as background for the descriptions, in subsequent chapters, of sounds in relation to Northstar activities, and the responses (or lack thereof) by nearby marine mammals. Overall Sequence of Northstar Development

Subsequent sections of this chapter provide details concerning the construction and initial operation of the Northstar Development. However, an introductory summary may be helpful in orienting the reader: • Activities began during the 1998–99 ice-covered period with construction of an ice road from a gravel mine in the Kuparuk River delta to the site of the proposed development at Seal Island. Because of regulatory problems, this ice road was abandoned before it could be used to haul gravel for construction of Northstar Island. • During the winter of 1999–2000, the ice roads to the island were re-constructed and gravel was hauled by truck to the island site. A second ice road was constructed from the shoreline near Point McIntyre to the Northstar site and used for installation of the subsea pipelines. • During the 2000 open-water period, construction activities on Northstar Island continued, and various modules and a drilling rig were delivered by sealift. • During the 2000–01 ice-covered season, an ice road was constructed from West Dock to Northstar, construction and installation of facilities continued, and the first wells were drilled. • During the 2001 open-water period, processor modules, a warehouse, gas-turbine compressors, and other equipment were delivered via sealift and installed on the island, and the power source for the island was changed from diesel to gas-turbine powered generators. Drilling was suspended during this open-water season. • Oil production began on 31 Oct. 2001, and continued through the following ice-covered season and the 2002 open water season. Well drilling also occurred on most days during these periods, although drilling during the 2002 open-water season was limited to depths above the oil reservoir. Subsequent sections provide more details concerning all of these activities. Seasonal Transportation Needs

For the purpose of describing BP’s activities consecutively during construction and operation of the Northstar project, the operational year is divided into two periods: the ice-covered period from 1 Nov. through 15 June, and the open water period from 16 June through 31 Oct. During each ice-covered period since the winter of 1998–99, BP has constructed and used ice roads. These ice roads provide access to Northstar/Seal Island by land-based vehicles and equipment, thus facilitating construction operations and transportation of personnel. During the open-water periods since 2000, crew vessels and barges have been used to transport 2-4 Monitoring at Northstar, 1999-2002 personnel, facilities for personnel and oil production, and to re-supply the island. Helicopters are used during the transition seasons, when ice and water conditions do not permit the use of either land-based equipment or vessels. Because the transportation methods were generally similar from year to year, these methods are described here, prior to the subsequent year-by-year account of BP’s activities. Ice Roads In order to use standard land-based equipment for construction and operation of the Northstar project during the ice-covered period, thickened ice roads have been constructed each winter since 1998– 99 to provide a stable surface for this equipment. The number, locations and uses of the ice roads have varied among years depending on the particular needs for each year. However, the timing and methods used for ice road construction were similar from one year to the next. Ice road construction began in November or December when surveying crews established the alignment of the roads using reflective delineators along the centerline. Snow machines, pickup trucks fitted with mat-tracks, rolligons, and Hägglunds tracked vehicles (Fig. 2.1) were used to travel on the sea ice prior to the completion of the ice roads. Floating ice must be artificially thickened in order to support heavy equipment safely. Thickening of the ice road was accomplished by flooding the road with sea water and allowing the water to freeze. Holes were drilled through the ice and pump units were used to flood the routes with seawater. Four to six Blue Bird rolligon-type vehicles, each with a powered auger and pump (Fig. 2.2), worked together as a unit along one side of the road corridor. The pumpers flooded a section of the road surface, whereupon they moved to a non-adjacent area while the first area was allowed to freeze. Flooding was repeated many times, with different areas generally being flooded in the same sequence, until each area was the appropriate thickness for the planned transportation or construction activities. Once the entire length of the ice road was thickened adequately, tanker trucks hauled fresh water to cap and finish the road. The fresh water was obtained from permitted sources, which were accessed either by ice roads or via the existing gravel roads in the Prudhoe Bay oil field. Once adequate ice thickness was achieved, accumulated snow was cleared and roads were maintained using small snow blowers, front-end loaders with a snow blower attachment, and D-3 and D-4 Caterpillar bulldozers. Caterpillar 14G and 16H graders were used to scrape and maintain the ice roads. Bottom-fast ice required only grading although in some cases ice was thickened to ensure that it was bottom-fast. Transition and Open-Water Seasons During spring break-up when ice roads could no longer be used by standard vehicles, Hägglunds vehicles (Fig. 2.1) were used for transportation to and from the island. Typically, the Hägglunds were used for one to two weeks until conditions on the ice worsened. Then Bell 212 helicopters (Fig. 2.3), generally two of them, were the primary method used to transport crews and materials to and from Northstar. Round-trip helicopter flights began in late May or early June, traveling from the West Dock Base of Operations (WDBO) or Deadhorse Airport to Northstar, depending on weather and helicopter availability (Fig. 2.4). All personnel on Northstar were exchanged approximately every 12 hours. Crew changeover took ~2–3 hours, and typically required up to 6 round trips per shift change during peak construction periods. One-way flight time to Northstar was ~15 min from WDBO and 30 min from Deadhorse airport. Helicopter flights became less frequent after break-up, at which time a crew boat began to be used for crew changes. Chapter 2: Description of BPs Activities 2-5

FIGURE 2.1. Photograph of a Hägglunds tracked vehicle and personnel carrier. Powered by a turbo- charged diesel engine capable of 143 hp at 4600 rpm; total weight of the two cars is ~4536 kg (10,000 lb).

FIGURE 2.2. Blue Bird rolligon-type vehicle with 12” ice auger (stowed) and 10.2” pump auger (in use). 2-6 Monitoring at Northstar, 1999-2002

FIGURE 2.3. Bell 212 helicopter used for transportation to and from Northstar especially during break-up and freeze-up, and to a lesser degree during the open-water season.

During the open water season, crew boats (Fig. 2.5) and barges became the primary method to transport crews and materials to and from Northstar, but in some years helicopter flights continued in summer on a less frequent basis. The vessels usually began operating by mid-July between Dock 2 at West Dock and Northstar. The standard route was similar to that used by the helicopters (Fig. 2.4). Vessel operations were re-routed when ice was present along the standard route. Transits typically required about 40–45 minutes, one-way. Typically, seven to eight round trips by the crew vessels were scheduled daily during summer and autumn operations. Crew boat and barge operations were usually terminated in October as sea ice formed. Thereafter the Bell 212 helicopters were used to transport crews and supplies until the sea ice became stable enough for transportation by Hägglunds. Helicopter and vessel routes were negotiated among the U.S. Fish and Wildlife Service (FWS), NMFS, and BP to minimize impacts on waterfowl and marine mammals. The Letter of Authorization (LoA) issued by NMFS stated that helicopter flights to support Northstar must be limited to a corridor from Northstar Island to the mainland and, except when taking off, landing or limited by weather, must maintain a minimum altitude of 305 m (1000 ft). During poor weather or emergency conditions, pilots followed FAA altitude regulations and BP safety policy. Chapter 2: Description of BPs Activities 2-7

FIGURE 2.4. Route flown by helicopters (purple line) between Deadhorse airport, the West Dock base of operations (WDBO), and the Northstar Development at Seal Island. After break-up the crew vessel and barges traveled between the island and West Dock along a similar route. 2-8 Monitoring at Northstar, 1999-2002

FIGURE 2.5. Crew-boat Hawk used for transportation between West Dock and Northstar during the open- water season. Length 18.7 m (61.5 ft); dual 200 hp Detroit diesel engines; 4-bladed propellers; average speed ~12 knots.

INITIAL CONSTRUCTION

Ice Road Construction, Early 1999

BP’s original plans called for re-constructing Seal Island during early 1999, and then building the pipelines to shore during the following winter of 1999–2000. Ice roads to Seal Island were built in early 1999, but construction did not go ahead that winter because of delays in the EIS schedule. The key delay resulted from the U.S. Fish and Wildlife Service’s elevation (to higher authority) of the decision to issue the required Section 404 permit. Two ice roads intended to support the construction of Northstar were surveyed in December 1998 and built between January and early April 1999 (Fig. 2.6). The main ice road was built from the Kuparuk River gravel mine site to the submerged location of Seal Island. An alternate road was built parallel to the northern portion of the main road, starting at a planned gravel reload area ~7 km (4.3 mi) southwest of Seal Island and extending northeast to the island. The alternate road was built on the west side of the main road. The centerline of the alternate road was ~366 m (1200 ft) west of the centerline of the main road. The alternate road was built to allow uninterrupted gravel hauling to Seal Island during maintenance of the main road. Construction of the ice roads was a single-shift, seven days per week operation. The crews were housed in the Prudhoe Bay area and bused to their work sites. The main ice road was ~91.5 m (300 ft) wide and was completed by early April 1999. However, final fresh water capping of the alternate road had not been completed when the decision was made to postpone construction of the Northstar production island for a year. From 8 April until 4 June 1999, BP personnel made weekly trips in a pickup truck along the entire main ice road to examine the movement of the ice roads and extent of deterioration of the ice as the summer approached. The deterioration of the ice was estimated by using a gas-driven auger to drill holes Chapter 2: Description of BPs Activities 2-9

FIGURE 2.6. Map of the ice roads constructed by BP in January–April 1999 in support of the planned re-construction of Seal Island. Island re-construction did not go ahead until early 2000. in the ice road at regular intervals to measure the thickness of the ice. No further activities were conducted at the Northstar site until ice-road construction began again in fall 1999. Ice Road Construction, Winter 1999-2000

Ice roads were again constructed during the winter of 1999–2000. These were positioned to allow re-construction of a gravel island at the location of the old Seal Island, and installation of pipelines between Northstar Island and shore. This required two ice road complexes (Fig. 2.7). (1) The ice road complex for gravel hauling was built from the Kuparuk River gravel mine site to the submerged remnants of Seal Island. An alternate road similar to the alternate road built in 1999 was constructed parallel to the northern portion of the main road. (2) The other ice road complex was built along the 533 m (1750 ft) wide pipeline right-of-way from West Dock to Seal Island. Surveying for two offshore ice road complexes began in November 1999 and the roads were built from early December 1999 to mid-March 2000. Unlike the previous year, the ice roads were constructed during double-shifts (each 12 hours long) on a seven-days-per-week schedule. In addition, an ice road from West Dock to the Kuparuk River mine site was built parallel to the coast (Fig. 2.7) prior to activities further offshore. This road was surveyed and flooded in November, and completed in December. 2-10 Monitoring at Northstar, 1999-2002

FIGURE 2.7. Northstar construction area northwest of Prudhoe Bay, Alaska, showing ice roads and ice dump areas established during the winter of 1999–2000. Depth soundings are in feet (3.3 ft = 1 m). Chapter 2: Description of BPs Activities 2-11

The ice road to be used for gravel hauling from the Kuparuk River Delta to the Seal Island site was completed on 13 Feb. 2000. The ice road for pipeline construction and installation was completed on 14 March 2000. By the latter date, the natural ice thickness was at least 1.8 m (6 ft) and the ice roads were thickened to at least 9 ft (2.8 m). Access roads between the two road complexes were cleared of snow (but not flooded) to provide transit routes around newly flooded areas of both ice road complexes. Island and Pipeline Construction, Early-Mid 2000

During this initial winter of construction, trucks hauled a large quantity of gravel from the mainland across the landfast ice to the old Seal Island site in order to rebuild the island and make it suitable as a platform for the Northstar oil production operations. Also, late in the winter, two 10-inch pipelines were buried below the seafloor between the island and the mainland, and additional construction work occurred on the newly rebuilt island. During January 2000, the accumulated rubble ice over the eroded remnants of the old Seal Island was removed and transported to the ice dump north of the island (Fig. 2.7). A hole was cut in the ice above the location where Northstar was to be built to allow gravel dumping on top of the remnants of Seal Island. A Ditchwitch R100 was used to cut the ice, and ice blocks were hauled to the ice-dump by Kenworth Maxihauls. As the gravel accumulated above the surface, the hole was enlarged and additional ice was removed and placed in the ice dump. A gravel mine in the Kuparuk Delta was opened on 6 Feb. 2000, and this served as the source of gravel for island re-construction. Gravel was hauled from the mine to the reload area by Euclid B-70 dump trucks (capacity 70 tons), and from the reload area to the Northstar site in Kenworth Maxihauls (capacity 30 cu. yd.). Gravel hauling to the reload area was initiated on 8 Feb. 2000 and continued until the reload area was cleared of excess gravel in April 2000. Gravel hauling over the floating ice to construct Northstar Island was initiated on 13 Feb. 2000 and completed by April 2000. A total of 547,735 cubic yards of gravel was transported to the island over the 13 Feb. to early April period, involving about 18,312 round trips by Maxihauls. During periods of sustained construction operation, Maxihauls operated ~20 hours per day. Maxihauls operating on the ice roads traveled at ~56 km/h (35 mph). Caterpillar D8N and D8K bulldozers were used to move gravel on the island. A sheetpile wall was built to surround the entire working surface on the island in order to provide protection of island facilities in case of extreme ice ride-up during the winter. Sheet pile installation at Northstar Island began on 7 March and continued through 29 May 2000, using both vibratory and impact techniques. A Manitowoc 888 crane lifted and placed the sheetpiles at the desired locations. The sheetpiles were driven using an American Piledriving Equipment (APE) model 200A vibratory pile driver and, at times, a DELMAG D62-22 impact hammer. The sheetpiles were driven to a depth of approximately 10 m. The APE 200A vibrated the sheetpiles at a range of 400–1400 vibrations per minute and weighed 13,000 lb (5900 kg; Fig 2.8). The DELMAG D62-22 impact hammer was used to drive the sheet piles through the frozen surface of the original island, which the vibratory piledriver could not penetrate. The D62-22 drove the sheetpiles at 36–50 blows per minute. Its hammer weighed 28,500 lb (12,950 kg). The impact hammer was used to drive sheetpiles intermittently from April through May. Installation of the pipelines began on 14 March and was completed by 29 May 2000. For installation of each pipeline segment, three activities occurred in sequence: ice cutting, seafloor trenching, and pipeline installation. Each of these three operations occurred continuously, moving progressively northward from the shore to the island site. A Ditchwitch R100 was used to cut a 1.5–1.8 m (5–6 ft) wide slot 2-12 Monitoring at Northstar, 1999-2002

FIGURE 2.8. Vibratory pile driver (American Pile Driving Equipment model 200) driving steel sheetpiles through the surface of Seal Island. in the ice for installation of the pipelines. A Caterpillar 330 backhoe was used to remove ice from the slot. A Hitachi EX-450 backhoe then moved along the slot, excavating the pipeline trench in the seafloor (Fig. 2.9). The depth of the seafloor trench (below the sea bottom) ranged from 1.8 to 2.4 m (6 to 8 ft) in different portions of the pipeline route, as called for in the engineering plans. A Euclid R-25 and a Volvo A-30 dump truck were used to haul cuttings from the trench. The R-25 and A-30 have a 25- and 30-ton capacity, respectively. The cuttings were placed on the ice until the pipeline was installed, and then backfilled into the trench as needed. Lengths of pipe were transported onto the ice by a standard tractor trailer. The 10-inch oil pipeline and the 10-inch gas pipeline were laid side by side in the same trench. Caterpillar 583K side-boom tractors were used to lay the pipeline through the ice into the trench. Figure 2.10 shows the pipeline on the ice before it was lowered into the trench in the seafloor. Chapter 2: Description of BPs Activities 2-13

FIGURE 2.9. Hitachi EX-450 backhoe excavating the pipeline trench in the seafloor through a slot in the ice.

FIGURE 2.10. Installation of the Northstar gas and oil pipelines was initiated on 14 March 2000 and was completed by 29 May 2000. (Note: Three Blue Bird rolligon-type vehicles idling in the background.) 2-14 Monitoring at Northstar, 1999-2002

It was necessary to set a line of buried charges across Stump Island to facilitate excavation of the pipeline trench across the island. This blasting occurred on 11 March and 2 April 2000. On 11 March, a total of 378 charges were placed at the bottoms of drill holes from 2.1 to 2.7 m (7 to 9 ft) deep. They were detonated in rapid sequence, over a total duration of just a few seconds, to loosen an area ~380 m (1250 ft) in length by 3 m (10 ft) wide. The sea ice was frozen to the bottom in the lagoon south of the southernmost charge, and in the Beaufort Sea north to 1 km (0.6 mi) north of the northernmost charge. Sounds from blasting were measured as part of the physical acoustic measurements program.

LATER CONSTRUCTION PHASES AND INITIAL DRILLING

2000 Break-up and Open-Water Season

During this season (nominally 16 June through 31 Oct. 2000), construction of basic facilities on Northstar Island continued, the well conductor pipes were driven into the island, and equipment was delivered by sealift barge. There was frequent vessel traffic to and from the island, including a crew boat and barges. Round-trip helicopter flights (using two Bell 212 helicopters) began on 1 June 2000 and became less frequent after break-up, at which time a crew boat began to be used for crew changes. The crew boat Hawk (Fig. 2.5), an 18.7 m (61.5 ft) vessel, began operating on 20 July 2000 between West Dock 2 and Northstar. The vessel had dual 200 hp Detroit diesel engines equipped with 4-bladed propellers, and traveled at an average speed of ~12 knots. Crew boat operations ceased on 7 Oct. 2000 as sea ice formed. During the open-water season, the Hawk made ~1337 departures from West Dock 2. Other vessels not associated with Northstar often operated in the region during this period, but no detailed records of those vessels or their dates of operation were obtained. After 7 0ct., the Bell 212 helicopters were used at increased frequency to transport crews and supplies. Capping of the sheet pile wall on Northstar Island was completed on 11 June 2000. Grading of the island slopes began on 1 June 2000 and was completed in early August 2000. Placement of filter fabric and concrete slope protection (Fig. 2.11) began on 11 June 2000 and was completed in early Sept. 2000. Foundation blocks for the modules, pipe racks, and drilling rig were placed intermittently from 1 June 2000 until 1 Aug. 2000. Impact hammering of the well conductor pipes (conductors) began on 20 June 2000. The conductors were 51 cm (20 in) in diameter and were hammered to a depth of 36.5 m (120 ft). In total, 25 conductor pipes were driven in a north-south line parallel to and 43 m (141 ft) from the water line (MLLW) on the eastern side of the island. Five well insulator pipes, each 107 cm (42 in) in diameter, were driven concentrically around five of the conductor pipes. This provided space for insulation around these conductor pipes. The insulator pipes were driven to a depth of 21.3 m (70 ft). Initially, both sizes of pipes were driven with the vibratory hammer (Fig. 2.8) until the substrate of the original island limited further downward progress. The pipes were then driven to their final depth with a diesel-powered impact hammer. Hammering of the pipes was completed during the last week of July 2000. Equipment used during island construction during this period included a Caterpillar 966 front-end loader, American crane model 11320, Manitowoc 888 crane, APE model 200A vibratory pile driver, DELMAG D62-22 impact hammer, Hitachi backhoe, Caterpillar D8 bulldozer, mechanic’s box truck, and two Polaris 6-wheelers. Chapter 2: Description of BPs Activities 2-15

FIGURE 2.11. Installation of concrete slope protection mats during 2000, with two 20” well conductor pipes in background awaiting piledriver.

Two self-powered flexi-float barges, the Stryker and Garrett, were used to support off-loading operations at Northstar throughout the open-water season (Fig. 2.12). The barges were each powered by two 220 hp Detroit diesel engines, each with a 4-bladed propeller operating between 1000 and 1400 rpm. These vessels served as mobile docks or pontoons for crew and equipment transfer from other vessels to the shore of the island. Other tugs and barges periodically traveled to Northstar bringing equipment. A sealift barge with the permanent living quarters (PLQ), utility module, and pipe racks arrived at the island on 9 August 2000. The modules and pipe racks were off-loaded to Northstar from 14 to 20 Aug. 2000 using a Scheuerle trailer model MPEK 5200 and two Manitowoc 888 cranes. The Scheuerle trailer had four 454 hp (1816 hp total) Deutz diesel engines. Installation of modules began immediately and the living quarters were ready for occupancy on 24 Sept. 2000. The “emergency” generator in the utility module became operational by the week of 27 Aug. 2000. This generator was powered by a 4- cycle, 16-cylinder diesel engine operating at 1800 rpm. This diesel generator supplied power to the PLQ and other island activities until late 2001, when the main source of electrical power was replaced by a gas-turbine system. 2-16 Monitoring at Northstar, 1999-2002

FIGURE 2.12. Self-propelled barge pushing against Northstar Island, summer 2000.

Alaska Clean Seas (ACS) conducted spill drills during the open-water season of 2000. Nine days in early July and three days in early Oct. were dedicated to training Northstar personnel in spill response techniques. Most of the activities occurred 6 to 8 miles north-northeast of Northstar Island. Vessels involved in the activities included the barge Arctic Endeavor, two Point-class tugs, four Bay-class work boats, two Munson work vessels, two landing craft, and two outboard-powered tow vessels. Spill drills were also conducted during the open-water periods of subsequent years. 2000–2001 Ice-covered Period

During this ice-covered season (nominally 1 Nov. 2000 through 15 June 2001), two offshore ice roads plus one alongshore ice road were constructed, additional construction and installation activities occurred on the island, and drilling began. Also, more gravel was placed over certain portions of the oil and gas pipelines to shore. Ice-Roads and Transportation Three ice roads were built during the 2000–01 ice-covered season (Fig. 2.13). The primary ice road was built for transport of personnel, equipment, and construction material between the Prudhoe Bay facilities and Northstar Island. This ice road was along a different route than any of the ice roads during the preceding two winters. A second offshore ice road was built along the pipeline alignment. This was necessary to provide access to sites along the pipeline route where additional gravel fill was needed over the sub-sea pipeline that had been installed the previous winter. A third ice road, on bottom-fast ice, was constructed along the coastline from the Point McIntyre #2 facility on the West Dock causeway to the shore crossing of the pipeline. This road provided ground access to the valve pad at the pipeline landfall and to the backfill sites south of Stump Island. Chapter 2: Description of BPs Activities 2-17

FIGURE 2.13. Northstar construction area northwest of Prudhoe Bay, Alaska, showing ice roads established during the winter of 2000–01, approximate area of artificially thickened ice, test area for the ARKTOS evacuation craft, and approximate travel route for on-ice equipment tests.

The primary offshore ice road for transportation to Northstar Island was built between West Dock and Northstar (Fig. 2.13) from late Nov. 2000 to mid-March 2001. Almost 100% of the primary ice road was built on floating ice, which was artificially thickened during the December to March period. The perimeter of the island was also flooded out to ~100 m (328 ft) offshore of the island. During late Nov. when flooding began, the ice was too thin to support heavy snow removal equipment. Thus, the snow within the primary road right-of-way was not cleared initially, contrary to the practice in previous years. The road was only cleared when snowdrifts impeded traffic or flooding operations. After construction, the road was maintained until the first week of June 2001 when it was closed to all traffic. 2-18 Monitoring at Northstar, 1999-2002

The second offshore ice road was cleared along the pipeline alignment and lightly flooded to facilitate access to sites along the pipeline route. Access was required in order to place additional gravel fill over the sub-sea pipeline to meet specifications of the U.S. Army Corps of Engineers permit. This ice road extended from Northstar Island to ~100 m north of Stump Island and from ~100 m south of Stump Island to the pipeline landfall (Fig. 2.13). The route was cleared of snow in January 2001, and flooding began during the third week of February. The third “offshore” ice road was constructed from late Nov. to mid-March, in the same manner as the primary ice road except it was located parallel to the coast. This ice road was an oversize, two-lane road, ~8 km (5 mi) long. The average width of the flooded ice of this road was ~60 m (200 ft). The ice road was maintained until the first week of June 2001 when it was closed to all traffic. Helicopters were used for transportation to and from Northstar during the freeze-up period in late 2000. Crew boat operations had ceased on 7 Oct. 2000. Two Bell 212 helicopters were used to transport crew and materials to and from Northstar Island until 19 Nov. 2000. By then, the ice road from the West Dock base of operations (WDBO) had been marked with reflective delineators and became the primary means of transportation to the island. Hägglunds tracked vehicles were used to transport personnel and materials from West Dock to Northstar Island from 20 Nov. 2000 to late Jan. 2001. Hägglunds operations ceased during the last week in January 2001 after the ice road surface permitted standard vehicle traffic. Heavy loads were not transported on the ice road until March 2001. During the majority of the ice-covered season, from late Jan. 2001 to late May, standard SUVs, crew-cab pick-ups, and buses were the main primary method of transportation for Northstar personnel. The Hägglunds were used again during the first week of June to transport crews during marginal ice conditions after standard vehicular traffic ceased. All travel on the primary ice road ceased on 8 June 2001. Helicopter flights resumed on 9 June 2001 as break-up progressed. Activities On and Near Northstar Island Numerous construction activities occurred on the island during the ice-covered season in late 2000 and early-mid 2001. Major construction activities included completing the assembly of the drilling rig, pipe rack, permanent living quarters, and grind and inject module, along with dock improvements. Installation of the mini-injection effluent skid and of the foundation blocks for modules housing the processing plant, compressor, and garage also occurred. A Manitowoc 888 crane lifted and placed the pilings for the dock face. The APE model 200A vibratory pile driver was used to drive piles for dock improvements. These included areas for the crew boat and piles for the dock face that had been damaged by ice. A total of five wells were drilled during the 2000–01 ice-covered season. Drilling for the Under- ground Injection Control (UIC) well began on 14 Dec. 2000 and was completed during the first week of January 2001. On 26 Jan. the UIC well was commissioned for disposal of permitted muds and cuttings. Drilling of well NS-26 was initiated during the first week of February 2001, but it was abandoned due to collapsing walls. The drilling of wells NS-27, NS-29, and NS-31 occurred from January 2001 to 13 June 2001. Prior to 13 June, wells were drilled to the production zone and prepared for production. (However, oil production did not begin until late autumn of 2001.) After 13 June, drilling was suspended until late 2001 to satisfy regulatory requirements associated with the limitations that broken ice would impose on any oil-spill response should a spill occur. BP agreed in subsequent years not to drill into oil-bearing Chapter 2: Description of BPs Activities 2-19 strata during the period from 13 June until 18 inches of continuous ice had formed out to 1.5 miles in all directions from the island (NMFS 2001). Equipment testing occurred periodically on the sea ice and island throughout the season. The drilling rig engines and “emergency” diesel generators (the primary generators until the gas turbine generator was installed in late summer 2001) were tested for emissions during mid-March 2001. An equipment exercise on the sea ice was conducted along a linear track on 1 and 2 June 2001 near Northstar Island to help determine capabilities of several types of equipment in softening ice conditions (Fig. 2.13). Equipment that was tested included a wide-tracked backhoe, a Hägglunds tracked vehicle with a trailer, a Blue Bird “rolligon”, and a Tucker tracked vehicle. Two articulated ARKTOS vehicles (Fig. 2.14) were used as emergency escape vehicles. They were driven to the island and tested during the first two weeks of December 2000 and tested again in mid- April 2001. Each ARKTOS is powered by two Cummins 260 hp diesel engines and weighs 29,484 kg (65,000 lb). Propulsion while in the water is by two 14-inch-diameter water jets. Each ARKTOS evacuation craft is 15 m (50 ft) long and 3.9 m (13 ft) high with a beam of 3.8 m (12 ft 9 in). The maximum speed is 16 km/h (10 mph) on land or ice, and 6 knots in water. Each ARKTOS is capable of carrying 52 people, 24 in the front section and 28 in the rear section. Backfill Along Pipelines Additional gravel fill was placed over the pipelines at nine sites between Northstar Island and landfall (▲ symbols in Fig. 2.13). Seven backfill sites were concentrated 2.1 to 3.6 km (1.3 to 2.2 mi) north of the barrier islands in waters >1.5 m (5 ft) deep. The remaining two sites were ~500 m (1650 ft) north of the landfall of the pipeline in waters <1.5 m (5 ft) deep. Engineering plans called for the pipeline to be buried by at least 1.8 m (6 ft) of material, and additional fill was required to achieve this.

FIGURE 2.14. Articulated ARKTOS evacuation craft, used as an island emergency escape vehicle. 2-20 Monitoring at Northstar, 1999-2002

The placement of backfill along the pipeline alignment occurred between mid-March and mid- April 2001. Holes 2.4 m (8 ft) wide were cut through the ice above the pipeline with a Ditchwitch R100 and a Caterpillar 330 backhoe was used to remove ice from the hole. Gravel was dumped through the holes where additional fill was needed. Gravel was transported from the Kuparuk River Oxbow gravel mine site to the fill locations using Volvo A-30 dump trucks. The total length of the pipeline that required backfill placement was 314 m (1030 ft). A total of 130 truckloads (3640 cubic yards) of stockpiled gravel was used to backfill the nine locations along the pipeline. A Caterpillar 966 loader was used to place the extra gravel through the opening, covering the pipeline by at least 1.8 m. 2001 Break-up and Open-water Season

The major activity during the season from 16 June through 31 Oct. 2001, aside from the usual helicopter and vessel operations for routine island support, was the arrival of the main production equipment via sealift during August, followed by its offloading, installation, and initial testing. Oil production commenced on 31 Oct. 2001. Transportation and Sealift As in previous years, Bell 212 helicopters (Fig. 2.3) were used as transportation to and from North- star Island during the break-up and freeze-up periods, and crew vessels were the primary method of transportation during the open water season. Helicopter flights began on 4 June, with two Bell 212 helicopters being used. Two 61.5 ft (18.7 m) crew vessels, the Hawk (Fig. 2.5) and the Arctic Express, began transits between West Dock 2 and Northstar on 23 July 2001. Crew vessel operations ceased on 7 Oct. 2001 as sea ice formed. During the period of vessel operations, the crew boats completed 824 round-trips between West Dock and Northstar Island. During freeze-up, helicopter flights again became the primary method of travel to and from Northstar. During the nominal “open-water” period in 2001 (16 June to 31 Oct.), there were ~989 round- trip helicopter flights, not counting additional helicopter trips in the 4–15 June period. Tugs and barges periodically traveled to and from Northstar, delivering equipment and fuel. There were 69 round-trips by barges. Eleven vessels were used for transport of diesel fuel to the island, and 17 vessels were used for cargo transport to and from the island. Eleven Alaska Clean Seas (ACS) tugs and barges were deployed around Northstar for training and fuel transfers on 23 days during the 2001 open- water season. Other vessels not associated with Northstar operated near Northstar and in the general Prudhoe Bay area. This non-Northstar vessel traffic included vessels associated with a pipeline and shallow hazards survey from West Dock to the east, a seismic program in Simpson Lagoon, fuel barges serving Nuiqsut and Kaktovik, and removal of the CIDS. In addition to “local” tug and barge traffic, three sealift barges arrived in the Prudhoe Bay area from southern Alaska on 10 Aug. 2001, towed by two Ocean-class tugs (the Bulwark and Navigator), and assisted by three River-class tugs (the Kavik River, Sag River and Toolik River). The Point-class tugs Pt. Barrow and Pt. Oliktok were also used when required by ice and weather conditions. The three barges carried the following facilities: • Barge 420: the gas-turbine compressor module and pumphouse. • Barge 400: North and South processor modules; these two buildings constituted by far the largest shipment (Fig. 2.15). • Barge 411: the warehouse, flare boom, and some pipe rack sections. Barges 420 and 411 arrived first, followed by Barge 400. Offloading of the first barge started on 12 Aug- ust and continued until 20 Aug. 2001. During this period there was much maneuvering of the sealift Chapter 2: Description of BPs Activities 2-21

FIGURE 2.15. Sealift Barge 400, with the N and S process modules, upon its arrival in Prudhoe Bay on 10 Aug. 2001. The Ocean-class tug Navigator can be seen holding the barge on one side. barges by tugs, and the sounds from these operations were measured. Offloading involved the use of a Scheuerle MPEK 5200 trailer (as in 2000). Ocean-class tug Bulwark departed with Barge 420 on 16 Aug. Ocean-class tug Navigator departed with Barges 400 and 411 on 30 Aug. Island Activities Equipment used for module off-loading and installation was similar to that used during earlier stages of island construction. Equipment used for these and other purposes on the island included the following: Scheuerle MPEK 5200 trailer, Caterpillar 966 loader, Volvo 150 front-end loader, Volvo A30 end dump, Manitowoc 888 crane, Caterpillar 330 backhoe, Caterpillar D5 bulldozer, mechanic’s box truck, 80 ft and 125 ft mobile aerial lifting platforms, 3 to 4 light plants, various plate and drum-roller compactors, Caterpillar 988B forklift, and Polaris 6-wheeler. Installation of modules began immediately after they were moved onto the island, and the “A” gas compressor was tested and began operating on 24 Oct. 2001. The compressors are powered by two GE LM-2500 gas-turbine engines that rotate clockwise at 8400–9800 rpm. Depending on ambient temperature they can operate at up to 32,000 hp (22,370 kW). The “B” gas compressor was not tested or operated during the summer or autumn of 2001. 2-22 Monitoring at Northstar, 1999-2002

The diesel-powered “emergency” generator in the utility module, as installed during 2000, was used for primary power until 24 Oct. 2001. At that time, the source of power became a series of three Solar gas-turbine-powered generators. Each generator is driven by a 13,000 hp (9700 kW) gas turbine rated at 10,780 rpm and operating at 9500 rpm. Drilling was temporarily suspended from 13 June 2001 until November 2001, and the drilling rig as a whole was shut down until October when the generators were again operational. However, the grind and injection module continued injecting seawater and miscellaneous Class 1 permitted wastes into the disposal well throughout the open water and broken ice periods. Minor welding repairs were made on the rig during the summer. Oil production from Northstar began on 31 Oct., initially at a low rate from a single well.

INITIAL PRODUCTION AND CONTINUED DRILLING

2001–2002 Ice-Covered Season

During this season (nominally 1 Nov. 2001 through 15 June 2002), oil production was underway, along with continued drilling of wells. Power was now being produced from gas-turbine generators, and additional gas turbines engines were operating to compress and inject gas. The gas flare operated intermittently. Transportation to and from the island was by helicopter at the start and end of the ice- covered season, and by ice-road during the remainder of that season. Ice Roads and Transportation General information about transportation procedures to and from Northstar during the winter and spring are given in the earlier section entitled “Seasonal Transportation Needs”. Specific details concerning transportation during the 2002 winter-spring period are described here. One offshore ice road was constructed this winter, from West Dock to Northstar, along the same 12 km (7.4 mi) route as the primary ice road built in 2000–01 (cf. Fig. 2.13). The majority of this ice road was built on floating ice, which was artificially thickened. Flooding of the ice road began on 27 Nov. 2001 and continued through 7 Feb. 2002. Ice thickness at the onset of flooding was at least 0.6 m (2 ft). The average width of the ice flooded for the ice road was ~61 m (200 ft) with a 30 m (100 ft) vehicle surface. After construction, the road was maintained until the first week of June 2002 when it was closed to all traffic. The ice around the perimeter of the island was not artificially thickened, in contrast to the practice in previous years. However, two ramps leading on and off the island (Fig. 2.16) were flooded to a width of ~61 m (200 feet) to provide stability between the island and the adjacent floating ice before both ramps met the ice road. Helicopter use began in October and initially ceased on 1 Nov. 2001 when ice thickness was adequate for Hägglunds tracked vehicles (Fig. 2.1) to transport crews and supplies along the route of the future ice road. A late autumn storm fractured the ice on 6 Nov. 2001 and helicopters were used again from 6 until 24 Nov. 2001. Hägglunds vehicles were then used until 15 Dec., when the ice road surface permitted standard vehicle and van traffic. When used, Hägglunds vehicles made ~14 round trips per day between Northstar and West Dock. Hägglunds were also used (intermittently) for crew transport during the 10 days from 28 May until 6 June 2002, when meltwater accumulating on the ice road prevented standard vehicles from safely transiting to and from the island. During spring break-up in 2002, helicopter flights began on 24 May 2002 and ceased on 26 July 2002. Two Bell 212 helicopters (Fig. 2.3) were used during both freeze-up and break-up. Chapter 2: Description of BPs Activities 2-23

FIGURE 2.16. Northstar Island, showing ice road entrances to the island, approximate area of artificially thickened ice, exercise area for the ARKTOS evacuation craft, and locations of oil spill drill activities, as established during the winter of 2001–02.

Standard vehicles, including SUVs, crew-cab pick-ups, and buses, were the main method of transportation from 15 Dec. 2001 to late May 2002. Heavy loads were not transported on the ice road until March 2002. An average of 25 to 30 vehicle round-trips per day occurred from December 2001 to May 2002. All travel on the ice road ceased on 6 June 2002. Activities On and Near Northstar Island The 2001–02 ice-covered season was the first winter season subsequent to the changeover of the island’s power generation from diesel electric generators to gas turbines. It was also the first season with oil production and gas injection. By November 2001, the diesel electric generators of previous years, which generated power for the island, had been replaced by three Solar gas-turbine-powered generators. As noted previously, each generator has a 13,000 hp (9700 kW) gas turbine rated at 10,780 rpm and operating at 9500 rpm. The 2-24 Monitoring at Northstar, 1999-2002 emergency diesel generators were also used a number of times during the 2002 season as back-up to the gas turbine generators. Five gas turbines are located at Northstar Island: the aforementioned three Solar generators for power generation, and two GE LM-2500 engines powering high-pressure compressors for gas injection. One high-pressure compressor (HP-A) was on-line by November 2001, and the second (identical) compressor (HP-B) was brought on-line in February 2002. There was also a low-pressure compressor driven by a 5000 hp (3730 kW) electric motor running at a constant speed of 3600 rpm. At various times, each compressor was shut down for maintenance. Drilling resumed in November 2001 after being suspended since mid-June 2001. A total of 7 additional wells were drilled during the 2001–02 ice-covered season: • NS 13: 16 days (17 Nov. – 2 Dec. 2001) • NS 9: 34 days (4 Dec. 2001 – 7 Jan. 2002) • NS 14: 37 days (7 Jan. – 12 Feb. 2002) • NS 8: 28 days (13 Feb. – 13 Mar. 2002) • NS 15: 34 days (13 Mar. – 15 Apr. 2002) • NS 7: 32 days (16 Apr. – 18 May 2002) • NS 6: 21 days (18 May – 7 June 2002) No major island-construction or island-maintenance work was done on Northstar Island during November 2001 through June 2002. However, various test, training, and inspection activities occurred in the area on an intermittent basis through the winter and spring: Equipment testing occurred periodically throughout the season. Training exercises with an ARKTOS escape vehicle (Fig. 2.14) were conducted on 17 and 24 Dec. 2001, 13 and 20 May 2002, and 3 and 4 June 2002. The exercises were conducted on the sea ice west/southwest of the island and along the ice road west and southwest ~305 m (1000 ft) away (Fig. 2.16). There were three oil spill exercises/drills during the 2001–02 ice-covered season. Two were for the containment of oil in water and were held on 11 and 18 March 2002. The third exercise, for the detection of oil in water under the ice, was conducted on 8 April 2002. The spill drills were located in an area ~61 m by 61 m (200 × 200 ft) just west of the island (Fig 2.12). Snow machines and 4 wheeled all-terrain vehicles were used to access the area. The spill drills included use of chain saws to cut ice slots ~0.5 m (1.5 ft) wide by ~6 m (20 ft) long. Ice auger holes were also drilled. Snow was pushed off to the west edge of the drill area with a Bobcat loader and snow blower to allow access to the ice for cutting. Two portable generators were used to power light plants at the drill site. The pipeline route from Northstar to the mainland was inspected for traces of oil under the sea ice every 30 days, as required. Five inspections were conducted from 1 Dec. 2001 to 1 May 2002, as conditions permitted. Crews accessed the drill sites along the pipeline route using a Hägglunds tracked vehicle. Boreholes were drilled through the ice along the entire length of the subsea pipeline route, a distance of ~9.5 km (6 mi). These consisted of 5–7 cm (2–3 in) diameter holes drilled with a hand-held electric auger powered by a portable generator. Water that came up through the hole as the auger was pulled out of the ice was inspected visually for oil. At ~61 m (200 ft) intervals along the pipeline route, a pair of holes was drilled ~4.5 m (15 ft) apart, one on each side of the pipeline. Approximately 300 holes were drilled during each inspection, 150 on each side. Chapter 2: Description of BPs Activities 2-25

Some gas-flaring events occurred during this ice-covered season, and there was one reportable Northstar- related spill onto the sea ice as a result of design problems with the flare cone. Hydrocarbon condensate and water droplets were expelled from the flare tip during flaring events. Fourteen events occurred, and these released a total of ~45 L (12 gal) of condensate. Approximately 37 L (10 gal) were released during the first event and 0.9 L (0.25 gal) were released per subsequent events. In consultation with the Alaska Department of Environmental Conservation (ADEC) and the U.S. Coast Guard (USCG), these incidents were classified as one spill. At times, small amounts of this material settled onto the ice and snow, with the location depending on wind direction. Following each event, crews cleaned up the surface by collecting the settled condensate. The entire area affected was ~334 m2 (400 yd2) surrounding the flare. The flare was modified during the spring of 2002 in two ways to eliminate this condensate: (1) grouting was added in the flare tip, and (2) equipment was installed to maintain the temperature in the flare line above the dew-point of the flare gas. In addition, there were 10 reportable spill events ≥37 L (10 gal) of crude oil, fuel, or other hydrocarbon-based substances (range: 37 L to 795 L [10 gal to 210 gal]) on Northstar Island during the 2001–02 ice-covered season. These were contained within the confines of the island and did not contact the sea ice or water. 2002 Break-up and Open-Water Season

This season (nominally 16 June through 31 Oct. 2002) was the first open-water season after the start of oil production and gas injection at Northstar. It was also the first open-water season when the island’s power source consisted of gas-turbine generators. The usual helicopter and vessel operations took place for routine island support but there was no major sealift of equipment to Northstar in 2002, in contrast to 2000 and 2001. Transportation Helicopters or a crew boat were used for crew changes, as in previous open-water and transitional seasons. Helicopter flights to and from Northstar (using two Bell 212 helicopters) began intermittently on 24 May and continued until 27 July. On that date, one crew boat, the Arctic Express, began to operate. Arctic Express was an 18.7 m (61.5 ft) vessel similar to that shown in Figure 2.5. Crew vessel operations ceased on 2 Oct. 2002 as sea ice formed. During the period of vessel operations in 2002, the crew boat completed 469 round-trips between West Dock and the island. Helicopter flights resumed on 1 Oct. During the 16 June through 31 Oct. 2002 period, there were 477 round-trip flights from 16 June to 27 July, none from 28 July to 1 Oct., and 878 from 2 to 31 Oct. In addition, barges made 63 round-trips to the island during the period 7 July to 10 Oct. 2002. Most barge trips originated at West Dock. Two tugs and two barges periodically traveled to Northstar bringing equipment. The barge used to transport fuel to the island was ~61 m (200 ft) in length and the barge used to transport cargo was ~49 m (160 ft) long. The tugs were ~20 m (65 ft) in length. The Sag River, which was used to pull the cargo barge, was powered by 3 diesel engines (1095 total hp). The Sinuk, used to pull the fuel barge, was powered by 2 diesel engines (1300 total hp). BP expects to reduce transportation to and from Northstar as drilling and final construction are completed. Also, testing of a hovercraft is planned for 2003. If successful, the hovercraft may be used as an alternative to regular vessels and helicopters, with helicopter use for emergencies only. Island Activities The five gas turbines described previously (three generators and two high-pressure compressors) continued to be used during the 2002 open-water period, along with the electrically powered low pressure compressor. At least 2 of the 3 gas-turbine generators were running continuously, and a diesel generator was used infrequently (8 hours of total operation). Two compressors powered by natural gas were also 2-26 Monitoring at Northstar, 1999-2002 operated during the period. Compressor HP-A was operated infrequently during July through September, but averaged 17.56 hours of use per day during October. Compressor HP-B was operated almost continually during June and July, averaged over 19 hours of use per day during August and September, and averaged 10.94 hours of use per day for October. Two crude stabilizer pumps averaged 21.85 hours of use per day each. Two crude sales pumps averaged 20.56 hours of use per day each. Two water injection pumps averaged 1.03 hours of use per day each. All of this equipment is electrically powered. Construction and maintenance equipment that was used on the island during the 2002 open-water period was similar to that used in previous years. Equipment that was used daily included the Volvo 150 front end loader, mechanic box truck, and Polaris 6-wheeler. Equipment that received intermittent or occasional usage included Caterpillar 966 loader, Volvo A30 end dump, Manitowoc 888 crane, Caterpillar 330 backhoe, Caterpillar D-8 bulldozer, mobile aerial lifting platform, light plants, compactors, and Caterpillar 988B forklift. Drilling activities were conducted on 130 days between 16 June and 28 Oct. 2002. Drilling did not occur on the island from 2 to 4 Sept. when a compactor was used to repair subsidence on the island. Also, in accordance with BP’s undertaking not to drill into oil-bearing formations during the break-up and open water periods, all drilling during this period was above the reservoir depths. Drilling occurred at eight well locations on Northstar Island: • NS 12: 12 days • NS16: 17 days • NS18: 22 days • NS24: 17 days • NS17: 27 days • NS14RWO: 1 day • NS8RWO: 2 days • NS20: 32 days Seawater and drilling wastes approved for “Class 1 well injection” were injected into the NS-10 disposal well. Spill response drills were conducted by Alaska Clean Seas (ACS) intermittently during the open- water period, normally on Monday afternoons from 16:00 to 17:00 hours local daylight time. Eleven spill drills were conducted during the season, including 3 in July, 1 in August, 3 in September, and 4 in October. Vessels used as part of the training included 8 zodiacs, 2 Kiwi Noreens, and 4 Bay-class boats. Spill response vessels ranged in size from 3.7 to 13.7 m (12 ft to 45 ft) in length. One barge was used for a fuel transfer of ~14 hours duration during the open-water period.

OTHER MAJOR OFFSHORE ACTIVITIES

Vessel traffic not associated with BP’s Northstar Development occurred in the general Prudhoe Bay/Northstar area during each of the open-water seasons discussed above (2000 through 2002). This included vessel traffic to and from other industrial sites and North Slope communities, seismic and shallow-hazards surveys and associated vessel traffic (2000 and 2001 only), a pipeline route survey (2001), and boat traffic associated with subsistence whaling based at Cross Island. Chapter 2: Description of BPs Activities 2-27

During the 2002 whaling season (28 Aug. to 19 Sept.), vessel traffic information for the Prudhoe Bay/Northstar area was obtained from logs maintained by the whaling communications center. Other vessel traffic unrelated to the Northstar project occurred near Northstar Island on at least 84 occasions during the 2002 whale hunt period (Table 2.1). Much of this traffic involved barge transport to Barter Island and Point Lonely that originated in the Prudhoe Bay area, or vessel traffic from West Dock to Point Thomson, Badami, or Endicott. The values in Table 2.1 include vessel movements between Prudhoe Bay and Cross Island, but do not include the activities of whaling vessels operating from Cross Island. Other vessel traffic not related to the Northstar project occurred before and probably after the 28 Aug. to 19 Sept. 2002 period for which records are available.

TABLE 2.1. Approximate number of one-way vessel trips east and west of the West Dock/ Prudhoe Bay area from 28 Aug. through 19 Sept. 2002, exclusive of those related to Northstar activities and subsistence whaling vessels.

Location Æ West East Cross Isl. Endicott Vessel Trips Æ 41 25 10 8

Vessel traffic not associated with Northstar has occurred in all years during the 2000–02 period, probably at rates similar to the rate observed in 2002. However, we have only been able to obtain vessel traffic statistics through the whaling communications center for 2002. Corresponding data were not arch- ived for previous years. During all three of the open water seasons discussed above, drilling caissons were moved into or away from the general Northstar area by other companies not associated with BP. In 2000, the Concrete Island Drilling System (CIDS) was towed by tugboats to a location about 1.1 km (0.6 n. mi) northeast of Northstar Island on 29 Aug. It subsequently was ballasted down onto the seafloor at that location for long-term storage. The CIDS was not a part of the Northstar operation, and was not associated in any way with BP. However, its presence in close proximity to Northstar during the latter part of the 2000 open-water season is relevant to the sound measurements and whale monitoring conducted for the Northstar project. There was continuing activity on and around the CIDS until early Sept. 2000, after which CIDS was dormant and unmanned for about one year. While dormant, CIDS (and SDC—see below) would produce no industrial sounds, but presumably would result in some wave slap sounds. In 2001, the CIDS was re-activated, re-floated, and (on 31 Aug.) towed out of the area to the west. During the weeks preceding its departure, two small tugs and a large ice-breaking tug were present near CIDS. Intermittent helicopter flights were observed between CIDS and Deadhorse airport. Again, these operations were unconnected with BP, but occurred in close proximity to Northstar. In 2002, the SDC (Steel Drilling Caisson) arrived in the Prudhoe Bay area from the west. On 4 Aug. 2002, it arrived at the McCovey prospect, about 18 km (11 mi.) east, and slightly north, of North- star. By 14 Aug., the SDC had been ballasted down and cold stacking of the caisson was completed. The SDC was present but inoperative for the remainder of the 2002 open-water season. Marine mammal and acoustical monitoring studies were conducted by BP near Northstar during each year from 1999 through 2002 in order to meet monitoring commitments called for in • the North Slope Borough (NSB) zoning ordinance for Northstar, • incidental take authorizations issued by NMFS, 2-28 Monitoring at Northstar, 1999-2002

• BP’s monitoring plans, as agreed to amongst BP, NMFS, the NSB, and the Alaska Eskimo Whaling Commission (AEWC) and, • stipulations by the U.S. Fish and Wildlife Service within the U.S. Army Corps of Engineers permit. The monitoring activities that have been conducted by BP in the Northstar area have been summarized in a general way in Chapter 1, and most of them are described in detail in later chapters of this report.

CONCLUSIONS

BP began construction of oil-production facilities for the Northstar Development on an offshore, man-made island in early 2000. BP’s activities during the construction and initial operational phases of the Northstar development have been described for each of two main periods within each year: the ice covered period, nominally extending from 1 Nov. through 15 June, and the open-water period, nominally from 16 June through 31 Oct. BP’s activities during these two parts of the year have differed, especially during the construction phase, because of the influences of ice vs. open-water on transportation and construction. Also, during early summer and late autumn, there are transition periods, each typically ~30 days in duration, when thin or broken ice does not allow either on-ice or vessel activities. Activities during the ice-covered period are largely adaptations of terrestrial transportation and construction techniques once the sea ice can support the weight of the equipment. For light vehicles, this is typically by mid-Nov. in an average year. Activities during the ice-covered period begin with the use of all-terrain tracked vehicles until the sea ice is artificially thickened and smoothed, which allows the use of regular vehicles, typically within 30–45 days of initiation of ice road construction. Open-water activities typically begin by mid-late July and continue until early Oct., when Northstar is supported by barge and vessel operations. Northstar activities prior to early 2000 were limited to ice-road construction during the winters of 1998–99 and 1999–2000. Major construction activity occurred during the mid-late winter of 2000, when the island was re-built and the pipelines constructed. Thereafter, construction and facility-installation continued through the open-water seasons of 2000 and 2001, and the intervening winter. Major facilities were delivered by sealift in August of 2000 and 2001, and drilling began during late 2000. During the late autumn of 2001, power generation transitioned from diesel to gas-turbine generators, and oil production began. Oil production and drilling have continued since then, supported by the usual types of transportation activities, but with no further heavy construction.

ACKNOWLEDGEMENTS

We thank Dr. Bill Streever, Allison Erickson, and Wilson Cullor of BP’s Environmental Studies Program for their specific support of the 2001 and 2002 monitoring work, and Dr. Ray Jakubczak for overall support of the Northstar monitoring efforts. Julie Arin, Pam Pope, and Bob Lipchak, BP HSE advisors for the Northstar project, and Russ Vogel of Alaska Interstate Construction, helped compile the 2001–02 data for this report. Marko Radonich supplied information on vessel traffic during the 2002 whale hunting period. Thanks to Drs. Susanna Blackwell and Bill Burgess of Greeneridge Sciences for providing valuable details for this report. We also thank everyone who contributed to our previous reports on BP’s activities in 1999–2001, as listed in the Acknowledgement sections of those reports. Much of that information has been incorporated into the present report. Dr. W. John Richardson of LGL reviewed and edited this report. Thanks also to Bill Streever of BP, who reviewed the report and Chapter 2: Description of BPs Activities 2-29 provided helpful comments. Additional Acknowledgements applicable to the report as a whole are included in Chapter 1.

LITERATURE CITED NMFS. 2001. Taking and importing marine mammals; taking marine mammals incidental to construction and operation of offshore oil and gas facilities in the Beaufort Sea. Fed. Regist. 66(246, 21 Dec.):65923-65935. Perham, C.J. 2002. Introduction and description of BP’s on-ice activities, late 2001 to mid-2002. p. 1-1 to 1-12 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of ringed seals, sounds and vibrations during operation of BP's Northstar oil development, Alaskan Beaufort Sea, winter and spring 2001-2002: 90-day report. LGL Rep. TA2707-1. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 86 p. Perham, C.J. and M.T. Williams. 2002a. Description of BP's on-ice activities, 2000-2001 ice-covered season. p. 2-1 to 2-10 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of industrial sounds, seals, and whale calls during construction of BP's Northstar oil development, Alaskan Beaufort Sea, 2001. [Oct. 2002 ed.] LGL Rep. TA2563-5. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 309 p. Perham, C.J. and M.T. Williams. 2002b. Description of BP's activities, break-up and open-water seasons, 2001. p. 6-1 to 6-8 In: W.J. Richardson and M.T. Williams (eds., 2002, as above). LGL Rep. TA2563-6. Williams, M.T. and C.J. Perham. 2001a. Description of BP’s activities, 1999-2000 ice-covered season. p. 2-1 to 2-8 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of industrial sounds, seals, and whale calls during construction of BP's Northstar oil development, Alaskan Beaufort Sea, 2000. LGL Rep. TA2432-6. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 316 p. Williams, M.T. and C.J. Perham. 2001b. Description of BP’s activities, break-up and open-water seasons, 2000. p. 6-1 to 6-11 In: W.J. Richardson and M.T. Williams (eds., 2001, as above). LGL Rep. TA2432-7. Williams, M.T. and W.J. Richardson. 2000. Introduction. p. 1-1 to 1-9 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of ringed seals during construction of ice roads for BP's Northstar oil development, Alaskan Beaufort Sea, 1999. LGL Rep. TA2348-1. Rep. from LGL Ltd., King City, Ont., and LGL Alaska Res. Assoc. Inc., Anchorage, AK, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 153 p. Williams, M.T. and R. Rodrigues. 2003. Introduction and description of BP activities. p. 1-1 to 1-11 In: W.J. Rich- ardson (ed.), Monitoring of industrial sounds and whale calls during operation of BP’s Northstar oil development, Alaskan Beaufort Sea, summer and autumn 2002: 90-day report. LGL Rep. 2707-3. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anch- orage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 66 p. CHAPTER 3:

DRILLING AND OPERATIONAL SOUNDS FROM AN OIL PRODUCTION ISLAND IN THE ICE-COVERED BEAUFORT SEA1

Susanna B. Blackwell a , Charles R. Greene, Jr. a and W. John Richardson b

a Greeneridge Sciences, Inc. 1411 Firestone Road, Goleta, CA 93117 (805) 967-7720; [email protected]; [email protected]

and

b LGL Ltd., environmental research associates 22 Fisher Street, POB 280, King City, Ontario L7B 1A6 (905) 833-1244; [email protected]

for

BP Exploration (Alaska) Inc. Dept. of Health, Safety & Environment 900 East Benson Blvd, P.O. Box 196612 Anchorage, AK 99519-6612

LGL Report TA2704-2

December 2003

1 Chapter 3 In: W.J. Richardson and M.T. Williams (eds.). 2003. Monitoring of industrial sounds, seals, and bowhead whales near BP’s Northstar Oil Development, Alaskan Beaufort Sea, 1999–2002. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 3-2 Monitoring at Northstar: Ice-Covered Season

ABSTRACT

Recordings of sounds underwater and in air, and of iceborne vibrations, were obtained in March 2001 and February–March 2002 at Northstar Island, an artificial gravel island built by BP for oil production at a location 5 km (3 mi) offshore in the Beaufort Sea and ~10 km (6 mi) northwest of Prudhoe Bay, Alaska. The purpose of the recordings was to document the levels, characteristics, and range-dependence of sounds and vibrations produced by Northstar-related industrial activities, such as drilling and oil production, when Northstar is surrounded by shore-fast ice during winter. The measurements were needed to help interpret data on occurrence of ringed seals near Northstar during winter and spring. This information also helped satisfy BP’s desire to understand and minimize impacts from development, as well as requirements of permits and authorizations issued to BP. Drilling sounds were readily identifiable underwater and produced the highest broadband (10–10,000 Hz) levels, with a maximum of 124 dB re 1 µPa 1 km (0.6 mi) from the drill rig. Drilling led to increased underwater levels, mainly at 700–1400 Hz. In contrast, drilling did not increase broadband levels in air or in the ice over those during oil production. The contribution of production operations to overall sound levels at Northstar was not very prominent, but during production levels underwater increased somewhat in the 125–160 Hz band and at certain lower frequencies associated with power generation. Repeated measurements at increasing distances from the sound source allowed computing propagation loss. In all three media, broadband sound levels decreased by ~20 dB / tenfold change in distance. Background levels underwater were reached by 9.4 km (5.8 mi) during drilling and 3–4 km (1.9–2.5 mi) without. In air, background levels were reached 5–10 km (3.1–6.2 mi) from Northstar, depending on the wind but irrespective of drilling. In the ice, background levels were reached 2–10 km (1.2–6.2 mi) from the island depending on the geophone channel and the wind, but irrespective of drilling. A comparison of the recorded sounds with harbor and ringed seal audiograms showed that sounds were probably audible to seals, at least intermittently, out to ~1.5 km in water and ~5 km in air.

INTRODUCTION

Background

During the winter of 1999–2000, BP Exploration (Alaska) Inc. began construction of its Northstar oil production facilities in nearshore waters of the Alaskan Beaufort Sea. The Northstar facilities are located at 70.49ºN, 148.70ºW, on an artificial gravel island built in a water depth of 11 m (36 ft). This artificial island is about 5 km (3 mi) offshore of Long Island, one of the barrier islands northwest of Prudhoe Bay (Fig. 3.1). The Northstar facilities were built on the eroded remains of Seal Island, an artificial island constructed by Shell Oil Company in 1982, used in 1984 and 1985 for exploratory drilling, and subsequently abandoned. Each winter since 1999–2000, ice roads have been constructed from shore to Seal/Northstar Island to facilitate winter construction and, since late 2001, oil production. The main island-reconstruction work was done during early 2000. By March 2001, when the first of the acoustic measurements reported here were obtained, reconstruction of the island itself was nearly completed, permanent living quarters and a diesel generator had been installed, the pipelines to shore had been completed, and drilling had begun (Fig. 3.2A). During summer 2001, a sealift brought modular buildings, gas turbine generators, and other equipment needed to complete the construction of the oil processing facilities on the island. The first oil was produced on 31 October 2001, although drilling of additional wells continued thereafter. During February–March 2002, when additional acoustic measurements were obtained, both oil production and drilling were underway. Chapter 3: Drilling & Operational Sounds, Winter 2001 & 2002 3-3

FIGURE 3.1. Location of study area in Alaska and acoustic recording locations close to Northstar Island, Northstar Development Area, Beaufort Sea, March 2001 plus February and March 2002. Northstar (Seal Island) is indicated by the star. 3-4 Monitoring at Northstar: Ice-Covered Season

FIGURE 3.2. (A) Northstar Island surrounded by land-fast ice, seen from the southwest in early February 2001. The drill rig is the tallest structure on the right side of the picture. (B) Recording setup next to all- terrain vehicle during the 2002 measurements. m = microphone and wind shield; h = hydrophone cable disappearing into the augered hole in the ice; g = geophones planted in the ice (one was used as a backup).

Numerous ringed seals (Phoca hispida) inhabit the area around Northstar year-around. Ringed seals maintain breathing holes through the land-fast ice during the winter and spring. During the ice-covered seasons, these seals spend time on the ice, under the snow, and in the water below the ice. Ice-based and aerial surveys have shown that ringed seals continued to overwinter near Northstar during construction (Williams et al. 2002; Moulton et al. 2002, 2003) and production phases (see Chapters 4 and 5). Measurements of underwater and airborne sounds near Northstar, and of iceborne vibrations, were needed in order to help interpret the data on occurrence of ringed seals near Northstar during winter and spring. Also, winter acoustical measurements were required to satisfy BP’s desire to understand and minimize impacts from development, and to satisfy requirements of permits and authorizations issued to BP. These included the North Slope Borough zoning ordinance for Northstar, and Letters of Authorization (LoA) issued by the National Marine Fisheries Service authorizing potential disturbance of seals during Northstar construction and oil production. Sounds associated with the initial island reconstruction work in early 2000 are described by Greene and McLennan (2000) and, briefly, Moulton et al. (2003). Here we describe and compare the winter sounds during the later stages of construction (early 2001) and initial oil production (early 2002). Chapter 3: Drilling & Operational Sounds, Winter 2001 & 2002 3-5

The specific objective of the present work was to measure and document the levels, characteristics and range-dependence of sounds and vibrations produced by Northstar-related industrial activities (mainly construction, drilling, and oil production) occurring during the winter-early spring of 2001 and 2002. The planned approach involved making recordings of sounds in air and in the water below the ice, along with recordings of vibrations in the ice, at a series of increasing distances from Northstar Island. Parameters that were of interest included broadband levels of sounds and vibrations from various industrial activities, as a function of distance, plus the frequency composition of those signals. The auditory sensitivity of seals depends strongly on frequency (e.g., Terhune and Ronald 1975; Richardson et al. 1995; Kastak and Schusterman 1998). Thus, it was important to document the sounds and vibrations in a form that will allow frequency weighting appropriate to the animals of concern. One-third octave data are useful for this purpose.

METHODS

Recordings were obtained on 6, 8 and 9 March 2001, and 28 Feb. and 1 March 2002 (Fig. 3.1). All recordings were made at stations on the land-fast ice. Stations were accessed using an all-terrain vehicle for transport of equipment and personnel (Hägglunds in 2001, Tucker Sno-Cat 1600 Terra in 2002), and a Blue Bird all-terrain vehicle equipped with a powered ice auger to drill the holes through which underwater recordings were made (Table 3.1). Equipment

The sensors included a hydrophone, a microphone, and a 3-axis geophone (Fig. 3.2B), all calibrated. The hydrophone was an International Transducer Corporation (ITC) model 6050C, which includes a low-noise preamplifier next to the sensor and a 30 m (98 ft) cable. The hydrophone cable was attached with cable ties to a fairing to minimize interference from strumming. Prior to recording, the hydrophone signals were amplified with an adjustable-gain postamplifier. In 2001, the omnidirectional microphone was an ACO model 7013 condenser microphone with a 4012 preamplifier. In 2002, we used a G.R.A.S Sound and Vibration ½” prepolarized free-field microphone model 40AE with an ICP preamplifier model TMS426C01. The 3-axis geophone was a GeoSpace model 20 with critical damping on the two horizontal axes (H1 and H2, perpendicular to each other) and the vertical axis (V). The geophone provided separate data for the three channels; signals were amplified with an adjustable-gain postamplifier. Hydrophone, geophone (3 channels), and microphone signals were recorded on five channels of a SONY model PC208Ax instrumentation-quality digital audio tape (DAT) recorder, at a sampling rate of 24 kHz. Quantization was 16 bits, providing a dynamic range of >80 dB between an overloaded signal and the quantization noise. A memo channel on the tape recorder was used for voice announcements, and the date and time were recorded automatically. Field Procedures

In late February and early March of 2001 and 2002, Northstar Island was surrounded by land-fast ice 1.4–1.5 m (4.5–5 feet) thick. Near the island and on the ice road south of the island, the ice had been artificially thickened and sometimes exceeded 2.5 m (8 feet) in depth. The recording procedure was identical on all days of recording. The first measurements were made as close as possible to Northstar, with subsequent recordings at progressively increasing distances, either eastward (6 and 8 March 2001) or north and northwestward (9 March 2001, 28 Feb. and 1 March 2002, see Fig. 3.1). In both years, the recording locations were constrained by the presence of a pressure ridge 3.5–5 km (2–3 mi) north of the island. The decision to make recordings north and northwest of the island was based mainly on the fact that water depth 3-6 Monitoring at Northstar: Ice-Covered Season

TABLE 3.1. Circumstances of recordings, including distance from the drill rig, water depth, ice thickness, mean wind speed, and wind direction for all acoustic recording locations on the ice, 6–9 March 2001 and 28 Feb.–1 March 2002. The general direction of the recordings in relation to Northstar, and the recording times (local), are given in parentheses. 2001 2002 Ice Ice Water thick- Wind Water thick- Wind Dist. depth ness Speed Direction Dist. depth ness Speed Direction Station (m) (m) (m) (m/s) Station (m) (m) (m) (m/s) 6 March (East, 16:00–17:00) 28 Feb. (North & Northwest, 11:16–15:14)

1 200 11.5 2.1 4.5 NE 9 470 12.0 1.3 0.6 E 2 460 11.5 1.2 “ “ 11 970 12.5 1.5 0.1 E 8 March (East, 11:00–16:00) 13 1970 13.5 1.5 1.3 SE 1 200 11.0 2.1 7.6 ENE 17 5000 11.2 1.6 1.0 SSE 2 460 11.8 1.2 “ “ 20 7620 13.7 1.4 0.8 SE 3 990 11.5 1.7 7.8 “ 1 March (North & Northwest, 8:46–14:13) 4 2030 11.3 1.4 “ “ 7 220 12.0 1.5 2.0 SSW 5 4020 10.0 1.4 8.0 “ 9 470 12.0 1.3 4.7 SW 6 6050 9.0 1.5 “ “ 11 970 12.5 1.5 3.8 SW 9 March (North & Northwest, 8:45–15:45) 13 1970 13.5 1.5 4.3 WSW 8 260 11.8 2.0 6.3 ENE 17 5000 11.2 1.6 1.7 W 10 490 11.8 1.5 “ “ 20 7620 13.7 1.4 1.7 W 12 990 11.5 1.4 7.3 “ 21 9400 13.5 1.5 0.7 SSW 14 1990 12.0 1.4 “ “ 15 3530 12.8 1.4 8.4 “ 16 4100 13.6 1.4 “ “ 18 5910 14.0 1.4 9.4 “ 19 7320 14.0 1.5 “ “ increases in those directions and therefore seals are more likely to be found there rather than to the south and west, where shallower water, barrier islands, and land occur. The drill rig was located on the eastern side of the island but drilling was directional and oriented toward the northwest. Suitable recording locations were chosen using a hand-held GPS receiver (Garmin model 12XL) or, for close distances (<800 m), a laser rangefinder (Bushnell model # 20-0880). The drill rig was used as the reference at all stations. The Blue Bird augered a 12” hole through the ice, after which it moved at least 500 m away from the recording site and remained idling (this was a mandatory safety requirement). In a comparable recording situation, Greene and McLennan (2000) did not find any differences in sound levels or spectra with and without a rolligon idling ¼ n.mi. (460 m) from their recording site. The recording equipment was set up in the all-terrain vehicle. The hydrophone was lowered through the ice hole and was positioned about 1 m above the bottom. Water depth increased slowly with increasing distance northward from Northstar (Fig. 3.1, Table 3.1). The geophone was planted in the ice and positioned with H1 pointing directly toward Northstar. After the geophone was properly positioned, water was pour- ed around the “foot” to freeze it in place. The microphone was placed 1.5–2 m (5–6 ft) above the ice with Chapter 3: Drilling & Operational Sounds, Winter 2001 & 2002 3-7 an unobstructed path to the island. It was shielded from the wind as much as possible by the all-terrain vehicle and was fitted with a windscreen. Recordings from all sensors commenced after all sound- generating devices on the all-terrain vehicle (engine and heater) had been turned off. An average of 5.7 min of recordings were obtained at each station, for a total of 2 hr 45 min. In 2001, hourly wind speed and direction were obtained from the Northstar weather station, accessed on the website http://www.resdat.com/mms. In 2002, wind speed and direction were recorded over a period of 5 min at each station with a Kestrel 2000 Pocket Thermo Wind Meter (Nielsen Kellerman, Chester, PA, 19013). This allowed us to keep better track of variations between recording stations, which were sometimes not evident in hourly data from nearby weather stations. Island Activities During Recordings

After each day of recording, island personnel provided us with detailed records of the drilling activities that took place on the island. The Nabors rig 33E on Northstar Island uses a Varco brand top-drive (250 RPM max, 1000 hp continuous). The rig is either powered by its own five diesel electric generators or by the island’s power supply (see below). On one recording day each year (6 March 2001 and 28 Feb. 2002) no drilling took place. On the other days the drill rig was operational, which included several procedures that likely varied in the amount of noise they produced. Therefore, the category “drilling”, as presented in this paper, includes only periods of time during which the drill bit was boring through the ground. Activities such as adding 27.5 m (90 ft) of drill pipe at the surface of the wellbore are included in the “no drilling” category. In 2001, two diesel electric generators (4-cycle, 16-cylinder engines running at 1800 rpm) provided the island’s power. In 2002 these had been replaced by five gas turbines plus an electrically-powered low- pressure compressor. Of those, the following units were functioning during the 2002 acoustic recordings: (1) “Solar” generator “B”, with a 13,000 hp = 9700 kW gas turbine rated at 10,780 rpm and operating at 9500 rpm with shaft rate = 158 Hz. (2) High-pressure compressor “B”, with a GE model LM-2500 gas turbine of 30,000 hp = 22,370 kW, rated at 10,000 rpm and operating at 9000–9400 rpm; the speed varied with the gas injection rate; shaft rate = 150–157 Hz. (3) Low-pressure compressor driven by a 5000 hp (3730 kW) electric motor, running at a constant speed of 3600 rpm; shaft rate = 60 Hz. Oil production had not begun in March 2001, but took place on both days of recording in 2002. In 2002, gas turbine compressors were used for gas injection into the formations. Other activities that took place during recordings in both years included flooding of the island perimeter and ice road to thicken the ice, transport of goods and personnel (in SUVs, crew-cab pick-up trucks, vans and buses), and some hauling of mud and cuttings from the drilling activities. No flaring of gas, piledriving, or helicopter flights occurred during the recording periods in either year. Signal Analysis

The recorded, digitized hydrophone signals were transferred directly to a computer hard drive as time series. They were then equalized and calibrated in units of sound pressure with flat frequency response over the data bandwidth, 4–500 Hz for the geophone and 4–10,000 Hz for the hydrophone and microphone. Analysis was done using MATLAB (The MathWorks, 3 Apple Hill Drive, Natick, MA 01760-2098) routines and custom programs. For each recording, a sound pressure time series (waveform) was generated, and the sound was played via a speaker to help the analyst match notes from the field with the recorded sounds. The sound waveform was used to select representative samples for further analysis. If the overall sound pressure levels (SPLs) varied little with time, at least two (but usually more) 8.5-s samples were selected from the recording and analyzed; these samples were evenly spaced throughout the 3-8 Monitoring at Northstar: Ice-Covered Season recording. If the sound waveform showed fluctuations in the SPL, then samples were taken from both the stronger and the weaker sections of the recording. Frequency composition was determined by calculating the sound pressure spectral density by Fourier analysis, using the Blackman-Harris minimum three term window (Harris 1978). The averaging time for such measurements was 8.5 s. The transform length was one second. With windowing, the spectral resolution was 1.7 Hz with 1-Hz bin separation. Transforms were overlapped by 50% and therefore 16 power spectral densities were averaged for each 8.5-s measurement. To show how received levels varied with distance from the activity, root-mean-square (rms) broadband levels were plotted against range from the dominant source. These plots are based on series of sound recordings at varying distances along a given transect. Interpretation of these data is complicated by variability of the sources within and between recordings, and by the possible presence of sound from more than one sound source. Nevertheless, the “received level vs. range” plots give an estimate of the range of levels received at several distances during the activity studied. Where appropriate, a spectral analysis of the sound is included. A tone was identified when the sound pressure spectral density level (SPSDL) for a given frequency was greater than the SPSDL for both adjacent frequencies, and at least 5 dB above the nearest minimum SPSDL at a lower frequency. Microphone and geophone data were transcribed to disk files and analyzed in the same way as hydrophone data. The 3-axis geophone senses particle velocity in three orthogonal directions; results from these three channels are presented separately. Microphone data are expressed in dB re 20 µPa and geophone data are expressed in dB re 1 pm/s (picometer per second). When appropriate, a propagation model was fitted to the broadband data in order to predict the propagation loss underwater, in air, and through the ice. The model used was based on logarithmic spreading loss appropriate for sound loss vs. distance:

Received Level (RL) = A–B × log(R), where R is range in m. Eq. (1)

RL is in dB re 1 µPa underwater, dB re 20 µPa in air, and dB re 1 pm/s in the ice. When fitting the model to the data, recordings were included (at increasing distances from the sound source) until the broadband levels flattened out.

RESULTS Underwater Sound

Broadband Levels Received broadband (10–10,000 Hz) levels of underwater sound are shown in Figure 3.3 for four sets of recording conditions: No drilling, no production and No drilling, with production (Fig. 3.3A); Drilling, no production and Drilling, with production (Fig. 3.3B). The highest broadband SPL (124 dB re 1 µPa) was recorded during drilling, 1 km east of Northstar. The lowest broadband SPLs (~75 dB re 1 µPa) were reached at stations 5–8 km northwest of Northstar, both during and without drilling, and correspond to background levels. The sound propagation model represented by equation (1) was fitted by least-squares to the recordings obtained during production but no drilling, for which four stations of data were available, exclusive of distant stations where received levels had apparently reached background values. The resulting regression line and equation are shown in Figure 3.3A. The spreading loss term, B in Eq. (1), was 22.0 dB/tenfold change in distance. Received levels during drilling (Fig. 3.3B) tended to be higher, but were extremely variable. It Chapter 3: Drilling & Operational Sounds, Winter 2001 & 2002 3-9

FIGURE 3.3. Broadband (10–10,000 Hz) levels of underwater sound as a function of distance from the drill rig on Northstar Island, for conditions of (A) No drilling, with and without production; and (B) Drilling, with and without production. Data collected on 6, 8 and 9 March 2001, plus 28 Feb. and 1 March 2002. Station numbers are indicated above their respective data sets. E = east, N = north. Also shown are the logarithmic spreading loss models (see text for details). 10,000 m = 10 km = 6.2 mi. was therefore not possible to fit the sound propagation model to a complete data set in a reasonable way. Instead, we fitted the model to the lowest broadband level recorded at each northern station (ranges 220, 490, 990, 1990, 4100 and 5000 m), regardless of whether production was taking place or not (Fig. 3.3B). This allowed us to get an estimate of the change in SPLs as a function of distance (i.e., propagation loss) 3-10 Monitoring at Northstar: Ice-Covered Season despite large variations in SPLs that were likely due to other factors (i.e., variable activity at Northstar). The spreading loss term obtained was 21.5 dB/tenfold change in distance. The distances at which broadband values leveled out in the absence of drilling (Fig. 3.3A) were ~1 km without production and 3–4 km during oil production at Northstar. During drilling (Fig. 3.3B), these values were ~4–6 km without production and 5–9.4 km with production taking place. Spectral Characteristics Figure 3.4 shows mean underwater one-third octave band levels for three sets of recording conditions: (A) Without drilling or oil production; (B) Without drilling but with oil production; and (C) With both drilling and oil production taking place. Elevated range-dependent received levels at frequencies below 100 Hz are visible in all three plots and can likely be attributed to baseline island sounds such as power generation. Another peak at 125–160 Hz is only prominent in the data recorded in 2002 (Fig. 3.4B and C) and may therefore be linked to oil production and/or operation of gas turbines. Finally, a peak centered at 1 kHz is only prominent during drilling (Fig. 3.4C). According to the drilling records, the drilling performed during recordings at station 21, 9.4 km from Northstar, was very similar to that performed during recordings at stations 13 and 17 (2 and 5 km from Northstar, respectively). However, the peak at 1 kHz (the “drilling peak”) was undetectable at the 9.4 km station. In this and some other plots, received levels at increasing ranges sometimes did not diminish progressively. Other activities, aside from the one of specific interest in each graph, often started, stopped or changed in an uncontrolled manner during the intervals between our recordings at different distances. Also, the activity of interest often varied within our series of recordings. For example, during drilling, torque, bit rpm, bit depth, and substrate condition varied from time to time. All of these factors contribute variability to the acoustic (and vibration) data. The “drilling peak” near 1 kHz is emphasized in Figure 3.5, which compares narrowband spectra (5– 10,000 Hz) of underwater sound with and without drilling. These data represent two 8.5-s samples recorded in 2002 at station 13, 2 km from the drill rig, on two consecutive days: one with and one without drilling. Oil production was taking place on both occasions. During drilling, received levels were on average 13 dB higher (range 4.5–38.5 dB) across all frequencies; the difference was particularly marked for the 60–250 and 700– 1400 Hz bands. Certain tones were evident on both days, including those at 20, 40, 51, 60, 80 and 140 Hz, and were probably related to power generation on the island. Power generation tones at 60 and 120 Hz were common in all the recordings; in addition 60 Hz corresponded to the shaft rate for one of the compressors. The “drilling peak” was also found in drilling samples from 2001, without production taking place. In-Air Sound

Broadband Levels Received broadband (10–10,000 Hz, unweighted) levels of airborne sound are shown in Figure 3.6. To minimize wind contamination in the data, only recordings obtained with wind speeds <5 m/s (11 mph) were used. This eliminated all recordings from 8 and 9 March 2001. Both the highest (80 dB re 20 µPa) and lowest (44 dB re 20 µPa) broadband levels were obtained in 2002, at the closest (220 m) and farthest (9.4 km) stations, respectively (Fig. 3.6). Broadband levels close to the island were 10–15 dB lower in 2001, before oil production had started, than in 2002, despite the generally lower wind speeds in 2002 (Fig. 3.6). However, the measurements are not directly comparable, as they were obtained east of the island in 2001 and north in 2002. Chapter 3: Drilling & Operational Sounds, Winter 2001 & 2002 3-11

FIGURE 3.4. Received underwater sound pressure levels in one-third octave bands (5–8000 Hz center frequencies), for three sets of recording conditions. Data collected on 6 March 2001 in (A), and 28 Feb. and 1 March 2002 in (B) and (C).

The sound propagation model represented by equation (1) was fitted by least-squares to the 2002 data (during production) with and without drilling, and the resulting regression lines and equations are shown in Figure 3.6. Spreading loss terms (20.5 versus 19.6 dB/tenfold change in distance, respectively) and constant terms (A in Eq. (1), 128.6 and 126.7 dB, respectively) were very similar for both days. On 1 March 2002, which included all the recordings during drilling, broadband levels continued decreasing until the farthest station, 9.4 km from Northstar. During 28 Feb. 2002, the day with the least wind and also a day with no 3-12 Monitoring at Northstar: Ice-Covered Season

FIGURE 3.5. Narrowband spectra (5–10,000 Hz) of underwater sound recorded from the same location (station 13, 2 km from the drill rig) on two consecutive days in 2002, one with drilling and one without. Oil production was occurring at both times. drilling, Northstar was still faintly audible to the field crew at the 5 km station. During that recording, fluid was being inserted and emptied down the annulus while an 8.5” bottom hole assembly was being made up on the rig floor. Spectral Characteristics Figure 3.7A compares one-third octave band levels for two stations at which recordings were made on two consecutive days, with and without drilling (in 2002, while production was taking place). Band levels were similar with vs. without drilling, consistent with the close similarity found in broadband levels with vs. without drilling (cf. Fig. 3.6). Distinct peaks can be seen at the one-third octave bands centered at 40 and 80 Hz; these peaks are present at 2 km both with and without drilling. • A tone at 42–45 Hz accounts for the peak in the one-third octave band centered at 40 Hz. This tone was evident at 11 out of 12 stations sampled in 2002; the only location at which it was absent was the 7.6 km station (#20) on 28 Feb. The tone was found in 53 of 65 samples analyzed (82%) and can be seen in Figure 3.7B, showing narrowband spectra from the closest and farthest stations on 1 March 2002 (during drilling). This tone is likely part of the island’s power generation. The sound propagation model (Eq. 1) was fitted by least squares to the SPL for the tone as received at every station over both days (Fig. 3.7C). The spreading loss term was 22.3 dB/tenfold change in distance. The greatest difference between the two spectra, about 60 dB, was found at about 350 Hz (Fig. 3.7B). Chapter 3: Drilling & Operational Sounds, Winter 2001 & 2002 3-13

FIGURE 3.6. Broadband (10–10,000 Hz, unweighted) levels of airborne sound as a function of drilling, production, and distance from the drill rig on Northstar Island. Data collected on 6 March 2001 (eastward, gray triangles), and 28 Feb. and 1 March 2002 (northward, open and filled circles, respectively). Station numbers are indicated above their respective data sets. Also shown are the logarithmic spreading loss models (see text for details). 10,000 m = 10 km = 6.2 mi.

• The peak in the one-third octave band centered at 80 Hz was found only in three recordings: on 28 Feb. at the 1 and 2 km stations (#11 and 13, respectively) and on 1 March at the 2 km station (#13). In the latter case, the peak was present in only part of the recording, suggesting that it was probably linked to a particular activity or piece of machinery. Activities reported at the rig during these three recordings did not shed any light on the possible origin of the tone. Higher frequencies (>2 kHz) leveled off 2–5 km from Northstar, whereas lower frequencies (<630 Hz) continued decreasing with increasing frequency out to the farthest station, 9.4 km from Northstar. Iceborne Vibrations

Broadband Levels Mean broadband (10–500 Hz) particle velocity (PV) levels, as recorded by the geophone channels, are shown in Figure 3.8A in the absence of drilling, and in Figure 3.8B during drilling. Both plots display 2002 data. In 2001, geophone measurements were only made on 9 March. Those recordings were affected by wind and showed high variability, particularly for the vertical and H2 channels. The less-affected H1 data are shown in Figure 3.8B for comparison with 2002. Broadband levels for the H1 channel during drilling were very similar in 2001 with no production and no gas turbines (gray filled circles in Fig. 3.8B) and in

2002 with production and gas turbines (black circles). However, in 2001 the H1 values leveled off at 2 km. 3-14 Monitoring at Northstar: Ice-Covered Season

FIGURE 3.7. Spectra and received levels of airborne sounds in relation to distance. (A) One-third octave band levels (12.5–8000 Hz center frequencies) for two stations where recordings were made on two consecutive days, one with and one without drilling (28 Feb. and 1 March 2002; production occurring at all times). (B) Narrowband spectra (10–10,000 Hz) from the closest (0.2 km) and farthest (9.4 km) stations on 1 March 2002, during drilling and production. One representative 8.5-s sample was chosen from each station. (C) Unweighted received level as a function of distance from Northstar for the 42–45 Hz tone on 28 Feb. and 1 March 2002. Chapter 3: Drilling & Operational Sounds, Winter 2001 & 2002 3-15

FIGURE 3.8. Mean broadband (10–500 Hz) particle velocity levels as a function of distance from the drill rig, for all three geophone channels and two production and drilling conditions in 2002 (28 Feb. and

1 March). Values for channel H1 in 2001 (9 March, during drilling but no production) are also shown in (B). Otherwise as in Figure 3.3.

Beyond that distance, wind probably became the dominant source. In 2002, H1 broadband levels continued to decrease until the farthest station, at 9.4 km. The sound propagation model represented by equation (1) was fitted by least-squares to the 2002 data for H1, with (Fig. 3.8B) and without (Fig. 3.8A) drilling. The resulting regression lines and equations are shown in Figure 3.8. As for the in-air data, the spreading loss terms (19.2 versus 19.6 dB/tenfold change in distance) and constant terms (159.6 and 160.5 dB re 1 pm/s) were nearly identical with and without drilling. Spreading loss terms of ~20 log(R), where R is distance from Northstar Island, suggest that the island was the dominant source of the measured iceborne vibrations. 3-16 Monitoring at Northstar: Ice-Covered Season

Spectral Characteristics

Figure 3.9 presents mean one-third octave band levels for the H1 geophone channel (horizontal, pointing toward sound source) for three sets of recording conditions: (A) Drilling but no production; (B) Production but no drilling; and (C) Production and drilling. All three plots show elevated values in the one-third octave bands centered at 20–40 Hz, but the peak is more distinct and more closely related to range from Northstar during production (Fig. 3.9B,C). During drilling and production, this peak is small but still evident 9.4 km from the island (Fig. 3.9C). During production (Fig. 3.9B,C) there is also an increase in received one-third octave levels with increasing frequency above 100 Hz, particularly for the closer stations.

FIGURE 3.9. Received iceborne H1 particle velocity levels for one-third octave bands (6.3–400 Hz center frequencies), for three sets of production and drilling conditions. Data collected on 9 March 2001 in (A) and 28 Feb. and 1 March 2002 in (B) and (C). Chapter 3: Drilling & Operational Sounds, Winter 2001 & 2002 3-17

Figure 3.10 shows narrowband spectra (5–400 Hz) for the geophone vertical channel at four distances on 1 March 2002, during drilling and production. The largest peaks, at ~27 and 43 Hz, were present in all four recordings. Though the received levels diminished with increasing range, both peaks remained relatively prominent out to the farthest station, 9.4 km from Northstar. A smaller range- dependent peak at ~60 Hz, a power generation frequency, can also be distinguished out to 9.4 km. Above 200 Hz, differences in received levels at 1 and 9.4 km are small, on average <6 dB. Received levels at ~7 Hz were the same at 0.5, 1 and 9.4 km. A peak at this frequency has been noted in nearly all our recordings since the beginning of Northstar construction (see Greene and McLennan 2000). It does not seem to vary as a function of distance to the island or activity taking place, and has been noted independently by other observers (i.e. Shepard et al. 2001). Most likely, this peak is related to the physical environment of the ocean and bottom. Summary of Results

Sound levels were documented during four different sets of recording conditions, with and without drilling and oil production at Northstar. Results are summarized in Table 3.2 and in the three paragraphs below. Underwater, broadband levels were quite variable and changed considerably with time and activity on the island as well as with distance from the island. Spectral characteristics of underwater sound revealed baseline island sounds (power generation etc.) to be low in frequency, generally below 100 Hz. Production alone led to an increase in received levels at 125–160 Hz. Drilling resulted in an average increase in received levels of 13 dB across all frequencies (range 4.5–38.5 dB); the increase was particularly marked for the frequency bands 60–250 Hz and 700–1400 Hz. The peak centered at 1 kHz was detectable at the 5 km station, but not at the 9.4 km station.

FIGURE 3.10. Narrowband spectra (5–400 Hz) for geophone vertical channel at four different distances from Northstar, during drilling and production. 3-18 Monitoring at Northstar: Ice-Covered Season

TABLE 3.2. Summary of main results: highest broadband value recorded and distance (in m) at which it was recorded; broadband value at 1000 m from the source; “background” (i.e. lowest) value and distance at which this background value was reached; center frequency (f) of the main peaks in the one-third octave and narrowband data; spreading loss terms, when available. Underlined center frequencies are the more important ones.

Broadband Received Levels Spreading “Background” loss term Highest level Level level (closest Center (dB / tenfold recorded at distance with frequency change in (distance) 1000 m that level) of peak(s) distance)

Underwater sound (in dB re 1 µPa) a No drilling, no production (N) 98 (500 m) 83 83 (1 km) 7, 40 No drilling, production 102 (220 m) 91 76 (3–4 km) 7, 20, 50, 125 22.0 Drilling, no production (N) 116 (250 m) 92 77 (4–6 km) Drilling, no production (E) 124 (1000 m) 124 80 (5 km) Drilling and production 104 (220 m) 99 76 (5–9.4 km) 7, 50, 125, 1k 21.5

In-air sound (in dB re 20 µPa) a No drilling, production 74 (470 m) 69 52 (5 km) 20, 40, 80 19.6 Drilling and production 80 (220 m) 67 45 (9.4 km) 20, 40, 80 20.5

Iceborne vibrations (H1) (in dB re 1 pm/s) b Drilling, no production 114 (250 m) 104 96 (2–4 km) 25–40 No drilling, production 111 (470 m) 101 86 (5 km) 25 19.6 Drilling and production 114 (220 m) 104 80 (9.4 km) 20–31.5 19.2 a Bandwidth 10–10,000 Hz for underwater and in-air sound (unweighted). b Bandwidth 10–500 Hz for iceborne vibrations (unweighted).

In air, drilling had no obvious effect on SPLs, as received levels and spreading loss terms obtained during production were very similar with and without drilling. Spectral characteristics of in-air sounds confirmed this similarity. A tone at 42–45 Hz was present both with and without drilling, and in quiet conditions was detectable in the air at the farthest station (9.4 km). Iceborne vibration levels were elevated within 2 to (at least) 9.4 km of Northstar, depending on conditions. Results for the H1 channel (the horizontal channel pointing directly toward Northstar) are summarized in Table 3.2. Values for the vertical channel were generally higher and values for the H2 channel (perpendicular to a line from the sound source) were generally lower. As seen in the airborne measurements, drilling had no obvious effect on ice vibrations. One-third octave bands centered at 20–40 Hz showed elevated values, relative to other bands, both with and without oil production, but the effect was more pronounced during production.

DISCUSSION

Sounds during Drilling

Drilling sounds were readily identifiable underwater (see Fig. 3.5), with a marked increase in received levels at 60–250 Hz and 700–1400 Hz relative to no-drilling times. The higher-frequency peak, Chapter 3: Drilling & Operational Sounds, Winter 2001 & 2002 3-19 which was distinct enough to be used as a drilling “signature”, was still clearly detectable 5 km from the drill rig, but had fallen to background values by the 9.4 km station. The lower-frequency peak straddled the range of frequencies involved in power generation on the island—60 Hz and multiples thereof, but most likely also 10 Hz and its harmonics (i.e., 20, 30, 40, 50, 70 and 90 Hz), which have been common in recordings since the beginning of construction at Northstar (see Greene and McLennan 2000). During drilling it is reasonable to expect an increase in the level of sound and vibration from any equipment having to work harder, such as the machinery for power generation or drilling. In air it was not possible to distinguish drilling sounds from overall island sounds based on spectral characteristics (Fig. 3.7A) or on broadband levels (Fig. 3.6). The logarithmic spreading loss models obtained with and without drilling were very similar. There was no temporal variation in SPLs within stations during drilling on 1 March 2002 (unlike the underwater measurements), and the associated r-value for received levels vs. distance was very high. In addition, the field crew could not distinguish drilling sounds from overall island sounds by listening in the field, even at the stations closest to Northstar (220–470 m away).

A similar result was found for recordings from geophone channel H1: broadband levels of iceborne vibrations with or without drilling were indistinguishable (Fig. 3.8), and the elevated levels for one-third octave bands centered at 20–40 Hz during production (Fig. 3.9B,C) were the same or only marginally higher during drilling (see, for example, values for the 1 and 5 km stations in Fig. 3.9C). Thus, the results (see Table 3.2) show that airborne sounds and iceborne vibrations were not strong enough during drilling to have much influence on overall Northstar sounds, in contrast to underwater sounds, which were notably higher during drilling. Previous studies of underwater (i.e., under-ice) noise from winter drilling on icebound gravel islands and ice pads have indicated that drilling noise is clearly detectable at close ranges, but is generally confined to low frequencies and has low received levels. Drilling sounds transmit poorly from the drill rig machinery through gravel into the water (Richardson et al. 1995). • Malme and Mlawski (1979) measured the underwater sound associated with drilling rigs operating on two icebound gravel islands (one natural and one man-made) near Prudhoe Bay during March. They found that most drilling sound was below 200 Hz and that broadband levels reached ambient values ~1.5 km from the islands; with low ambient noise, low-frequency tones were measurable to ~9.5 km, but with high ambient noise they were undetectable beyond ~1.5 km. • Richardson et al. (1990) reported broadband (20–1000 Hz) drilling sounds from a rotary-table drill rig on an ice pad grounded in very shallow water (6–7 m). Again, drilling sounds were confined to low frequencies, <350 Hz. At 2 km drilling sounds reached 85 dB re 1 µPa and were barely detectable; this is equivalent to the values obtained in this study during drilling with no production (Fig. 3.3A). • Greene (1997) measured sounds from a rotary-table drilling operation on Tern Island, an artificial gravel island in the Liberty Prospect, east of Prudhoe Bay. Again water depth was 6–7 m, shallower than at Northstar. A comparison of spectra from the two sites, obtained during drilling 1160 m from Tern Island and 970 from Northstar, showed them to be very similar. SPLs in the Northstar recording were a few dB higher, particularly at low frequencies, but both spectra displayed prominent tones at 20, 40, 50, 60 and 120 Hz and heightened levels at ~30–200 Hz. Results from the present study indicate that winter drilling can produce underwater sound at somewhat higher frequencies, e.g. at 700–1400 Hz. Recording conditions were unusually quiet (low wind speeds) in 2002, when the “drilling peak” centered at ~1 kHz was detectable up to at least 5 km from Northstar. Inspection of unpublished higher-frequency data from Tern Island (from the work of Greene 1997) showed that the spectrum there also contained a “drilling peak” at 800–2000 Hz, similar to the results from Northstar (see Fig. 3.5). 3-20 Monitoring at Northstar: Ice-Covered Season

Sound levels from drilling on icebound caissons with concrete or metal sides have not been documented very extensively, but are expected to be somewhat higher than those from gravel islands given the (presumed) better coupling to the bottom and the water. The CIDS (Concrete Island Drilling System) is a self-contained floatable concrete structure that is ballasted down onto the bottom. Hall and Francine (1991) made under-ice recordings of the sounds produced by the CIDS during winter drilling with a rotary- table rig. Broadband (20–1000 Hz) levels 1370 m from the source were ~90 dB, which is ~4 and 2 dB above values obtained in this study without and with drilling, respectively, for the same frequency band. Hall and Francine found that CIDS also produced a much stronger infrasonic tone at 1.375–1.5 Hz, outside the bandwidth considered during this and other studies of drilling sound. The infrasonic tone was probably associated with the rotary table drillrig aboard CIDS, which turned at 75–110 rpm. Of all the offshore drilling operations in the Arctic, drillships and their support vessels apparently produce the highest sound levels. The drillship hull contains generators, the drill rig and other machinery, and is well coupled to the water. Also, support vessels are usually present near drillships. Greene (1987) measured sounds during drilling from the ice-strengthened drillships Explorer II and Kulluk. Broadband (20–1000 Hz) levels at 1, 2 and 5 km were nearly 40 and 50 dB higher, respectively, than levels recorded at the same distances during drilling at Northstar. The drillship recordings were made in open water, with a number of support ships present. A number of variables during drilling can potentially account for some of the temporal variation in broadband levels that is visible in Figure 3.3B. These include (1) composition of the layer being drilled through, which was gravel on 8 March 2001 and clay on 9 March 2001; (2) depth of the drill bit, which varied by an order of magnitude during our measurements (255–2880 m); (3) rate of penetration, ranging from ~6–150 m (~20–500 feet) per hour during our measurements; and (4) bathymetry, which was shallower east than north of the island. In addition, other sources of sound were operating on the island and in the Prudhoe Bay area, and could contribute to overall SPLs. On the island itself, the type and identity of the generators and turbines involved in power generation varied on a daily—and sometimes hourly—basis. Nevertheless, despite this variation in the data, fitting a logarithmic propagation loss model to the minimum values recorded during drilling resulted in a spreading loss term very close to that obtained with no drilling, i.e. about 20 dB/tenfold change in distance. Sounds during Production

Production operations were underway in early 2002 but not in early 2001. There was little evidence that oil production at Northstar during wintertime led to higher levels of underwater sound. Without drilling, underwater broadband levels recorded north of the island were similar with and without production (Fig. 3.3A, black and open circles). Likewise during drilling, levels north of the island were again similar with and without production (Fig. 3.3B, black and open circles). Although the broadband underwater levels did not seem to be affected appreciably by the occurrence of production, one-third octave analysis (Fig. 3.4) revealed a peak at 125–160 Hz that could be due to production. This peak was no longer detectable 5 km from the island, either with or without simultaneous drilling. Thus, oil production at Northstar did not cause any substantial increase in overall levels of underwater sound relative to the levels in the absence of production. However, production probably caused a change in frequency composition. This was to be expected for two reasons: (1) “No production” recordings were obtained while diesel generators provided the island’s power source (2001), whereas “production” recordings were obtained after the island had shifted to gas turbines (2002). (2) Production implies the use of compressors, which were a new sound source. There were not enough good quality in-air recordings to determine the effect of oil production on airborne levels. In the absence of drilling, measurements made north of the island with production (open Chapter 3: Drilling & Operational Sounds, Winter 2001 & 2002 3-21 circles in Fig. 3.6) had higher broadband levels than did measurements at similar distances east of the island with no production (triangles in Fig. 3.6). Although consistent with the expectation that in-air sounds would be stronger with than without production, the apparent difference could also be a function of recording location and weather conditions. There was no evidence that ice vibration levels, on a broadband (10–500 Hz) basis, were affected by the start of oil production or the associated change in power source. Geophone recordings north of the island during 2001 and 2002 yielded nearly identical values up to 2 km from Northstar (Fig. 3.8B; gray vs. black circles). At 2 km, particle velocity levels in 2001 flattened out at ~97 dB re 1 pm/s, most likely because of the higher wind speeds during those recordings. However, levels in the 20–40 Hz band, which were already somewhat elevated with drilling and no production, were further elevated with production and drilling combined (Fig. 3.9A,C). This was probably a result, at least in part, of the increased demand for power generation during production. This peak at 20–40 Hz was still evident 9.4 km from Northstar (Fig. 3.9C, 3.10). Sounds from Power Generation

The power source at Northstar Island consisted of diesel generators during acoustical measurements in early 2001 and gas turbines in early 2002. Also, compressors were in use in 2002 to compress and inject natural gas. Overall, the transition did not seem to have resulted in detectable changes in broadband levels of island sounds in the water or in the ice. In-air levels might have increased by a few dB, though the evidence for this (Fig. 3.6) was also subject to other interpretations. Tones expected from the generators and turbines could either not be found or appeared unpredictably in time and space. The sounds from power generation were low frequency in nature and contained a number of tones (and harmonics thereof) at frequencies below 100 Hz. These tones were also apparent during summer measurements around Northstar (see Blackwell and Greene 2001). Background Levels

Underwater background levels, as recorded at the farthest stations, were ~80 dB re 1 µPa in 2001 and ~75 dB in 2002 (broadband, 10–10,000 Hz band). The year-to-year difference can probably be explained by the difference in mean wind speeds, 4.5–9.4 m/s in 2001 vs. 0.1–4.7 m/s in 2002. The spectrum levels of these background sounds were well below Knudsen et al.’s (1948) sea state 0 curve, and fall within the range of other under-ice ambient noise measurements (e.g., Greene and Buck 1964; Milne and Ganton 1964). In-air sound levels recorded during drilling and production (filled circles in Fig. 3.6) decreased to as low as 44 dB re 20 µPa (broadband, 10–10,000 Hz) at the farthest station (9.4 km). However, because the values had not leveled off, we cannot state with certainty that these were background levels. If A-weighted, these data convert to broadband levels of 20 dBA re 20 µPa. This value falls at the lower range of Kinsler et al.’s (1982) “wilderness” category (given in dBA), corresponding to some of the quietest environments found in nature. They are also the lowest in-air broadband levels we have recorded in the area since Northstar construction began in 2000. In 2002, broadband (10–500 Hz) PV levels measured in the ice during drilling and production, considering geophone axis H1 (pointing toward Northstar), continued decreasing until the farthest station (9.4 km), reaching a low of 80 dB re 1 pm/s. Greene and Buck (1964) reported spectrum levels of vertical velocity (25–1000 Hz band) as received by a seismometer (geophone) in thin (15 in. or 38 cm) ice far offshore in the Beaufort Sea. Recording conditions during our measurements were exceptionally quiet, so we compared vertical data from our most distant station (9.4 km from Northstar, Fig. 3.10; ice thickness 1.5 m or 59 in.), with the lowest values found by Greene and Buck (1964). Broadband levels 9.4 km from 3-22 Monitoring at Northstar: Ice-Covered Season

Northstar probably were close to background values. Minimum values measured by Greene and Buck (1964) were about 5 dB lower than those in the present study at 25 Hz, but higher by 9–11 dB at 50, 100, 200 and 400 Hz. Greene and Buck’s measurements were made during April in the central Arctic ice pack, a more dynamic environment (in terms of ice movement) than the shore-fast ice of our study. This could explain their higher values at frequencies above 30 Hz. Also, their study area was completely removed from industrial activity, whereas our data at 9.4 km from Northstar still contain evidence of industry-related vibrations, such as the peaks at 24 and 43 Hz (Fig. 3.9). The shallow water depth around Northstar will tend to limit the transmission of low frequencies. This results in a “dip” in received levels, mainly between 10 and 20 or 30 Hz, which can be seen in data from both the hydrophone (Fig. 3.4, 3.5) and the geophone (Fig. 3.9, 3.10). The lower received levels at ~10–30 Hz than at 30–50 Hz could therefore be primarily a result of low-frequency cut-off in propagation.

AUDIBILITY RANGES AND CONCLUSIONS

The ranges at which Northstar sounds reached background levels varied among the three media (water, air, and ice) and are summarized in Table 3.2. These ranges are strictly the distances at which broadband levels flattened out; the stronger frequency peaks or tones produced by Northstar may be detectable beyond these distances. The ranges were naturally also a function of the recording conditions, determined principally by wind speed. Wind affected all sensors in 2001, but particularly the microphone and the geophone. The ranges at times with the best recording conditions and the highest source levels yield conservative (high) estimates of island audibility. Therefore we mainly based our ranges on data collected in 2002, during production. Northstar sounds then reached background levels underwater by 9.4 km (5.8 mi) with drilling and 3–4 km (1.9–2.5 mi) without. In air, background levels were reached 5–10 km (3.1–6.2 mi) from Northstar, depending on the wind, irrespective of drilling. In the ice, background vibration levels were reached 2–10 km (1.2–6.2 mi) from the island, depending on the geophone channel and the wind speed. At times with high background noise (e.g., windy periods), Northstar sounds disappeared below ambient levels at closer distances. Northstar is located in shallow Arctic waters and was surrounded by land-fast ice during this study in early 2001 and early 2002. The island was also mainly in a drilling and/or production phase, rather than an active construction phase. [Greene and McLennan (2000) provide corresponding data for the initial period of heavy construction during the winter of 2000.] We conclude that the sounds produced during the activities studied here reached background levels by 10 km (~6 mi) in quiet ambient conditions, and by 2–5 km (~1–3 mi) in noisier ambient conditions. Drilling produced a wide range of SPLs, depending on the type of drilling activity. Overall, sound levels from Northstar during drilling and production were low in comparison with drillships or drilling from caissons, and comparable to drilling from grounded ice pads or other artificial islands. Sounds recorded from all three types of sensors showed a rather consistent spreading loss of about 20 dB/tenfold change in distance over the range of distances studied (0.2 to 9.4 km, or 0.1 to 5.8 mi). The maximum detection distances presented above are based on data extending, on a flat-weighted basis, from 10 to 10,000 Hz for underwater and airborne sounds, and from 10 to 500 Hz for iceborne vibrations. However, seal hearing is not flat weighted (Richardson et al. 1995, p. 211-16; Kastak and Schuster- man 1998). The auditory sensitivity of ringed seals, which occupy the land-fast ice near Northstar, has been documented only at frequencies ≥1 kHz underwater (Terhune and Ronald 1975), and not at all in air. More information exists for harbor seals, Phoca vitulina (Møhl 1968; Terhune and Turnbull 1995; Kastak and Schusterman 1998; Schusterman et al. 2003), which are close relatives of ringed seals. To assess the audibility of Northstar sounds to seals, we compared one-third octave band levels of recorded sounds to available audiogram data for both species, both underwater (Fig. 3.11A) and in air (Fig. 3.11B). These comparisons Chapter 3: Drilling & Operational Sounds, Winter 2001 & 2002 3-23

FIGURE 3.11. Comparison of one-third octave band levels of industrial sound with hearing threshold levels for harbor and ringed seals. (A) Underwater: ringed seal audiogram from Terhune and Ronald (1975); “composite” audiogram for harbor seals based on three sets of measurements (Møhl 1968; Terhune and Turnbull 1995; Kastak and Schusterman 1998). (B) In air: open circles show a “composite” audiogram for harbor seals based on the same studies as in (A); closed circles show recent unpublished data from Schusterman et al. (2003). D = drilling, E = east, N = north, P = production. assume that the approximate filter bandwidth for seal hearing is one-third octave. Given that, plus the limited available data on auditory sensitivity of seals, especially at low frequencies, these comparisons should be recognized as first approximations. Underwater received levels 200 and 990 m to the east were above threshold for many frequencies, whereas levels at 1970 m (and to the north) were well below the auditory thresholds. Maximum detection distance for a seal would be ~1.5 km. In air received levels were compared with two different sets of threshold levels: (1) a “composite” audiogram based on three published studies (open circles in Fig. 3.11B), and (2) recent unpublished audiometric data by Schusterman et al. (2003) (filled circles in Fig. 3.11B). 3-24 Monitoring at Northstar: Ice-Covered Season

• Comparison with the composite (older) audiogram data shows that received levels at 220 and 470 m were above the harbor seal thresholds, whereas levels at 970 and 1970 m were below threshold at all measured frequencies. Maximum detection distance would be ~0.75 km. • In contrast, one-third octave band levels at some frequencies remained above the more recent audiogram out to a distance of ~5 km. The 5 km station was the closest one where received levels were below the “new” hearing threshold for the harbor seal at frequencies under 1.6 kHz. Near and above 1.6 kHz, one-third octave levels were identical at the 5, 7.6 and 9.4 km stations, indicating that island sounds were no longer contributing measurably to these frequencies and the seals were unlikely to hear components of island sound above 1.6 kHz. Thus, sounds were probably audible to seals, at least intermittently, out to ~1.5 km in water and nearly 5 km in air. The 5 km value assumes that ringed seals hear as well in air as suggested by the new harbor seal data of Schusterman et al. (2003). However, given the wide variation in received levels, the sounds would not always be audible to seals this far away. For example, sounds recorded underwater 220 m north of the island during production and drilling (Fig. 3.11A) probably would have been barely audible to a seal at that location, whereas those recorded 200 m east of Northstar with neither production nor drilling were probably quite audible. In addition, during February and March ringed seals spend much of their time on the ice inside snow lairs (Williams et al. 2002). Snow has an important dampening effect on in-air sounds, on the order of ~40 dB (broadband) per meter of snow thickness (Blix and Lentfer 1992). This would reduce considerably the range of detectability of in-air sounds to a ringed seal in its lair. Finally, the detectability of ice vibrations to seals cannot at this point be discussed as there is no information on the response thresholds of phocid seals to ice vibrations.

ACKNOWLEDGEMENTS

This study was funded by BP Exploration (Alaska) Inc. to fulfill BP’s policy of documenting and minimizing environmental impacts and to satisfy monitoring requirements associated with the Northstar oil development. We thank Jonah Leavitt from LGL Alaska for help in the field, and B. Dunbar, D. Otton, C. Hermans and C. Lester with Alaska Interstate Construction (AIC) for being our all-terrain vehicle drivers. M. Williams, S. Schwenn, and C. Perham from LGL Alaska helped with logistics. For logistical support and problem solving at Northstar we thank J. Huey of BP, R. Greenway and R. Vogel of AIC, and W. Cul- lor and A. Erickson from BP’s Environmental Group. J. Huey provided the picture of Northstar in Figure 3.2. G. Kidd, J. Polya, J. McDade, S. Carboy and M. Johns provided us with information on drilling activities and machinery schedules. Dr. Bill Burgess, of Greeneridge Sciences, helped with Figure 3.1. We thank Dr. Ron Schusterman for permission to use the recent harbor seal audiometric data. Finally, we thank Dr. Ray Jakubczak, Dr. Bill Streever of BP, and Dave Trudgen of Oasis Environmental for their support, and Dr. Bill Streever of BP and Robert Suydam of the North Slope Borough for their critical reviews.

LITERATURE CITED

Blackwell, S.B. and C.R. Greene Jr. 2001. Sound measurements, 2000 break-up and open-water seasons. p. 7-1 to 7-55 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of industrial sounds, seals, and whale calls during construction of BP's Northstar oil development, Alaskan Beaufort Sea, 2000. [April 2001 ed.] LGL Rep. TA2429-2. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 316 p. Chapter 3: Drilling & Operational Sounds, Winter 2001 & 2002 3-25

Blix, A.S. and J.W. Lentfer. 1992. Noise and vibration levels in artificial polar bear dens as related to selected petroleum exploration and developmental activities. Arctic 45(1):20-24. Greene, C.R., Jr. 1987. Characteristics of oil industry dredge and drilling sounds in the Beaufort Sea. J. Acoust. Soc. Am. 82(4):1315-1324. Greene, C.R., Jr. 1997. Underice drillrig sound, sound transmission loss, and ambient noise near Tern Island, Foggy Island Bay, Alaska, February 1997. Rep. from Greeneridge Sciences Inc., Santa Barbara, CA, and LGL Alaska Research Associates Inc., Anchorage, AK, for BP Explor. (Alaska) Inc., Anchorage, AK. 24 p. Greene, C.R. and B.M. Buck. 1964. Arctic Ocean ambient noise. J. Acoust. Soc. Am. 36(6):1218-1220. Greene, C.R., Jr., and M.W. McLennan, with R.W. Blaylock. 2000. Sound and vibration measurements during North- star construction in early 2000. p. 4-1 to 4-31 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of ringed seals and sounds during construction of BP's Northstar oil development, Alaskan Beaufort Sea, winter and spring 1999-2000: 90-day report. LGL Rep. TA2426-1. Rep. from LGL Ltd., King City, Ont., and LGL Alaska Res. Assoc. Inc., Anchorage, AK, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 107 p. Hall, J.D. and J. Francine. 1991. Measurements of underwater sounds from a concrete island drilling structure located in the Alaskan sector of the Beaufort Sea. J. Acoust. Soc. Am. 90(3):1665-1667. Harris, F.J. 1978. On the use of windows for harmonic analysis with the discrete Fourier transform. Proc. IEEE 66(1):51-83. Kastak, D. and R.J. Schusterman. 1998. Low-frequency amphibious hearing in pinnipeds: methods, measurements, noise, and ecology. J. Acoust. Soc. Am. 103(4):2216-2228. Kinsler, L.E., A.R. Frey, J.V. Sanders and A.B. Coppen. 1982. Fundamentals of acoustics, 3rd edition. Wiley, New York, NY. 480 p. Knudsen, V.O., R.S. Alford and J.W. Emling. 1948. Underwater ambient noise. J. Mar. Res. 7(3):410-429. Malme, C.I. and R. Mlawski. 1979. Measurements of underwater acoustic noise in the Prudhoe Bay area. BBN Tech. Memo. 513. Rep. from Bolt Beranek & Newman Inc., Cambridge, MA, for Exxon Prod. Res. Co., Houston, TX. 74 p. Milne, A.R. and J.H. Ganton. 1964. Ambient noise under Arctic-sea ice. J. Acoust. Soc. Am. 36(5):855-863. Møhl, B. 1968. Auditory sensitivity of the common seal in air and water. J. Aud. Res. 8(1):27-38. Moulton, V.D., W.J. Richardson, T.L. McDonald, R.E. Elliott and M.T. Williams. 2002. Factors influencing local abundance and haulout behaviour of ringed seals (Phoca hispida) on landfast ice of the Alaskan Beaufort Sea. Can. J. Zool. 80(11):1900-1917. Moulton, V.D., W.J. Richardson, M.T. Williams and S.B. Blackwell. 2003. Ringed seal densities and noise near an icebound artificial island with construction and drilling. Acoust. Res. Let. Online 4(4):112-117. Richardson, W.J., C.R. Greene, Jr., W.R. Koski, C.I. Malme, G.W. Miller, M.A. Smultea and B. Würsig. 1990. Acoustic effects of oil production activities on bowhead and white whales visible during spring migration near Pt. Barrow, Alaska–1989 phase. OCS Study MMS 90-0017; LGL Rep. TA848-4. Rep. from LGL Ltd., King City, Ont., for U.S. Minerals Manage. Serv., Herndon, VA. 284 p. NTIS PB91–105486. Richardson, W.J., C.R. Greene, Jr., C.I. Malme and D.H. Thomson. 1995. Marine mammals and noise. Academic Press, San Diego, CA. 576 p. Schusterman, R.J., D. Kastak, B.L. Southall, C. Reichmuth Kastak and M.M. Holt. 2003. Noise-induced temporary threshold shift in pinnipeds: effects of exposure medium, intermittence, duration, and intensity. Presentation at ECOUS (Environmental Consequences Of Underwater Sound) meeting, 12-16 May 2003, San Antonio, TX. 3-26 Monitoring at Northstar: Ice-Covered Season

Terhune, J.M. and K. Ronald. 1975. Underwater hearing sensitivity of two ringed seals (Pusa hispida). Can. J. Zool. 53(3):227-231. Terhune, J. and S. Turnbull. 1995. Variation in the psychometric functions and hearing thresholds of a harbour seal. In: R.A. Kastelein, J.A. Thomas and P.W. Nachtigall (eds.), Sensory systems of aquatic mammals. De Spil Publ., Woerden, Netherlands. Shepard, G.W., P.A. Krumhansl, M.L. Knack and C.I. Malme. 2001. ANIMIDA phase I: ambient and industrial noise measurements near the Northstar and Liberty sites during April 2000. BBN Tech. Memo. 1270; OCS Study MMS 2001-0047. Rep. from BBN Technologies Inc., Cambridge, MA, for U.S. Minerals Manage. Serv., Anchorage, AK 66 p. Williams, M.T., T.G. Smith and C.J. Perham. 2002. Ringed seal structures in sea ice near Northstar, winter and spring of 2000-2001. p. 4-1 to 4-33 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of industrial sounds, seals, and whale calls during construction of BP’s Northstar Oil Development, Alaskan Beaufort Sea, 2001. LGL Rep. P557-2. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. CHAPTER 4:

FIXED-WING AERIAL SURVEYS OF SEALS NEAR BP’S NORTHSTAR AND LIBERTY SITES, 2002 1

Valerie D. Moulton a, Robert E. Elliott b and Michael T. Williams c

a LGL Ltd., environmental research associates 388 Kenmount Rd., POB 13248 Stn. A, St. John’s, Nfld. A1B 4A5 (709) 754-1992; [email protected]

b LGL Ltd., environmental research associates 22 Fisher St., POB 280, King City, Ont. L7B 1A6 (905) 833-1244; [email protected]

c LGL Alaska Research Associates, Inc. 1101 E. 76th Ave, Suite B, Anchorage, AK 99518 (907) 562-3339; [email protected]

for

BP Exploration (Alaska) Inc. Dept of Health, Safety & Environment 900 East Benson Blvd., POB 196612 Anchorage, AK 99519-6612

LGL Report TA2702-2

May 2003

1 Chapter 4 In: W.J. Richardson and M.T. Williams (eds.). 2003. Monitoring of industrial sounds, seals, and bowhead whales near BP’s Northstar Oil Development, Alaskan Beaufort Sea, 1999–2002. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 4-2 Monitoring at Northstar, Ice-Covered Season

ABSTRACT

This chapter documents ringed seal (Phoca hispida) densities and local distribution in the area of landfast ice that surrounded BP’s Northstar oil development in spring 2002. Aerial surveys for seals were conducted during 30 May–7 June 2002 in a manner consistent with surveys done during each spring since 1997. There had been construction activities at Northstar from early 2000 through late 2001, and oil production and drilling at Northstar from then through the time of the aerial surveys in the spring of 2002. These industrial activities had the potential to affect the local abundance and distribution of ringed seals. The survey design provided for intensive and repeated coverage of the study area within each spring season. In 2002, the surveys covered 3618 km2 or 1396 mi2 of landfast ice habitat. The overall observed density in 2002 was 0.72 seals/km2. Excluding areas where the summer water depth was <3 m, where seals were rarely seen, the overall observed density in 2002 was 0.83 seals/km2, as compared with 0.43 to 0.63 seals/km2 in 1997–2001. In 2002, sightings were most common in shallower portions (5–15 m or 16–49 ft) of the landfast ice zone, and in the zone 10–15 km or 6–9 mi shoreward of the fast ice edge. Seal sightings may have been most common at these intermediate distances from the ice edge because there were numerous cracks in the ice in this zone, and these provided suitable habitat for hauling out to molt. This may also explain, at least in part, why the overall ringed seal density for 2002 was higher than during previous survey years. Univariate analysis provided no conclusive evidence that Northstar industry activities significantly affected the observed ringed seal density in 2002. Chapter 5 describes a multivariate analysis of the combined 1997–2002 data, taking account of the confounding effects of weather, ice, and temporal factors that are known or expected to influence seal haulout and distribution. That analysis indicated that Northstar industry activities did not significantly affect ringed seal densities in 2002, or in 2000–01.

INTRODUCTION

This chapter is the sixth in a series of annual reports describing results of intensive, site-specific aerial surveys for ringed seals, Phoca hispida (Fig. 4.1), on landfast ice surrounding BP’s Northstar oil development and the possible Liberty development. The surveys during the spring seasons in 1997, 1998, 1999, 2000, 2001, and 2002 provided intensive and replicated survey coverage of this area within each spring season (Miller et al. 1998; Link et al. 1999; Moulton et al. 2000a, 2001, 2002a,b). When this study began in 1997, oil development had not begun at either Northstar or Liberty, and this study was designed to evaluate the effects on seals of future development at either site. The 1997 and 1998 surveys were pre-development “control” surveys (although some limited offshore industrial activities occurred around Liberty in early 1997 and east of Liberty in early 1998). Subsequently, development went ahead at Northstar but not at Liberty; however, the survey design remained unchanged. The 1999 survey followed limited industrial activity at Northstar, the 2000 survey followed intensive activities associated with island construction and pipeline installation at Northstar, and the 2001 survey followed limited construction and drilling. During the most recent ice-covered season in late 2001 and early 2002, oil production and gas injection were ongoing, an ice road to Northstar was re-built, several wells were drilled, and there were intermittent tests, inspections and drills near the island and along the pipeline corridor (see Chapter 2). This report describes results from the aerial surveys conducted in spring 2002, near the end of the period when the aforementioned activities had been occurring. The high intensity of aerial survey coverage within a 75 × 40 km (47 × 25 mi, or 40 × 22 n.mi.) area was designed to assess possible changes in seal density close to the industrial site(s) after oil devel- Chapter 4: Aerial Surveys of Seals, 2002 4-3

FIGURE 4.1. A ringed seal (Phoca hispida). Photograph by I. Stirling, Canadian Wildlife Service. opment began. The within-season replication was designed, in part, to allow detection and quantification of distributional effects that might be quite localized. Also, the replication was designed to assess and distinguish the relative contributions of industrial effects and various natural factors (e.g., date in season, ice conditions, bathymetry, weather) on numbers of seals hauled out at different places and times. These surveys include the essential elements of a Before–After/Control–Impact (BACI) study design, although there are other complicating factors. These surveys were initiated Before the Northstar and Liberty oil developments began, and have continued After Northstar development began. The surveys include areas potentially Impacted by oil development and other Control areas a substantial distance away. The present study design provides for spatial replication, at least of the “control” areas. BACI designs are considered optimal for field studies of environmental impact (Green 1979; Underwood 1992). The present study in May and June 2002 was a continuation of the work begun in 1997–2001, using the same survey design and methodology. It was the third year of surveys following the start of full-scale construction activities at Northstar in 2000, and the first year of surveys following completion of construction and the start of oil production. Some disturbance and localized displacement effects on seals overwintering near Northstar were a possibility in early 2002. This study was designed to test for these potential effects, with allowance for other factors (natural and otherwise) that might influence seal sightings within the study area. No industrial activities are known to have occurred at Liberty (or McCovey) during the winter of 2001–02, so the 2002 surveys provide an additional year of pre- development “control” data for those areas. In 2002, as in 1998–2001, a ringed seal study conducted by the University of Alaska occurred near Northstar during the aerial survey period (Kelly et al. 2002; 4-4 Monitoring at Northstar, Ice-Covered Season

Harding et al. 2003). Their research activities had some potential to affect numbers of seals counted by aerial observers. The chapter is a slightly modified version of the “90-day report” on the 2002 surveys (Moulton et al. 2002b), and as such provides a description and basic analysis of the 2002 survey results. This chapter is supplemented by Chapter 5 of this report. Chapter 5 provides a more comprehensive analysis of the combined 1997–2002 data, including multivariate and power analyses. Chapter 5 is intended for publication in a peer-reviewed scientific journal and builds on several publications, including • Moulton et al. (2002a): draft final report of the corresponding treatment of combined BP/LGL aerial survey data for 1997 through 2001. • Moulton et al. (2002c): journal paper describing the relationships between “natural” factors and observed ringed seal densities, based on data collected in 1997–99, before intensive Northstar construction. • Moulton et al. (2003): journal paper providing a preliminary account of ringed seal densities vs. distance from Northstar during 2000–02 as compared with pre-construction data (1997–99). The following subsections provide background information on seals in the study area, followed by a description of the specific objectives of this study.

Observed Ringed Seal Densities in the Area

The Alaska Department of Fish and Game (ADF&G) began conducting studies of ringed seal distribution and abundance in the central Alaskan Beaufort Sea in the early 1970s (Burns and Harbo 1972), and has conducted such surveys periodically up to 1999 (Frost and Lowry 1988, 1998, 1999; Frost et al. 1988, 1997, 2002). Recent ADF&G survey results (mid 1980s to 1999) have been summarized by Frost et al. (1997, 2002) and Frost and Lowry (1998, 1999). The survey effort by ADF&G in the 1996– 99 surveys was similar to that obtained during previous ADF&G studies conducted in the mid- to late 1980s (Frost et al. 1997, 2002). In zones considered to be industrial areas, parallel north-south transects spaced as closely as 1.85 km (1.15 mi; 1 n.mi.) apart were surveyed once each spring season. In other areas the ADF&G transects were 3.7 km apart and at least 60% of the transects were randomly selected to be flown once per year. Relevant density data from those reports are shown in Table 4.1. These studies have included aerial surveys of both fast ice and pack ice habitats. The Northstar and Liberty sites are well within the landfast ice zone during late May and early June when aerial surveys are usually conducted. The study area for the present BP project is entirely within sector B3 as defined by ADF&G. Sector B3 extends from Oliktok Point (149º51'W) to Flaxman Island (146º03'W). Considering only fast ice habitats in sector B3, observed ringed seal densities in 1985–87 and 1996–99 ranged from 0.54 to 2.41 seals/km2. Thus, seal densities observed on fast ice in this sector varied by about a factor of five among these seven years of ADF&G surveys. These density estimates include seals sighted both at holes and at cracks. Sector B3 has been sub-divided by ADF&G into an “industrial prospect area” and a “non-industrial area”. The industrial prospect area includes the proposed Northstar development area but not the Liberty development area. Lower ringed seal densities were observed in the “industrial prospect area” than in other fast ice areas within Sector B3 in 2 of 3 years for which data were presented (1986–87 but not 1985; Table 4.1). Observed ringed seal densities in adjacent sector B4 (Flaxman Island to Barter Island, 143º45'W) were generally higher than in Sector B3 in six of the seven years for which data are available, ranging from 0.60 to 3.00 seals/km2. Chapter 4: Aerial Surveys of Seals, 2002 4-5

2 TABLE 4.1. Observed seal densities (seals/km ) in fast ice for ADF&G sectors B3 and B4, excluding water depths <3 m (from Frost et al. 1997, 2002; Frost and Lowry 1998, 1999).

Sector B3a Sector B4b

Year Industrial Prospect Non-Industrial Overall Overall

1985 1.44 0.81 1.16 0.60 1986 1.21 1.26 1.28 2.70 1987 2.48 3.11 2.41 3.00 1996c 0.54 0.69 1997 0.77 1.26 1998 0.76 1.36 1999 0.87 1.93 a Extends from Oliktok Point (149º51'W) to Flaxman Island (146º03'W). b Extends from Flaxman Island to Barter Island (143º45'W). c Industrial vs. Non-Industrial data not available separately for 1996-99.

Green and Johnson (1983) found average seal densities of 0.74 seals/km2 in their Seal Island (Northstar) study area in June 1982, following island-construction activities during February–April 1982. Densities observed in a control area centered about 23 km (14 mi) west of the Seal Island survey grid averaged 0.66 seals/km2. Both the industrial and control areas were in the landfast ice zone, and the density calculations excluded areas of predominantly rough ice, areas inside the barrier islands, and areas with water <5.5 m (18 ft) deep. The 1997, 1998, 1999, 2000, and 2001 phases of the present study provided specific information about densities of ringed seals observed on the landfast ice near Northstar and Liberty in those years. Considering all surveyed areas >3 m (10 ft) deep, the observed densities in 1997–2001 were 0.43, 0.39, 0.63, 0.47, and 0.54 seals/km2, respectively. In 1997, 1998, 2000, and 2001, densities were highest in shallower water depths. Based on surveys conducted during late May–early June 1997, the highest ringed seal density was observed in depths of 5 to 10 m (0.51 seals/km2; Miller et al. 1998). Similarly, in 2000 and 2001, densities were highest in water depths ranging from 3–10 m; with the highest density observed in the 5–10 m stratum (0.69 and 0.76 seals/km2, respectively; Moulton et al. 2001, 2002a). In late May 1998, ringed seal densities varied significantly with water depth, and the highest densities were found in water depths of 10 to 15 m (0.59 seals/km2; Link et al. 1999). In 1999, more seals were sighted in deeper water; highest densities were observed in water depths ranging from 10–20 m and 30–35 m (0.88– 1.40 seals/km2; Moulton et al. 2000a, 2002c). All five surveys found very low seal densities in water <3 m deep (0.01–0.13 seals/km2) as compared with densities in areas ≥3 m deep (0.39–0.63 seals/km2). Much of the area <3 m deep is frozen to the bottom by late winter, and the remainder of this area has very little water below the ice. Ringed seal densities in the pack ice are more variable from year to year than are those in the fast ice (Frost et al. 1988). In the Beaufort Sea, seal densities in pack ice generally decrease with increasing distance from the ice edge out to about 37 km (23 mi) north of the ice edge. Pack ice habitat ≥18.5 km (12 mi) beyond the fast ice edge typically supports observed seal densities of about 0.29 seals/km2. 4-6 Monitoring at Northstar, Ice-Covered Season

During the six years for which ADF&G data are available (1986–87 and 1996–99), observed seal densities in the pack ice ranged from 0.84 to 1.47 seals/km2 in sector B3, and from 0.60 to 1.59 seals/km2 in sector B4 (Frost et al. 2002). Prior to the 1997–2001 phase of this study, there were few data on seal densities in lagoons such as the Stefansson Sound/Foggy Island Bay area where the Liberty Development is planned. Previous spring surveys of seals on the ice in the central Alaskan Beaufort Sea have concentrated on the landfast ice zone seaward of the barrier islands. Some surveys extended south into the more northerly parts of the lagoons, but generally did not attempt to survey the shallower parts of the lagoons. In 1997–2001, BP/LGL seal surveys found (as expected) very low densities of ringed seals in areas <3 m deep (Miller et al. 1998; Link et al. 1999; Moulton et al. 2000a, 2001, 2002a). However, in 1997–98 and 2000–01, lagoon areas ≥3 m deep had seal densities similar to those on landfast ice seaward of the barrier islands. The potential Liberty Development, in lagoon water 6.4 m (21 ft) deep, is an area where ringed seals do occur in late winter/spring. Factors Influencing Numbers of Seals Counted

Aerial surveys for ringed seals are usually flown in late May and June in the Alaskan Beaufort Sea when ringed seals haul out on the ice to molt and are therefore most easily counted from the air. However, ringed seals are not uniformly distributed on the ice and not all ringed seals haul out at the same time. Many habitat factors are known or suspected to influence ringed seal abundance and distribution. Similarly, temporal and weather factors likely influence the proportion of ringed seals hauled out and hence available to be counted by observers. Observed densities of ringed seals are also affected by variation in the effectiveness of the observers, which can involve variable observer experience, fatigue, seat position, and sighting conditions. Comparisons of seal counts by observers on the left vs. right sides of the survey aircraft, have found no significant differences (Frost et al. 1988). However, some observer differences are inevitable and the proportion seen will vary with observer abilities and alertness, visibility, and glare. Furthermore, even an attentive observer will see less than 100% of the seals on the ice along the flight track. As compared with many other marine mammals, ringed seals hauled out on the ice are relatively easy to count. They are usually conspicuous because of their dark appearance, which contrasts with the white background, and they usually occur either singly or in groups small enough to be counted accurately. Using simultaneous counts by primary and experienced backup observers, Frost et al. (1988) calculated that a single experienced observer sees 83% of the groups and 82% of the single seals hauled out on the ice within the aerial survey coverage. Factors Affecting Seal Abundance and Distribution Ringed seals inhabit a dynamic physical environment that makes between-year, and even within-year, comparisons of abundance and distribution problematic. Variable levels of ice deformation and snow conditions influence where ringed seals occur during their winter occupation of breathing holes and subnivean lairs (Smith and Stirling 1975, 1978). Frost et al. (1988, 2002) speculated that ringed seals prefer flat ice on which to bask, because they are better able to detect predators in areas of unobstructed vision. Breakup of the sea ice occurs earlier in some years than in others as a result of storm events or mild temperatures. Ice deterioration usually involves the formation of cracks and increased levels of melt water. The formation of cracks during spring break-up can result in an early influx of ringed seals as some individuals may move from unstable pack ice into adjacent areas of landfast ice (Finley 1979). Chapter 4: Aerial Surveys of Seals, 2002 4-7

Ringed seal densities observed during spring surveys, especially after the ice starts to deteriorate, may be biased upward by the presence of seals that spent the winter elsewhere (Smith and Harwood 2001). It is thought that pack ice seals may have biased the 1999 BP/LGL survey results upwards as concentrations of seals were observed near cracks extending from the pack ice into the landfast ice (Moulton et al. 2000a, 2002c). This influx of pack ice seals may increase group size as well (Moulton et al. 2000a). However, group size at holes is also known to increase in stable fast ice as the molt season progresses (Smith and Hammill 1981), probably in response to increased numbers of seals hauling out as molt proceeds and the decrease in availability of suitable haul-out holes due to melt water draining into them. Seals probably avoid hauling out in areas with higher levels of melt water because it is important that their skin be dry and warm to promote the growth of new hair during the molt (Feltz and Fay 1966). Water depth, which is relatively constant from year-to-year, may also influence seal abundance and distribution. Previous aerial surveys conducted by LGL in the Northstar area have shown that ringed seals are observed most frequently in water depths ranging from 3 to 20 m (10 to 66 ft; Miller et al. 1998; Link et al. 1999; Moulton et al. 2000a, 2001, 2002a,c). Factors Affecting the Proportion of Seals Hauled Out Both temporal and weather factors influence the proportion of ringed seals that haul out on the ice at any given time. As for other phocids that associate with ice, these factors are known or suspected to include: date within the spring season, time of day, solar radiation, cloud cover, temperature, wind speed, and wind chill (Finley 1979; Smith and Hammill 1981; Thomas and DeMaster 1983; Kovacs 1987; Lydersen and Kovacs 1993; Moulton et al. 2000b; Born et al. 2002). The effects and interactions of these variables are not fully understood. However, the proportion of ringed seals hauled out and the observed seal densities are usually found to be negatively correlated with wind speed (Finley 1979; Smith and Hammill 1981). Multivariate analysis of results from the 1997–2001 BP/LGL surveys indicates that the numbers of seals seen were positively related to cloud cover and air temperature, negatively related to wind speed, and not strongly related to heat loss (Moulton et al. 2002a). Frost et al. (2002) found inconsistent relationships between weather variables and ringed seal densities. More ringed seals haul out around mid-day than at other times of day (Finley 1979; Smith and Hammill 1981; Kelly and Quakenbush 1990). The mid-day period is approximately 11:00–17:00 h in Alaska, where solar noon occurs around 14:00. The multivariate analysis of the 1997–2001 BP/LGL surveys, which were usually conducted between 10:00 and 18:00 h, did not indicate a consistent relationship between time of day (within that range of hours) and seal sighting rate (Moulton et al. 2002a). Similarly, no consistent relationship between time of day and seal sighting rate was found by Frost et al. (2002). They suggested that a diel pattern in haul-out behavior was not obvious from their data because their surveys were flown only during the hours of peak haulout (Frost et al. 2002). Some of the effects of these variables on seal censuses can be minimized by standardizing aerial survey procedures, e.g., flying surveys at the same time of day and minimizing survey effort during extremely cold or windy conditions. However, even moderately different weather conditions on different days, or at different times during the same day, may result in different proportions of ringed seals hauling out on the ice, and thus differences in the observed numbers of seals. Effects of Industrial Activity on Ringed Seal Distribution

Several studies have attempted to measure the impacts of industrial activities on the distribution and densities of ringed seals in the Beaufort Sea: 4-8 Monitoring at Northstar, Ice-Covered Season

• Reduced numbers of seals have been reported within 3.7 km (2.3 mi) of artificial islands during some but not all years with industrial activity (Frost and Lowry 1988; Frost et al. 1988). In 1985, two artificial islands were being constructed (the original Northstar island and Sandpiper island), and drilling was underway at Seal Island (now referred to as Northstar). Aerial surveys showed a 50–70% reduction in seal density within 3.7 km as compared to 3.7–7.4 km (2.3–4.6 mi) away from the three islands in 1985. In 1986, in contrast, ringed seal density during one survey was 60% higher within 3.7 km vs. 3.7–7.4 km of Sandpiper Island, the only island active when surveys were flown in 1986. In 1987, all three artificial islands were inactive and seal densities were 12–30% lower closer to the islands. All of these results should be treated with caution, as the small amount of survey coverage limits meaningful comparisons of densities relative to distance from the island. Also, analyses did not account for factors like water depth that varied within 3.7 km vs. 3.7–7.4 km from the islands. • Green and Johnson (1983) found that, based on densities of ringed seal breathing holes, ringed seals avoided the immediate area of Seal Island during the winter that the island was constructed. The radius of discernible effects was not precisely determined but it was apparently on the order of a few kilometers. • Seal densities in an area of Vibroseis activity east of Tern Island in 1998 were significantly lower than those in an adjacent non-Vibroseis area in 1998. No such difference was evident between those same two areas during the previous year when there was no Vibroseis east of Tern Island (Link et al. 1999). Over larger areas, no changes in ringed seal distribution or numbers have been seen with respect to industrial activities of any type, including on-ice seismic surveys as well as artificial islands (Frost and Lowry 1988; Kelly et al. 1988). Green and Johnson (1983) concluded that, despite some apparent localized avoidance, the overall effects of the construction of Seal Island in the winter of 1982 on seal distribution and densities were insignificant. A study of ringed seal numbers and distribution in the Canadian Beaufort Sea, Amundsen Gulf, and Prince Albert Sound in 1984 found no correlation between ringed seal densities and proximity to industrial sites that had been active the previous year (Kingsley 1986). The studies summarized above were not designed specifically to control for the potential effects of other factors known to influence seal abundance and haul-out behavior—weather, local ice conditions, water depth, etc. Moulton et al. (2002a, 2003) used a multivariate statistical approach to identify and account for those factors that are known or expected to affect ringed seals, based on data from the 1997– 2001 phases of the present study. They found no indication of a reduced ringed seal density near the Northstar development in 2000 and 2001 after allowance for other factors.2 In fact, in 2001 higher seal densities were found close to Northstar than elsewhere in the immediate Northstar area. We originally hypothesized that the potential effects of the Northstar development on densities of ringed seals close to the Northstar facilities would be greatest in 2000, and smaller in 2001 and 2002, when there would be less ice-road traffic and less construction activity than during 2000. Given the 2000–01 results, any reduction in observed seal densities near Northstar in 2002 were expected to be small and localized, if present at all. In most of the previous studies, with the exception of BP/LGL surveys in 1997–2001, the survey coverage and number of seal sightings close to the industrial operations were small. This limited the ability to detect and quantify any avoidance effect that might exist. The present 1997–2002 study includ-

2 That analysis also showed no evidence of industry effects on seal density near Liberty in 1997, or in and near the vibroseis area in 1998, after allowance for other factors influencing numbers of seals (Moulton et al. 2002a). Chapter 4: Aerial Surveys of Seals, 2002 4-9 ed intensive site-specific survey coverage, replicated within each year, around the Northstar (and Liberty) oil development sites. This approach was used in order to obtain larger sample sizes in areas close to the development sites. In addition, the study was designed as a multi-year study. A more comprehensive analysis is possible by combining results from different years into a single overall analysis. This comprehensive multi-year analysis is included in Chapter 5. Chapter 5 also includes an analysis of the statistical power of this survey approach. Objectives of BP Aerial Survey

Aerial surveys of seals during the spring of 2002 were a required part of the monitoring necessary to meet the conditions of the Letter of Authorization issued by NMFS for 2002 (NMFS 2001). The purposes for the aerial survey task were summarized in more detail in NMFS (2000, p. 34029): “Continue an ongoing (since the spring, 1997) Before–After/Control–Impact Study on the distribution and abundance of ringed seals in relation to development of the oil and gas resources in the Central Beaufort Sea. Collection and analysis of data before and after construction is expected to provide a reliable method for assessing the impact of oil and gas activities on ringed seal distribution in the Northstar construction area.” BP’s monitoring plan for 2002 (BPXA 2001) stated that “The overall objective of [this task] is to document the local distribution and abundance of ringed seals in the central Alaskan Beaufort Sea before and after construction of the Northstar Project. These data will be used to evaluate whether there is any localized change in seal distribution and abundance attributable to Northstar and, if so, to quantify the extent of this effect…. The initial surveys (1997–98) were to provide data on baseline distribution and density prior to construction of offshore oil production facilities. The subsequent surveys (1999–2002) were to provide comparative data during and after construction. Aerial surveys in 2002 will be conducted in the same manner as in 1997–2001, to ensure that “baseline”, “construction”, and “post-construction” results are directly comparable.… The 2002 survey will be the final survey in the series for the study (1997–2002), and will provide “post-construction” data. The specific objectives we considered necessary to address the above requirements include the following: 1. Conduct repeated surveys of the study area during spring 2002, consistent with those conducted in 1997–2001, to obtain estimates of the relative abundance and observed density of ringed seals on the fast ice. 2. Determine the observed densities of ringed seals in the study area in relation to weather, time, and habitat variables in 2002. 3. Compare seal densities at varying distances from industrial sites active during early 2002 (especially Northstar). 4. Update previous multivariate analyses (Moulton et al. 2002a, 2003) to determine which factors (environmental and industrial) are related to observed densities of ringed seals on the fast ice in 2002 and 1997-2002 combined, and to characterize the nature of those relationships. 5. Based on the results of (2), (3) and (4), establish whether observed differences in densities at varying distances from Northstar and other industrial sites are likely related solely to habitat, temporal and weather effects, or whether differing exposure to industrial activities also has a significant effect. 4-10 Monitoring at Northstar, Ice-Covered Season

6. Based on the results of (5), review the adequacy of the study approach, and recommend any changes or improvements to the study design, methods or analyses that would improve any future assessment of industry effects (if any) on ringed seal abundance and distribution. This chapter addresses objective 1 and (in part) objectives 2 and 3. The objectives not addressed in this chapter are addressed in Chapter 5 of this report.

METHODS

Survey Design

In 2002, as in 1997–2001, two “grids” of aerial survey transects were flown between longitudes 147º06'W and 149º04.5'W, an east-west extent of about 75 km or 40 n.mi. (46 mi). Each grid consisted of 40 north-south transects spaced 1.85 km or 1 n.mi. (1.15 mi) apart. Each transect extended from the Beaufort Sea shoreline to roughly 37 km (23 mi) offshore or to the edge of the landfast ice if it was encountered and recognizable <37 km offshore (Fig. 4.2). One of the grids that was surveyed included some of the same transects flown by ADF&G during their wider-ranging ringed seal surveys in previous years. The second or alternate grid was offset from the first by 0.9 km (0.6 mi) to the east. In this report, we define a survey replicate as a complete survey of the 80 unique transects. In 2002, as in 1998, 1999, 2000, and 2001, two full survey replicates were completed.3 Survey replicate 1 was flown on 30 May– 2 June 2002 and replicate 2 was flown on 3–7 June 2002. In total, 4538 linear kilometers (2820 mi) of surveys were flown over landfast ice by LGL during the nine-day survey period. A 40-transect grid usually required two days to complete and an 80-transect survey replicate took four days to complete. Ideally, the 20 odd-numbered transect lines from one grid were flown on one day, and the 20 even-numbered lines from that grid were flown on the next day. The odd and even numbered lines from the alternate grid were then (ideally) flown on the third and fourth days. In this way, each day’s flight was designed to sample 20 of the 80 distinct transects spanning the study area, rather than sampling the eastern portions one day and the western portions the next. Thus, the entire study area was to be sampled four times during each replicate survey, and eight times during each year. In 2002, fog or low cloud prevented the timely completion of 20 full transects during two survey days (30 May, 5 June). The coverage of those 20-transect groups was completed on another day. In 2002, the survey took nine days to complete. The northern ends of repeated transects varied somewhat from day to day. Northbound transects were usually terminated when we had flown at least 37 km (23 mi) or when it was apparent that we had reached the northern edge of the fast ice. In 2002, it was quite difficult to discern the ice edge (see ICE CONDITIONS, later). The southern ends of transects were usually defined by the coastline. However, we sometimes avoided flying over narrow nearshore bands of deteriorated ice. Near the Endicott production facilities, we started or ended some transects 1–2 km (0.6–1.2 mi) north of Howe Island to avoid flying close to bird colonies located there.

3 In 1997, only one complete survey replicate (Replicate 1: 80 unique transects) was surveyed. Forty of the 80 transects were surveyed two additional times, referenced here as “Replicate 2” and “Replicate 3”. Thus, for 1997, Replicate 1 was complete but Replicates 2 and 3 were only 50% complete (Miller et al. 1998). Chapter 4: Aerial Surveys of Seals, 2002 4-11 tion to e east was also was e east . ) 2002 ( tar (Seal Island) and Liberty (Liberty and Tern Islands) in and rela Islands) Tern (Liberty and Liberty (Seal Island) tar grid of 40 transects offset 0.5 n.mi. (0.58 mi; 0.93 km) to th offset 0.5 n.mi. (0.58 mi; 0.93 km) of 40 transects grid ed here is the one evaluated in MMS the one evaluated is ed here pp site ma y ecific Libert ecific p 4.2. Central Alaskan Beaufort Sea showing locations of Norths locations Sea showing 4.2. Central Alaskan Beaufort IGURE flown twice. The s flown twice. aerial survey transects flown twice in May-June 2002. A similar 2002. in May-June flown twice transects survey aerial F

4-12 Monitoring at Northstar, Ice-Covered Season

Survey Procedures

The 2002 surveys, like the 2000 and 2001 surveys, were flown in a Turbo Commander 690A operated by Commander Northwest Ltd. of Wenatchee, Washington (Fig. 4.3). This twin-engine high- wing aircraft was equipped with turbo-prop engines, a GPS navigation system, and large bubble windows installed in the emergency exits. Two pilots were present during the entire survey.

FIGURE 4.3. The Turbo Commander aircraft used to fly the survey in 2002 (and 2000–01).

The survey procedures in 2002, as in 1997–2001, generally followed those of Frost and Lowry (1988). We used strip transect methodology, which has been standard for previous aerial surveys of ringed seals in Alaska. Surveys were conducted at an altitude of 91 m (300 ft) above sea level (ASL) and a ground speed of 222 km/h (120 knots). We surveyed transect strips 411 m (1350 ft) in width on each side of the aircraft. These strips extended from 135 m (443 ft) to 546 m (1791 ft) from the centerline. Strip boundaries were marked on the aircraft’s windows with tape at the appropriate inclinometer angles, which were 9.5º and 34º below the horizontal for surveys at 91 m altitude. Sightings of seals inside 135 m or beyond 546 m were recorded as off-transect sightings. For consistency with previous ringed seal surveys, we have not attempted to adjust the strip boundaries or calculated densities to take account of the “earth curvature” corrections described by Lerczak and Hobbs (1998). The transect width used in the BP/LGL surveys is likely underestimated but the error associated with inclinometer angles >5º is negligible. Also, the use of consistent survey methodology across years permits meaningful comparisons of survey results. Chapter 4: Aerial Surveys of Seals, 2002 4-13

The two primary observers occupied the right seat behind the co-pilot and the left seat behind the pilot. A third observer operated a computerized data logger and recorded weather conditions and polar bear sightings. This third observer did not record seal sightings and was positioned behind the right observer. The two primary observers sat at large bubble windows that allowed greater downward visibility than standard windows. The surveys were usually flown between 10:00 and 18:00 h Alaska Daylight Time, when numbers of seals hauled out on the ice were expected to be highest. (In the Prudhoe Bay area, the sun reaches its highest elevation near 14:00 h.) When sightability was severely impaired (usually by fog) for more than five of the 1-min time periods along a transect, that transect was re- surveyed later, and the data from the initial incomplete survey were discarded. Data Recording Procedures

The data logging system used in 2002 consisted of a portable computer, GPS unit (Garmin, model 12XL), custom written data logging software, and a data acquisition card and ribbon cable. A custom written software program (in VisualBasic, version 5.0) automatically recorded time and aircraft position (latitude and longitude) at 2-s intervals throughout the flights. This program also logged aircraft altitude data from the aircraft's radar altimeter via a data acquisition card and ribbon cable. As in previous years, the two primary observers recorded the time, visibility (n.mi.), ice cover (%), ice deformation (%), meltwater (%), sunglare (none, moderate, or severe), and overall sightability conditions (ranging from “excellent” to “impossible”) onto audio tape at the end of each 1-min (~3.7 km or 2.3 mi) time period. An electronic timer signaled the observers at 1-min intervals. Ice deformation was estimated by the observers on each side of the aircraft. At the end of each 1-min interval, the observers estimated the percent of the on-transect ice surface surveyed during the preceding minute that was deformed rather than smooth ice. The ice deformation estimates were categorized by intervals of 10%. Cracks and leads in the ice were also noted by the observers at the specific times when seen, allowing their locations to be extracted subsequently from the data log files. Environmental parameters were recorded by the computer operator (with the assistance of the pilots) at the start of each transect. These variables included cloud cover (in tenths), ceiling height (ft), visibility (n.mi.), wind speed (knots), wind direction (ºT), and air temperature (ºC). Wind data were acquired from the aircraft’s GPS and air temperature from a thermometer mounted externally on the aircraft. For each seal sighting, the observer dictated onto audio tape the species, number, habitat (hole or crack), and behavior (look, move, dive, or none) of the seal(s), and noted whether the sighting was on or off transect. When cracks extended outside the transect, only on-transect seals were considered in determining number. When polar bears were sighted, the observer recorded size/age/sex class when this was determinable, behavior, and direction of movement. The observers recorded the times of any sightings of industrial sites or activity, including ice roads, areas with evidence of seismic surveys, or artificial islands. Observers also recorded the time of any sighting of research activity on the ice associated with a study of ringed seal haul-out behavior by researchers from the University of Alaska. Analysis Procedures

The location of each seal sighting was determined by matching the time of the sighting with the position recorded for that time in the GPS logs. Time periods with severely impaired sightability conditions were excluded from all analyses. Each sighting was also linked to the environmental variables 4-14 Monitoring at Northstar, Ice-Covered Season recorded for the corresponding one minute (3.7 km) time period. The fast-ice edge was subjectively located by mapping open leads, shear zones, and areas where the observers classified the ice as pack; these areas were excluded from analyses. Water depth contours were developed based on all available depth soundings. Sounding data, obtained on CD-ROMs from NOAA, included Hydrographic Survey Data, Vol. 1, vers. 3.1, and Marine Geophysical Data/Bathymetry, Magnetics, Gravity, vers. 3.2. The 3-m, 5-m, and additional contours by 5-m intervals out to 45 m were derived using Vertical Mapper for MapInfo. These depth contours were used as a GIS layer. MapInfo was used to calculate the surveyed areas within each contour interval. Seal sightings were overlaid onto the GIS layer for water depth, and densities for both on-transect sightings and individual seals were calculated. Five kilometer “bins” of distance as measured from the ice edge shoreward were also plotted and used as a GIS layer. The on-transect surveyed area in each bin was calculated. In the same manner as described above, seal sightings were overlaid onto this layer, and seal sightings/km2 and individuals/km2 were calculated for each 5-km interval. We examined seal sightings in relation to distance from the Northstar development area. Ten 1-km “bins” of distance from the edge of the development area were plotted and used as a GIS layer (Fig. 4.10, later). In defining the edge of the Northstar development area, we included the ice road and Seal/North- star Island as part of that area. We did not include, as part of the “development area”, an area near Northstar where emergency equipment was occasionally tested, or the area along the pipeline route where monitoring for oil spills was done intermittently (Chapter 2). These industrial activities were considered to be limited in geographic extent and too infrequent (i.e., five or fewer operations during the 2001–02 ice-covered season) for there to be any likelihood of measurable effects on seal density, given the very limited evidence of such effects after much more intensive industrial activities in 2000–01 (Moulton et al. 2001, 2002a, 2003; Williams et al. 2002). The on-transect area per 1-km bin was calculated, and the seal sighting data were overlaid onto this layer to permit density calculations relative to distance from on-ice activities. The results from survey replicates 1 and 2 were combined because of the relatively small areas and numbers of sightings involved in this localized analysis. Water depths <3 m were excluded from these calculations. Statistical Tests

We used the chi-square (χ2) goodness-of-fit test to assess the significance of observed differences in ringed seal densities with respect to variables such as water depth and distance from industrial site. Strata were defined for each explanatory variable (see “Results” for stratum boundaries). Simultaneous Bonferonni-corrected 95% confidence intervals (CIs) were calculated for the observed proportions by strata. An expected proportion (based on available survey area) falling outside the CI for the observed proportion for that stratum was considered significantly different (Manly et al. 1993). Strata with very small surveyed areas of <30 km2 are excluded from analyses. All tests were conducted based on numbers of seal sightings (singletons or groups) rather than numbers of individual seals. Individual seals at the same hole were considered as one group or sighting. When seals were distributed along a crack, the total number of seals along a single crack (and within the transect strip width) typically comprised one group or sighting. The different seals within a closely spaced group are not statistically independent. The expected numbers of seal sightings in the various strata (if seal density were unrelated to the variable in question) were assumed proportional to the surveyed amounts of landfast ice within those strata. Chapter 4: Aerial Surveys of Seals, 2002 4-15

Although the statistical tests were always conducted on the basis of seal sightings (total number of singletons or groups seen), we discuss the results in terms of observed seal densities (individuals/km2). For univariate comparisons of seal densities with respect to physical factors that did not change during the course of the study, such as water depth, we considered the survey replicates to be non- independent. At any given location along each of those transects, these variables would be the same during each survey, and some of the same seals may have been seen repeatedly. To avoid pseudoreplica- tion problems associated with the lack of independence of these “repeated measures”, we examined each survey replicate (group of 80 unique transects) separately whenever possible. It should be noted that the location of survey lines varied somewhat from replicate to replicate. The statistical analyses in this chapter are limited to basic descriptive and univariate approaches, considering only the 2002 data. Chapter 5 includes a multivariate analysis of the combined 1997–2002 data, and considers the simultaneous influences of industrial and natural factors (temporal, weather, ice, etc.) known or expected to influence number of seals sighted. Chapter 5 also evaluates the statistical power of the aerial surveys and associated analyses as applied in this multi-year study.

DESCRIPTION OF THE STUDY AREA

Chapter 2 describes the specific area around BP’s Northstar offshore oil development, including the ice road, Seal/Northstar Island, and areas where various tests, inspections and drills were occasionally conducted during the winter of 2001–02. The following paragraphs describe the broader study area where the fixed-wing aerial surveys were conducted in 2002. This is the same study area as used for the corresponding 1997–2001 surveys. Water Depth

The study area is about 75 km (47 mi) in east-west extent, and extends about 40 km (25 mi) offshore (Fig. 4.2). Within this area, maximum water depth reaches about 30 m (98 ft) near the north ends of the survey transects. Barrier islands occur across much of the study area (Fig. 4.2). West of Prudhoe Bay, these barrier islands are fairly close to the mainland shore (2–7 km or 1–4 mi). However, in the generally shallower waters near and east of Prudhoe Bay, the barrier islands tend to be farther offshore, with some being as much as 20 km (12 mi) from shore. Waters inside the barrier islands are shallow. West of Prudhoe Bay, maximum depths of about 4.5 m (14.8 ft) occur in the narrow lagoons formed by the barrier islands. In the broader areas inside the barrier islands east of Prudhoe Bay (e.g., Foggy Island Bay), water depths reach a maximum of about 9 m (30 ft). The water depth is 12 m (39 ft) at the Northstar development site (formerly known as Seal Island), and 6.4 m (21 ft) at the potential Liberty development site in Foggy Island Bay. In this report, references to the Liberty site refer to the specific location considered in MMS (2002). Ice Conditions

The study area in the spring of 2002 extended beyond the edge of the landfast ice. As in previous years, survey coverage of pack ice north of the landfast ice has been excluded from our analyses. Thus, fast ice is the only habitat considered here. In spring, first-year sea ice in the Beaufort Sea is generally about 2 m (6.6 ft) thick. From the shore out to a water depth of about 2 m, the ice is frozen to the bottom, forming the bottom-fast ice zone. The ice in deeper parts of the landfast ice zone floats on seawater, with occasional grounded ridges. 4-16 Monitoring at Northstar, Ice-Covered Season

Sea ice forms in September or October, typically starting along shore where water is less saline and where wave action is often reduced. Initially the water is covered with slush and pancake ice, which gradually thickens into ice sheets. If storms occur during the early stages of freeze-up, the smooth sheet of ice can be broken into blocks, forming a chaotic mass of ice. These storm events are less severe inside the barrier islands and the landfast ice there tends to be stable and is usually very smooth. Offshore of the barrier islands, the ice is more subject to storm events and interactions with drifting pack ice during freeze up. These storm events result in the formation of rough, deformed ice with high (up to 10s of meters) ridges of ice rubble. The landfast ice in the study area formed between 3 and 10 October 2001 (http://www.natice.noaa.gov), but a storm in early November disrupted the ice (Chapter 2). As in 1997– 2001, rougher (more deformed) ice was very evident offshore of the barrier islands based on the data collected by aerial observers during the spring of 2002 (Fig. 4.4). In addition, during 2002 many cracks and leads were present in the ice within 5 km or 3 miles north of the barrier islands, particularly in the western portion of the study area. Cracks and leads were more prevalent in that area in spring 2002 than during 1997–2001. Hence, there was more open-water relatively close to shore in 2002 than in previous years. This made determining where the edge of the landfast ice was located quite difficult in 2002. The edge of the landfast ice in the Beaufort Sea typically occurs in a region of shear zones between the seasonal pack ice and polar pack ice (Kovacs and Mellor 1974). The extent of the seasonal pack ice can be highly variable and made up of a number of shear zones. The edge of the landfast ice moves seaward as the winter progresses and may extend many kilometers seaward by May and June in any given year, depending on the prevailing winds (Kovacs and Mellor 1974). In the Alaskan Beaufort Sea, the edge of the landfast ice generally coincides (approximately) with the 18 m depth contour. However, the ice edge can deviate greatly from this depth contour depending on the extent of ice shelves attached to the seaward side of grounded pressure ridges (Stringer 1974). It is generally recognized that the edge of the landfast ice extends just beyond the maximum depth (18–27 m) where grounded ridges can form (Wad- hams 2000). During previous LGL/BP surveys (1997–2001), shear zones have been concentrated along the 25 m depth contour, ~35–40 km from shore. We have noted in previous years that pack ice and flaw leads are more evident seaward of the second prominent shear zone, and we used the second prominent shear zone to indicate the edge of the landfast ice in 2002. Breakup of the sea ice is initially characterized by melting of the surface layer of snow, usually beginning in June and finishing by July (Wadhams 2000). The accumulation of meltwater on the ice surface and sometimes under the ice thins the ice from the top and bottom until a hole is formed (Wadhams 2000). In 2002, breakup was not far advanced even at the end of the seal survey (7 June), despite the fact that leads and cracks were unusually frequent rather close to the barrier islands. By 3 June 2002, we began to observe melt water on the ice, and by 6 June larger pools of melt water had formed, as a result of temperatures not falling below freezing overnight. Melt water increased and covered some large portions of the study area on 7 June, but was absent in other areas. In 2002, we observed significant open-water areas (leads and cracks) just offshore of the barrier islands; these persisted throughout the survey period. We observed large nearshore and offshore cracks and leads on the first day of surveying (30 May 2002) and also on the last day (7 June 2002). The locations of these open- water areas did not shift greatly during the survey period, indicating that these were consistent features of the landfast ice during at least the later half of the spring and not an indication of break-up. Chapter 4: Aerial Surveys of Seals, 2002 4-17 June e edge in the Central Alaskan Beaufort Sea study area, 30 May - 7 30 May area, Sea study in the Central Alaskan Beaufort e edge 4.4. Percent ice deformation and location of the landfast ic location of the landfast and 4.4. Percent ice deformation IGURE 2002. F 4-18 Monitoring at Northstar, Ice-Covered Season

RESULTS Ringed Seal Abundance and Distribution

Ringed seals were widely distributed throughout the study area during the BP/LGL aerial surveys in the spring of 2002 (Fig. 4.5, 4.6). A total of 2591 ringed seals (1879 sightings) were seen on-transect in fast ice habitat during the two complete replicates of the 80 transects. These surveys covered a total of 3618 km2 of fast ice habitat (~1800 km2 per replicate). These values exclude the survey coverage and associated seal sightings in areas of close pack ice habitat north of the landfast ice edge depicted in Figures 4.5 and 4.6. (Total survey coverage, including both fast and pack ice, totaled 5766 linear km or 4621 km2.) The observed overall density of seals on fast ice in 2002 was 0.66 seals/km2 during the first survey replicate and 0.77 seals/km2 during the second replicate. Combining the results from the two replicates yielded an observed seal density of 0.72 seals/km2. These averages include the part of the study area where water depth is <3 m. If that area (largely bottom-fast ice) is excluded, the overall observed densities in areas ≥3 m deep during replicates 1, 2, and both combined were 0.77, 0.88 and 0.83 seals/km2, respectively. All of these densities are “raw” densities of seals observed per km2 of fast ice surveyed, with no adjustments for the estimated proportions of seals not hauled out (availability bias) or on the ice but missed (detection bias). Although ringed seals were observed in most parts of the study area, there was again in 2002 an obvious tendency for lower densities in the shallowest parts of the lagoons (Fig. 4.5–4.7). However, from Prudhoe Bay eastward, many seals were sighted in the area seaward of the 3 m depth contour in the northern parts of lagoons. West of Prudhoe Bay, very few seals were seen in the lagoons, which are narrower there than farther east. North of the barrier islands, seals were concentrated in areas inshore of the 15 m depth contour, with lower sighting densities farther offshore. However, north of the 15 m contour there was a tendency for seals to occur in larger groups (≥3 individuals), especially during replicate 2 (Fig. 4.5, 4.6). There was an obvious tendency (Fig. 4.5, 4.6) for the densities in many parts of the study area to be similar during replicates 1 and 2. Places with high (or low) densities during replicate 1 also tended to have high (or low) densities during replicate 2. During the second survey replicate in 2002, 29.1% of the sightings were of groups consisting of ≥3 seals. Approximately 22% of the sightings during replicate 1 were of groups with ≥3 seals. The majority of ringed seals were observed at holes rather than at cracks. During replicate 1, 95.1% of the sightings and 85.8% of the individuals were at holes. During replicate 2, 93.9% of the sightings and 82.0% of the individuals were at holes. Observed Ringed Seal Behavior Most ringed seals did not exhibit any obvious “negative” response to the aircraft as it flew overhead. Many ringed seals (37.0%), including those seen on pack ice as well as fast ice, showed no obvious response (“None”), whereas 40.8% were observed to look up at the aircraft (Table 4.2). Only 1.5% (55 seals) of the observed ringed seals were seen to dive into their holes or cracks in apparent response to the aircraft. An additional 1.1% (39 seals) were recorded as moving on the ice without diving into the water (Table 4.2). Bearded Seal Sightings

Nine single bearded seals were recorded during the 2002 aerial surveys (Fig. 4.8). Four sightings were in the pack ice north of the ice edge and five sightings were in the landfast ice. Of the bearded seals sighted in the landfast ice, three sightings were relatively close to the ice edge but two sightings were well within the landfast ice, south of the barrier islands. Chapter 4: Aerial Surveys of Seals, 2002 4-19 rvey replicate 1 (transects 1–80), 30 May–2 May–2 June 2002. 30 1–80), (transects 1 replicate rvey 4.5. Distribution of ringed seal sightings during su during sightings seal 4.5. Distribution of ringed IGURE F 4-20 Monitoring at Northstar, Ice-Covered Season Chapter 4: Aerial Surveys of Seals, 2002 4-21 sightings during survey replicate 2 (transects 1–80), 3–7 June 2002. June 1–80), 3–7 2 (transects replicate survey during sightings 4.6. Distribution of ringed seal 4.6. of ringed Distribution IGURE F 4-22 Monitoring at Northstar, Ice-Covered Season Chapter 4: Aerial Surveys of Seals, 2002 4-23 nalysis of replicates 1 and 2 combined, 30 May - 7 June 2002. - 7 June 30 May 2 combined, 1 and of replicates nalysis 4.7. Ringed seal densities derived from trend surface a surface from trend derived 4.7. Ringed seal densities IGURE F 4-24 Monitoring at Northstar, Ice-Covered Season

TABLE 4.2. Number of ringed seals observed showing various behaviors during survey replicates 1 and 2 in 2002. Both fast ice and pack ice observations are included, as well as seals sighted both on and off transect.

Behavior Fast Ice Pack Ice Total

Dive No. of Seals 38 17 55 % of Total 1.3 2.1 1.5 Move No. of Seals 27 12 39 % of Total 0.9 1.5 1.1 Look No. of Seals 1215 289 1504 % of Total 42.3 35.4 40.8 None No. of Seals 1152 213 1365 % of Total 40.1 26.1 37.0 Unknown No. of Seals 438 286 724 % of Total 15.3 35.0 19.6

Total 2870 817 3687

Two bearded seals were observed near cracks and seven bearded seals were observed at holes. None of these seals were observed exhibiting a dive or other obvious reaction to the aircraft: four were recorded as looking at the aircraft, two showed no apparent reaction, and the behavior of the remaining bearded seals was recorded as unknown. Density calculations and further analyses were not conducted because of the small number of bearded seals seen. Several bearded seals were sighted in the study area during 1999, 2000, and 2001 but none were seen during 1997–98 (G. Miller, LGL Ltd., pers. comm.).

Factors Affecting Ringed Seal Abundance and Distribution We examined the observed density of ringed seals on landfast ice in 2002 in relation to two habitat parameters that may affect seal abundance and distribution: water depth and distance from the ice edge. The results of these analyses are described below. (Seal densities in relation to percent ice deformation, percent melt water, weather, and other factors are examined in Chapter 5 of this report .) The observed average densities in various strata are expected to underestimate the actual densities. Not all seals are hauled out at any one time, and aerial observers miss some proportion of the seals that are hauled out (Frost et al. 1988; Kelly and Quakenbush 1990). However, the observed densities in different strata are believed to be meaningful indicators of the relative utilization of various depths and distances from the ice edge. Chapter 4: Aerial Surveys of Seals, 2002 4-25 h survey replicates (transects 1-80), 30 May–7 June 2002. May–7 30 1-80), (transects replicates h survey 4.8. Distribution of bearded seal sightings during bot during seal sightings 4.8. Distribution of bearded IGURE F 4-26 Monitoring at Northstar, Ice-Covered Season

Water Depth Average densities of ringed seals observed on landfast ice within different water depth strata during the two survey replicates in 2002 ranged from 0.02 to 1.15 seals/km2 (Fig. 4.9B). The water depth strata used in this analysis were based on the depth contours shown in Figure 4.2. The sighting rates depended strongly on water depth during both survey replicates (replicate 1: χ2 = 244.2; replicate 2: χ2 = 233.4; P < 0.001 and df = 6 in each case). The observed seal density in the 0–3 m zone was 0.04 and 0.02 seals/km2 during replicates 1 and 2, respectively. As in previous years, the density in the 0–3 m zone was much lower than densities observed in most of the deeper strata, where the observed seal densities in 2002 ranged from 0.49 to 1.15 seals/km2. The low seal density in water 0–3 m deep was expected. Most of the 0–2 m portion of the 0–3 m zone would be frozen solid in spring and could not be used by seals. The 2–3 m portion would be marginal habitat at best. Statistical tests were based on “independent” sightings (singletons and groups treated equally; Fig. 4.9A), not numbers of individuals as shown in Fig. 4.9B. In 2002, numbers of seal sightings did not always parallel numbers of individual seals when plotted relative to water depth strata (Fig. 4.9A vs. B). This divergence occurred because seal group size was larger in deeper waters, most notably in the 15– 20 m and 20–25 m depth strata (Figs. 4.5, 4.6). If the areas <3 m deep are excluded from the analysis, the observed densities of seal sightings in the remaining depth categories still differed significantly (replicate 1: χ2 = 90.0; replicate 2: χ2 = 91.2; P < 0.001 and df = 5 in each case). During replicate 1, the overall significant difference among strata was attributable mainly to the relatively high densities of seal sightings in the 5–10 m and 10–15 m depth strata and lower than expected sighting densities in the 20–25 m strata (Fig. 4.9A). Similarly, during replicate 2, the overall significant difference among strata was attributable mainly to the higher than expected numbers of sightings in shallower water (5–15 m deep), and lower than expected numbers in water 20–25 m deep (Fig. 4.9A). During both replicates, the lowest individual densities in the 3–30 m zone were observed in the 3–5 m stratum (Fig. 4.9B); individual densities were highest in the 25–30 m zone during replicate 1 and in the 15–20 m zone during replicate 2. The 0–3 m depth category has been excluded from all of the following analyses because of the very low numbers of seals in waters <3 m deep. This eliminated 501 km2 of surveyed area and 16 seals (14 sightings) from further consideration.

Distance from Ice Edge Observed ringed seal densities varied significantly with distance from the ice edge during both survey replicates in 2002 (Fig. 4.9C,D). There were more sightings 10–15 km from the ice edge, and fewer sightings within 5 km from the ice edge, than would be expected if seal distribution were uniform (replicate 1: χ2 = 117.4; replicate 2: χ2 = 103.3; P < 0.001 and df = 5 in each case). The stratum of highest individual density during each replicate was the one 5–10 km inshore of the ice edge (0.97 and 1.21 seals/km2 in replicates 1 and 2, respectively). The stratum of lowest individual density during each replicate was the one 25–30 km inshore of the ice edge (0.32 and 0.37 seals/km2 in replicates 1 and 2, respectively). As for the water depth analyses, the numbers of seal sightings did not parallel the numbers of individual seals relative to distance from the ice edge (Fig. 4.9C vs. D). This divergence occurred because seal group size was larger close to the ice edge, most notably in the 5–10 km distance bin (Figs. 4.5, 4.6). Chapter 4: Aerial Surveys of Seals, 2002 4-27

FIGURE 4.9. Observed density of ringed seals during survey replicate 1 (30 May–2 June 2002) and 2 (3–7 June 2002) in relation to water depths (A, B) and distances from the ice edge (C, D). A and C show sighting densities; B and D show densities of individual seals. In A and C, P-values are significance levels for overall χ2 test; solid symbols represent strata where the number of sightings was significantly lower or higher than expected based on the overall average sighting rate (α=0.05; chi-square test with Bonferroni adjustment). Data from water depths 0–3 m are excluded from C and D. Observed Ringed Seal Densities Near Development Sites Northstar

Figure 4.10 shows the locations of both the on-transect and the off-transect seal sightings in the area close to Northstar during both of the 2002 survey replicates. Only the on-transect sightings are considered in the remainder of this subsection.4 No seals were observed within the relatively small area (1.2 km2 over water depths >3 m, both survey replicates combined) that was considered the Northstar development zone. This development zone encompasses the area covered by the ice road and Northstar island (see the narrow strip shaded dark gray in Fig. 4.10). There were 21 sightings (involving 22 seals) within 1 km of the development zone, i.e., within an area of 40.0 km2; and an additional 32 sightings (37 seals) at locations 1–2 km from the development zone (44.6 km2). Several seals were observed within 250 m (820 ft) of the

4 Off-transect sightings are excluded because their positions were not recorded with sufficient precision to be useful in this fine-scale analysis. Sighting locations mapped in Figure 4.10 (and elsewhere) are the locations of the survey aircraft at the specific time that the sighting was recorded. The outer edge of the transect strip was 0.55 km from the aircraft. Recorded times could differ by as much as 5 s from the time when the aircraft was closest to the seal, and the aircraft traveled ~0.3 km along the transect during 5 s. Thus, plotted positions could be as much as 0.62 km, i.e. √(0.552 + 0.32), from actual positions, though most will be closer than this. 4-28 Monitoring at Northstar, Ice-Covered Season

FIGURE 4.10. Distribution of ringed seal sightings during replicate 1 (30 May–2 June 2002) and replicate 2 (3–7 June 2002) within 10 km of the edges of the Northstar development zone. Successive 1-km distance categories from the edges of the ice road and artificial island (the “Northstar development zone”) are shown. 1 km = 0.62 mi = 0.54 n.mi. Chapter 4: Aerial Surveys of Seals, 2002 4-29 ice road, and one seal was estimated to be ~60 m (197 ft) from the road (Fig. 4.10). The closest seal sighting to Northstar island was an estimated 440 m (1444 ft) away. The observed seal density within 10 km of the Northstar development zone (total surveyed area of 550 km2) was 0.93 seals/km2, based on replicates 1 and 2 combined. The lowest seal density in any of the ten 1-km “distance from Northstar” categories was found within 1 km of the development zone (0.55 seals/km2; Fig. 4.11). The extent to which this was attributable to an industry effect (vs. sampling variation or confounding environmental variables) was addressed with a multivariate approach, taking account of other variables such as water depth, ice deformation, etc., along with data from other years (see CHAPTER 5). The multivariate analysis showed no significant effect of Northstar-related industrial activities on ringed seal densities in 2002, or indeed in any year (see CHAPTER 5).

Liberty Ringed seals were seen near the potential Liberty site and Tern Island during 2002 (Fig. 4.5, 4.6). No industrial activities are known to have occurred near Liberty during the winter of 2001 and 2002. The survey data from the Liberty area in 2002 (and 1997–2001) are available for possible future use if oil development goes ahead at Liberty.

DISCUSSION

Observed Ringed Seal Densities, Abundance and Distribution

The overall observed ringed seal density based on our spring 2002 surveys of landfast ice habitat was 2 2 0.72 seals/km , or 0.83 seals/km if areas <3 m deep are excluded. The overall density in areas ≥3 m deep was higher in 2002 (0.83 seals/km2) than found during previous BP/LGL surveys in 1997–2001 (0.43, 0.39,

FIGURE 4.11. Observed density of ringed seals during survey replicates 1 and 2 combined (30 May–7 June 2002) at differing distances from the edge of the Northstar development zone. The P-value is the significance level for the overall χ2 test. There were no distance categories at which the number of sightings was significantly lower or higher than expected based on the overall average sighting rate (α=0.05; chi- square test with Bonferroni adjustment). Results from areas with water depths 0–3 m are excluded. 4-30 Monitoring at Northstar, Ice-Covered Season

0.63, 0.47, and 0.54 seals/km2, respectively; cf. Miller et al. 1998; Link et al. 1999; Moulton et al. 2000a, 2001, 2002a). During 1996, 1997, 1998, and 1999, ADF&G recorded densities of 0.54, 0.77, 0.76, and 0.87 seals/km2 in their sector B3, which includes our entire study area plus additional areas to the east and west (Frost et al. 1997, 2002; Frost and Lowry 1998, 1999). These recent density estimates are low compared to seal densities recorded in the same general area in the 1980s. During 1985–87, observed seal densities in fast ice portions of ADF&G's sector B3 ranged from 1.16 to 2.41 seals/km2 (Table 4.1). These variable densities are not especially surprising given what is known about inter-annual variations of ringed seal body condition and reproduction in the Beaufort Sea in earlier years (Stirling et al. 1977; Smith and Stirling 1978; Smith 1987; Kingsley and Byers 1998; Harwood et al. 2000). Recent BP/LGL (1997–2002) and ADF&G (1996–99) survey results are consistent with the possibility that the central Alaskan Beaufort Sea population of ringed seals may be smaller than in the early 1980s. However, the lower densities obtained during many of the recent surveys relative to surveys in the early 1980s may also be attributable, at least in part, to differences in the dates when the surveys were flown, along with year- to-year differences in the timing of lair abandonment and basking (Frost et al. 2002; Kelly et al. 2002). Many of the recent ADF&G and BP/LGL surveys were conducted a week or more earlier than surveys in the 1980s and may have occurred before the peak period of ringed seal haulout. The observed density of 0.72 seals/km2 based on our spring 2002 results (0.83 seals/km2 excluding areas <3 m deep) is an overall average density based on all surveys flown. As such, it is comparable to the densities determined by the other studies cited above. However, it may be biased upwards by the fact that the ice edge was closer to shore in 2002 and this restricted the amount of survey coverage in deeper waters where lower seal densities are typically observed. Regardless of this, observed densities of ringed seals are underestimates of the actual densities of seals present. The 2002 density increases to 2.05 seals/km2 (or 2.36 seals/km2 if we exclude areas <3 m deep) if adjusted to allow for • seals hauled out but not sighted by observers (× 1.22, based on Frost et al. 1988), and • the proportion of ringed seals not hauled out during surveys (× 2.33, based on Kelly and Quaken- bush 1990). The University of Alaska Southeast has recently conducted telemetry-based studies of ringed seals within our study area (Harding et al. 2003). Their results, when available, are expected to provide more precise and year-specific correction factors for the proportion of seals not hauled out. Our surveys of the Northstar area during each year were spread over 7–9 days of flying. This is expected to reduce year-to- year variability in the proportion of the seals detected, as compared with the variability that could occur if the area of interest was surveyed on fewer days (or only one day) in a given year. In 2002, ringed seals were observed in most parts of the study area, but there was a tendency for higher densities in the deeper parts of the lagoons from Prudhoe Bay eastward, and seaward of the barrier islands but inshore of the 15 m depth contour. Similarly, ringed seals were common in deeper parts of the nearshore lagoons in 1997, 1998, 2000, and 2001 (cf. Miller et al. 1998; Link et al. 1999; Moulton et al. 2001, 2002a). However, in 1999, there was a tendency for lower densities inside the lagoons as seals were concentrated farther offshore in waters 15–25 m deep (Moulton et al. 2000a). The reasons for the various high-density areas, and the interannual variation in their locations, are not entirely known but are related in part to preferences for particular types of ice habitat (Moulton et al. 2002a,c; CHAPTER 5). Local variability in prey density and availability, though not documented in this area, is probably also a factor. Chapter 4: Aerial Surveys of Seals, 2002 4-31

Factors Affecting Observed Ringed Seal Density

Both water depth and distance from ice edge were significantly related to the observed density of ringed seals on the landfast ice. Seals generally preferred waters 5–15 m deep, and densities were higher 10–15 km inshore of the fast ice edge than near the pack ice. A more comprehensive analysis of habitat, temporal, weather, and industrial activity factors known or expected to influence the abundance and/or haulout behavior of ringed seals is included in Chapter 5. Water Depth Effects As in the 1997–2001 BP/LGL surveys, densities in 2002 were significantly higher in areas with water depths exceeding 3 m than in areas with depths <3 m. This is expected as most of the 0–2 m portion of the 0–3 m zone would be frozen solid in late May or June, and could not be used by seals, and the 2–3 m portion would be marginal habitat at best. During both replicate surveys in 2002, significantly more seal sightings were observed in waters 5– 15 m deep, and fewer seals were observed in waters 15–25 m deep, than would be expected if seal density on the fast ice were uniform. These results were very similar to those in 1997 and 1998, but differed somewhat from those found during 2000 and 2001 when densities were highest in water depths ranging from 3 to 10 m. In 1999, more seals were sighted in deeper water near the fast ice edge (Moulton et al. 2000a, 2002c). Distance from Ice Edge The highest density of ringed seals observed during 2002 occurred 5–15 km shoreward of the landfast ice edge during both survey replicates. Significantly fewer seals were observed within 5 km of the ice edge and relatively low densities were observed 20–30 km shoreward of the ice edge. The opposite trend was observed in most other survey years when higher than average densities were found farther from the ice edge in 1997 (10–20 km from ice edge), 1998 (25–35 km), 2000 (25–30 km), and 2001 (25–30 km). In 1999, more seals were observed within 10 km of the fast ice edge but only during replicate 2 (8–13 June 1999); fewer seals were closer to shore. It appears that ringed seals from pack ice may have moved into the adjacent fast ice as our 1999 survey progressed. It is possible that this occurred in 2002. However, during both survey replicates, significantly fewer seal sightings were noted adjacent (0–5 km) to the ice edge than farther inshore. Thus, the influx of seals from the pack ice was likely not as prevalent as in 1999. In 2002, seal sightings may have been most abundant at intermediate distances from the ice edge because the numerous cracks in the ice in this zone provided suitable habitat for hauling out to molt. This may also explain why the overall ringed seal density for 2002 was higher than previous survey years. Observed Seal Densities Near Northstar

One of the main objectives of this study is to determine whether (and to what extent) industrial activity at Northstar influences the local distribution and abundance of ringed seals. During the ice- covered season of late 2001 and early 2002, oil production and gas injection were conducted routinely, an ice road to Northstar was built and used, several wells were drilled, and various tests, inspections and drills were conducted near the island and along the pipeline route. These activities had the potential to disturb and displace ringed seals from the area, and allow us to address the above objective. No seals were observed within the relatively small area (1.2 km2) that was considered the Northstar development zone, but that area was sufficiently small that one would not have expected more than 1 or 2 seals in that area in the absence of Northstar. Also, several seals were observed at an estimated 250 m (820 ft) from 4-32 Monitoring at Northstar, Ice-Covered Season the ice road, and one seal was sighted at an estimated 60 m (197 ft) from the road. Overall, the density of seal sightings in various 1-km increments of distance from the development zone in 2002 was somewhat lower close to Northstar. It is possible that some ringed seals may have been displaced by activities at Northstar. However, other factors that are not accounted for in the simple analyses in this chapter may also have made the 0–1 km area somewhat less suitable for ringed seals. A multivariate analysis of potential influences of Northstar activities in 2001–02 (and other years) is provided in Chapter 5. That Poisson regression analysis of 1997–2002 data, in addition to drawing together the results from different years, also takes account of other factors that influence the numbers of seals seen. The Poisson regression indicates that, after allowance for confounding variables, there is no evidence of reduced seal densities close to Northstar during the springs of 2000, 2001, and 2002 (see CHAPTER 5). A simplified statistical power analysis was conducted after the 2000 surveys (McDonald and Nations 2001). The results showed that, if the intensive aerial surveys continued for an additional two years after 2000, they would have substantial (at least 80%) power to detect a reduction in seal density of 30–40% within 4 km of the Northstar facilities, if such a reduction occurred. The univariate and multivariate results from 2000 and 2001, and the preliminary univariate results from 2002, indicate that, if there is a reduction in observed density close to Northstar, it was not of that magnitude. An updated power analysis has been done based on the 1997–2002 results and on the specific type of multivariate analysis (Poisson regression) that was applied. Results of this new analysis are given in CHAPTER 5.

ACKNOWLEDGEMENTS

We thank Inupiat observer Jonah Leavitt for his enthusiastic participation as an aerial observer (along with V. Moulton and M. Williams) during the 2002 BP/LGL aerial surveys. We are grateful to Ralph Aiken, Charles Pike, and Tom Lampee of Commander Northwest Ltd. for piloting the survey aircraft and providing other logistical support. We thank Shelly Schwenn of LGL for providing logistical support and Craig Perham of LGL for mapping assistance. This study was designed in 1997 by W.R. Koski, G.W. Miller, and Dr. W.J. Richardson of LGL. We are grateful to Dr. C. Herlugson of BP for initiating this project, Dr. R. Jacubczak and D. Trudgen of BP for their support in previous years of the project, and Dr. B. Streever of BP for his support and encouragement of the 2002 phase. Thanks also to participants in the peer/stakeholder review group that has commented periodically on project plans and results. B. Streever of BP, plus W.J. Richardson of LGL, reviewed drafts of this chapter. Anne Wright formatted and helped produce this chapter. We thank Alicia Haskell, Meike Holst, Wayne King and Arturo Serrano, along with Anne Wright, all of LGL, for helping with data entry. The 2002 aerial surveys were conducted under Scientific Research Permit 481-1623-01 issued by the National Marine Fisheries Service to LGL Ltd. on 12 February 2002. We thank the Permits Division of the NMFS Office of Protected Resources for processing the permit request.

LITERATURE CITED Born, E.W., J. Teilmann and F. Riget. 2002. Haul-out activity of ringed seals (Phoca hispida) determined from satellite telemetry. Mar. Mamm. Sci. 18(1):167-181.

BPXA. 2001. Monitoring and reporting plan. Section 13 In: Request for a Letter of Authorization ... covering taking of marine mammals incidental to construction and operation of offshore oil and gas facilities in the U.S. Beaufort Sea, Alaska (50 C.F.R. Part 216, Subpart R). Request from BP Exploration (Alaska) Inc., Anchorage, AK, to Nat. Mar. Fish. Serv. [Silver Spring, MD]. Prepared by LGL Alaska Res. Assoc. Inc., Anchorage, AK. Chapter 4: Aerial Surveys of Seals, 2002 4-33

Burns, J.J. and S.J. Harbo, Jr. 1972. An aerial census of ringed seals, northern coast of Alaska. Arctic 25(4):279- 290. Feltz, E.T. and F.H. Fay. 1966. Thermal requirements in vitro of epidermal cells from seals. Cryobiology 3(3):261-264. Finley, K.J. 1979. Haul-out behaviour and densities of ringed seals (Phoca hispida) in the Barrow Strait area, N.W.T. Can. J. Zool. 57(10):1985-1997. Frost, K.J. and L.F. Lowry. 1988. Effects of industrial activities on ringed seals in Alaska, as indicated by aerial surveys. p. 15-25 In: W.M. Sackinger et al. (eds.), Port and Ocean Engineering Under Arctic Conditions. Volume II. Symposium on Noise and Marine Mammals. Geophysical Inst., Univ. Alaska, Fairbanks, Fairbanks, AK. 111 p. Frost, K.J. and L.F. Lowry. 1998. Monitoring distribution and abundance of ringed seals in northern Alaska/Inter- im report, April 1997-March 1998. Rep. from Alaska Dep. Fish & Game, Fairbanks, AK, for U.S. Minerals Manage. Serv., Anchorage, AK. 48 p. Frost, K.J. and L.F. Lowry. 1999. Monitoring distribution and abundance of ringed seals in northern Alaska/Inter- im report, April 1998-March 1999. Rep. from Alaska Dep. Fish & Game, Fairbanks, AK, for U.S. Minerals Manage. Serv., Anchorage, AK. 44 p. Frost, K.J., L.F. Lowry, J.R. Gilbert and J.J. Burns. 1988. Ringed seal monitoring: relationships of distribution and abundance to habitat attributes and industrial activities. U.S. Dep. Commer., NOAA, OCSEAP Final Rep. 61 (1989):345-445. NTIS PB89-234645. Frost, K.J., L.F. Lowry, S. Hills, G. Pendleton and D. DeMaster. 1997. Monitoring distribution and abundance of ringed seals in northern Alaska. Rep. from Alaska Dep. Fish & Game, Fairbanks, AK, for U.S. Minerals Manage. Serv., Anchorage, AK. 42 p. Frost, K.J., L.F. Lowry, G. Pendleton and H.R. Nute. 2002. Monitoring distribution and abundance of ringed seals in northern Alaska. OCS Study MMS 2002-043. Rep. from Alaska Dep. Fish & Game, Juneau, AK, for U.S. Minerals Manage. Serv., Anchorage, AK. 106 p. Green, J.E. and S.R. Johnson. 1983. The distribution and abundance of ringed seals in relation to gravel island construction in the Alaskan Beaufort Sea. p. 1-28 In: B.J. Gallaway (ed.), Biological studies and monitoring at Seal Island, Beaufort Sea, Alaska 1982. Rep. from LGL Ecol. Res. Assoc. Inc., Bryan, TX, for Shell Oil Co., Houston TX. 150 p. Green, R.H. 1979. Sampling design and statistical methods for environmental biologists. John Wiley, New York, NY. 257 p. Harding, O.R., B.P. Kelly and M. Kunnasranta. 2003. Behavior of ringed seals and re-interpretation of aerial survey results. p. 18 In: Presentation Summaries, Information Transfer Meeting, Anchorage, AK, 10-12 March 2003. U.S. Minerals Manage. Serv., Alaska OCS Region, Anchorage, AK. Harwood, L.A., T.G. Smith and H. Melling. 2000. Variation in reproduction and body condition of the ringed seal (Phoca hispida) in western Prince Albert Sound, NT, Canada, as assessed through a harvest-based sampling program. Arctic 53(4):422-431. Kelly, B.P. and L.T. Quakenbush. 1990. Spatiotemporal use of lairs by ringed seals (Phoca hispida). Can. J. Zool. 68(12):2503-2512. Kelly, B.P., J.J. Burns and L.T. Quakenbush. 1988. Responses of ringed seals (Phoca hispida) to noise disturbance. p. 27-38 In: W.M. Sackinger et al. (eds.), Port and Ocean Engineering Under Arctic Conditions. Vol. II. Symposium on Noise and Marine Mammals. Geophysical Institute, Univ. Alaska, Fairbanks, Fairbanks, AK. 111 p. Kelly, B.P., L.T. Quakenbush and B.D. Taras. 2002. Correction factor for ringed seal surveys in northern Alaska. p. 3-15 In: Annu. Rep. 8, Coastal Marine Inst., Univ. Alaska Fairbanks. OCS Study MMS 2002-001. 4-34 Monitoring at Northstar, Ice-Covered Season

Kingsley, M.C.S. 1986. Distribution and abundance of seals in the Beaufort Sea, Amundsen Gulf, and Prince Albert Sound, 1984. Environ. Stud. Revolv. Funds Rep. 25. Dep. Fish. & Oceans, Winnipeg, Man. 16 p. Kingsley, M.C.S. and T.J. Byers. 1998. Failure of reproduction in ringed seals (Phoca hispida) in Amundsen Gulf, Northwest Territories in 1984-1987. p. 197-210 In: M.P. Heide-Jørgensen and C. Lydersen (eds.), Ringed seals in the North Atlantic. Tromsø, Norway: NAMMCO Scientific Publications, Vol.1. Kovacs, A. and M. Mellor. 1974. Sea ice morphology and ice as a geological agent in the southern Beaufort Sea. p. 113-161 In: J.C. Reed and J.A. Sater (eds.), The coast and shelf of the Beaufort Sea. Arctic Institute of North America, Arlington, VA. 750 p. Kovacs, K.M. 1987. Maternal behaviour and early behavioural ontogeny of harp seals, Phoca groenlandica. Anim. Behav. 35(3):844-855. Lerczak, J.A. and R.C. Hobbs. 1998. Calculating sighting distances from angular readings during shipboard, aerial, and shore-based marine mammal surveys. Mar. Mamm. Sci. 14(3):590-599. Link, M.R., T.L. Olson and M.T. Williams. 1999. Ringed seal distribution and abundance near potential oil development sites in the central Alaskan Beaufort Sea, spring 1998. LGL Rep. P-430. Rep. from LGL Alaska Res. Assoc. Inc., Anchorage, AK, for BP Explor. (Alaska) Inc., Anchorage, AK. 58 p. Lydersen, C. and K.M. Kovacs. 1993. Diving behaviour of lactating harp seal, Phoca groenlandica, females from the Gulf of St. Lawrence, Canada. Anim. Behav. 46(6):1213-1221. Manly, B.F.J., L.L. McDonald and D.L. Thomas. 1993. Resource selection by animals. Chapman & Hall, New York, NY. 177 p. McDonald, T.L and C. Nations. 2001. Statistical power of intensive aerial surveys to assess Northstar effect on ringed seal densities. p A-1 to A-11 In: W.J. Richardson and M.T. Williams (eds., 2001, q.v.). Miller, G.W., R.E. Elliot and W.J. Richardson. 1998. Ringed seal distribution and abundance near potential oil development sites in the central Alaskan Beaufort Sea, spring 1997. LGL Rep. TA2160-3. Rep. from LGL Ltd., King City, Ont., for BP Explor. (Alaska) Inc., Anchorage, AK. 36 p. MMS. 2002. Liberty Development and Production Plan/Final Environmental Impact Statement. 4 vols. OCS EIS/EA MMS 2002-019. U.S. Minerals Manage. Serv., Anchorage, AK. Var. pag. Moulton, V.D., T.L. McDonald, W.J. Richardson, R.E. Elliott and M.T. Williams. 2000a. Fixed-wing aerial surveys of seals near BP’s Northstar and Liberty sites in 1999 (and 1997-99 combined). p. 3-1 to 3-71 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of ringed seals during construction of ice roads for BP’s Northstar oil development, Alaskan Beaufort Sea, 1999. LGL Rep. TA2349-3. Final Rep. from LGL Ltd., King City, Ont., and LGL Alaska Res. Assoc. Inc., Anchorage, AK, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 153 p. Moulton, V.D., E.H. Miller and H. Ochoa-Acuña. 2000b. Haulout behaviour of captive harp seals (Pagophilus groenlandicus): incidence, seasonality, and relationships to weather. Appl. Anim. Behav. Sci. 65(4):367-378. Moulton, V.D., R.E. Elliott, W.J. Richardson and T.L. McDonald. 2001. Fixed-wing aerial surveys of seals near BP’s Northstar and Liberty sites in 2000 (and 1997-2000 combined). p. 5-1 to 5-66 In: W.J. Richardson and M.T. Williams (eds., 2001, q.v.). LGL Rep. TA2428-3. Moulton, V.D., R.E. Elliott, W.J. Richardson and T.L. McDonald. 2002a. Fixed-wing aerial surveys of seals near BP’s Northstar and Liberty sites in 2001 (and 1997-2001 combined). p. 5-1 to 5-60 In: W.J. Richardson and M.T. Williams (eds., 2002, q.v.). LGL Rep. TA2570-3. Moulton, V.D., R.E. Elliott and M.T. Williams. 2002b. Fixed-wing aerial surveys of seals near BP’s Northstar and Liberty sites, 2002. p. 3-1 to 3-34 In: W.J. Richardson and M.T. Williams (eds.) 2002. Monitoring of ringed seals, sounds and vibrations during operation of BP’s Northstar Development, Alaskan Beaufort Sea, winter and spring 2001-2002: 90-day report. LGL Rep. TA2702-1. Rep. from LGL Ltd., King City, Ont., and Chapter 4: Aerial Surveys of Seals, 2002 4-35

Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 86 p. Superseded by the present report. Moulton, V.D., W.J. Richardson, T.L. McDonald, R.E. Elliott and M.T. Williams. 2002c. Factors influencing local abundance and haulout behaviour of ringed seals (Phoca hispida) on landfast ice of the Alaskan Beaufort Sea. Can. J. Zool. 80(11):1900-1917. Moulton, V.D., W.J. Richardson, M.T. Williams and S.B. Blackwell. 2003. Ringed seal densities and noise near an icebound artificial island with construction and drilling. Acoust. Res. Let. Online. “In press”. NMFS. 2000. Taking marine mammals incidental to construction and operation of offshore oil and gas facilities in the Beaufort Sea: final rule. Fed. Regist. 65(102, 25 May): 34014-34032. NMFS. 2001. Taking and importing marine mammals; taking marine mammals incidental to construction and operation of offshore oil and gas facilities in the Beaufort Sea. Fed. Regist. 66(246, 21 Dec.):65923-65935. Richardson, W.J. and M.T. Williams (eds.). 2001. Monitoring of industrial sounds, seals, and whale calls during construction of BP’s Northstar Oil Development, Alaskan Beaufort Sea, 2000. [April 2001 ed.] Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 316 p. Richardson, W.J. and M.T. Williams (eds.). 2002. Monitoring of industrial sounds, seals, and whale calls during construction of BP’s Northstar Oil Development, Alaskan Beaufort Sea, 2001. [Oct. 2002 ed.] Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 309 p. Smith, T.G. 1987. The ringed seal, Phoca hispida, of the Canadian western Arctic. Can. Bull. Fish. Aquat. Sci. 216: 81 p. Smith, T.G. and M.O. Hammill. 1981. Ecology of the ringed seal, Phoca hispida, in its fast ice breeding habitat. Can. J. Zool. 59(6):966-981. Smith, T.G. and L. A. Harwood. 2001. Observations of neonate ringed seals, Phoca hispida, after early break-up of the sea ice in Prince Albert Sound, Northwest Territories, Canada, spring 1998. Polar Biol. 24(3):215-219. Smith, T.G. and I. Stirling. 1975. The breeding habitat of the ringed seal (Phoca hispida). The birth lair and associated structures. Can. J. Zool. 53(9):1297-1305. Smith, T.G. and I. Stirling. 1978. Variation in the density of ringed seal (Phoca hispida) birth lairs in the Amund- sen Gulf, Northwest Territories. Can. J. Zool. 56(5):1066-1071. Stirling, I., W.R Archibald and D. DeMaster. 1977. Distribution and abundance of seals in the the eastern Beaufort Sea. J. Fish. Res. Board Can. 34(7):976-988. Stringer, W.J. 1974. Sea ice morphology of the Beaufort shorefast ice. p. 165-172 In: J.C. Reed and J.A. Sater (eds.), The coast and shelf of the Beaufort Sea. Arctic Institute of North America, Arlington, VA. 750 p. Thomas, J.A. and D.P. DeMaster. 1983. Diel haul-out patterns of Weddell seal (Leptonychotes weddelli) females and their pups. Can. J. Zool. 61(9):2084-2086. Underwood, A.J. 1992. Beyond BACI: the detection of environmental impacts on populations in the real, but variable, world. J. Exp. Mar. Biol. Ecol. 161(2):145-178. Wadhams, P. 2000. Ice in the ocean. Gordon & Breach Sci. Publ., Amsterdam, The Netherlands. 351 p. Williams, M.T., T.G. Smith and C.J. Perham. 2002. Ringed seal structures in sea ice near Northstar, winter and spring of 2000-2001. p. 4-1 to 4-33 In: W.J. Richardson and M.T. Williams (eds., 2002, q.v.). LGL Rep. P557-2. CHAPTER 5:

EFFECTS OF NORTHSTAR ON LOCAL ABUNDANCE AND DISTRIBUTION OF RINGED SEALS (PHOCA HISPIDA) OF THE ALASKAN BEAUFORT SEA 1

Valerie D. Moulton a, W. John Richardson b, Trent L. McDonald c, Robert E. Elliott b, Michael T. Williams d and Chris Nations c

a LGL Ltd., environmental research associates 388 Kenmount Rd., POB 13248 Stn. A, St. John’s, Nfld. A1B 4A5 (709) 754-1992; [email protected] b LGL Ltd., environmental research associates 22 Fisher St., POB 280, King City, Ont. L7B 1A6 (905) 833-1244; [email protected] c Western EcoSystems Technology, Inc. 2003 Central Ave., Cheyenne, WY 82001 (307) 634-1756; [email protected] d LGL Alaska Research Associates, Inc. 1101 E. 76th Ave, Suite B, Anchorage, AK 99518 (907) 562-3339; [email protected]

for

BP Exploration (Alaska) Inc. Dept of Health, Safety & Environment 900 East Benson Blvd., POB 196612 Anchorage, AK 99519-6612

LGL Report TA2702-4 December 2003

1 Chapter 5 In: W.J. Richardson and M.T. Williams (eds.). 2003. Monitoring of industrial sounds, seals, and bowhead whales near BP’s Northstar Oil Development, Alaskan Beaufort Sea, 1999–2002. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 5-2 Monitoring at Northstar, Ice-Covered Season

ABSTRACT

This study investigates how densities of ringed seals, Phoca hispida, were affected by construction and operational activities at Northstar Island, an artificial island built by BP in the nearshore central Alaskan Beaufort Sea. Intensive and replicated aerial surveys of seals on landfast ice were conducted during three spring seasons before island construction began (1997–1999); after a winter of intensive island construction (2000); and after more limited construction plus drilling (2001) and drilling plus oil production (2002). A Poisson regression model examined seal densities relative to distance from Northstar after allowance for environmental covariates known or suspected to influence seal distribution and haulout behavior. A post hoc power analysis indicated that the study design and Poisson regression approach (with covariates) had high power to detect small-scale changes in seal densities near Northstar if such changes had occurred. However, seal densities near Northstar during spring were not significantly affected by industrial activities in 2000, 2001 and 2002. Other human activities, including Vibroseis surveys, that occurred in other portions of the study area during one or more survey years apparently did have detectable negative effects on seal densities. Also, various habitat, temporal, and weather factors had significant and strong effects on seal densities. This study shows that effects of an offshore oil development on the local distribution of basking ringed seals are no more than slight, and are small in comparison to the effects of natural environmental factors. An understanding of the latter effects is essential when assessing potential impacts of industrial activity on ringed seals.

INTRODUCTION

This study investigates whether local abundance and distribution of ringed seals (Phoca hispida) on landfast ice of the central Alaskan Beaufort Sea were affected by construction, drilling and initial production operations at an artificial island, BP’s Northstar oil development. This site is 9.5 km from the mainland and 4.8 km seaward of the closest barrier island. Northstar island is in 12 m of water, which is covered by landfast ice from November to early July; it was built on the submerged remnants of Seal Island, which was constructed in 1982, used for exploration drilling during the 1980s, and subsequently abandoned. There is limited previous evidence that some ringed seals may be displaced from areas within a few kilometers of active artificial islands (summary in Richardson et al. 1995, p. 280-1). However, results of those previous studies are not always consistent, and generally did not take account of natural factors that are known to influence the numbers of ringed seals seen. Ringed seal hearing in water (Terhune and Ronald 1975) and presumably in air is probably similar to that of other phocinid seals (Richardson et al. 1995). During winter and spring, Northstar produces underwater and airborne sound, and iceborne vibrations, that are sometimes detectable by instruments to distances of 2–10 km (Greene and McLennan 2000; see also CHAPTER 3). Samples of some of these underwater sounds can be heard in the online paper of Moulton et al. (2003). Drilling and production sounds from Northstar are probably audible to ringed seals, at least intermittently, out to ~1.5 km in water and ~5 km in air (see CHAPTER 3). During winter and spring, ringed seals occur in both landfast and pack ice of the Beaufort Sea. Within the landfast ice zone, ringed seals maintain breathing holes and create subnivean lairs (snow caves) in which they haul out (Smith and Stirling 1975). They give birth in lairs from mid-March through April, and mate in late April and May (Smith 1973; Hammill et al. 1991; Lydersen and Hammill 1993). From about mid-May through early June, ringed seals haul out intermittently on top of the ice and snow where they can be surveyed. Most quantitative surveys of ringed seal abundance and distribution in the Chapter 5: Northstar Effect on Seal Distribution 5-3

Alaskan Beaufort Sea, including this study, are conducted during the late May–early June period, when large numbers of molting seals are basking on the ice. Near Northstar, springtime aerial surveys of seals were conducted before island construction began (1997–1999), after a winter of intensive island construction and pipeline installation (2000), after a winter with more limited construction plus drilling (2001), and after a winter with drilling and oil production (2002). During the winter preceding the 2000 aerial survey, industry activities included construction of ice roads, transport of ~18,300 truckloads of gravel from shore to Northstar, and installation of pipelines to shore. We hypothesized that seal density close to Northstar would be reduced in 2000, and to a lesser extent in 2001 and 2002, as a result of displacement of seals disturbed by the activities during the preceding winter and spring. Although the primary focus of this study was to investigate industry effects from Northstar, numerous environmental factors are known to affect ringed seal distribution and haul-out behavior. Previously- reported analyses of the “pre-construction” aerial survey data (1997–1999) included in this study showed that habitat factors like water depth and degree of ice deformation influence the numbers of ringed seals seen on the landfast ice (Moulton et al. 2002). Also, temporal and weather factors influence the haul-out behavior of ringed seals and hence influence the numbers of seals available to be counted by aerial observers (Kelly and Quakenbush 1990; Moulton et al. 2002). It is critical to account for these otherwise- confounding factors when investigating potential influences of industry on ringed seal densities. This study uses a multivariate statistical approach to do this. Other industry activities, different in characteristics and of shorter duration than activities at North- star, occurred within small parts of our study area at various times during 1997–2002. These activities included on-ice seismic exploration via the Vibroseis technique, exploration drilling from an artificial island, construction of ice roads, and a shallow-hazards survey. Some or all of these activities might have had local effects on ringed seal abundance and haul-out behavior, so these industry activities were also treated as potential confounding factors in our analyses.

METHODS

Survey Design and Procedures

Each spring (late May to early June) from 1997 to 2002, intensive aerial surveys were flown between longitudes 147º06'W and 149º04.5'W, an east-west extent of about 75 km (Fig. 5.1). We flew two survey replicates each year. A replicate consisted of 80 north–south transects spaced 0.9 km apart; for legibility, only half of these transects are shown in Figure 5.1. Each year’s surveys typically required 8 days (4 per replicate) spread over a 7–12 day period. Survey transects extended from the Beaufort Sea shoreline to roughly 37 km offshore or to the edge of the landfast ice. Weather permitting, 20 transects spaced 3.6 km apart were flown each day so all parts of the study area were sampled daily for ~8 days. Most surveys were at altitude 91 m above sea level and ground speed 222 km/h. We surveyed transect strips 411 m in width on each side of the aircraft. Survey coverage of the landfast ice averaged 4140 km2 per year in 1997–2002 (Table 5.1). Within 10 km of the Northstar industry zone, coverage averaged 521 km2 per year in 1997–2002 (Table 5.1). Two observers, one on each side of the plane, collected data on seal occurrence, locations of industrial sites, and environmental variables. Observers recorded environmental conditions like percent of the surface (intervals of 10%) where the ice was deformed and where melt water was present. These and other frequently- changing variables were recorded at 1-min (~3.7-km) intervals along every transect. Cloud cover was recorded at the beginning of each transect, by 10% intervals. Hourly (or more frequent) data on air temperature and wind speed, as recorded at the Deadhorse airport (Fig. 5.1), were acquired from the National 5-4 Monitoring at Northstar, Ice-Covered Season

FIGURE 5.1. Central Alaskan Beaufort Sea showing (in red) half of the aerial survey transects flown each year in 1997–2002. A similar grid of 40 transects was offset 0.9 km to the east, and each grid was flown twice per year. See Table 5.2 for specific configuration of Northstar facilities each year. Other industrial sites active in one or more of the six study years are also shown, in green.

Climatic Data Center, Asheville, N.C. Heat loss, an index of wind chill in W/m2, was estimated from wind speed and air temperature data (Siple and Passel 1945). More details about survey procedures can be found in Moulton et al. (2002) and CHAPTER 4. Analysis Procedures

The response variable in this study was the number of seal sightings (singletons or groups) in a transect segment flown over landfast ice. Segments were nominally 500 m in length, but the last segment of a transect was often shorter. A 500-m segment length was selected to allow detection of small-scale changes in ringed seal densities while limiting the potential for autocorrelation. To reduce concerns about spatial autocorrelation, adjacent segments were combined if all covariates for those transects were identical. Log (segment area) was used as an offset variable to allow for variation in segment length. Data Chapter 5: Northstar Effect on Seal Distribution 5-5

TABLE 5.1. Summary of survey coverage and field procedures during aerial surveys in the central Alaskan Beaufort Sea, spring 1997–2002. Year 1997 1998 1999 2000 2001 2002 Survey Data Dates Flown 26-29, 31 May; 1- 23, 25-27 May 4-8 June 31 May-3 28-29, 31 May; 30 May-2 Rep 1 2 June June 1,4 June June Dates Flown 27-29, 31 May a 27-30 May 8-10, 12-13 4-7 June 4-8 June 3-4, 6-7 June Rep 2 June Dates Flown 3-4 June a ------Rep 3 Total Area (km2) b 4510 4079 3980 4503 4147 3618 Linear Dist. (km) b 6089 6350 5486 5581 5154 4538 Area < 10 km from 470 515 462 503 628 550 Northstar (km2) b, c Survey Procedure d

Aircraft Cessna 185 Shrike Twin Otter Turbo Turbo Turbo Commander (DHC-6) Commander Commander Commander Bubble Windows No Yes Yes Yes Yes Yes GPS Trimble Trimble Garmin Trimble Garmin Garmin

a In 1997, only one complete survey replicate (Replicate 1: 80 unique transects) was surveyed. Forty of the 80 transects were surveyed two additional times; referenced here as "Replicate 2" and "Replicate 3". Thus, for 1997, Replicate 1 was complete but Replicates 2 and 3 were only 50% complete. b Total survey area or length, including only on-transect survey effort over the landfast ice (i.e., excludes pack ice). c In 1997-1999, this area is based on the 2000 Northstar development zone. d In each survey year, a data logger program automatically recorded time and aircraft position at 1-s (in 1997-1998, 2000-2001) or 2-s (in 1999, 2002) intervals throughout the flights. collected over pack ice and during conditions of poor sightability were excluded from the data set. Data collected over water depths <3 m were excluded from statistical analyses. Numbers of sightings in the transect segments were examined relative to three types of potential influences described below: (1) Northstar development zone, (2) other industry sites, and (3) “natural” variables that influence seal distribution and haul-out behavior. Industrial activities at Northstar and other sites within the study area during the winter and spring are summarized in Table 5.2. (1) Data collected <10 km from Northstar in 2000, 2001, and 2002 were categorized by 1-km bins measured from the edge of the Northstar development zone. That zone was defined by the extent of the physically altered or flooded sea ice. In 2000, that zone included Northstar Island, the pipeline route, and ice roads (Fig. 5.4A, later). In 2001, it included the island, ice roads, subsea pipeline, and an area where emergency equipment was tested (Fig. 5.4B). In 2002, the development zone included the island, an ice road, and an equipment testing area near the island (Fig. 5.4C). Data collected from 1997–1999 (Fig. 5.5, later) were considered “pre-industry” data, although ice roads to the Northstar site were constructed during the winter preceding the 1999 survey (Fig. 5.5C). Data collected >10 km from Northstar in 2000– 2002, and all data collected in 1997–1999, were categorized as “Distance from Northstar = 11”. 5-6 Monitoring at Northstar, Ice-Covered Season 1999. 97–2002. c Activities Activities ng the ice-covered seasons of 19 seasons ng the ice-covered On-ice seismic surveys in winter 1997-1998. in winter seismicOn-ice surveys Shallow-hazards survey at the McCovey site in April 2000. Involved borehole core core borehole Involved in April site 2000. McCovey at the survey Shallow-hazards sampling, penetration cone testing,of and use multibeam side-scan sonar; plus a route. transit the clear to used was bulldozer Ice road construction (Dec. 1999 to March 2000); island construction: 18,300 round- 18,300 construction: island 2000); March to 1999 (Dec. construction road Ice truckstrips to Northstar,of gravel impact to April driving ice 2000); pile (Feb. 2000). May to (March pipelines of installation and trenching, cutting, Ice road construction (Nov. 2000 to Feb. 2001); island maintenance; drilling (Dec. (Dec. drilling maintenance; island 2001); Feb. to 2000 (Nov. construction road Ice 2000 to2001); June pipeline maintenance (March to April 2001); emergency equipment testing (Dec. 2000 and April 2001). roadIce construction (Nov. 2001 to March 2002); operation of gas turbine drilling to for oil compressors 2001 continued (Nov. and and production generators June to 2001 Dec. (periodically testing equipment emergency limited 2002); June 2002). Exploratory drilling, ice road construction, reinforcing islandwith ice, very limited 1996-1997. winter in program drilling (soils) geotechnical and Vibroseis b sites within the study area duri within sites Location Area extending from extending Area shore to 10-m depth contour near eastern boundary of studyarea Ice roads to Northstar, subsea pipeline from shore to to from shore pipeline subsea Northstar, to roads Ice Northstar, and Northstar Island Ice roads to Northstar, subsea pipeline from shore to to from shore pipeline subsea Northstar, to roads Ice area testing and Island, Northstar Northstar, Tern Island Tern Tern Island Limited geotechnical surveys in winter 1997-1998. a ) 2 d d Area (km 2000 0.16 Transit route and shallow-hazards survey area 2001 6.13 "2001" 2001 0.46area Nearshore east of Endicott On-ice seismic surveys in winter 2000-2001. Drilling Drilling 2002 0.71 Island Northstar and Northstar to road Ice Drilling 1997 "Eastern" 1998 150.60 "Prudhoe Bay" 1998 4.92 near Area Prudhoe Bay On-ice seismic surveys in winter 1997-1998. 5.2. Summary of industrial activities at Northstar and other and activities at Northstar 5.2. Summary of industrial Island Construction 2000 2.93 Ice road construction road Ice 1999 0.76 Northstar to road Ice April early in abandoned was road 1999); April to (Jan. construction road Ice Geotechnical surveysGeotechnical 1998 Industrial Activity Year Represents total area of the industry zone over water depths > 3 m; not the actual area surveyed. See Figure 5.1. Distance from Tern in used Island analyses; was therefore, no on-ice area associated with this industry zone. See Chapter 2 for more details about industrial activities at Northstar. ABLE Limited Construction and a b c d Other Industry Shallow-hazard Survey Shallow-hazard Vibroseis Northstar T Chapter 5: Northstar Effect on Seal Distribution 5-7

(2) Similarly, data collected <10 km from other industry sites within the study area (e.g., Vibroseis, shallow-hazards survey, other industry; Table 5.2) were categorized by 1-km bins measured from the edges of those industry zones (Fig. 5.1). Data collected >10 km away, or in years without industrial activity, were categorized as “Distance from Shallow-hazards Survey = 11”, “Distance from Vibroseis = 11”, and “Distance from Other Industries = 11”. (3) Natural “habitat” covariates considered in analyses included ice deformation, melt water, water depth, and distance from the edge of the landfast ice (Moulton et al. 2002). A new habitat variable not considered in our previous analyses distinguished whether the location was within a lagoon (i.e., inshore of a barrier island) or not. Temporal variables considered included time of day, day of year, and year. Weather variables considered included air temperature, wind speed, cloud cover, and heat loss (an index of wind chill). Statistical Tests A Poisson regression model (Cameron and Trivedi 1998) was used to quantify effects of proximity to Northstar industrial activities on the number of sightings of seals on the ice, after allowing for other factors that also influence the number of seal sightings. The Poisson regression equation had the form [ ˆ ] EY= (area )exp b01122+ bx+ bx+ ... + bp xp where Y was the number of seal groups seen in a transect segment, (area) was a measure of the effective area of the transect segment (i.e., offset variable), bi were estimated regression coefficients, and xi were covariates. As noted above, there were three categories of predictor variables. (1) The variable of most direct interest, “Distance from Northstar”, was scaled from 0 to 11. For the 2000–2002 surveys, it coded whether the transect segment was within the development zone (0), 0–1 km away (1), 1–2 km away (2), …, 9–10 km away (10), or >10 km away (11). All data from 1997–1999, when there were no intensive construction or drilling activities at Northstar, were coded “11”. (2) Three similar variables were constructed to represent the distances from “Vibroseis”, “Shallow-hazards Survey”, and “Other Industry” (if active at the time). (3) The model also considered 13 environmental and temporal variables as potential covariates, consistent with Moulton et al. (2002) except for the addition of the “lagoon” variable and omission of “visibility” and “time from solar noon”. The first step in the analysis involved building a “base” model that only included natural covariates, which accounted for much of the variation in seal sightings (see Moulton et al. 2002). Most natural variables considered for inclusion in the Poisson regression model were treated as continuous; the exceptions were “survey replicate number”, “year”, and “lagoon”. Variables that were organized in ordered bins (e.g., distance from ice edge) were considered continuous and not discrete, resulting in a model that was easier to interpret. To investigate possible non-linear relationships to natural covariates, quadratic terms were included for water depth, ice deformation, distance from edge of the landfast ice, heat loss, wind speed, air temperature, cloud cover, time of day, and date. All “natural” continuous variables, and the quadratic terms for continuous variables, plus the interaction of year with each of these variables, were considered for inclusion in the model. Survey replicate was nested within year and within all interactions between year and each of the continuous covariates. Main effects were not considered for elimination from the model if they were involved in statistically-significant interactions. Forward model selection (Rawlings et al. 1998) was employed to derive the “base” model. At each step, the covariate added to the model was the one that would, if included, be most effective in “explaining” the remaining variability in seal numbers. We employed the Bayesian information criterion (BIC) to assess how many natural covariates should be included in the “base” model (Venables and 5-8 Monitoring at Northstar, Ice-Covered Season

Ripley 1999). When values of the BIC increased rather than decreased from one model step to the next, indicating that the model fit was not improving, the “base” model was limited to this number of covariates. To test for the effects of industry on seal densities, each industry variable was forced into the “base” model of natural covariates. Significance of all terms (natural and industry) in the model was assessed by approximate F-tests, accounting for overdispersion of sightings via a quasi-likelihood approach. “Distance from Northstar” and other industry variables were considered statistically significant based on the α=0.05 criterion. A three-step approach was used when forcing in industry variables. For example, to test for the effects of Northstar on number of seal sightings within a transect segment, (1) the linear and quadratic terms for “Distance from Northstar” were forced into the base model. If this was not significant based on the α=0.05 criterion, then (2) just the linear term was forced into the base model. (3) The interaction term between each natural covariate included in the base model (Table 5.3, later) and “Distance from Northstar” was also forced into the model and corresponding P-values were assessed. If the P-value was <0.05, the interaction of the natural covariate and “Distance from Northstar” was included in the final model. This three-step procedure was repeated for “Vibroseis”, “Shallow-hazards Survey”, and “Other Industry”. In summary, the final Poisson model consisted of the (1) natural covariates selected based on the BIC, (2) any significant (α=0.05 criterion) linear or quadratic terms for industry variables, and (3) any significant interactions between industry variables and natural covariates from (1). Moran’s I (Moran 1950) and deviance residuals were calculated to examine autocorrelation and model fit. Calculations were done with S-Plus Version 6.1 (Venables and Ripley 1999). Power Analysis A post hoc power analysis of the 1997–2002 data was done to assess the size of a “Northstar” effect (change in sighting density relative to distance from Northstar) that could have been detected if it occurred. Statistical power is defined as the probability of correctly rejecting a false null hypothesis. Power was assessed by simulating numerous datasets based on an assumed effect size, followed by regression estimation and hypothesis testing for each simulated dataset, and then repeating for different assumed effect sizes. Power was estimated as the proportion of these hypothesis tests that indicated rejection of the null hypothesis. The power analysis conducted here is based on the same data used in the Poisson regression analysis and as such considers habitat, temporal, and weather covariates. An adverse impact of the Northstar development on seals would have manifested itself as a positive coefficient for the variable “Distance from Northstar”. That is, we would expect fewer seals close to the Northstar development zone and more seals farther away from the zone. Thus, the power analysis focused on hypothetically positive coefficients for “Distance from Northstar”. To determine statistical power, we used the natural covariates in the fitted Poisson regression model (see Table 5.3, later) plus a hypothetical term for “Distance from Northstar” with a coefficient represented by β*. Here β* is the assumed strength of the Northstar effect on the sighting density close to Northstar in units “percent change in density per km from Northstar”. A sequence of hypothetical values for β*, specifically 0, 0.0015, 0.0030, 0.0045, …, 0.0300, was assessed. For each of these β* values, we calculated the predicted number of seal sightings in every transect segment that was surveyed. The predicted ˆ th value, Yi , represents the mean or expected number of seal sightings in the i transect segment given both the observed covariate values for that segment and the regression relationship (the fitted coefficient estimates plus the assumed coefficient for “Distance from Northstar”). Chapter 5: Northstar Effect on Seal Distribution 5-9

The regression model assumes that the observed number of seal sightings would vary as an overdis- * ˆ * ˆ persed Poisson random variable with mean = Yi and variance = φYi , where φ = 1.09 is the estimate of the dispersion parameter (based on the fitted Poisson regression model; see Table 5.3, later). Simulation of overdispersed Poisson variates was based on a relationship that expresses the negative binomial distribution as a mixture of Poisson and gamma distributions (Hausman et al. 1984). Therefore, in each transect segment, we used the expected number of seal sightings to generate a single negative binomial random variate, i.e., a simulated count. A new Poisson regression equation was fitted to these simulated counts, obtaining an estimate for β*. For a simulated dataset, regression results included a Wald t-test (t = coefficient / standard error, adjusted for overdispersion) for the null hypothesis β* = 0 versus the alternative hypothesis β* > 0. The test indicated that the “Distance from Northstar” coefficient was significantly greater than 0 when the observed t-statistic was greater than the critical value tdf,1-α, where α was either 0.05 or 0.10. This process (generation of random variates, regression estimation, and hypothesis testing) was repeated 1000 times for each hypothetical value of β*. Statistical power was estimated as the proportion of times (out of the 1000 simulations for each value of β*) that the calculated t-statistic was greater than the critical value for some specified α-level. If the null hypothesis was true (β* = 0), the estimated power of the test should be close to α. If the null hypothesis was false (β* > 0), the estimated power should be greater than α. Furthermore, as the true (simulated) value of β* increases, power should increase. A second power analysis was conducted to investigate the influence of environmental covariates on the power of the analysis to detect a Northstar effect. As in the first power analysis, for each hypothetical value of β* (specifically 0, 0.0075, 0.0090, 0.0105, …, 0.1050), the number of seal sightings in every transect segment was predicted based on the observed covariate values for that segment, the fitted regression relationship, and the assumed coefficient for “Distance from Northstar”. These predicted numbers of seal sightings were used to generate simulated counts from negative binomial (i.e., overdispersed Poisson) distributions. However, estimation of the Poisson regression equation differed from that in the first power analysis in that the new equation fitted to the simulated counts did not contain covariates. This mimicked a situation where environmental covariates influence seal counts, but those covariates are unknown or unmeasured and are not accounted for in the regression equation. This process (generation of random variates, regression estimation excluding covariates, and hypothesis testing) was repeated 1000 times for each hypothetical value of β*. Power was calculated the same way as in the previous power analysis (i.e., power = % iterations with t-statistic > critical tdf,1-α).

RESULTS

Observed Seal Densities and Distribution

The overall observed ringed seal density on landfast ice, calculated as total seals observed in all replicates / total area surveyed in all replicates, was 0.39, 0.35, 0.56, 0.42, 0.49, and 0.72 seals/km2, in 1997 to 2002, respectively. The six-year average density was only 0.07 seals/km2 in areas where the summer water depth is <3 m; in these areas, much or all of the water column is frozen during spring. If those shallow areas are excluded, observed densities were 0.43, 0.39, 0.63, 0.47, 0.54, and 0.83 seals/km2 in 1997 to 2002, respectively. 5-10 Monitoring at Northstar, Ice-Covered Season

Ringed seals were observed in all parts of the study area >3 m deep. Ringed seals were common in deeper parts of the nearshore lagoons inshore of barrier islands in all survey years except 1999. In 1999, there was a tendency for lower densities inside the lagoons, and seals were concentrated farther offshore in waters 15–25 m deep (Moulton et al. 2002). In 1998 and 1999, higher densities of ringed seals were observed in the western portion of the study area, but this was not evident in 1997 or 2000–2002. Environmental Factors

The “base” Poisson model of natural covariates included numerous habitat, temporal, and weather factors that accounted for significant variation in ringed seal densities (Table 5.3). The influences of environmental factors on ringed seal densities observed in 1997–2002 were generally consistent with previously-reported results based on the 1997–1999 data (Moulton et al. 2002). Factors Affecting Local Abundance Based on the Poisson regression, number of seal sightings tended to peak around water depths of 5–20 m (Fig. 5.2A). There was a significant interaction between year and distance from the landfast ice edge, indicating that the relationship between seal density and distance from the ice edge varied from year to year. After allowance for other factors, peak densities tended to occur relatively close to the ice edge in 1997, 1998, 1999, and 2002, and well inshore away from the ice edge in 2000 and 2001 (Fig. 5.2B; Table 5.3). The edge of the landfast ice was closer to shore in 2002 than in other survey years: 15.8 km north of Northstar Island in 2002 vs. an average distance of 25.4 km (± st.dev. 1.75 km) in 1997–2001. Significantly lower seal densities occurred in areas with high levels of ice deformation and melt water (Fig. 5.2C,D). The model suggested that, after allowance for other factors, significantly more seal sightings occurred inside lagoons than outside (Table 5.3). Factors Affecting Proportion of Seals Hauled Out The sighting rate was significantly related to survey date during each of the six survey years. After allowance for other factors, sighting rates peaked near the middle of the 1999 and 2000 surveys, near the mid- to-late portion of the 1997 survey, and near the end of the 1998, 2001, and 2002 surveys (Fig. 5.2E; Table 5.3). The Poisson model indicated that, after allowance for other covariates, time of day was not significantly related to seal sighting rate within the range of times when our surveys were conducted (10:00–18:00 local time). Based on the Poisson regression, a significant negative relationship occurred between numbers of seal sightings and heat loss (Fig. 5.2F; Table 5.3). A significant positive relationship was found between numbers of seal sightings and cloud cover (Fig. 5.2G; Table 5.3). The Poisson model indicated that, after allowance for other covariates, air temperature and wind speed were not significantly related to seal sighting rate. Northstar Effects

Seal Densities Observed near Northstar Raw sighting and density data, prior to allowance for other covarying factors, showed no evidence of reduced seal densities close to Northstar during the springs of 2000 or 2001, and inconclusive evidence for 2002 (Fig. 5.3, 5.4). Some of the highest observed densities in 2000 and 2001 occurred within the Northstar development zone and 0–1 km outside that zone. In 2002, there was a slight tendency for lower observed densities within 2 km of the development zone, but the overall observed density within 10 km of the Northstar development was higher than during all previous survey years. No seals were sighted within the Northstar development zone in 2002, but surveys in that zone encompassed only a small area (1.2 km2) where Chapter 5: Northstar Effect on Seal Distribution 5-11

TABLE 5.3. Number of ringed seal sightings in 1997–2002 in relation to distance from industry and environmental variables, as estimated by a Poisson regression model. “Distance from Northstar” was not significantly related to number of seal sightings.

% Change in sighting nos. 95% CI for for one unit percent change Standard F -value increase of Covariate Model Term Coefficient Error (approx.) P -value covariate Lower Upper Unit

Intercept 1997 -444.72 45.80 1998 -435.02 50.29 1999 -490.07 51.87 21.16 <0.0001 2000 -476.21 49.76 2001 -474.75 48.95 2002 -471.67 49.06 Main Effects a

Water Depth 0.0370 0.0092 16.35 <0.0001 b Water Depth2 -0.0014 0.0002 34.52 <0.0001 Ice Deformation -0.0162 0.0009 322.30 <0.0001 -1.60 -1.78 -1.42 10% Melt Water -0.0114 0.0014 70.07 <0.0001 -1.13 -1.41 -0.86 10% Lagoon 0.5013 0.2255 4.70 0.0302 c Date 5.8835 0.6086 95.17 <0.0001 b Date2 -0.0195 0.0020 Heat Loss -0.0007 0.0001 66.88 <0.0001 -0.07 -0.09 -0.05 100 W/m2 Cloud Cover 0.0015 0.0004 17.51 <0.0001 0.15 0.08 0.22 10% Distance from Vibroseis 0.0196 0.0087 5.17 0.0230 1.97 0.26 3.69 1 km

Interactions d Distance from Ice Edge-1997 -0.0260 0.0044 -2.56 -3.41 -1.70 5 km Distance from Ice Edge-1998 -0.0183 0.0049 -1.81 -2.77 -0.85 " Distance from Ice Edge-1999 -0.0489 0.0055 -4.77 -5.85 -3.68 " 36.97 <0.0001 Distance from Ice Edge-2000 0.0194 0.0043 1.95 1.11 2.80 " Distance from Ice Edge-2001 0.0137 0.0045 1.38 0.50 2.26 " Distance from Ice Edge-2002 -0.0173 0.0043 -1.71 -2.56 -0.86 " Date2-1997 5.8835 0.6086 Date2-1998 5.8192 0.5897 Date2-1999 6.1773 0.6434 20.21 <0.0001 b Date2-2000 6.0828 0.6275 Date2-2001 6.0762 0.6236 Date2-2002 6.0614 0.6201 Dist. Fr. Shall.-haz. Survey-outside Lagoon 0.0389 0.0115 3.96 1.68 6.24 1 km 9.05 0.0026 Dist. Fr. Shall.-haz. Survey-in Lagoon -0.0260 0.0188 -2.56 -6.30 1.18 "

Note: df = 60234; Overdispersion = 1.09; residual correlation (Pearson's r ) = -0.01. a The relationships of seal sightings to most of these covariates were the same across years; i.e., there was no significant interaction between a variable and most of these covariates. The exception was the quadratic term for "Date". b This value and the corresponding CI were not calculated because it is difficult to define a covariate unit for quadratic terms. c This value and the corresponding CI were not calculated because "Lagoon" is a factor vs. a continuous variable. d F -values indicate a significant interaction between the covariate and "Year", or in the case of "Distance from Shallow-hazards Survey", with "Lagoon"; i.e., coefficients changed significantly from year to year or outside vs. inside a lagoon. The coefficients presented here represent the actual slopes for each year or outside vs. inside lagoon . 5-12 Monitoring at Northstar, Ice-Covered Season

FIGURE 5.2. Number of ringed seal sightings in relation to distance from industry zones and environmental variables in 1997–2002 as estimated by a Poisson regression model. “Distance from Northstar” was not significantly related to number of seal sightings. P-value in (I) represents interaction between “Distance from Shallow-hazards Survey” and “Lagoon”. The relationship in (I) was not significant inside the lagoon. Chapter 5: Northstar Effect on Seal Distribution 5-13

FIGURE 5.3. Observed density of ringed seals within and near the Northstar development zone in 2000, 2001, 2002, and in 1997–1999 relative to the 2000 zone. No seals were sighted in the “Within” distance category in 2002 when only a small “Within” area was surveyed (1.2 km2). only ~1 seal might be expected. Raw sighting data during the springs of 1997–1999 (Fig. 5.5A,B,C), prior to re-construction of the island, showed no evidence of higher seal abundance near Northstar relative to years with intensive construction and drilling. Average observed seal densities were lower near Northstar in 1997– 1999 than in years with intensive construction and drilling (Fig. 5.3). Seal Densities with Allowance for Other Factors The Poisson regression analysis indicated that, after allowance for other predictor variables, there was no evidence of reduced seal densities close to Northstar during the springs of 2000, 2001, and 2002. When “Distance from Northstar” was forced into a “base” model that included nine natural covariates, no significant relationship between sighting probability and distance was evident (F = 0.53, P = 0.47 for linear term; F = 0.52, P = 0.47 for quadratic term). Also, there were no significant interactions between “Distance from Northstar” and those natural covariates included in the “base” model. After allowance for covariates, there was no evidence that seal densities near Northstar were reduced in 2000–2002, when there was industrial activity at and near Northstar, relative to 1997–1999, when there was no intensive industrial activity there. Power of Poisson Regression Analysis The post hoc power analysis indicated that the Poisson regression, including allowance for environmental covariates, had high power to detect small-scale changes in seal sighting density relative to distance from Northstar. If β* = 0.03, power was estimated to be essentially 1.0 (0.999 for α = 0.05 and 1.000 for α = 0.10; Fig. 5.6). Here, the asterisk indicates that the regression coefficient β is an assumed one (see “Methods”). More specifically, if the number of seal sightings in a transect segment within 10 km of Northstar increased by a factor of 1.0305 (= exp[0.03]), or about 3%, with each additional kilometer away from the Northstar development zone, the effect would almost certainly have been 5-14 Monitoring at Northstar, Ice-Covered Season

FIGURE 5.4. Distribution of ringed seal sightings near Northstar during (A) 31 May – 7 June 2000, (B) 28 May – 8 June 2001, and (C) 30 May – 7 June 2002. Ten successive 1-km distance categories from edges of Northstar development zone are shown. Chapter 5: Northstar Effect on Seal Distribution 5-15

FIGURE 5.5. Distribution of ringed seal sightings near the abandoned Seal Island site (now Northstar) during (A) 26 May – 4 June 1997, (B) 23–30 May 1998, and (C) 4 – 13 June 1999. No industrial activities took place near Northstar in 1997 and 1998. In 1999, ice roads to Seal Island were constructed but abandoned before aerial surveys commenced. Ten successive 1-km distance categories from the edges of Northstar ice roads are shown in (C). 5-16 Monitoring at Northstar, Ice-Covered Season detected. This would represent a 35% reduction in seal density within 1 km of the development zone relative to that >10 km away. At β* = 0.015, power was still high (0.72 for α = 0.05 and 0.83 for α = 0.10). In other words, with a 1.5% increase in the number of seal sightings per kilometer of distance (out to 10 km) there would have been an 83% chance of detecting this effect at the α = 10% level. As expected, power was estimated to be close to the nominal α level when β* = 0 (0.037 for α = 0.05 and 0.086 for α = 0.10). Exclusion of the environmental covariates from the Poisson regression dramatically decreased the * power of the analysis. Power to detect 0.03 < β < 0.08 was very close to 0.0 when covariates were not * included in the model (Fig. 5.6). When covariates were included, power to detect 0.03 < β < 0.08 was very close to 1.0 (Fig. 5.6). Other Anthropogenic Effects

The Poisson regression model suggested that seal sighting density was significantly reduced (P = 0.023) close to areas where Vibroseis had occurred during the preceding winter (Table 5.3; Fig. 5.2H). Similarly, sighting density was apparently significantly reduced close to an area (McCovey) where a shallow-hazards survey had occurred earlier in the winter (P = 0.0026; Table 5.3; Fig. 5.2I). Interpre- tation of this latter effect was complicated by an interaction with an environmental variable (“Lagoon”— see Table 5.3), and by the presence of other human activities in the same area during the aerial survey period (see Discussion). The covariate “Other Industry” was not significantly related to sighting density

DISCUSSION

Observed Seal Densities

Based on our surveys, the overall observed ringed seal density on landfast ice in the central Alaskan Beaufort Sea ranged from 0.35 to 0.72 seals/km2 in 1997–2002, or 0.39 to 0.83 seals/km2 if areas with water <3 m deep are excluded. Observed densities varied by a factor of ~2× in our six-year study period. Comparable and larger inter-annual variations in ringed seal densities have been commonly reported by other researchers in the Alaskan Beaufort Sea (Frost et al. 2002) and elsewhere in the Arctic (Stirling et al. 1977; Smith and Stirling 1978; Lunn et al. 1997; Kingsley and Byers 1998). These variations have been attributed to actual changes in seal abundance from year to year (Frost et al. 2002) and to differences in the proportion of seals hauled out. The latter is related to the timing of molt (Lunn et al. 1997; Frost et al. 2002; Kelly et al. 2002) and to varying weather and ice conditions (Burns and Harbo 1972; Finley 1979; Moulton et al. 2002). Results of our Poisson regression analysis indicate that there was inter-annual variation in seal numbers even after accounting for a variety of habitat, weather and temporal factors. Varying annual densities during this study are likely the result of some combination of changes in actual seal abundance and the proportion of seals basking relative to the timing of molt. Northstar Effects

There was no indication that ringed seal densities during spring were significantly reduced by industrial activities at Northstar in 2000, 2001, and 2002. When “Distance from Northstar” was forced into the model, either without or with an associated quadratic term to allow for a possible non-linear relationship to “Distance”, the relationship between sighting density and distance from the Northstar Chapter 5: Northstar Effect on Seal Distribution 5-17

FIGURE 5.6. Power of the Poisson regression to detect various sized Northstar effects quantified by the regression coefficient for “Distance from Northstar”. β represents the change in sighting density for a 1 km change in distance from Northstar (for distances ≤10 km). Power was assessed both including and excluding the natural covariates in the analysis. development zone was far from significant (P = 0.47 for the linear term). Also, overall seal densities in the study area were not reduced in 2000–2002 with Northstar development relative to 1997–1999 with no intensive industrial activities at Northstar. In 2000, when on-ice construction activities were most intense, observed seal densities were, if anything, slightly higher within 1 km (and within 3 km) of the development zone as compared with areas somewhat farther away (Fig. 5.3). Also, on-ice surveys conducted with trained dogs on 2–8 Dec. 1999 and again on 19–21 May 2000 found 19 active ringed seal structures (breathing holes and lairs) within ~1 km of Northstar. One pupping lair was located between two ice roads, as close as 140 m from the nearest road (Williams et al. 2001). Our aerial surveys commenced 10 days after the May dog-assisted searches. The aerial surveys detected four seals (2 sightings) within the zone identified as physically- disturbed or flooded (Northstar development zone) as well as other seals within ~1 km (Fig. 5.4A). During the Feb.–May period preceding the 2000 aerial surveys, ~18,300 truckloads of gravel were transported along the ice roads to Northstar, a pair of pipelines were constructed from Northstar to shore, and many other construction activities took place. These findings suggest that some individual ringed seals, including breeding females, tolerate much construction activity, and that ringed seal densities near Northstar were not significantly reduced by the intense construction activities in 2000. Results from previous less-detailed studies of ringed seal densities near offshore construction sites have been mixed. Green and Johnson (1983) studied the effects on seals of construction of the precursor of Northstar Island (i.e., Seal Island) in the winter of 1982. They found indications of local displacement, 5-18 Monitoring at Northstar, Ice-Covered Season but concluded that the overall effects on seal distribution and densities were insignificant. However, during some but not all years, seal densities within 3.7 km (2.3 mi) of artificial islands were lower than those 3.7–7.4 km (2.3–4.6 mi) away. The reduction was 50–70% within 3.7 km of islands under construction or with drilling activity (Frost and Lowry 1988; Frost et al. 1988). These previous results should be treated with caution, as the small amount of survey coverage limits meaningful comparisons of densities relative to distance from the island. Also, previous analyses did not account for factors like water depth that varied within 3.7 km vs. 3.7–7.4 km from the islands, thereby confounding interpretation of the results. Kingsley (1986) reported that, after accounting for habitat preferences, ringed seal density was not reduced in areas with industry activity during the previous winter. In 2001, there was no evidence that seal numbers were reduced near the Northstar development zone. In 2001, observed seal densities generally were lower 3–10 km from the industry zone than <3 km away. On-ice surveys with trained dogs, similar to those done the previous year but over a larger area, were done in late 2000, and in March and May 2001 (Williams et al. 2002). Once again, numerous seal structures were located near the Northstar development zone. The dog-assisted study indicated that during May (before aerial surveys began), the density of active structures was higher near Northstar (0–2.5 km away) than farther away (2.5–3.5 km). These results suggest that any negative effects on seals are likely minor and very localized. These finer scale results corroborate the findings of the aerial survey study. In 2002, industrial activities at Northstar included drilling and production, with use of gas-turbine engines rather than the diesels used in previous years. These activities were probably audible to ringed seals, at least intermittently, out to ~1.5 km in water and ~5 km in air (see CHAPTER 3). In 2002, the raw density data provided some indication (Fig. 5.3) that observed densities were lower within 2 km of the development zone relative to densities somewhat farther away. However, the Poisson regression did not detect a significant “Distance from Northstar” effect, despite having high statistical power to do so if such an effect occurred. Other Anthropogenic Effects

Vibroseis surveys that occurred in portions of the study area (see Fig. 5.1) during early 1998 and 2001, prior to our survey dates, apparently did have a negative affect on observed seal densities. The area with Vibroseis surveys in 1998 was larger than the area included in the Northstar development zone in 2000–2002, the years with construction and operational activities. Burns and Harbo (1972) concluded that ringed seals in the Prudhoe Bay area were not appreciably displaced by intensive on-ice seismic activity. However, Link et al. (1999), based on univariate analysis of the same 1998 data that are included in the present multivariate analysis, did find that fewer seals were sighted within and near the Vibroseis site. The present multivariate model suggests that this apparent effect was not an artifact. The multivariate model also suggested that, after allowance for other factors, observed seal densities in 2000 might have been lower near an area where an on-ice shallow-hazards survey had been done several weeks earlier (cf. Coltrane et al. 2001). However, seal researchers were working on the ice in the same general area prior to and during the aerial survey period (Kelly et al. 2000). The relative contributions of the previous industrial activity and the ongoing seal research in reducing observed seal numbers in that area are unknown. This study was not designed to assess the effects of Vibroseis or other anthropogenic activities (aside from Northstar) on seals within our study area. Thus, the above results, although statistically significant, should be considered only suggestive. Chapter 5: Northstar Effect on Seal Distribution 5-19

Environmental Factors

Habitat, temporal, and weather factors had significant effects on densities of ringed seals observed during this six-year study. Higher seal densities occurred in areas with low levels of ice deformation and little melt water. This is consistent with findings by other researchers who conducted surveys in a larger portion of the central Alaskan Beaufort Sea (Frost et al. 2002). Areas with rough ice may not be suitable habitat during molt as seals basking on top of the ice and snow would likely be more vulnerable to predation by polar bears (Stirling and Archibald 1977; Smith 1980; Frost et al. 1988, 2002). Although some degree of ice deformation is required for sufficient accumulation of snow to create lairs, it is possible that some seals may redistribute themselves before or during the molt period to areas with flat ice where detection of polar bears is easier. Areas with lower amounts of melt water may be favored because exposure to melt water would conduct heat away from the skin, thereby inhibiting the molt process (Feltz and Fay 1966; Fay 1982). During 1997–2002, after allowance for other variables, the highest densities of seals tended to occur in waters 5–20 m deep, which is a slightly shallower range of preferred depths than found based on analysis of the 1997–1999 data alone (Moulton et al. 2002). It appears that, after excluding data from areas <3 m deep and allowing for other variables, higher densities of seals occurred within the deeper parts of lagoons than outside lagoons. This is an unexpected result, perhaps related to habitat or prey availability. The relationship between sighting density and distance from the ice edge differed significantly among years. The highest sighting densities tended to occur closer to the ice edge in 1997– 1999 and 2002 than in 2000 and 2001. Densities of seals near the ice edge are related to the stage of ice break-up; more seals occur near the edge as break-up advances (Frost et al. 1988; Moulton et al. 2002). Break-up was not far advanced in 2000 and 2001 even at the end of the aerial surveys. No significant diel pattern in seal haul-out behavior was detected by the Poisson regression within the range of times when our surveys were conducted (10:00–18:00 local time). Frost et al. (2002) also did not find a consistent relationship between seal density and time of day across a similar range of times (11:00–17:00 local time). This lack of apparent relationship with time is not surprising as aerial surveys were purposefully conducted during the times of day when peak haulout is thought to occur (Smith 1973; Kelly and Quakenbush 1990). The effect of date on sighting density differed significantly from year to year. Sighting rates peaked near the middle of the 1999 and 2000 survey periods, in the mid-to-late portion of the 1997 survey, and near the end of the 1998, 2001, and 2002 surveys. These results suggest that the dates when the largest proportion of seals were hauled out, and hence available to be counted, varied from ~29 May in 1998 to 8 June in 1999. Over the six years, the mean day of year when maximal densities of basking seals occurred on the ice, after allowance for other variables, was 154.5 (~3–4 June). More ringed seal sightings occurred when much or all of the sky was covered by clouds, based on the 1997–2002 data. Seal sighting rates were negatively related to heat loss, indicating that a combination of cold air temperatures and high wind speeds inhibited ringed seal haul-out behavior. This is consistent with observations that ringed seals are more likely to escape into their breathing holes when aircraft fly overhead during colder wind chill conditions (Born et al. 1999) and that sighting densities increased with increasing air temperatures in 1997–1999 (Moulton et al. 2002). 5-20 Monitoring at Northstar, Ice-Covered Season

Effectiveness of Intensive Aerial Surveys

Survey Design The intensive and replicated design of the aerial surveys conducted for this study were intended to detect changes in ringed seal densities near BP’s Northstar development. Other researchers have acknow- ledged that broad-scale aerial surveys are not suitable for detecting small-scale changes in seal densities in restricted areas (Frost et al. 1988). Previous broader-scale surveys off northern Alaska have typically used transects spaced 1.8 km apart and flown once or at most twice in a spring season, sampling any given area on 1–2 days each spring. Our surveys involved lines spaced 0.9 km apart, each of which was flown twice per season, with coverage of any given area on ~8 days each spring (nominally 20 transects per day, spread across the study area). The high intensity of aerial survey coverage within our study area was designed to assess possible small-scale changes in seal density close to the industrial site(s) after oil development began. The within-season replication was designed, in part, to allow detection and quantification of distributional effects that might be quite localized. Also, the replication was designed to assess and distinguish the relative contributions of industrial effects and various natural factors (e.g., date in season, ice conditions, bathymetry, weather) on numbers of seals hauled out at different places and times. Our surveys of the Northstar area during each year were spread over 7–9 days of flying, and the area near Northstar was sampled almost daily. This approach is expected to reduce year-to-year variability in the proportion of the seals detected, as compared with the variability that could occur if the area of interest was surveyed on fewer days (or only one day) in a given year. The collection of environmental data over a six-year study period, and the inclusion of these data in the multivariate analysis as covariates, greatly increased the power of the statistical analysis to detect any effects of Northstar (or any other specific factor) on seal abundance and distribution (Fig. 5.6). Analysis Approach As discussed in Moulton et al. (2002), the use and interpretation of covariates in assessing changes in ringed seal densities is complicated. Given the correlations among some covariates, it is not always possible to isolate the effects of one variable from those of another related variable. Despite these complications, our analyses provide insight into factors, or groups of related factors, most likely to have direct effects on observed seal density. The Poisson regression analysis can help determine whether nearby industrial activities affect observed seal density after allowance for the effects of all natural factors combined. We strongly recommend that future surveys account for the influences of natural factors on seal abundance and haul-out behavior when assessing industry effects. Statistical Power The post hoc power analysis indicated that changes in ringed seal densities near Northstar probably would have been detected by the Poisson regression analysis if they were present. If the sighting density within 10 km of Northstar changed by 3% per km (35% overall), it almost certainly would have been detected. A change by 1.5% per km might well have been detected. The high statistical power of this study is a consequence not only of the large sample size resulting from intensive data collection over a six-year study period, but also the inclusion of natural covariates in the multivariate analysis. The power analysis verifies that natural covariates need to be accounted for before assessing effects of industrial activities on ringed seal densities. Power is much higher when this is done. Chapter 5: Northstar Effect on Seal Distribution 5-21

CONCLUSIONS

Based on intensive and replicated aerial surveys during the spring of 1997–1999 (“pre-industry”) and 2000–2002 (industry), there was no evidence that construction, drilling, and production activities at BP’s Northstar oil development affected local ringed seal distribution and abundance. Seal densities were not significantly reduced near Northstar in 2000–2002 either in comparison with more distant areas or in comparison with the same area in pre-industry years. The study design and Poisson regression approach (with covariates) had high power to detect small-scale changes in seal densities near Northstar if such changes had occurred. Some other industry activities, e.g., Vibroseis and perhaps shallow-hazards surveys, that occurred in other portions of the study area during one or more survey years appeared to be associated with locally reduced seal densities. However, certain environmental factors such as bathymetry, degree of ice deformation, date, and percent cloud cover, had more conspicuous and statistically- significant effects on seal sighting rates. This study shows that effects of an offshore oil development on the local distribution of basking ringed seals are no more than slight, and are small in comparison to the effects of natural environmental factors. An understanding of the latter effects is essential when assessing potential impacts of industrial activity on ringed seals.

ACKNOWLEDGEMENTS

This study was funded by BP Exploration (Alaska) Inc. as part of an effort to satisfy its corporate goal of minimizing and documenting impacts associated with the Northstar oil development, and to fulfil regulatory requirements. We thank W.R. Koski and G.W. Miller of LGL, who helped design the project, and the numerous members of a peer and stakeholder group convened annually by the National Marine Fisheries Service (NMFS), who provided comments on study design and results. We also thank B. Adams and J. Leavitt (of Barrow, AK), and S.R. Johnson, G.W. Miller and L. Noel of LGL, for their participation as aerial observers (along with VDM and MTW). G.W. Miller, M.R. Link and T.L. Olson of LGL did much of the original analysis of the 1997–98 data. We are grateful to W. Lentsch and V. Domke of Tamarack Air Ltd. (1997), V. Aberly and D. Lowell of Corporate Air (1999), and the late T. Blaesing (2000, 2001), R. Aiken (2002), D.C. Weintraub (1998, 2001), K. Sutton (2001), C. Pike (2002), and T. Lampee (2002) of Commander Northwest, for piloting the survey aircraft and providing other logistical support. A. Haskell, K. Hester, M. Holst, W. King, M. Moulton, C. Perham, S. Schwenn, A. Serrano and A. Wright of LGL helped in various ways. We are grateful to C. Herlugson, R. Jakubczak and B. Streever of BP, and D. Trudgen of OASIS Environmental on behalf of BP, for their support during the study. We also thank B. Streever of BP and R. Suydam of the North Slope Borough Dept. of Wildlife Management for reviewing a draft of this chapter. Additional Acknowledgements applicable to the entire report appear in Chapter 1. The 1997–2001 aerial surveys were conducted under the General Authorization for scientific research on non-endangered species, MMPA §104(c)(3)(C) and 50 C.F.R. §216.45, per NMFS Letters of Confirmation 481-1382 and 481-1626. The 2002 aerial surveys were conducted under NMFS Scientific Research Permit 481-1623-01.

LITERATURE CITED

Born, E.W., F.F. Riget, R. Dietz and D. Andriashek. 1999. Escape responses of hauled out ringed seals (Phoca hispida) to aircraft disturbance. Polar Biol. 21(3):171-178. 5-22 Monitoring at Northstar, Ice-Covered Season

Burns, J.J. and S.J. Harbo, Jr. 1972. An aerial census of ringed seals, northern coast of Alaska. Arctic 25(4):279- 290. Cameron, A.C. and P.K. Trivedi. 1998. Regression analysis of count data. Cambridge Univ. Press, New York, NY. 412 p. Coltrane, J., M.T. Williams and C.J. Perham. 2001. On-ice location of ringed seal (Phoca hispida) structures at the McCovey Site, central Alaskan Beaufort Sea, spring 2000. LGL Rep. P498. Rep. from LGL Alaska Res. Assoc. Inc., Anchorage, AK, for Western Geophysical, Anchorage, AK, and Nat. Mar. Fish. Serv., Anchor- age, AK, and Silver Spring, MD. 22 p Fay, F. H. 1982. Ecology and biology of the Pacific walrus, Odobenus rosmarus divergens Illiger, 1811. N. Am. Fauna No. 74. Feltz, E. T. and F.H. Fay. 1966. Thermal requirements in vitro of epidermal cells from seals. Cryobiology 3(3):261-264. Finley, K.J. 1979. Haul-out behaviour and densities of ringed seals (Phoca hispida) in the Barrow Strait area, N.W.T. Can. J. Zool. 57(10):1985–1997. Frost, K.J. and L.F. Lowry. 1988. Effects of industrial activities on ringed seals in Alaska, as indicated by aerial surveys. p. 15-25 In: W.M. Sackinger et al. (eds.), Port and Ocean Engineering Under Arctic Conditions. Volume II. Symposium on Noise and Marine Mammals. Geophysical Inst., Univ. Alaska, Fairbanks, Fairbanks, AK. 111 p. Frost, K.J., L.F. Lowry, J.R. Gilbert and J.J. Burns. 1988. Ringed seal monitoring: relationships of distribution and abundance to habitat attributes and industrial activities. U.S. Dep. Commer., NOAA, OCSEAP Final Rep. 61 (1989):345-445. NTIS PB89-234645. Frost, K.J., L.F. Lowry, G. Pendleton and H.R. Nute. 2002. Monitoring distribution and abundance of ringed seals in northern Alaska. OCS Study MMS 2002-043. Final rep. from the Alaska Dept. of Fish and Game, Juneau, AK, for U.S. MMS, Anchorage, AK. 66 p. + Appendices. Green, J.E. and S.R. Johnson. 1983. The distribution and abundance of ringed seals in relation to gravel island construction in the Alaskan Beaufort Sea. p. 1-28 In: B.J. Gallaway (ed.), Biological studies and monitoring at Seal Island, Beaufort Sea, Alaska 1982. Rep. from LGL Ecol. Res. Assoc. Inc., Bryan, TX, for Shell Oil Co., Houston TX. 150 p. Greene, C.R., Jr., and M.W. McLennan, with R.W. Blaylock. 2000. Sound and vibration measurements during Northstar construction in early 2000. p. 4-1 to 4-31 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of ringed seals and sounds during construction of BP's Northstar oil development, Alaskan Beaufort Sea, winter and spring 1999-2000: 90-day report. LGL Rep. TA2426-1. Rep. from LGL Ltd., King City, Ont., and LGL Alaska Res. Assoc. Inc., Anchorage, AK, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 107 p. Hammill, M.O., C. Lydersen, M. Ryg and T.G. Smith. 1991. Lactation in the ringed seal (Phoca hispida). Can. J. Fish. Aquat. Sci. 48:2471-2476. Hausman, J.A., B.H. Hall, and Z. Griliches. 1984. Econometric models for count data with an application to the patents – R and D relationship. Econometrica 52:909-938. Kelly, B.P. and L.T. Quakenbush. 1990. Spatiotemporal use of lairs by ringed seals (Phoca hispida). Can. J. Zool. 68(12):2503-2512. Kelly, B.P., L.T. Quakenbush and B.D. Taras. 2000. Correction factor for ringed seal surveys in northern Alaska. p. 51-64 In: Annu. Rep. 6, Coastal Marine Institute, Univ. Alaska Fairbanks. OCS Study MMS 2000-046. Kelly, B.P., L.T. Quakenbush and B.D. Taras. 2002. Correction factor for ringed seal surveys in northern Alaska. p. 3-15 In: Annu. Rep. 8, Coastal Marine Institute, Univ. Alaska Fairbanks. OCS Study MMS 2002-001. Chapter 5: Northstar Effect on Seal Distribution 5-23

Kingsley, M.C.S. 1986. Distribution and abundance of seals in the Beaufort Sea, Amundsen Gulf, and Prince Albert Sound, 1984. Environ. Stud. Revolv. Funds Rep. 25. Dep. Fish. & Oceans, Winnipeg, Man. 16 p. Kingsley, M.C.S. and T.J. Byers. 1998. Failure of reproduction in ringed seals (Phoca hispida) in Amundsen Gulf, Northwest Territories in 1984-1987. p. 197-210 In: M.P. Heide-Jørgensen and C. Lydersen (eds.), Ringed seals in the North Atlantic. Tromsø, Norway: NAMMCO Scientific Publications, Vol.1. Link, M.R., T.L. Olson and M.T. Williams. 1999. Ringed seal distribution and abundance near potential oil development sites in the central Alaskan Beaufort Sea, spring 1998. LGL Rep. P-430. Rep. from LGL Alaska Res. Assoc. Inc., Anchorage, AK, for BP Explor. (Alaska) Inc., Anchorage, AK. 58 p. Lunn, N.J., I. Stirling and S.N. Nowicki. Distribution and abundance of ringed (Phoca hispida) and bearded (Erignathus barbatus) in western Hudson Bay. Can. J. Fish. Aquat. Sci. 54:914-921. Lydersen, C. and M.O. Hammill. 1993. Diving in ringed seal (Phoca hispida) pups during the nursing period. Can. J. Zool. 71:991-996. Moran, P. A. P. 1950. Notes on continuous stochastic phenomena. Biometrika 37:17-23. Moulton, V.D., W.J. Richardson, T.L. McDonald, R.E. Elliott and M.T. Williams. 2002. Factors influencing local abundance and haulout behaviour of ringed seals (Phoca hispida) on landfast ice of the Alaskan Beaufort Sea. Can. J. Zool. 80(11):1900-1917. Moulton, V.D., W.J. Richardson, M.T. Williams and S.B. Blackwell. 2003. Ringed seal densities and noise near an icebound artificial island with construction and drilling. Acoust. Res. Let. Online 4(4):112-117. Rawlings, J.O., S.G. Pantula and D.A. Dickey. 1998. Applied regression analysis: a research tool, 2nd ed. Springer- Verlag, New York, NY. 657 p. Richardson, W.J., C.R. Greene, Jr., C.I. Malme and D.H. Thomson. 1995. Marine mammals and noise. Academic Press, San Diego, CA. 576 p. Siple, P.A. and C.F. Passel. 1945. Measurements of dry atmospheric cooling in subfreezing temperatures. Proc. Am. Phil. Soc. 89:177-199. Smith, T.G. 1973. Censusing and estimating the size of ringed seal populations. Fish. Res. Board Can. Tech. Rep. 427: 22 p. Smith, T.G. 1980. Polar bear predation of ringed and bearded seals in the land-fast sea ice habitat. Can. J. Zool. 58:2201-2209. Smith, T.G. and I. Stirling. 1975. The breeding habitat of the ringed seal (Phoca hispida). The birth lair and associated structures. Can. J. Zool. 53(9):1297-1305. Smith, T.G. and I. Stirling. 1978. Variation in the density of ringed seal (Phoca hispida) birth lairs in the Amundsen Gulf, Northwest Territories. Can. J. Zool. 56(5):1066-1071. Stirling, I. and W.R. Archibald. 1977. Aspects of predation of seals by polar bears. J. Fish. Res. Board Can. 34: 1126-1129. Stirling, I., W.R Archibald and D. DeMaster. 1977. Distribution and abundance of seals in the eastern Beaufort Sea. J. Fish. Res. Board Can. 34(7):976-988. Terhune, J.M. and K. Ronald. 1975. Underwater hearing sensitivity of two ringed seals (Pusa hispida). Can. J. Zool. 53(3):227-231. Venables, W.N. and B.D. Ripley. 1999. Modern applied statistics with S-Plus, 3rd ed. Springer-Verlag, New York, NY. 501 p. Williams, M.T., J.A. Coltrane and C.J. Perham. 2001. On-ice location of ringed seal structures near Northstar, December 1999 and May 2000. p. 4-1 to 4-22 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of 5-24 Monitoring at Northstar, Ice-Covered Season

industrial sounds, seals, and whale calls during construction of BP’s Northstar Oil Development, Alaskan Beaufort Sea, 2000. (April 2001). LGL Rep. P485-2. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 316 p. Williams, M.T., T.G. Smith and C.J. Perham. 2002. Ringed seal structures in sea ice near Northstar, winter and spring of 2000-2001. p. 4-1 to 4-33 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of industrial sounds, seals, and whale calls during construction of BP’s Northstar Oil Development, Alaskan Beaufort Sea, 2001. (April 2002). LGL Rep. P557-2. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 286 p.

IN MEMORY OF TOM BLAESING

We dedicate this chapter to the late Tom Blaesing who passed away suddenly in November 2001. Tom was the head pilot and “CEO” of Commander Northwest. He piloted the survey plane with great expertise for this study in 2000 and 2001. Tom was also key in helping LGL locate a survey plane for the 1999 survey and in providing aircraft services in 1998; his co-workers provided excellent service in 2002. Tom devoted much effort to the modification and outfitting of aircraft to make them suitable for specialized survey and related work in remote areas. This has been to the great benefit of many researchers and many projects in Alaska and elsewhere. We all appreciated Tom’s professionalism and accommodating nature—our surveys went smoothly largely due to his efforts. Tom was a great guy—fun, thoughtful, and energetic. For those of us who flew with him, our thoughts often wander his way and he is sadly missed.

Ringed seal aerial survey crew in spring 2001. From left to right: Tom Blaesing, Val Moulton (LGL), Mike Williams (LGL), and Kevin Sutton (co-pilot). CHAPTER 6:

SOUND MEASUREMENTS, 2002 OPEN-WATER SEASON1

a Susanna B. Blackwell

Greeneridge Sciences, Inc.

a 1411 Firestone Road Goleta, CA 93117 (831) 685-2773; [email protected]

for

BP Exploration (Alaska) Inc. Dept. of Health, Safety & Environment 900 East Benson Blvd, P.O. Box 196612 Anchorage, AK 99519-6612

LGL Report TA2705-2

December 2003

1 Chapter 6 In: W.J. Richardson and M.T. Williams (eds.). 2003. Monitoring of industrial sounds, seals, and bowhead whales near BP’s Northstar oil development, Alaskan Beaufort Sea, 1999–2002. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 6-2 Monitoring at Northstar, Open-Water Season

ABSTRACT

The primary objective of the acoustic measurement program at Northstar during the 2002 open- water season was to document underwater and airborne sounds during the first open-water season with oil production operations. Greeneridge Sciences Inc. made sound measurements during three days in Sep- tember 2002 at seven distances (¼ to 20 n.mi., or ½–37 km) north of Northstar Island. In addition, two near-island cabled hydrophones and (for backup) autonomous seafloor recorders, all located about ¼ n.mi. (420 m) off the north shore of Northstar, recorded underwater island sounds continuously for 31 days from 31 Aug. to 1 Oct. in 2002. Measurements at the seven distances showed that, in the absence of vessels, both underwater and in-air broadband levels were the lowest recorded in the three years for which these types of measurements have been made (2000–02). Background levels (lowest recorded values) were 87–90 dB re 1 µPa underwater and 37–40 dBA re 20 µPa in air at the more distant stations. Maximum levels reached 116 dB re 1 µPa underwater and 56 dBA re 20 µPa in air. While both drilling and oil production were taking place, underwater and airborne sounds reached background levels ~3.5 km (2.2 mi) from Northstar. This is comparable to values obtained in similar conditions in other studies. When the crew boat or other operating vessels were present, sound levels were higher, up to 128 dB re 1 µPa at 3.7 km. The cabled hydrophone showed that broadband (10–1000 Hz) levels recorded in 2002 spanned a narrower range than in 2001: The 95th percentile was lower in 2002 (117.2 dB re 1 µPa) than in 2001 (122.8 dB). The 5th percentile was higher in 2002 (94.8 in 2002) than in 2001 (87.8 dB). Median values were comparable in the two years: 103.0 dB in 2002 vs. 102.6 dB in 2001. Drilling and oil production, present in September 2002 but absent in September 2001, may have been responsible for the higher 5th percentile values in 2002. Vessels were an important but intermittent sound source around Northstar during both years, and many “spikes” in broadband levels could be attributed to the crew boat and barge traffic. The ISI (Industrial Sound Index) is an index used to quantify the amount of industrial noise emanating from Northstar over specified time intervals. The index is used in analyzing how Northstar sounds affect the migration path of bowhead whales (see Chapters 7–11). The index is introduced in this chapter because it is a measure of Northstar sound. Five different methods were tested to determine which frequency bands should be used in defining the ISI for the period 31 Aug.–23 Sept. 2002, when bowhead migration was studied. One method provided equivocal results, but four methods supported use of the five one-third octave bands centered at 31.5–80 Hz to define the ISI in 2002.

INTRODUCTION

Background

During the winter of 1999–2000, BP Exploration (Alaska) Inc. began construction of its Northstar production unit in nearshore waters of the Alaskan Beaufort Sea. Northstar is built on the eroded remains of the original Seal Island, an artificial island constructed by Shell Oil Company in 1982 about 5 km (3 mi) offshore of Long Island (one of the Return Islands), northwest of Prudhoe Bay. It was used in 1984 and 1985 for exploratory drilling, and subsequently abandoned. The artificial island originally named Northstar was built northwest of Seal Island by Amerada-Hess in 1985 for exploratory drilling. However, BP’s facilities at the old Seal Island location are now widely known as Northstar Island. The Northstar Development addressed in this report includes the gravel island itself, living quarters, 30 wells Chapter 6: Sound Measurements, 2002 Open-Water Season 6-3

(some still to be drilled), and associated oil production facilities on the island. Also included are two buried subsea pipelines that connect the island to the existing production infrastructure in Prudhoe Bay. Construction of the island, emplacement of production and support facilities, and drilling of the initial production wells were completed by the end of the 2001 open-water season. Also, during the autumn of 2001 the island's main power source transitioned from diesel generators to gas turbines. The first oil was produced on 31 Oct. 2001. Initial oil production along with drilling of additional wells occurred during the ice-covered season in late 2001 and early-mid 2002. BP’s Northstar activities are described in Chapter 2. During the summer and early autumn of 2002, the period of interest in this report, additional wells were drilled (above formation depth) while production was taking place. This was the first open-water season when Northstar was in a production rather than a construction mode, and the first open-water season when gas- turbine engines were operating on Northstar to provide electrical power and to operate gas compressors. The work described in this chapter is a continuation of the open-water acoustics work described in Blackwell and Greene (2001, 2002). This chapter provides a detailed account of the specific measurements in 2002, complementary to the just-cited reports on the corresponding 2000 and 2001 measurements. The present chapter is an expanded version of one that appeared in the 90-day report on monitoring during the 2002 open-water season (Blackwell 2003); the present chapter supersedes that earlier version. The specific acoustic monitoring plans for the 2002 open-water season were described in BP’s request to NMFS (dated 15 May 2001) for a Letter of Authorization (LoA) covering the first year of oil production at Northstar. The sound measurements during the open-water season of 2002 complement the measurements obtained during the first and second summers of Northstar construction in 2000 and 2001 (Blackwell and Greene 2001, 2002). This chapter provides a detailed account of the specific measurements in 2002, complementary to the just-cited reports on the corresponding 2000 and 2001 measurements. The three reports combined will result in at least one journal manuscript to be completed in 2003 and submitted for publication. Objectives and Approach

The LoA applicable to the first year of oil production called for continued acoustic measurements to document production sounds and sounds of types that had not been measured in previous years. “New” types of sounds that were recorded during summer 2002 consisted of those related to oil production (with continued drilling), and those related to the shift in power source from diesel generators to gas turbines. These types of sounds had been recorded in the winter of 2001–02 (Chapter 3), but it was important to repeat those measurements during the open-water season when sound propagation conditions are different. As described in the LoA Request of 15 May 2001, received levels, spectral characteristics, specific sources, and transmission loss were to be determined. Two methods were to be used during the 2002 open-water period to obtain sound measurements: (1) a hydrophone and a microphone were to be deployed from a boat at various distances from Northstar; and (2) a fixed, bottom-mounted hydrophone was to be deployed 300–500 m (984–1640 ft) seaward of Northstar Island and connected to the island by a cable. This was to obtain continuous data at one distance over a period of a week or more, primarily to document temporal variability. This second effort was to be continued for all of September as part of the acoustic monitoring of the bowhead whale migration (see Chapters 7–10). Backup sensors and recorders, and frequent checks on the quality of the incom- ing data, were implemented to ensure that these data were acquired continuously for the required period in September 2002. The boat-based samples were intended primarily to provide information on propagation characteristics. The cabled-hydrophone data were to document a variety of operational sounds over an extended period, 6-4 Monitoring at Northstar, Open-Water Season allowing assessment of source variability. Analysis of the data from the fixed recording site also allows us to recognize and quantify any changes in sound characteristics among the different periods when boat-based recordings were acquired. Propagation characteristics are unlikely to change significantly over periods of hours. In combination, the data from the brief recordings at various stations plus the “reference” data from the fixed hydrophone should be sufficient to characterize propagation, at least on the day of the measurements. Boat-based hydrophone and microphone data were acquired in the Northstar area on three dates in September 2002. Uninterrupted cabled hydrophone data were obtained for all of September, including the three days when boat-based open-water recordings were obtained, and the entire period when the bowhead whale migration was being monitored by acoustic methods (see Chapters 7–10).

METHODS

Several different measurement units appear in this report. The following conversions can be used: • 1 m ≅ 3.3 feet • 1 km = 1000 m ≅ 0.54 n.mi. (nautical mile) ≅ 0.62 mi • 1 n.mi. ≅ 1.85 km ≅ 1.15 mi Boat-based Recordings in Open Water

Boat-based open-water recordings of island sounds were obtained on 5, 6 and 12 Sept. 2002 (Table 6.1; Fig. 6.1), to allow comparisons with similar measurements obtained during the summers of 2000 and 2001 (Blackwell and Greene 2001, 2002). Recordings were made north of the island, at 7 nominal distances: ¼, ½, 1, 2, 5, 10 and 20 n.mi. (0.46–37 km or 0.29–23 mi). Unlike previous years, we did not reach the pack ice even at the farthest station (20 n.mi.). During the morning of 5 Sept., Nuiqsut whalers (based at Cross Island) struck a bowhead whale. Because of their concerns about possible interference, we did not proceed to the offshore stations on that date, and only made recordings at the three proximal stations (¼, ½ and 1 n.mi. or 0.46–1.85 km from Northstar). On 6 and 12 Sept. all of the stations were sampled (Table 6.1; Fig. 6.1). On 4 Aug. 2002, the SDC (Steel Drilling Caisson) arrived in the Prudhoe Bay area and was positioned in the McCovey prospect, about 10 n.mi. east (and slightly north) of Northstar. On 14 Aug., cold stacking of the caisson was completed. It therefore did not generate industrial noise during the subsistence whaling season (early September to early October), which includes the days of recording reported on in this chapter. However, waves slapping onto the steel hull no doubt produced some sound. The SDC has no connection with BP, and is mentioned here only because of its proximity to Northstar. Equipment The sensors included a hydrophone and a microphone, both calibrated. The hydrophone was an International Transducer Corporation (ITC) model 6050C (S/N 617), which includes a low-noise preamplifier next to the sensor and a 98 ft (30 m) cable. The hydrophone cable was attached with cable ties to a fairing to minimize strumming. Prior to recording, the hydrophone signals were amplified with an adjustable-gain postamplifier. The omnidirectional microphone was a G.R.A.S. Sound and Vibration ½-inch prepolarized free field microphone model 40AE with an ICP preamplifier model TMS426C01 and a windscreen. Prior to recording, the microphone signals were amplified with an adjustable-gain postamplifier. Chapter 6: Sound Measurements, 2002 Open-Water Season 6-5

TABLE 6.1. Distance from Northstar (Seal) Island and water depth for acoustic recordings on 5, 6 and 12 Sept. 2002. Some stations have more than one entry. A second recording was obtained at a particular station if sea state had notice- ably decreased or if a certain sound source (i.e., crew boat), present during the first recording, was no longer present.

Distance Water Depth Date and Station km mi feet m

5 Sept. 2002 ¼ n.mi. 0.50 0.31 46 14.0 ½ n.mi. #1 0.94 0.58 47 14.3 “ #2 1.00 0.63 47 14.3 1 n.mi. 1.96 1.22 50 15.2

6 Sept. 2002 ¼ n.mi. 0.53 0.33 46 14.0 ½ n.mi. 0.95 0.59 47 14.3 1 n.mi. 1.82 1.13 50 15.2 2 n.mi. #1 3.57 2.22 48 14.6 “ #2 3.41 2.12 48 14.6 5 n.mi. 9.27 5.76 66 20.1 10 n.mi. #1 18.51 11.50 80 24.4 “ #2 18.50 11.49 80 24.4 20 n.mi. #1 37.06 23.03 120 36.6 “ #2 37.05 23.02 121 36.9

12 Sept. 2002 ¼ n.mi. #1 0.59 0.37 42 12.8 “ #2 0.54 0.33 44 13.4 ½ n.mi. #1 0.98 0.61 44 13.4 “ #2 0.94 0.58 46 14.0 1 n.mi. #1 1.99 1.24 46 14.0 “ #2 1.88 1.17 47 14.3 2 n.mi. #1 3.77 2.35 45 13.7 “ #2 3.74 2.32 45 13.7 5 n.mi. 9.18 5.71 63 19.2 10 n.mi. 18.60 11.56 77 23.5 20 n.mi. 37.08 23.04 118 36.0

Hydrophone and microphone signals were recorded on two channels of a SONY model PC208Ax instrumentation-quality digital audio tape (DAT) recorder. The recorder was run at regular speed for 2-channel recording, providing a frequency response that was nearly flat from <4 to 20,000 Hz on each channel. Both sensors were calibrated from 4 to 20,000 Hz. Quantization was 16 bits, providing a dynamic range of >80 dB between an overloaded signal and the instrumentation noise. A memo channel on the tape recorder was used for voice announcements, and the date and time were recorded automatically. 6-6 Monitoring at Northstar, Open-Water Season

FIGURE 6.1. Acoustic recording locations north of Northstar (=Seal) Island, 5, 6 and 12 Sept. 2002. Northstar is indicated by the yellow square and the cold-stacked SDC (Steel Drilling Cais- son) is indicated by the green triangle. On 5 Sept., only the ¼, ½ and 1 n.mi. stations were sampled, whereas all stations were sampled on 6 and 12 Sept. 1 n.mi. = 1.15 mi and 1.85 km.

Field Procedures All recordings were obtained using the Alaska Clean Seas (ACS) vessel Harrison Bay of length 12.8 m (42 feet) as a recording platform. After selecting an appropriate recording location, the hydrophone and spar buoy to which it was connected were lowered into the water with hydrophone depth = 10.5 m. A depth reading was taken with the recording vessel's depth sounder before all sound-generating devices (engines, generator, depth sounder) on the vessel were turned off. A microphone was positioned on the deck of the vessel in such a way that it had an unobstructed path to the sound source at all times. During recording, the recording vessel drifted with the current. A GPS position (Garmin model 12XL) was obtained at the beginning and end of each recording. These positions were used to calculate a mean distance of each station from the northern shore of Northstar (= Seal) Island. A laser rangefinder (Bush- nell model # 20-0880) was used for distances <800 m (i.e., at the 0.25 n.mi. station). Wind speed, wind direction, and temperature were recorded over a period of 3–4 minutes at each station with a Kestrel 2000 Chapter 6: Sound Measurements, 2002 Open-Water Season 6-7

Pocket Thermo Wind meter (Nielsen Kellerman, Chester, 19013 PA); these data are shown in Table 6.2. In addition, notes on wave height were made at each recording station in order to determine sea state. A total of 122.5 min of boat-based recordings were obtained. The main sound producing activities on Northstar Island during the open-water measurements included the drilling of new wells and oil production (i.e., pumping of crude oil and injection of pressurized gas), in addition to the usual boat traffic associated with the island. Recordings were made with no foreknowledge of the sound-producing activities taking place on and around the island. How- ever, the following information was collected from island personnel on the schedule of various sound sources during the recordings: 1. Schedule of the crew boat Arctic Express 2. Departure and arrival times at West Dock and Northstar for the daily island barge, if applicable 3. Minute by minute drilling record and summary of drill rig operations 4. Island Rotating Equipment report, which indicates which generators, compressors and pumps were running 5. Daily Operations Report, which details overall island activities. Wind speed information is also on hand from the Northstar Island anemometer. However, the sensor was at a level below the levels of the processor plants and will not be accurate for wind with a northerly component. Cabled Hydrophones and ASARs

The cabled hydrophone (C.H.) system consisted of two identical cabled hydrophones (i.e., a duplicate set), deployed about 420 m from the north shore of Northstar Island (Fig. 6.2). In addition, three Autonomous Seafloor Acoustic Recorders (ASARs), deployed a total of four times, were used as backups (Fig. 6.2). The system was designed to provide a continuous record of sounds emanating from Northstar (plus ambient noise) over an extended period, including the period in September when bowhead whale migration was to be monitored acoustically (see Chapters 7–10). As it turned out, the two cabled hydrophones provided an uninterrupted record of island sounds from 30 Aug. until 1 Oct., a total of 31 days. They rendered unnecessary any use or examination of the data collected by the three ASARs that were deployed as redundant systems. No further mention will therefore be made of the ASAR recorders (for a description, see Blackwell and Greene 2002). For the present task, the cabled hydrophone data were used to characterize variation in island noise while open-water recordings were obtained successively at distances of ¼–20 n.mi. from Northstar. Equipment An ITC model 8212 hydrophone, which contains a low-noise preamplifier, was spliced to the conductors of an 1800 ft (550 m) double-armored well-logging cable to form a "cabled hydrophone". Inside the "Whale House" (see below), a battery was connected to two of the conductors to power the preamplifier in the hydrophone. The hydrophone signals were connected to electronics similar to those used in the DASARs (Directional ASARs) described in Chapters 7–10. (The C.H. recording system had been upgraded relative to that used in 2000.) This equipment provided for signal conditioning and recording on 10 and 30 GB disks. Sampling was at 2000 samples/second, twice as fast as in the DASARs (see Chapter 8), to permit recording sounds at frequencies up to 1000 Hz. Quantization was 16 bits. The disk capacity and battery life were sufficient to record without interruption for over 40 days. 6-8 Monitoring at Northstar, Open-Water Season

TABLE 6.2. Mean and maximum wind speed, temperature and sea state at each of the recording stations on 5, 6 and 12 Sept. 2002. Wind Speed (mph) Date and Station Temperature Sea State* mean max (ºC)

5 Sept. 2002 ¼ n.mi. 2.1 2.3 10.5 0 ½ n.mi. #1 2.0 2.2 10.4 ½ “ #2 2.0 2.2 10.5 ½ 1 n.mi. 2.1 2.4 10.4 ½

6 Sept. 2002 ¼ n.mi. 2.3 2.5 8.2 0 ½ n.mi. 1.7 2.2 8.9 0 1 n.mi. 2.2 2.4 8.1 0 2 n.mi. #1 2.1 2.8 8.3 ½ “ #2 2.1 2.8 8.3 ½ 5 n.mi. 1.8 2.3 8.6 ½ 10 n.mi. #1 2.8 3.2 5.8 1 “ #2 3.0 3.3 5.9 ½ 20 n.mi. #1 3.0 4.4 5.7 2 “ #2 1.5 1.7 5.9 ½

12 Sept. 2002 ¼ n.mi. #1 8.5 9.8 4.2 2 “ #2 8.5 9.8 4.2 2 ½ n.mi. #1 5.8 6.9 3.9 2 “ #2 2.3 2.8 4.6 2 1 n.mi. #1 5.4 6.4 4.1 2 “ #2 3.5 3.8 4.6 2 2 n.mi. #1 6.8 8.1 4.3 2 “ #2 4.9 6.0 4.6 2.5 5 n.mi. 8.9 10.1 4.2 2.5 10 n.mi. 6.0 7.0 4.0 2.5 20 n.mi. 7.4 8.6 3.8 2.5

*determined from wave height, which was also recorded

Field Procedures The hydrophone was mounted on a small iron stand (C.H. #2) or a larger stand made out of PVC tubing with weights attached to it (C.H. #1). C.H. #1 was deployed 440 m north of the north shore of Northstar Island on 31 Aug. (Fig. 6.2). C.H. #2 was deployed to the east of C.H. #1 on 30 Aug., 417 m from the north shore, in a location very close to that of C.H. #2 during the summer 2001 (see Fig. 6.2). Water depth at both locations was 12–13 m. The shore end of the cable was placed in the "cracks" between adjacent concrete blocks comprising the slope protection outside the island retaining wall. The cables extended to the “Whale House”, a small shack (about 3’ by 6’ and 7’ high) hanging on the north retaining wall of the island and containing the recorders themselves and the batteries powering them. (The “Whale House” corresponds to the “Bird House” of 2001, but the name was changed to avoid confusion with a twin structure used by researchers on Northstar for a bird study. In addition, the location of the structure was changed: in 2002 it was positioned very close to the northeastern corner of the island, further east Chapter 6: Sound Measurements, 2002 Open-Water Season 6-9

FIGURE 6.2. Locations of cabled hydrophones (C.H.) and autonomous seafloor acoustic recorders (ASARs) in relation to the “Whale House” (black square on x-axis). The Whale House was located on the north shore of Northstar Island (thick gray line) and housed the recorders for the C.H. systems. Lines projecting from ASAR symbols are cables connecting the recorders to their anchors. To allow retrieval, care had to be taken not to overlay lines. Dates on which the instruments started recording are indicated in parentheses. ASARs started recording at midnight on the day following their deployment. than in 2001.) Both C.H. recorders were started on 31 Aug. (at 11:06:33 for C.H. #1 and 11:12:56 for C.H. #2) and shut down at 11:00:00 on 1 Oct. The following events occurred during the 31-day recording period: • On 5 Sept., each recorder was turned off for about one hour to remove one of the hard disks and replace it with a new one. The removed disks were sent to Greeneridge Sciences in Goleta, CA, for a data quality check. The disk swap did not result in overall loss of data since the recorders were stopped sequentially, not simultaneously. • On 6 Sept., recorder #1 stopped recording for 12 minutes during a check of the hydrophone sounds, but recorder #2 was functioning throughout. • On 23 Sept., stormy weather resulted in currents that likely tipped over C.H. #1, leading to dis- torted sounds involving saturation. Data from C.H. #1 are therefore not useful after that date. In summary, the two C.H.s provided an uninterrupted record of island sounds for 31 days, from locations just above the seafloor a few hundred meters north of Northstar Island. 6-10 Monitoring at Northstar, Open-Water Season

Weather data (specifically hourly averages of wind speed) collected by an MMS weather station located on the roof of the Permanent Living Quarters (PLQ) on Northstar Island2 were used to assess the contribution of wind and wave sounds to background levels as recorded by the cabled hydrophone. The process module on Northstar is considerably taller than the PLQ building and therefore influences wind readings, particularly when the wind comes from the north (the main wind direction at Northstar is ENE). However, these data can still be used as a rough measure of wind speed over the month-long sampling period. Signal Analysis

Boat-based Recordings Underwater Sounds.— The recorded, digitized hydrophone signals were transferred as time series to a computer hard drive for processing. They were then equalized and calibrated in units of sound pressure with flat frequency response over the data bandwidth (4–20,000 Hz). Analysis was done using MATLAB (The MathWorks, 3 Apple Hill Drive, Natick, MA 01760-2098) routines and custom programs for analysis of both transient and continuous signals. For each recording, a sound pressure time series (waveform) was generated; an example is shown in Figure 6.3. In general, these plots showed varying levels as sound sources approached or receded, or started and stopped. The sound was played via a speaker to help the analyst identify the probable source of the sound and to facilitate matching notes from the field with the recorded sounds. The sound waveform was used to select representative samples for further analysis. If the overall sound pressure levels (SPLs) varied little with time, at least two 8.5-s samples were selected from the recording and analyzed. If the sound waveform showed fluctuations in the SPL (as in Fig. 6.3), then 8.5-s samples were taken from both the stronger and the weaker sections of the recording. Throughout this chapter, the term “sample” (of sound) refers specifically to these 8.5-s long slices of the time series unless specified otherwise. Frequency composition was determined by calculating the sound pressure spectral density by Fourier analysis. The averaging time for such measurements was 8.5 s. One-third octave band levels were derived from the narrowband spectral densities by summing the mean square pressures in all frequency cells between the lower and upper frequency limits for the one-third octave band in question. Proportional amounts were taken from the end cells as appropriate. To show how received levels varied with distance from the activity, root-mean-square (rms) broadband SPL values were plotted against range from the dominant source. These plots are based on the series of sound recordings at varying distances along a given transect. Interpretation of these data is complicated by variability of the source activities within and between recordings, and by the presence (at some recording stations) of significant sound from more than one sound source. Nevertheless, the “received level vs. range” plots give an estimate of the levels received at several distances during the various activities studied. In addition, where appropriate, spectral analyses of the dominant sounds are included. Airborne Sounds.—Microphone data were transcribed to disk files and analyzed in the same way as boat-based hydrophone data. Broadband microphone data were A-weighted (and expressed in dBA referred to 20 µPa) to allow comparisons with common airborne sounds described in the literature. During A-weighting, a frequency-dependent weighting factor is applied to the sound in accordance with the sensitivity of the human ear; therefore frequencies below 1 kHz and above 6 kHz are de-emphasized (Kinsler et al. 1982; Kryter 1985). One-third octave and narrowband data were not A-weighted and are expressed in dB re 20 µPa.

2 Weather data obtained at site http://www.resdat.com/mms/ Chapter 6: Sound Measurements, 2002 Open-Water Season 6-11

FIGURE 6.3. Typical sound pressure time series, showing received sound pressure vs. time. This 120-s recording was obtained on 6 Sept. 2002, about 1 n.mi. (1.82 km or 1.13 mi) north of the island while the crew boat approached the island from the southeast and passed behind Northstar, on its way to the western dock. The island served as a sound barrier between the recording vessel and the crew boat, decreasing the sound pressures by an order of magnitude.

Propagation Loss Modeling.—We fitted a simple propagation model to broadband levels received by both hydrophone and microphone in order to develop equations that characterize propagation loss underwater and in air. The model used was based on logarithmic spreading loss:

Received Level (RL, dB re 1 µPa(3)) = A – B × log(R), where R is range in m. Eq. (1)

In this equation, the constant term (A) is the hypothetical extrapolated received level at distance 1 m (dB re 1 µPa(3)) based on far-field measurements. The estimated “A” value is useful mainly as a basis for comparison with other sound sources operating in the same region. Expected values for the spreading loss term (B) are 20 dB/tenfold change in distance for spherical spreading (such as seen in the open ocean far from surface, bottom or other boundaries) and 10 dB/tenfold change in distance for cylindrical spreading (Richardson et al. 1995). Spreading-loss terms exceeding 20 dB/tenfold change are possible with shallow sources and/or receivers. Cabled Hydrophone Signals Data collected by C.H. #2 (see “Field Procedures” under “Cabled Hydrophones and ASARs”, above) were used to determine sound spectral densities (i.e., the sound levels at different frequencies)

3 The units dB re 1 µPa are used for underwater sound pressure levels; units for in-air sound pressure levels are dB re 20 µPa, applicable to all microphone data. When A-weighting is applied to in-air measurements, the sound pressure levels are cited as dBA re 20 µPa. 6-12 Monitoring at Northstar, Open-Water Season every 4.37 minutes (262 s), or ~330 times per 24-hr day. These data were used to determine broadband and one-third octave band levels in the 10–1000 Hz frequency range and to provide a continual record of the levels and spectral characteristics of low-frequency underwater sounds near Northstar during the period of interest in this chapter, 31 Aug. (12:00 hrs)–1 Oct. (11:00 hrs). Statistical Spectra.—Once for each 4.37-min period, a series of one second long segments, overlapped by 50% and spanning one minute, were analyzed at frequencies up to 1000 Hz. The resulting 1-Hz spectra were averaged over one minute to derive a single averaged spectrum for the 1-min period. There were 10,192 of these 1-min spectra for the 31-day study period and 7551 for the 23-day period 31 Aug.–23 Sept. (see below). For each of the frequency cells in these averaged spectra, the values were sorted from smallest to largest, and the minimum, 5th-percentile, 50th-percentile, 95th-percentile, and maximum values for that frequency cell were determined. Calculation of the ISI.—The ISI (Industrial Sound Index) is an index used to quantify the amount of industrial noise emanating from Northstar over a specified time interval. The index is used in the study of how Northstar sounds affect the migration path of bowhead whales (see Chapters 7–10), but the index is introduced in this chapter because it is a measure of Northstar sound. The ISI is only calculated for underwater sound as it is underwater sound that is of concern with respect to bowhead whales. As defined by Greene et al. (2002) for 2001, the ISI was a measure of the sound in three one-third octave bands, those centered at 63, 80, and 160 Hz. For those bands, the sounds received at the cabled hydrophone in August 2001 appeared to be dominated by industrial components. (The most appropriate definition of the ISI for the migration period of bowhead whales in 2001 is re-assessed in Appendix 6.1 following this chapter.) In general, the frequencies considered in deriving the ISI should exclude, as much as possible, those with strong contributions by non-industrial sounds such as those produced by wind and waves. Because some of the sound sources at Northstar were different in 2002 than in 2001, it was not assumed that the best index of industrial sound in 2002 would be the same as that for 2001. Five methods were used in selecting the definition of the ISI in 2002. In this process, only the cabled hydrophone data acquired up to 23 Sept. were used because the storm on that date overturned the instruments used to monitor whale migration, terminating effective monitoring of the 2002 bowhead migration. • In the first method, the frequencies to include in the ISI were assessed by examining the distribution of sound across frequency based on the percentile narrowband spectral densities over the entire study period (31 Aug.–23 Sept.). • The second method is similar to the first, but the percentile distribution of the one-third octave band levels was used instead of the narrowband data. • The third method involved examining the relationship between one-third octave band levels and wind speed. Wind speed was logged from a gauge on top of the permanent living quarters (PLQ) on Northstar Island. Although this gauge was below the level of the processing units to its north, it provided the best available measure of wind speed at the island, as wind recorded at mainland facilities was known not to be the same. One-third octave band levels were measured as described above, based on spectral analysis of 1-min samples of C.H. 2 sound every 4.37 min. It was expected that the one-third octave bands most influenced by island sounds would show the least dependence on wind speed. To measure this dependence, the following linear equation was fitted separately to the 20 one-third octave band levels from 10 to 800 Hz, and the associated coefficients of determination r2 were determined: Expected Band Level = A + (B × Wind Speed) Eq. (2) Chapter 6: Sound Measurements, 2002 Open-Water Season 6-13

The values of B and r2 were expected to be lowest for those bands least related to the wind speed and most dependent on industrial sounds. • The fourth method amounted to cross-correlating the measured levels in the various one-third octave bands. The expectation was that levels in the bands dominated by island sound would correlate well with one another, but would not correlate well with the levels in bands dominated by natural sea noise. • In the fifth method, percentile levels during periods of calm and low winds (by one-third octave band) were used as indicators of island sound. Background sea noise from wind and waves should be minimum when the winds are low, permitting the island-generated sounds to dominate. (However, when the wind speed drops after having been high for a period of hours, the wave noise may take time to diminish correspondingly.) Percentile levels were computed as in method 2 while only using cabled hydrophone data collected during periods of low wind speeds (<5 knots or 2.6 m/s).

RESULTS Boat-based Recordings Results are presented below for all three days of recordings, including both underwater and in-air measurements. Underwater Sounds Routine Island Sounds.— Three series of recordings were made at up to seven nominal distances north of Northstar Island in order to record overall sound levels from the island, compare them to data collected during the summers of 2000 and 2001, and characterize propagation loss. On 5 Sept., recordings were made ¼, ½ and 1 n.mi. (½–2 km) north of the island. On both 6 and 12 Sept., recordings were made at seven distances from ¼ to 20 n.mi. (½–37 km) north of the island (Table 6.1; Fig. 6.1). Mean broadband (10–10,000 Hz) SPLs vs. distance are shown in Figure 6.4 for all three dates. These data represent times when routine production and/or drilling operations were occurring on the island. Recordings when transient events were taking place (such as the arrival of the crew boat) were omitted from this plot. When more than one recording was obtained at a particular station but in different sea states, the recording made at a lower sea state was used, unless that recording contained a transient event. Sounds produced by waves hitting the hull of the boat increase in importance with increasing sea state, and these tend to mask any sounds that might otherwise be received from distant industrial activities. On 5 and 6 Sept., sea state was lower (0–½) than on 12 Sept. (2–2.5). This difference, visible in Figure 6.4, is likely responsible for much of the difference in broadband levels between 6 and 12 Sept. Oil production took place on 5 and 6 Sept., but not on 12 Sept. due to the repair of a 4th stage LP (low pressure) compressor seal. Drilling was taking place during at least some portion of all the recordings shown in Figure 6.4, except for the one at the 1 n.mi. (1960 m) station on 5 Sept. The highest broadband level was recorded on 12 Sept. (drilling but no production), and was 116 dB re 1 µPa, about 600 m from Northstar. This recording contained prominent industrial sounds. The lowest level, 85 dB re 1 µPa, was obtained from one sample, 20 n.mi. (37 km) from Northstar on 6 Sept. (mean values are shown in Fig. 6.4). On that date, broadband levels seemed to have diminished to (or close to) background values within 2 n.mi., and were similar at all distances from 2 to 20 n.mi. (3½–37 km or 2.3–23 mi). On 6 Sept., industrial sounds could not be heard on the recordings from the 10 and 20 n.mi. stations by listening with headphones. Because sound levels reached background values too close to the source and with too few (2) useable stations for a logarithmic fit, none of the data sets could be used to determine propagation loss. 6-14 Monitoring at Northstar, Open-Water Season

FIGURE 6.4. Mean broadband (10-10,000 Hz) levels of underwater sound during routine production and/or drilling (left Y-axis, filled symbols) as a function of distance from Northstar, 5, 6 and 12 Sept. 2002. Corresponding sea states are also shown (right Y-axis, open symbols). Higher sea states on 12 Sept. led to higher received levels. 10 km = 5.4 n.mi. or 6.2 mi.

Figure 6.5 shows broadband levels (10–10,000 Hz) for all open-water recordings made during the summers of • 2000 (30 Aug., 1 and 17 Sept., see Blackwell and Greene 2001), • 2001 (10 and 12 Aug., see Blackwell and Greene 2002), and • 2002 (5, 6 and 12 Sept.). The principal utility of this Figure is to illustrate the wide range of levels possible at different distances, and that factors other than distance influence the sound levels. It shows that, for every distance studied, the lowest broadband levels were obtained on 6 Sept. 2002. The highest levels were those obtained on 12 Aug. 2001, while three sealift barges and seven tugs were in the Northstar area. On several dates, received levels tended to “flatten out” or reached a minimum at distances less than the maximum recording distance. This suggests that, at those times and distances, industrial sound from the Northstar area was no longer the dominant sound. However, on other dates, most notably 1 Sept. 2000 and 12 Aug. 2001, received levels continued to diminish out to the maximum distances studied on those dates—about 29 and 20 km, respectively (15.5 and 11 n.mi. or 18 and 12 mi; Fig. 6.5). Numerous activities were ongoing at and near Northstar on 1 Sept. 2000, including a continuously operating tug, and several vessels were operating nearby on 12 Aug. 2001 as noted above. Figure 6.6 shows sound pressure density spectra for recordings with and without production (but with drilling in both cases). Production did not take place on 12 Sept.; the recording taken on that day at 0.5 n.mi. (gray line in Fig. 6.6) is compared to that taken from the same station on 6 Sept., during Chapter 6: Sound Measurements, 2002 Open-Water Season 6-15

30 Aug. 00 1 Sept. 00 17 Sept. 00 10 Aug. 01 12 Aug. 01 5 Sept. 02 6 Sept. 02 12 Sept. 02

0.54 n.mi. 5.4 n.mi. 150

140 Pa) µ 130

120

110

100

Received Level (dB re 1 90

80 100 1,000 10,000 100,000 Distance from Northstar (m)

FIGURE 6.5. Mean broadband (10–10,000 Hz) levels of underwater sound as a function of distance from Northstar, for eight comparable sets of open-water recordings north of Northstar over three summer seasons: 2000 (green triangles), 2001 (red squares) and 2002 (black circles). 10 km = 5.4 n.mi. or 6.2 mi. production (black line). On 6 Sept. (with production), many tones were prevalent between 12.5 and 16 kHz; these tones were present to some extent in every recording from 5 and 6 Sept., but in no recordings from 12 Sept. These tones may therefore be linked to the gas compressors or some other equipment that is used during production. Both wind speed and sea state were lower on 6 Sept. during drilling (Table 6.2). Crew Boat Sounds.—The crew boat Arctic Express ferried passengers from West Dock on the mainland to Northstar Island and ran eight daily round trips from 05:15 until 00:55 local time every day. Arctic Express was an 18.7 m (61.5 ft) vessel with dual 200 hp Detroit diesel engines equipped with 4-bladed propellers, and traveled at an average speed of ~12 knots. Figure 2.5 (in Chapter 2) shows a similar vessel. Several recordings were made while the crew boat was either at Northstar (loading and unloading passengers) or going to or from the island at a known distance from the recording vessel. Broadband levels (10–10,000 Hz) obtained during these recordings are shown in Figure 6.7. Figure 6.7A shows consecutive samples taken during 9 min of recording, spanning the arrival, docking (with loading / unloading of passengers), and departure of the crew boat. The recording vessel was ~1 n.mi. north of the island (1.85 km or 1.15 mi); the crew boat approached the island from the southeast, went south of the island (i.e., behind it in relation to the recording vessel; see Fig. 6.3), emerged from behind the island, and docked at the island’s western dock ~1980 m from the recording vessel. Upon departure, it followed the same route. The highest received broadband SPL (123 dB re 1 µPa) was recorded when the vessel left the Northstar dock and got 6-16 Monitoring at Northstar, Open-Water Season

FIGURE 6.6. Sound pressure density spectra with and without production, September 2002: 4–20,000 Hz spectra from 0.5 n.mi. recording station. On 6 Sept. (black line) production was taking place at Northstar whereas on 12 Sept. (gray line) it was not. Drilling was taking place during both recordings. underway on the return trip to West Dock. During those first few seconds of acceleration, propellers usually cavitate severely. Sound pressure levels were similar (121 dB re 1 µPa, see Fig. 6.7A) just before the crew boat passed behind Northstar upon arrival, when it was still cruising at full speed. Figure 6.7B shows broadband levels from three different distances while the crew boat was cruising. The bow aspect data (i.e., approaching crew boat) were recorded at a closer distance than were the stern aspect data. The highest SPL (128 dB re 1 µPa) was recorded as the crew boat was 3740 m away and receding. Note that received levels ~1 n.mi. offshore of the island on 6 Sept. when the crew boat was underway nearby (Fig. 6.7A) were notably higher than those near Northstar at other times on that date (Fig. 6.4). Figure 6.8 presents sound pressure density spectra for recordings with and without the crew boat. Both spectra were recorded at the same station (1 n.mi.), on 5 Sept. without the crew boat present at or near Northstar (gray line in Fig. 6.8) and on 6 Sept. while the crew boat was leaving the island (black line). Sea state and wind speed were nearly identical during the two recordings (Table 6.2). The presence of the crew boat led to two peaks, at 26 and 52–53 Hz. These peaks were present in all recordings with the crew boat within sight of the recording vessel, but disappeared when the crew boat passed behind Northstar on its way to the island’s western dock. We do not know what caused the heightened levels below 20 Hz on 5 Sept. Chapter 6: Sound Measurements, 2002 Open-Water Season 6-17

FIGURE 6.7. Broadband (10-10,000 Hz) levels of underwater sound, as recorded about 1 n.mi. (1.82 km or 1.13 mi) north of Northstar, during a crew change by the crew boat Arctic Express on 6 Sept. 2002. (A) Entire sequence, which lasted 9 minutes. Each data point represents the analysis of one 8.5-sec sample; the samples are shown consecutively in time but the time scale is discontinuous. (B) Received levels while the crew boat was cruising, seemingly at full speed, as received at three different distances.

Selected one-third octave band SPLs are shown in Figure 6.9 for recordings made on 6 Sept. Three stations (¼, 10 and 20 n.mi., or 0.5, 18.5 and 37 km) were sampled while the crew boat was neither at the island nor on its way to or from the island; corresponding one-third octave band received levels are shown in Figure 6.9A. Another four stations (½, 1, 2, and 10 n.mi., or 0.9–18 km) were sampled while the crew boat was at Northstar and visible from the recording vessel, or should have been at the island or on its way according to the crew boat schedule (Fig. 6.9B). In the absence of the crew boat or other vessels, received levels for all of the bands plotted were quite low (≤90 dB re 1 µPa), even at the closest station (Fig. 6.9A). With vessels present, sound levels in most bands were elevated at distances out to 2 or 3 km (1–1.6 n.mi. or 1.2–1.9 mi). Levels were not conspicuously elevated at greater distances (Fig. 6.9B). At the 5 n.mi. (~9 km) station, vessel sound could be heard clearly on the recording, but according to the schedule the crew boat was not present. This unseen vessel, at an unknown distance from the 5 n.mi. station, created elevated levels at frequencies from 100 to 1000 Hz, as exemplified by the increase in received levels for the one-third octave bands centered at 200 and 630 Hz (Fig. 6.9B). 6-18 Monitoring at Northstar, Open-Water Season

FIGURE 6.8. Sound pressure density spectra with and without the presence of the crew boat at the island, September 2002: 4–20,000 Hz spectra from 1 n.mi. recording station. On 6 Sept. (black line) the crew boat was at the island whereas on 5 Sept. (gray line) it was not. Drilling and production were taking place during both recordings.

Airborne Sounds A-weighted broadband SPLs are shown in Figure 6.10 for in-air measurements made on 5, 6 and 12 Sept. north of Northstar. The crew boat was at the island during the recording at the 1 n.mi. station on 6 Sept.; this station is shown with a different symbol (diamond). As with the underwater data, the recordings made on 6 Sept. were considered to be of the highest quality because of the low sea state and low wind speed throughout most of the day (see Table 6.2). On 12 Sept., sea state and wind were higher at the farther stations, and this was probably responsible for the increasing received levels with increasing distance from Northstar. The highest mean broadband levels recorded on all three days were at the stations closest to Northstar: 52–56 dBA re 20 µPa at 500–590 m. On 12 Sept. the mean broadband level at the farthest station was also high, but this was attributed to increasing sea state, not industrial noise. On 6 and 12 Sept., airborne sound levels diminished to background values by about 2–3 km (1–1.6 n.mi. or 1.2–1.9 mi) from the island. Background levels on 6 Sept., the calmer of the two days with recordings out to 20 n.mi., were 37– 40 dBA re 20 µPa. On 6 Sept., notwithstanding the “flattening out” of the broadband received levels beyond ~3.5 km (1.9 n.mi. or 7.2 mi), the acoustic personnel and ACS boat crew could hear the island faintly during the second recording at the 20 n.mi. station as well as at 10 n.mi. and closer stations. The sound propagation model represented by equation (1) was fitted by least-squares to mean broadband (10–10,000 Hz) A-weighted levels of in-air sound for all three days (Fig. 6.10). On 5 and 12 Sept., data from the first three stations, for which broadband levels diminished with increasing distance, were used in the model. On 6 Sept., the crew boat was at Northstar during the recording at the third station Chapter 6: Sound Measurements, 2002 Open-Water Season 6-19

FIGURE 6.9. Mean one-third octave band levels of underwater sound vs. range for 7 selected frequencies, 6 Sept. 2002. (A) Recordings made without crew boat or other vessel near Northstar (3 stations). (B) Recordings made while crew boat or other vessel was near the island, or was expected to be at the island according to the crew boat schedule (5 stations). (The schedule was used to classify stations too distant from Northstar to enable visual confirmation of the crew boat’s presence.) The frequency value indicated for each band corresponds to the center value for that band. 10 km = 5.4 n.mi. or 6.2 mi.

(1 n.mi. or 1.8 km), so only the values from the first (¼ n.mi. = 0.5 km), second (½ n.mi. = 0.95 km) and fourth (2 n.mi. = ~3.5 km) stations were used in the model. The resulting regression lines and equations are shown in Figure 6.10. The spreading loss terms obtained were similar on the three days: 21.6, 18.1 and 17.1 dB/tenfold change in distance. Figure 6.11 shows how A-weighted broadband levels of airborne sound measured on 6 Sept. 2002 compare to similar data collected in September 2000 and August 2001. As for the underwater data, in-air SPLs were lower in 2002 than during previous years. Figure 6.12 shows sound pressure density spectra for recordings at the 0.5 n.mi. station with production (6 Sept.) and without production (12 Sept.), but with drilling in both cases. Figure 6.6 is the equivalent Figure for underwater sound. During production, received levels were consistently 15–20 dB higher at all frequencies, even though wind speed was slightly lower (1.7 vs. 5.8 mph on average). To the extent that there is any wind contribution to the spectra shown in Figure 6.12, this would be greater without production. Thus, under comparable wind and sea conditions, the elevation of levels with production may be slightly greater than evident from Figure 6.12. 6-20 Monitoring at Northstar, Open-Water Season

5 Sept. 6 Sept. 12 Sept.

0.54 n.mi. 5.4 n.mi. 60

RL = 103.7 - 18.1 log(R )

Pa) r = -0.99 µ

RL = 99.9 - 17.1 log(R ) r = - 0.99 40

RL = 110.7 - 21.6 log(R ) r = - 0.99 Received Level (dBA re 20

30 100 1,000 10,000 100,000 Distance from Northstar (m)

FIGURE 6.10. Mean A-weighted, broadband (10–10,000 Hz) levels of airborne sound as a function of distance from Northstar, 5, 6 and 12 Sept. 2002. The logarithmic regression models shown were fitted to the data obtained within 4 km of the island (see text for details). R is in meters. The point shown as a diamond on 6 Sept. was obviously influenced by the proximity of the crew boat and was not included in the regression model. 10 km = 5.4 n.mi. or 6.2 mi.

The crew boat did not affect narrowband levels in air as much as it did underwater. In-air data are shown in Figure 6.13, which presents sound pressure density spectra for times with (6 Sept.) and without (5 Sept.) the crew boat, as recorded at the 1 n.mi. station. Figure 6.8 is the equivalent Figure for underwater sound. For most frequencies, the narrowband spectra in air do not differ by more than 10–15 dB. The 80 Hz tone visible in Figure 6.13 was present in all samples on 6 Sept. but was absent on 5 Sept., regardless of the presence or absence of the crew boat. It is not known where this tone originated. Figure 6.14 shows levels of airborne sound in six selected one-third octave bands, as measured at all seven recording stations on 6 Sept. The crew boat’s presence at Northstar during recording at the 1 n.mi. station (1.8 km or 1.1 mi) resulted in a slightly higher broadband level for the one-third octave band centered at 25 Hz. For other bands, there was little evidence of elevated levels at that distance. Summary of Boat-Based Recordings in Open Water Greeneridge Sciences made open-water recordings at seven nominal distances ¼ to 20 n.mi. north of Northstar Island on 5, 6, and 12 Sept. 2002. Oil production took place on 5 and 6 Sept. but not on 12 Sept., while drilling took place during all three days of recording. Recordings on 6 Sept. benefited from unusually low winds and sea state. Underwater Sounds.— Broadband levels obtained on 6 Sept. were the lowest of any measurements made during the summers 2000, 2001, and 2002. The highest mean broadband level reached 116 dB re 1 µPa and was recorded ~600 m from the island during drilling but without production. The lowest level, Chapter 6: Sound Measurements, 2002 Open-Water Season 6-21

1 Sept. 00 10 Aug. 01 12 Aug. 01 6 Sept. 02

0.54 n.mi. 5.4 n.mi. 70

65 Pa)

µ 60

40 Received Level (dBA re 20 Received Level (dBA 35

30 100 1,000 10,000 100,000 Distance from Northstar (m)

FIGURE 6.11. Mean A-weighted, broadband (20–10,000 Hz) levels of airborne sound as a function of distance from Northstar, for four comparable sets of open-water recordings north of Northstar over three summer seasons: 2000 (green triangles), 2001 (red squares), and 2002 (black circles). For 2002, the only recordings shown are those from 6 Sept., the day with the lowest wind speed. 10 km = 5.4 n.mi. or 6.2 mi.

No production (12 Sept.) Production (6 Sept.)

140

120 /Hz) 2 Pa)

µ 100

20 Received Level (dB re (20

0 0 5 10 15 20 Frequency (kHz) FIGURE 6.12. Sound pressure density spectra with and without production, September 2002: 10–18,000 Hz spectra from 0.5 n.mi. recording station. On 6 Sept. (black line) production was taking place at Northstar whereas on 12 Sept. (gray line) it was not. Drilling was taking place during both recordings. Wind speed was slightly greater without production, so the effect of production on sound levels could exceed that shown here (see text). 6-22 Monitoring at Northstar, Open-Water Season

With Crew Boat Without Crew Boat 100

90 /Hz) 2 80 Pa)

µ 70

Received Level (dB re (20 10

0 10 100 1,000 10,000 100,000 Frequency (kHz)

FIGURE 6.13. Sound pressure density spectra with and without the presence of the crew boat at the island, September 2002: 10–18,000 Hz spectra from 1 n.mi. recording station. On 6 Sept. (black line) the crew boat was at the island whereas on 5 Sept. (gray line) it was not. Drilling and production were taking place during both recordings.

25 63 200 630 2000 6300 Hz

100 0.54 n.mi. 5.4 n.mi. Pa) µ

40 One-third Octave Band SPL (dB re 20 20 100 1,000 10,000 100,000 Distance from Northstar (m)

FIGURE 6.14. One-third octave band levels of airborne sound vs. range for 6 selected frequencies, 6 Sept. 2002. The frequency value indicated for each band corresponds to the center value for that band. 10 km = 5.4 n.mi. or 6.2 mi. Chapter 6: Sound Measurements, 2002 Open-Water Season 6-23

87 dB re 1 µPa at 10 n.mi., corresponds to a background level. Propagation loss was not determined, as sound levels reached background values too close to the source (3.5 km on 6 Sept.). The crew boat was the main source of vessel sound at the island in September 2002. Broadband levels were obtained during the arrival, transfer of passengers, and departure of the crew boat. The highest received broadband level (123 dB re 1 µPa; range 1 n.mi. or 1.8 km) was recorded as the vessel left the Northstar dock and got underway. Levels were 10–15 dB lower while the crew boat backed against the dock during the transfer of passengers. A comparison of narrowband spectra obtained with and without the crew boat showed it to be the source of distinct low-frequency tones (26 and 52–53 Hz). The crew boat is also a source of broadband sound, as shown both by narrowband and one-third octave data. Airborne Sounds.— As seen for the underwater data, the recordings made on 6 Sept. yielded the lowest broadband levels of all measurements made during the open-water seasons of 2000, 2001 and 2002. The highest mean broadband levels, 52–56 dBA re 20 µPa, were recorded at the stations closest to Northstar (500–600 m). Background levels of 37–40 dBA re 20 µPa were reached at distances of 2–3.5 km and beyond. The sound propagation model was fitted by least-squares to mean broadband levels for all three days; spreading loss terms obtained were 21.6, 18.1 and 17.1 dB/tenfold change in distance. Narrowband spectra were 15 to 20 dB higher during production than without, at all frequencies, whereas narrowband spectra obtained with and without the presence of the crew boat were quite similar. Near-Island Cabled Hydrophone Recordings The cabled hydrophones provided an uninterrupted record of the sounds from the island and the vessels servicing the island for 31 days, from 31 Aug. until 1 Oct. On 23 Sept., water current or surge associated with a storm likely affected the positioning of C.H. #1 (probably by tipping it over), rendering the data it collected useless after that date. However, C.H. #2 continued working normally. C.H. #2 was deployed about 420 m north of the island, within ~9 m of its position in 2001 (Fig. 6.2), and the data it collected are presented in Figure 6.15. The storm that caused C.H. #1 to tip over on 23 Sept. is visible in Figure 6.15B as a large peak in wind speed. Percentile levels of broadband (10–1000 Hz) sound are presented in Table 6.3 for C.H. #2 in 2001 and 2002. In 2002, we chose to limit the comparison to the period ending with the arrival of the storm (on 23 Sept. 2002), as the associated bowhead monitoring effort effectively terminated on that date. No comparable storm occurred during the study period in 2001.

TABLE 6.3. Percentile levels of broadband (10–1000 Hz) underwater sound recorded near Northstar Island by C.H. #2 in 2002 (31 Aug.–23 Sept.) and 2001 (1–21 Sept.). All levels are in dB re 1 µPa.

2002 2001

Minimum 89.8 81.6 5th percentile 94.8 87.8 50th percentile 103.0 102.6 95th percentile 117.2 122.8 Maximum 135.3 140.9 6-24 Monitoring at Northstar, Open-Water Season

FIGURE 6.15. (A) Broadband (10-1000 Hz) sound pressure levels observed underwater from 31 Aug. to 1 Oct. 2002, as recorded by cabled hydrophone (C.H.) #2 located ~420 m north of Northstar. (B) Mean hourly wind speed during the same period, as recorded by a weather station located at Northstar on the roof of the permanent living quarters (PLQ). (C) Broadband (10-1000 Hz) sound pressure levels as observed in (A), but removing eight daily 35-min windows that are meant to include the arrivals and departures of the crew boat at Northstar, according to the crew boat schedule (see text). Chapter 6: Sound Measurements, 2002 Open-Water Season 6-25

Vessel Sounds Vessels operating near Northstar were responsible for many of the strongest sounds received by the cabled hydrophone. During the worst part of the storm (late on the 22nd until early on the 24th) neither the island barge nor the crew boat, nor any other vessels, were operating. This accounts for the complete lack of vertical spikes in SPL during that period (Fig. 6.15A). These spikes, as evident on other days, were caused by vessels around the island, in particular the crew boat that ran on a regular schedule and therefore created a regular daily spike pattern on the sound pressure time series (see Fig. 6.15A). The Northstar barge (200 series) traveled between West Dock and Northstar on a (quasi) daily basis (18 round trips during the 1–23 Sept. period in 2002), bringing any goods or construction materials needed on the island. This barge was pushed by the tug Sag River, a River-class tug powered by 3 diesel engines (1095 total hp). To illustrate the connection between “spikes” and crew boat runs, we plotted broadband cabled hydrophone data for 13 and 14 Sept. in Figure 6.16. (These two days had high rates of detection for bowhead whale calls, see Chapter 7.) The eight daily scheduled crew boat runs are shown as shaded slices, 35 min in duration. These time slices • began 10 min before the scheduled arrival time, • included the 15 min during which the crew boat remained at the island, maintaining itself at one of the docks by backing against it, and • ended 10 min after the scheduled departure.

Daily barge runs are plotted as two darker-gray bands; these correspond to the 15 min periods preceding the arrival and following the departure of the barge, respectively (we had the exact times of departure from and arrival at West Dock and Northstar for each round trip). Unlike the crew boat the barge did not run on a regular schedule; the times of arrival and departure from Northstar, as well as the length of stay at the island, varied on a daily basis. Island records indicate that, on each of the two days plotted in Figure 6.16, the barge made one round trip and the crew boat made seven round trips to the island (instead of the scheduled eight). From the broadband levels shown we can infer that the midnight trip was skipped on both days. On 14 Sept., the match between spikes in broadband levels and the crew boat and barge schedules is very good. On 13 Sept., five spikes match well with the crew boat and/or the barge schedule. The ACS vessel Harrison Bay stopped at Northstar between 16:00 and 17:00 hrs with the acoustics crew, and likely produced the pair of spikes occurring at that time. The weather was windy (see Fig. 6.15B) and it is conceivable that the crew boat was delayed, accounting for the slight mismatch at ~20:00 hrs. Finally, since the crew boat did make seven runs that day, the remaining two spikes at ~13:00 hrs and 22:00 hrs were likely rescheduled crew boat trips, accounting for the last two “unused” crew boat windows shown in Figure 6.16. The data from the two days shown in detail in Figure 6.16 suggest that, during September 2002, most of the sharp “spikes” in sound levels as received near Northstar were attributable to nearby vessels. Figure 6.15C shows the broadband time series depicted in Figure 6.15A, but with the crew boat’s contribution largely removed by deleting a 35 min time slice (see above) for each scheduled (but not necessarily real) visit by the crew boat to Northstar (eight per day). Numerous sound spikes were removed through this procedure, but a number remain (Fig. 6.15C). Some of these likely belong to the crew boat and were not removed because it was running off schedule or on supplemental trips. Delayed runs were not uncommon, particularly when sea state increased (C. Ledgerwood, AIC, pers. comm.). We know the daily number of crew boat runs, but we do not have records of the times of the cancelled or off- schedule runs. Thus, we do not have an objective basis for determining which of the remaining spikes are attributable to the crew boat. 6-26 Monitoring at Northstar, Open-Water Season

FIGURE 6.16. Broadband (10–1,000 Hz) levels as recorded by C.H. #2, ~420 m north of Northstar, during 13 and 14 Sept. 2002. Light-gray shading shows 35-min periods corresponding to the eight daily scheduled crew boat trips (see text). Dark-gray shading with diamond symbols shows two 15-min periods per day corresponding to the arrival and departure of the near-daily barge and associated tug. See text for details about the match (or lack of match) between crew boat trips and “spikes” of high-level sound.

To gauge the contribution of the crew boat to daily sound levels at Northstar we selected seven days during which every spike could be accounted for by our records of the crew boat, the daily barge, and our own trips with ACS vessels. These seven days were September 10, 11, 14, 15, 19, 22 and 29. Figure 6.17 shows mean percentile levels of broadband (10–1000 Hz) underwater sound for those seven days. Broadband levels were calculated for the entire dataset (gray bars) as well as for two subsets, the boat spikes Chapter 6: Sound Measurements, 2002 Open-Water Season 6-27

All Spikes Only Spikes Removed

140

130 Pa) µ 120

110

100 Broadband Level (dB re 1 90

80 Min 5th % 50th % 95th % Max Percentile Category

FIGURE 6.17. Mean percentile levels of broadband (10–1000 Hz) underwater sound for seven days in September 2002 (10, 11, 14, 15, 19, 22 and 29 Sept.). Values are shown for the entire dataset (gray bars), the boat spikes only (black bars), or the dataset with the boat spikes removed (white bars). Error bars are one standard deviation. only (black bars) and the dataset with the spikes removed (white bars). A comparison of the gray versus white bars in Figure 6.17 shows that the crew boat had notable effects only on the maximum and 95th percentiles, by 10–20 dB. This is also illustrated in Figure 6.18, showing “statistical” narrowband percentile levels levels versus frequency on 10 Sept. for the entire dataset (Fig. 6.18A) and the dataset with the boat spikes removed (Fig. 6.18B). Thus, the upper curves show, for each frequency, the highest levels observed in all of the daily 330 one-minute samples analyzed. The median levels (50th percentiles) indicate that half the values were smaller and half were larger at every frequency. The presence of tones in the narrowband spectra is significant. For example, tones at 26 and 52 Hz stand out in the maximum curve for the entire dataset (Fig. 6.18A). They can be attributed to the crew boat, in part because these tones have already been identified as a “signature” for that vessel (see Fig. 6.8), but also because they are found in the curve of maximum values, which is to a large extent made up of values generated by the crew boat. Overall Underwater Sound The minimum broadband level in Figure 6.15A is about 90 dB re 1 µPa, compared to 82 dB for the same period in 2001 (Table 6.3 and Fig. 8.11 in Greene et al. 2002). The 5th percentile broadband value showed the same trend, i.e., it was about 7 dB higher in 2002 compared to 2001 (~95 vs. 88 dB, respectively). In contrast, the 95th percentile broadband value was about 6 dB lower in 2002 than in 2001 (117 vs. 123 dB re 1 µPa). Median values in the two years were very similar, at ~103 dB re 1 µPa both years (see Table 6.3). These differences cannot be due to the instrumentation (i.e., differing noise floors) since C.H. #2 was the same exact instrument (same sensor and same electronics) during both years. 6-28 Monitoring at Northstar, Open-Water Season

FIGURE 6.18. Percentile sound spectral densities of underwater sound vs. frequency as recorded by C.H. #2 on 10 Sept. 2002. (A) Entire dataset. (B) Dataset with boat spikes removed. In both plots, the five curves show, for each frequency, the minimum, the 5th, 50th and 95th percentiles, and the maximum. Each statistical spectrum is based on 330 one-min measurements. Chapter 6: Sound Measurements, 2002 Open-Water Season 6-29

Figure 6.19 presents “statistical” (A) narrowband and (B) one-third octave band levels by frequency over the period 31 Aug.–23 Sept. (7551 samples analyzed). The presence of tones in the narrowband spectra indicates the regular occurrence of certain tones and harmonic families from machinery. If present most of the time, the tones tend to be more prominent when background levels are lower, e.g., they stand out more in the 5th than in the 95th percentile spectrum. For example, tones at 90 and 519 Hz stand out in the minimum and 5th percentile narrowband data, but are not detectable above that. Certain other tones, e.g. at ~20, 30, 60 and 80 Hz, were sufficiently consistent (and strong) to be evident in the 50th and 95th percentile spectra as well. The 30 and 60 Hz tones were probably associated with generation of 60-cycle power. These same four narrowband components were also evident in corresponding spectra for 31 July–22 Aug. 2001 (Fig. 7.19 in Blackwell and Greene 2002). During the summer of 2001, there were prominent tones at 147–150 Hz, but there was no indication of similar tones in September 2002. The 50th percentile one-third octave levels (Fig. 6.19B) were within a ±3 dB range (i.e., relatively consistent) for all center frequencies from 31.5 Hz to >800 Hz, with slight peaks near 40 Hz and 125–250 Hz. These results were similar to those during the preceding year. From 28 Aug. to 3 Oct. in 2001, the median one-third octave spectrum was within a ±3 dB range from 40 Hz to >800 Hz, with slight peaks at 63–80 Hz and at 125–200 Hz (Fig. 8.13 in Greene et al. 2002). The data collected by the C.H. close to Northstar give us an opportunity for comparison with the open-water recordings obtained simultaneously at different distances from the island. Figure 6.20 compares broadband levels recorded on 6 Sept. by C.H. #2 420 m from Northstar (line) with those obtained during the boat-based measurements 0.5–37 km from Northstar (symbols). The two types of measurements refer to the same frequency range (10–1000 Hz), but the analysis durations were different: 8.5 s for the samples from the boat-based recordings (with two analyzed samples being separated by as little as a few seconds) vs. 1 min every 4.37 min for the samples from the C.H. The comparison reveals the following: • Broadband levels obtained more than 15 km from Northstar (open symbols) were in general lower than levels recorded by the C.H., by as much as 40 dB. Also, note that the recordings obtained at 37 km at ~17:05 hrs are lower than those obtained at ~11:45 hrs, because sea state decreased in late afternoon. The same is true for the two recordings obtained at 18.5 km. • Boat-based broadband levels at 9.3 km were higher than expected, as they were identical to the values recorded by the C.H. at the same times. However, boat-based levels recorded ½–1 hr later at range 3.4 km were lower than the C.H. values for those times, as expected. • The recordings at 1.8 km were obtained while the crew boat arrived, docked and departed. Maxi- mum levels at that station were comparable to those at the C.H., but other boat-based measurements were lower than corresponding C.H. values. The discrepancies are probably caused, at least partly, by the aforementioned differences in averaging time. • The highest values obtained by the open-water recordings were at 0.9 km, while the crew boat was cruising back to West Dock, right after 15:00 hrs. At this time, the C.H. recording shows a “dip” in broadband levels. This dip could be the result of the island serving as an acoustic shield between the crew boat and the C.H. (see Fig. 6.7A and Fig. 6.3). In contrast, the island was not interposed between the crew boat and the boat-based hydrophone. The above comparisons are based on the same bandwidth for both the boat-based recordings and the C.H. (10–1000 Hz). However, the boat-based recordings also provided data for the 10–10,000 Hz bandwidth. Broadband levels for the 10–10,000 Hz band were essentially identical (on average higher by 1.3 dB) to those for 10–1000 Hz. This indicates that, on average, frequencies between 1 and 10 kHz contributed very little to the overall sound levels. 6-30 Monitoring at Northstar, Open-Water Season

FIGURE 6.19. Percentile sound spectral densities of underwater sound vs. frequency as observed ~420 m north of Northstar via C.H. #2, 31 Aug.–23 Sept. 2002. (A) Narrowband data. (B) One-third octave band data. In both plots, the five curves show, for each frequency, the minimum, the 5th, 50th and 95th percentiles, and the maximum. Each statistical spectrum is based on 7551 one-min measurements. Chapter 6: Sound Measurements, 2002 Open-Water Season 6-31

FIGURE 6.20. Broadband (10–10,000 Hz) levels during 7 hours on 6 Sept. 2002, as recorded by C.H. #2 located ~420 m north of Northstar (line), and simultaneously at stations 0.5–37 km from Northstar (symbols). Symbol color becomes lighter with increasing distance from Northstar.

Calculation of the ISI in 2002 The results from all five methods, presented below, were obtained using cabled hydrophone recordings from 31 Aug. to 23 Sept., when a storm overturned the DASARs and rendered the subsequent data unusable. • In the first method, the choice of frequencies to include in the ISI was guided by the distribution of sound in the percentile narrowband spectral densities shown in Figure 6.19A. It appears from that Figure that, for most of the time, much of the sound from the island was in the general range 30–80 Hz. • The second method was similar to the first, but in this case the distribution of one-third octave band levels (instead of narrowband levels) was examined. For the minimum and 5th percentiles in Figure 6.19B, peaks occur for the 31.5–63 Hz bands, but the median is broadly higher in the 31.5–80 Hz bands. Levels are also higher in some bands at and above 125 Hz, but correlations with wind speed at these higher frequencies (see below) suggest that those bands are influenced primarily by wind-generated sounds. Thus, the information in Figure 6.19B supports the choice of one-third octave bands centered at 31.5–80 Hz to represent the important Northstar-related components of the near island sounds. • A third perspective came from examining the relationship between one-third octave band levels and wind speed. We expected that the one-third octave bands most influenced by island sounds would show the least dependence on wind speed; therefore the values of B and r2 in Eq. (2) were expected to be lowest for those bands least related to the wind speed and most dependent on industrial sounds. Table 6.4 shows that the values for B were <0.5 for all bands centered at 10 through 80 Hz, and >0.5 for all bands centered at 100–800 Hz. This is consistent with the premise that, of the sound received 420 m from Northstar in 2002, more of the sound at frequencies ≤80 Hz than of that at frequencies ≥100 Hz was attributable to industrial activities. 6-32 Monitoring at Northstar, Open-Water Season

TABLE 6.4. Parameter values for regressions of band levels vs. wind speed, 31 Aug.–23 Sept. 2002. Columns show center frequency of one-third octave band, fitted values of A and B in Eq. (2), P-value for difference between B and zero, and coefficient of determination r2. See text for further details.

Center Freq. (Hz) ABP-value r2 10 89.782 0.169 0.0263 0.009 12.5 77.110 0.300 <0.0001 0.031 16 72.983 0.358 <0.0001 0.090 20 80.201 0.212 0.0008 0.020 25 82.734 0.076 0.1078 0.005 31.5 90.009 -0.002 0.9781 0.000 40 93.003 0.012 0.8572 0.000 50 90.910 0.228 0.0028 0.016 63 90.255 0.275 0.0001 0.027 80 87.595 0.434 <0.0001 0.040 100 82.418 0.644 <0.0001 0.072 125 86.522 0.585 <0.0001 0.047 160 86.901 0.630 <0.0001 0.059 200 85.600 0.668 <0.0001 0.062 250 85.481 0.710 <0.0001 0.074 315 82.705 0.883 <0.0001 0.104 400 86.234 0.583 <0.0001 0.050 500 85.091 0.703 <0.0001 0.088 630 84.924 0.618 <0.0001 0.078 800 85.012 0.543 <0.0001 0.071

Sounds at frequencies below ~30 Hz would not be expected to propagate well in the shallow water near and north of Northstar Island. The “low-frequency cut-off” for propagation in shallow water is expected on theoretical grounds (Officer 1958). Also, it has been demonstrated empirically during studies of the propagation of airgun sound pulses in waters near Northstar (Burgess and Greene 2000). Although airgun pulses contain considerable energy at frequencies as low as 10–15 Hz, that energy is rapidly attenuated as the pulses propagate in shallow-water areas. In the general Northstar area, at distances beyond a few hundred meters, most of the remaining energy is at frequencies above 30 Hz. Thus, components of Northstar sound below 30 Hz would not be expected to be very strong (if detectable at all) in the southern part of the bowhead migration corridor, which is generally a few kilometers offshore of Northstar. Thus, the regression results, combined with existing information about the low-frequency cutoff, support the choice of the one-third octave bands centered at 31.5– 80 Hz for the ISI in 2002. • The fourth method was the cross-correlation of the measured levels in the various one-third octave bands (Table 6.5). We expected the levels in the bands dominated by island sound to correlate well with one another, but not to correlate well with the levels in bands dominated by natural sea noise. Table 6.5 shows that levels in any one one-third octave band tended to be most closely correlated with those in adjacent bands, or in other bands at relatively similar frequencies. This pattern applied in both the 31.5–80 Hz and the 100–800 Hz range, although it was stronger in the latter range. Thus, this comparison provided results that were equivocal with regard to the choice of one-third octave bands to Chapter 6: Sound Measurements, 2002 Open-Water Season 6-33 define the ISI. This may result from the fact that the sources of island sound vary with time, and are not all active at the same times. • The percentile levels for periods of calm and low winds (by one-third octave band) may also be a useful indicator of island sound, providing a fifth basis for selecting the ISI. Background sea noise from wind and waves should be minimum when the winds are low, permitting the island-generated sounds to dominate. (However, when the wind speed drops after having been high for a period of hours, the wave noise may take time to diminish correspondingly.) Figure 6.21 shows the distribution of one- third octave band levels for wind speeds <5 knots (2.6 m/s). It shows that, for median sound levels, the bands from 31.5 to 80 Hz are higher than the levels for the bands <31.5 Hz and for the band at 100 Hz. However, the median levels in the bands centered at 125–250 and 400 Hz were similar to those at 31.5 to 80 Hz, just over 90 dB re 1 µPa. During quieter times, as evident from the 5th percentile and minimum levels, the highest levels were for the bands from 31.5 to 63 Hz. During noisier times (95th percentile), the levels for the bands from 40 Hz upward are broadly higher than the levels in bands <40 Hz. Calm wind does not necessarily translate into low sea noise. However, during times with near-calm wind and low noise, as reflected in the minimum and 5th-percentile curves for times with wind speed <5 knots, we can be confident that the data show any significant sounds coming from the island. That is, we can confidently conclude that island sounds contributed importantly to the total measured sounds in the 31.5, 40, 50 and 63 Hz bands. At somewhat noisier times (i.e., times with median and higher levels), the 80 Hz band also contains considerable sound energy, confirming our initial choice of the one-third octave bands centered at 31.5–80 Hz to define the ISI. Summary of Cabled Hydrophone Recordings Thirty-one days of cabled hydrophone recordings from 31 Aug. to 1 Oct. 2002 provided continuous information on island sounds as received 420 m seaward of Northstar Island. Minimum broadband (10– 1000 Hz) levels of about 90 dB re 1 µPa were ~10 dB higher than in the corresponding season of 2001, and maximum levels were ~6 dB lower; median values were comparable to those in 2001. Many “spikes” in broadband levels could be attributed to the crew boat. Removing the crew boat from the analysis of broadband levels on seven specific days for which good records on boat traffic were available decreased the values for the maximum and 95th percentile by 10–20 dB. “Statistical” narrowband spectra, calculated over the entire sampling period, were similar to those obtained in 2001. Broadband levels obtained simultaneously by the cabled hydrophone and during the open-water recordings at varying distances were compared; values obtained during the open-water recordings were usually lower than values obtained by the cabled hydrophone, except close to the island. Five different methods were used to help determine which frequencies to include in the Industrial Sound Index (ISI) for 2002: (1) inspection of the distribution of sound in the percentile narrowband spectral densities; (2) inspection of the distribution of sound in the percentile one-third octave band levels; (3) examination of the relationship between one-third octave band levels and wind speed; (4) cross-correlation of the measured levels in the various one-third octave bands; and (5) inspection of the distribution of sound in the percentile one-third octave band levels during periods when the wind speed was <5 knots. Method (4) yielded equivocal results, but the other four methods supported the choice, with some variation, of the one-third octave bands centered at 31.5–80 Hz to define the ISI in 2002. 6-34 Monitoring at Northstar, Open-Water Season 1 g 1 0.988 1 0.97 tes correlation tes 0.949 1 0.977 0.961 0.947 1 0.924 0.924 0.917 0.899 1 0.938 0.878 0.899 0.878 0.857 1 0.967 0.885 0.838 0.861 0.847 0.832 1 0.965 0.905 0.801 1 0.973 0.915 0.847 0.753 1 0.941 0.937 0.915 0.870 0.816 ents between 0.8 and 0.9, and gray type deno 0.8 and ents between 1 0.942 0.964 0.938 0.891 0.834 0.714 0.7590.731 0.709 0.746 0.7790.730 0.733 0.776 0.7860.718 0.726 0.766 0.781 0.716 0.754 420 m from Northstar, 31 Aug.–23 Sept. 2002. Dark gray shadin Dark gray Sept. 2002. 31 Aug.–23 m from Northstar, 420 1 0.639 0.753 0.886 0.883 0.859 0.832 1 0.677 0.771 0.662 0.726 0.835 0.801 0.718 0.642 0.731 0.696 0.742 0.656 0.719 0.808 0.815 0.714 1 0.621 0.688 1 026 0.244 0.351 0.506027 0.247 0.561 0.389 0.523031 0.263 0.575 0.398 0.538033 0.273 0.584 0.400 0.514 0.581 0.542 1 0.076 0.24 0.397 0.569 9, light gray coeffici 9, light gray shading indicates 1 160 0.127 0.379 0.573 124 0.121 0.329 0.530 0.291 0.291 0.537 0.246 0.248 0.499 0.279 0.218 0.529 0.254 0.201 0.467 1 0.616 1 -0.04 0.000 0.331-0.05 0. -0.02 0.335 0. 0.472 0.331 0.403 0.527 0.187 0.287 0.506 0.573 0.104 0.167 0.4040.042 0.586 0.097 0.545 0.4190.066 0.427 0.174 0.539 0.4990.069 0.427 0.112 0.441 0.394 0.038 0.102 0.42 0.054 0.106 0.362 0.015 0.078 0.374 0.005 0.029 0.3090.041 0.043 0.010.033 0.224 0.003 0.015 0.2350.054 0. 0.054 0.031 0. 0.2580.046 0.076 0.026 0. 0.260 0.090 0. 0.160 0.220 0.435 0.533 0.919 1016 1 20 25 40 50 63 80 6.5. Cross-correlation of measured one-third octave one-third band levels of measured 6.5. Cross-correlation 100 125 160 200 250 315 400 500 630 800 12.5 31.5 ABLE 1/3 Octave Bands 10 12.5 16 20 25 31.5 40 50 63 80 100 125 160 200 250 315 400 500 630 800 T indicates correlation coefficients above 0. coefficients correlation indicates coefficients <0.6.coefficients See text for details. Chapter 6: Sound Measurements, 2002 Open-Water Season 6-35

FIGURE 6.21. Percentile distribution of one-third octave band levels of underwater sound vs. center frequency as observed ~420 m north of Northstar via C.H. #2, 31 Aug. to 23 Sept. 2002, for wind speeds <5 knots. The five curves show, for each frequency, the minimum, the 5th, 50th and 95th percentiles, and the maximum. Each statistical spectrum is based on 1236 one-minute measurements.

DISCUSSION

Boat-based Recordings Underwater Sounds Routine Island Sounds.— Broadband levels recorded on 6 Sept. 2002, which were lower than any measurements made during the summers of 2000 and 2001, were a result of both exceptionally good recording conditions (sea state 0–½) and the absence of exceptional events such as the presence of sealift in 2001 (see Fig. 6.5). The highest mean broadband level recorded in 2002 in the absence of some specific source (e.g., crew boat) was 116 dB re 1 µPa, only 5 dB higher than the lowest mean broadband level recorded the previous year (111 dB re 1 µPa). The lowest broadband (10–10,000 Hz) level recorded, 87 dB re 1 µPa at 10 n.mi. on 6 Sept., is 7 dB below the previous lowest value (94 dB re 1 µPa at 1 n.mi. on 17 Sept. 2000). The corresponding value for the broadband range 20–10,000 Hz is 85 dB, which falls between the 5th and 50th percentile levels in Burgess and Greene’s (1999) measurements of ambient sound levels in the Beaufort Sea (for the same broadband range). In the absence of vessels, the broadband levels in 2002 at the closest recording station (¼ n.mi.) were 15–39 dB lower than recordings made in August and September 2000 at similar distances, but with 6-36 Monitoring at Northstar, Open-Water Season vessels present. Broadband values obtained on 6 Sept. indicate that, in the absence of vessels, background levels were attained underwater ~3.5 km from Northstar. Background levels were relatively low that day because of the near-calm sea, so underwater sounds were presumably detectable farther away than on an average day. The range of detectability for the island is difficult to assess for 12 Sept., as the broadband levels decreased only between the ¼ and ½ n.mi. stations, and then increased at every station thereafter. This most likely reflected the increasing sea state with increasing distance offshore on that date (Fig. 6.4). On 5 and 6 Sept., broadband levels at the two closest stations were the same and probably reflect “true” Northstar sound levels at that distance during drilling and production, with little or no contribution by background sound, as the sea was mirror-still during both recordings. Oil production took place on 5 and 6 Sept., but not on 12 Sept., whereas drilling took place at Northstar during most recordings. Figure 6.6 shows spectrum levels for two recordings that only differ by the presence or absence of production. Received levels during production were 15–20 dB higher across nearly all frequencies; in addition there was a group of tones in the 13–16 kHz frequency range that were not present without production. We infer that these tones may be linked to production itself or to some machinery that was used on the 6th but not on the 12th. For example, the Solar power generator A, the crude stabilizer pumps (A and B), and the crude sales pumps (A and B) all ran on the 6th but not at all on the 12th (see below). Nevertheless, the broadband levels received at close range (500–600 m) during boat-based recordings on 6 Sept. (with production) were lower than those on 12 Sept. (without production) (Fig. 6.4). This suggests that production was not necessarily the dominant source of sound even at those relatively close distances. The probable importance of sea state at all but the closest distances is emphasized in a comparison of the SPLs obtained on 6 and 12 Sept. (Fig. 6.4). Sea state was higher on the 12th and is potentially responsible for the higher broadband levels. However, there are a number of other differences between the two days that complicate explaining this difference. For instance, the following are consistent with higher broadband levels on 12 Sept: • There was no barge to the island on 6 Sept. but normal barge traffic (one barge) on 12 Sept. • The drill rig operated on 3 engines on the 6th, but 5 engines on the 12th; these engines are independent from the island’s main power source. • Emergency generator A ran for 0.4 hours on the 6th and 6.4 hours on the 12th.4 In contrast, the higher broadband levels obtained on 12 Sept. are not consistent with the following facts: • There was oil production on the 6th but not on the 12th, because of a 4th stage LP compressor seal repair. • The LP gas compressor, crude stabilizer pumps (A and B), and crude sales pumps (A and B) ran all day on the 6th, but not at all on the 12th.4 • The Solar power generator A ran for 5.2 hours on the 6th, but not at all on the 12th.4 Offshore drilling is usually done from natural or man-made islands, caissons emplaced on the bottom, platforms standing on legs, or drilling vessels. Of these categories, the underwater sounds associated with drilling from natural or man-made islands are generally the weakest (Richardson et al. 1995), and do not propagate beyond a few kilometers. Davis et al. (1985) made sound measurements in

4 The Island Rotating Equipment reports made available to us did not specify the times at which the rotating equipment was functioning, only the total number of hours per day. Chapter 6: Sound Measurements, 2002 Open-Water Season 6-37 open-water conditions at Seal Island, the artificial island that used to exist on the Northstar site. When nothing but power generation was taking place at the island, it could not be heard underwater 2.3 km away. During well-logging, the detection range increased to ~5 km. Johnson et al. (1986) measured drilling sounds from Sandpiper Island, in the same general area northwest of Prudhoe Bay. As in this study, the sound levels produced were low but notable. They reported tones at 20 and 40 Hz, which they attributed to power production. The tones, as well as broadband (20–1000 Hz) levels, increased during drilling by 6–24 and 8–10 dB, respectively. The tones were weakly detectable at 3.7 km during drilling. As they are for drilling, underwater sounds from production operations are apparently weaker during production from man-made islands than during production from most or all other types of offshore facilities (Richardson et al. 1995). Gales (1982) summarized the sounds received from 11 production platforms and one man-made island (all in temperate waters) and rated the sounds from the man-made Rincon production island as “very quiet”. He surmised that poor conduction of sound through the rock and fill island, and poor conduction of low-frequency sound in shallow water, both resulted in the low levels of production noise in the sea. Similar conditions occurred in the present study. Crew Boat Sounds.—In 2002, levels of underwater sound were elevated considerably when the crew boat was operating nearby. During the recording of the crew boat’s arrival, offloading of passengers, and departure (1 n.mi. station on 6 Sept. 2002, see Fig. 6.7), broadband levels varied by 31 dB (92– 123 dB re 1 µPa), depending on the presence and activity of the crew boat. The highest sound levels produced by the crew boat were, predictably, obtained while the vessel was either cruising at full speed or accelerating. Levels while it was pushing backward against the dock while loading/offloading passengers were 10–15 dB below the levels while accelerating, at a comparable distance. The crew boat and other vessels are important sources of underwater sound around Northstar. When present, they contribute substantially more underwater sound than do the routine production and drilling operations on the island. However, because they are transient sound sources, their contribution to overall sound levels is likely limited. This is further discussed below, in the section on “Near-island Cabled Hydrophone Recordings”. The narrowband spectrum that includes sound from the crew boat (Fig. 6.8) shows two large peaks at 26 and 52–53 Hz. These tones were present in all samples with the crew boat within sight, but disappeared from the sound record when the crew boat passed behind the island. With a maximum shaft rate of 900 rpm (or 15 Hz) and four-bladed propellers, we would expect a maximum blade rate of 60 Hz (rpm × number blades per propeller / 60) for the crew boat. If the shaft or one of the propeller blades of the vessel has damage (i.e., a blade nicked or a piece of propeller knocked off, or a shaft not quite straight), the shaft rate can become very prominent. It is likely that the 52–53 Hz peak is the crew boat’s blade rate and the 26 Hz peak is a harmonic of the shaft rate (in this case 13 Hz), which is at a frequency too low to propagate well in the shallow waters around Northstar. Figure 6.8 also shows that the crew boat, besides producing tones at 26 and 52–53 Hz, produced substantial broadband sound in the range 50– 2000 Hz. Propeller cavitation would account for this broadband sound. This broadband effect is confirmed in Figure 6.9, in which all one-third octave bands show heightened values as a result of the crew boat’s presence, particularly out to distances of ~3 km. Airborne Sounds This study (including its 2000–01 as well as 2002 phases) has provided some of the first available data on the airborne sounds associated with offshore oil industry activities in the arctic. Nonetheless, this study has given more attention to underwater sounds because of their importance to bowhead whales. Broadband in-air levels recorded during September 2002 were lower than any airborne levels obtained during the summers of 2000 and 2001 (Fig. 6.11). The highest values (52–56 dBA re 20 µPa) were obtained 6-38 Monitoring at Northstar, Open-Water Season

¼ n.mi. from Northstar and are comparable to the sounds produced by light road traffic at 30 m (100 feet; Kinsler et al. 1982). Despite differing sea state conditions, the spreading loss terms were very similar on all three days of recording: 21.6, 18.1 and 17.1 dB/tenfold change in distance, respectively (Fig. 6.10). On 6 Sept., the crew boat caused an increase in airborne broadband levels at a distance of about 1 n.mi. (1.9 km or 1.2 mi), but not at 2 n.mi. (3.7 km or 2.3 mi). In the crew boat’s absence, airborne broadband levels from Northstar reached background values 1–2 n.mi. (1.9–3.7 km or 1.2–2.3 mi) from Northstar. The A-weighted broadband levels obtained at the stations ≥2 n.mi. from Northstar (37–40 dBA re 20 µPa) were comparatively low and correspond to the sound levels of a quiet residential neighborhood during nighttime (Kinsler et al. 1982). However, on one occasion in 2002 airborne sounds from Northstar were nonetheless faintly audible by the acoustics field crew at a distance of 20 n.mi. (37 km or 23 mi). Narrowband spectra (Fig. 6.12) show that levels were about 20 dB higher during production than without production, and the difference was seen at all frequencies. This difference cannot be accounted for by wind, since wind speed was actually lower when production was taking place (1.7 mph) than during the recording on 12 Sept. without production (5.8 mph). In contrast to the underwater results (Fig. 6.8), narrowband spectra did not differ much as a function of the crew boat’s presence or absence 1 n.mi. away. However, the 26 Hz peak—possibly a harmonic of the shaft rate—is visible in Figure 6.13, as well as a faint 52 Hz tone (26 × 2). These results illustrate that the crew boat had a larger effect on the underwater sound field than on the in-air sound. The one-third octave data (Fig. 6.14) showed that, with increasing frequency, the band level diminished to background levels progressively closer to Northstar, i.e. 1 km for 6300 Hz5, 3.5 km for 630 and 2000 Hz5, and about 10 km for 200 Hz5 and below. Near-island Cabled Hydrophone Recordings

Broadband levels as recorded by the C.H. over 23 days in 2002 were less variable than during the corresponding 21-day period in 2001: the 5th percentile was higher in 2002 than in 2001 (95 vs. 88 dB re 1 µPa) and the 95th percentile was lower (117 vs. 123 dB). Median values in the two years were very close, ~103 dB both years (see Table 6.3). (The same trend was seen in the minimum and maximum values, but that trend is less meaningful as those values can depend on a single event, unlike the 5th and 95th percentiles.) The most likely explanation for the increase in the 5th percentile in 2002 is that drilling and production, which did not take place in September 2001, increased the baseline sound levels in 2002. The peak levels recorded by the cabled hydrophone were likely caused by similar types of sound sources during both years, i.e., vessels near or [especially] just north of Northstar and close to the hydrophone. Some of the boat-based measurements presented in this report seem inconsistent with the cabled hydrophone data, since the recordings of 6 Sept. yielded the lowest broadband values to date both underwater and in air despite the fact that both production and drilling were underway (Fig. 6.5, 6.11). However, at least some of these apparent discrepancies are understandable when one pays attention to the specific circumstances of the various recordings. For the underwater data shown in Figure 6.5, the high values of 2001 were all recorded in August, while the sealift vessels were present. In 2000, the two sets of recordings taken on 1 and 17 Sept. showed levels in the same general range as those documented in 2002. Finally, one factor that sets the recordings of 6 Sept. 2002 apart from all of our other 2000–02 open-water data near Northstar is the exceptionally low sea state that day, which nearly eliminated the contribution of wave slap to the broadband values.

5 Center of one-third octave band. Chapter 6: Sound Measurements, 2002 Open-Water Season 6-39

Vessel Sounds The gap in SPL “spikes” created by the lack of vessel traffic during the storm of 23 Sept. (Fig. 6.15A) emphasizes the role of vessels as an important sound source around Northstar. The crew boat is a more important sound source than the daily barge, because the crew boat typically makes eight round- trips daily (instead of 0–1). It also travels faster. “Removing” the contributions of vessels from the cabled hydrophone dataset on seven specific dates (Fig. 6.17) did not affect the minimum or the 5th or 50th percentiles, but lowered the 95th percentile by 10 dB and the maximum by 20 dB. This is visible in both Figures 6.17 and 6.18, and was expected because the crew boat is one of the stronger sound sources around Northstar. However, because of its transient nature, the overall effect on median broadband levels is likely small, on the order of 1 dB. It is important to remember that there are many other vessels in the Northstar area that are not linked to Northstar (see Table 2.1 and section “Other Major Offshore Activ- ities” in Chapter 2). The cabled hydrophone data, acquired 420 m from the island, are not likely to be substantially affected by sounds from these vessels because they do not come to Northstar. Most of the vessel noise that is prominent in the cabled hydrophone recordings represents Northstar-associated boat traffic—in 2002, mainly the crew boat and barge. During the period 1–19 Sept. 2002, the crew boat accounted for about 60% of the boat runs to and past Northstar. The match between the “crew boat windows”—determined from the schedule—and the spikes in broadband levels that most likely correspond to the crew boat or a similar vessel is very good on 14 Sept., less so on 13 Sept. (Fig. 6.16). As explained in the “Results” section, the discrepancies can in this case be attributed to changes in the crew boat schedule and to a visit by the ACS vessel Harrison Bay with the acoustics crew. Changes in the crew boat schedule likely explain many of the unaccounted-for spikes in Figure 6.15C, as such changes or delays were not uncommon. Other vessels, such as those from ACS, would create similar spikes. For example, the acoustics crew used ACS vessels to tend the DASAR array (see Chapters 7–8), for deployment and retrieval of instrumentation both near Northstar and in the DASAR array, and for all the open-water recordings. On these days, a visit to Northstar to check on the functioning of the cabled hydrophone recorders was almost always included. During the period 30 Aug.– 1 Oct., the acoustics crew visited Northstar 14 times on 12 different days, each time presumably producing a spike such as that depicted in Figure 6.16 (13 Sept., time ~16:30). During the 15 min that the crew boat remained at the island, the boat held itself against either the eastern or western dock by backing up against the dock, using the engines and propellers to maintain station. The barge spent much more time tied up at the island, sometimes hours, and the tug associated with the barge never shut down. In calm weather the tug just idled. However, in rough weather, the tug helped maintain the barge against the dock (C. Ledgerwood, AIC, and J. Huey, UOSS, pers. comm.). When the barge was tied up, it was at the island’s southern dock, which is on the opposite side of the island in relation to the C.H. (and the migration corridor of the bowhead whales). Northstar thus served as an acoustic shield between the tug and the C.H. (and migration corridor). This presumably explains why, on both days shown in Fig. 6.16, SPLs returned to normal between the barge’s arrival and departure. Overall Underwater Sound The statistical spectra shown in Figure 6.19 are similar to those obtained in 2001 (Fig. 8.13 in Greene et al. 2002), but with a few differences. The 2001 data revealed a peak level for the one-third octave band centered at 63 Hz, which was absent in 2002. The narrowband spectra for 2001 also showed marked peaks at nearly every multiple of 10 between 20 and 100 Hz; these are also absent in Figure 6.19A. This could be a result of the switch in the island’s power source from diesel generators (up to autumn 2001) to gas turbines. 6-40 Monitoring at Northstar, Open-Water Season

Sound Field Around Northstar

Estimated distances at which sounds from a specific source reach background levels vary with the variable source levels of the sound in question, with the ambient noise levels, and with the frequency band under consideration. Ambient levels depend primarily on sea state and wind. Estimated distances at which Northstar sounds reach background levels are presented in Figure 6.22 for September 2002, based on broadband data. • Natural ambient noise levels in the Beaufort Sea (away from industrial areas) during the open-water season were measured by Burgess and Greene (1999). Their broadband (20–1000 Hz) values for the 5th, 50th and 95th percentiles were 79, 99 and 114 dB re 1 µPa, respectively. These ambient noise levels are shown in Figure 6.22 as horizontal dashed lines. Ambient levels cannot be determined from the present 2002 data because these data often included industrial sounds. • The 5th, 50th and 95th percentile Northstar sounds at various distances are shown by the sloped lines in Figure 6.22. The values recorded 417 m N of the island from 31 Aug. to 23 Sept. 2002 (C.H. #2) are plotted on Figure 6.22 at the “R = 417 m” location. These values are for the 20–1000 Hz band, consistent with the ambient noise data of Burgess and Greene (1999). Approximate levels that would prevail at other distances are shown by the “15 log (R)” spreading-loss curves plotted through those three percentile values. The 15 log (R) rate was used because it is close to the propagation loss rate obtained from several sets of open-water measurements in both 2000 and 2001: 15 log(R) during the arrival of barge 400 in 2001, and 18.3 log(R) and 14.4 log(R) during open-water recordings on 30 Aug. and 1 Sept. 2000, respectively. • At times with median levels of island sounds as reported in September 2002 (i.e., 103 dB re 1 µPa at distance ~417 m), those sounds are expected to diminish below the median ambient level (99 dB) at a distance of about 730 m (~2400 ft) from the island. These median-level island sounds would diminish below the 5th percentile ambient level (79 dB) at a distance of ~15.8 km (8.5 n.mi. or 9.8 mi) from the island. These two cases are shown by the green and red circles (respectively) in Figure 6.22. • At times with higher (e.g., 95th percentile) levels of island sound, the distances at which those sounds would diminish below various possible levels of ambient noise are correspondingly higher, e.g., 6.7 km (3.6 n.mi. or 4.2 mi) for median noise levels (gray circle in Fig. 6.22). • At times when island sound was weak (e.g., 5th percentile), that sound would be below ambient beyond ~4.2 km (2.3 n.mi. or 2.6 mi), even at times with relatively low ambient noise. When island sound is weak, it would diminish below the median ambient level even before reaching the C.H. location 417 m from the island (Fig. 6.22). “Island sounds” as discussed here include sounds both from the island and from the boats that occurred around the island during September 2002. Removing the crew boat and other vessels from the sound record, as was done in Figures 6.17 and 6.18, allows us to estimate the range of detectability of the island alone after excluding the contribution by vessels directly linked to the island. Without the vessels, the 95th percentile of the broadband levels decreased by ~10 dB. At times with high (e.g., 95th percentile) levels of island sound, these sounds would reach median ambient levels at a distance of ~1440 m (4720 feet) excluding all vessel noise, as compared with 6700 m with the vessels included. At times with median levels of island sound, these sounds would reach median ambient levels at a distance of ~680 m (2230 feet) excluding all vessel noise, as compared with 730 m with the vessels included. Chapter 6: Sound Measurements, 2002 Open-Water Season 6-41

FIGURE 6.22. Approximate distances at which broadband Northstar sounds, as recorded in September 2002, would diminish below broadband ambient levels. Horizontal dashed lines represent " low", "average" and "high" ambient sound levels (5th, 50th and 95th percentile values; see text). Red symbols show 5th, 50th and 95th percentiles of 20–1000 Hz broadband island sounds as recorded 420 m north of the island in September. Sloping lines show estimated levels of island sounds at other distances, assuming a spreading loss rate of 15 log (R), as documented by 2000 and 2001 open-water measurements. Gray portions of sloping lines are extrapolated beyond the range of 2000 to 2002 measurements. Intersections of sloping Northstar sound lines with horizontal ambient noise lines show approximate distances at which broadband Northstar sounds would reach background levels underwater with various combinations of industrial and ambient levels. Distances corresponding to green and red circles are those where median Northstar sound would diminish below, respectively, the median and fifth percentile ambient noise. The distance corresponding to the gray circle is that where higher (95th percentile) levels of Northstar sound would diminish below median levels of ambient noise.

The above analysis is based on broadband (20–1000 Hz) data. Results would differ somewhat depending on the frequency band considered. Sounds at certain frequencies where the source level of industrial sound is strong and where ambient noise is not strong could be above ambient and audible to marine mammals somewhat farther away than suggested by Figure 6.22.

SUMMARY The primary objective of the acoustic measurement program at Northstar during the 2002 open- water season was to document underwater and airborne sounds near Northstar during the first open-water season with oil production operations. Greeneridge Sciences Inc. measured underwater and airborne sounds during three days in September 2002 at seven distances from Northstar Island. In addition, two near-island cabled hydrophones, located off the north shore of Northstar, recorded underwater island sounds continuously for 31 days from 31 Aug. to 1 Oct. 2002. 6-42 Monitoring at Northstar, Open-Water Season

Boat-based Recordings in Open Water

Boat-based recordings were obtained on 5, 6, and 12 Sept. 2002 to help characterize sound levels underwater and in air during drilling (all days) and production (5 and 6 Sept.). Recordings were made at seven nominal distances ranging from ¼ to 20 n.mi. (0.46–37 km or 0.29–23 mi) north of Northstar Island. Data recorded on 6 Sept. were of particularly good quality as sea state was low (0–½), all stations were sampled, and both drilling and production were taking place. In the absence of vessels, both underwater and in-air broadband levels were the lowest recorded in the three years for which these types of measurements have been made (2000–02). Measured values flattened out at background values of 87–90 dB re 1 µPa underwater and 37–40 dBA re 20 µPa in air at the more distant stations. The maximum recorded levels were at the station closest to the island (¼ n.mi.) and reached 116 dB re 1 µPa underwater and 56 dBA re 20 µPa in air. These are quite low levels. While both drilling and oil production were taking place at Northstar, underwater and airborne sounds reached background levels ~3.5 km (2.2 mi) from Northstar. A logarithmic sound propagation model was fitted to in-air data from 6 Sept. and resulted in a spreading loss term of ~18 dBA/ tenfold change in distance. Sounds produced by Northstar during drilling and production were comparable to those obtained in similar conditions in other studies. Overall, sounds from drilling and/or oil production on gravel islands are low compared to those produced by many other oil production structures such as platforms and drill ships. When the crew boat or other operating vessels were present in 2002, sound levels were higher. From a distance of ~1 n.mi. (1.9 km or 1.2 mi), the presence of the crew boat at the island increased broadband levels by as much as 30 dB. The highest levels were recorded when the vessel accelerated or was underway at full speed. Levels were 10–15 dB lower while the crew boat backed against the dock, with engine and propellers running, during the transfer of passengers. The highest recorded broadband level from the crew boat was 128 dB re 1 µPa at a distance of 3.7 km (2.3 mi). A comparison of narrowband spectra obtained with and without the crew boat showed it to be the source of distinct low-frequency tones at 26 and 52–53 Hz. The crew boat is also a broadband source of sound underwater, as shown both by narrowband and one-third octave band data. In contrast, the crew boat had little effect on narrowband spectra in air at range 1.8 km (1.1 mi or 1 n.mi.). Near-island Cabled Hydrophone Recordings

A cabled hydrophone located 420 m (1370 feet) north of the north shore of Northstar provided a continuous record of low-frequency (10–1000 Hz) sounds near the island from 31 Aug. to 1 Oct. 2002. The minimum broadband (10–1000 Hz) levels were about 90 dB re 1 µPa, which was ~10 dB higher than in the corresponding season of 2001, and the maximum levels in 2002 were ~6 dB lower than in 2001. However, median values were comparable. Drilling and oil production, which were not taking place in September 2001, may have been responsible for the higher minimum levels in 2002. Vessels were an important but intermittent sound source around Northstar in 2002 and many “spikes” in broadband levels could be attributed to the crew boat. When times with “spikes” of boat noise were excluded from a 7-day dataset, the maximum and 95th percentile values decreased by 20 and 10 dB, respectively. “Statistical” narrowband spectra, calculated over the entire sampling period, were similar to those obtained in 2001. Broadband levels obtained simultaneously by the cabled hydrophone and during the open-water recordings were compared; values obtained during the open-water recordings were usually lower than values obtained by the cabled hydrophone, except close to the island. Chapter 6: Sound Measurements, 2002 Open-Water Season 6-43

Five different methods were used to help determine which frequencies to include in the Industrial Sound Index (ISI) for 2002. One method yielded equivocal results, but the other four methods supported the choice of the one-third octave bands centered at 31.5–80 Hz to define the ISI in 2002.

ACKNOWLEDGEMENTS

I thank Jim Nevels, Eric Olsen, Walter Henry, Mark Theriault, Karl Pulliam, Charles Crabaugh, Gene Pennington and Gary Seims from Alaska Clean Seas for their cooperation, enthusiasm, and expert boating skills. From LGL Alaska, I thank Jonah Leavitt for help with fieldwork, Michael Williams and Shelly Schwenn for help with logistics, and Ted Elliott for creating Figure 6.1. For logistical support and problem solving on Northstar, I thank Jeff Huey, Julie Arin, Pam Pope and David Maguire. I also thank Allison Erickson and Wilson Cullor of BP’s Environmental Studies Group for their help with logistics. The security guards on the island were also supportive. Dr. W. John Richardson, LGL Ltd., provided program direction and guidance as well as helpful criticism of this report. Dr. Charles Greene of Greene- ridge Sciences reviewed and improved the manuscript. Dr. Bill Streever of BP’s Environmental Studies Group and Robert Suydam of the North Slope Borough provided valuable comments on the manuscript. Finally, we thank BP, specifically Dr. Ray Jakubczak and Dr. Bill Streever, and Dave Trudgen of OASIS Environmental, for their support of this project in 2002 and/or previous years.

LITERATURE CITED

Blackwell, S.B. 2003. Sound measurements, 2002 open-water season. p. 2-1 to 2-25 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of industrial sounds and whale calls during operation of BP’s Northstar Oil Development, Alaskan Beaufort Sea, summer and autumn 2002: 90-day report. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 66 p. Superseded by the present report. Blackwell, S.B. and C.R. Greene Jr. 2001. Sound measurements, 2000 break-up and open-water seasons. p. 7-1 to 7-55 In: W.J. Richardson and M.T. Williams (eds.). 2001. Monitoring of industrial sounds, seals, and whale calls during construction of BP’s Northstar oil development, Alaskan Beaufort Sea, 2000. [April 2001 ed.] Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 316 p. Blackwell, S.B. and C.R. Greene Jr. 2002. Sound measurements, 2001 open-water season. p. 7-1 to 7-39 In: W.J. Richardson and M.T. Williams (eds., 2002a, q.v.). LGL Rep. TA2572-2. Burgess, W.C. and C.R. Greene, Jr. 1999. Physical acoustics measurements. p. 3-1 to 3-65 In: W.J. Richardson (ed.), Marine mammal and acoustical monitoring of Western Geophysical’s open-water seismic program in the Alaskan Beaufort Sea, 1998. LGL Rep. TA2230-3. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for Western Geophysical, Houston, TX, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 390 p. Burgess, W.C. and C.R. Greene, Jr. 2000. Physical acoustics measurements, 1999. p. 3-1 to 3-45 In: W.J. Richardson (ed., 2000), Marine mammal and acoustical monitoring of Western Geophysical’s open-water seismic program in the Alaskan Beaufort Sea, 1999. LGL Rep. TA2313-4. Rep. from LGL Ltd., King City, Ont., and Greene- ridge Sciences Inc., Santa Barbara, CA, for Western Geophysical, Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 155 p. Davis, R.A., C.R. Greene and P.L. McLaren. 1985. Studies of the potential for drilling activities on Seal Island to influence fall migration of bowhead whales through Alaskan nearshore waters. Rep. from LGL Ltd., King City, Ont., for Shell Western E&P Inc., Anchorage, AK. 70 p. 6-44 Monitoring at Northstar, Open-Water Season

Gales, R.S. 1982. Effects of noise of offshore oil and gas operations on marine mammals–An introductory assessment. NOSC TR 844, 2 vol. U.S. Naval Ocean Systems Cent., San Diego, CA. 79 + 300 p. NTIS AD-A123699 + AD-A123700. Greene, C.R., M.W. McLennan, T.L. McDonald and W.J. Richardson. 2002. Acoustic monitoring of bowhead whale migration, autumn 2001. p. 8-1 to 8-79 In: W.J. Richardson and M.T. Williams (eds., 2002a, q.v.). LGL Rep. TA2573-2. Johnson, S.R., C.R. Greene, R.A. Davis and W.J. Richardson. 1986. Bowhead whales and underwater noise near the Sandpiper Island drillsite, Alaskan Beaufort Sea, autumn 1985. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for Shell Western E&P Inc., Anchorage, AK. 130 p. Kinsler, L.E., A.R. Frey, J.V. Sanders and A.B. Coppen. 1982. Fundamentals of acoustics, 3rd edition. Wiley, New York, NY. 480 p. Kryter, K.D. 1985. The effects of noise on man, 2nd edition. Academic Press, Orlando, FL. 688 p. LGL and Greeneridge. 2000. Technical plan for marine mammal and acoustic monitoring during construction of BP's Northstar Oil Development in the Alaskan Beaufort Sea, 2000-2001. [August 2000 ed.] Rep. from LGL Alaska Res. Assoc. Inc. (Anchorage), LGL Ltd. (King City, Ont.), and Greeneridge Sciences Inc. (Santa Barbara, CA), for BP Explor. (Alaska) Inc., Anchorage, AK. 80 p. Officer, C.B. 1958. Introduction to the theory of sound transmission with application to the ocean. McGraw-Hill, New York, NY. 292 p. Richardson, W.J., C.R. Greene, Jr., C.I. Malme and D.H. Thomson. 1995. Marine mammals and noise. Academic Press, San` Diego, CA. 576 p. Richardson, W.J. and M.T. Williams (eds.). 2002a. Monitoring of industrial sounds, seals, and whale calls during construction of BP’s Northstar Oil Development, Alaskan Beaufort Sea, 2001. [Oct. 2002 ed.] Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 309 p. Chapter 6: Sound Measurements, 2002 Open-Water Season 6-45

2 APPENDIX 6.1: DEFINITION OF INDUSTRIAL SOUND INDEX (ISI) IN 2001

This Appendix explains the choice of one-third octave bands to be used in defining an “industrial sound index” (ISI) most representative of the Northstar-related sounds measured underwater near Northstar Island during the autumn of 2001. The measured sound levels came from Cabled Hydrophone #2, located ~400 m north of the north edge of the island. This was the same instrument and the same hydrophone location as were used in 2002 to monitor island sounds (see Chapter 6). Blackwell and Greene (2002) and Greene et al. (2002) describe the measurements of industrial sounds obtained at this location during the summer and autumn of 2001. Based primarily on the sounds measured during August 2001, they developed an ISI that was based on the sound components in the one- third octave bands centered at 63, 80 and 160 Hz. For those August dates, sound components in those three frequency bands seemed to be most strongly affected by industrial sound emissions from Northstar or its support vessels, and least affected by background sounds. The average sound pressure level based on components in those three one-third octave bands averaged 5.8 dB less than that for the 10–1000 Hz band as a whole. For purposes of analyzing bowhead whale distribution relative to sounds from Northstar, it is more relevant to consider sounds near Northstar during the period 28 Aug. through 3 Oct. 2001, when bowhead whale migration was monitored (Greene et al. 2002; see also Chapters 9 and 10 of this report). Therefore, we here re-examine the sound measurements obtained ~400 m north of Northstar during that period in 2001. As demonstrated below, we find that the most appropriate definition of the ISI for that early- autumn period in 2001 is based on five one-third octave bands centered at frequencies from 31.5 to 80 Hz. This is the same definition that has been found most appropriate for September 2002 (see Chapter 6). As described in Greene et al. (2002) and Chapter 6, the 2001 recordings were analyzed for sound pressure spectral densities (SPSD) by averaging the spectral densities of 1-s long segments, overlapped by 50%, spanning one minute. Such one-minute averages were computed every 4.37 minutes (262 seconds). The entire data set was so analyzed from 28 Aug. until 3 Oct. 2001, yielding 11,486 spectral analyses. The individual cells in the spectra were sorted from smallest to largest, and the minimum, 5th percentile, 50th- percentile, 95th-percentile, and the maximum values were stored for each cell. Then, composite spectra were assembled as shown in Figure 6.23. For example, viewing the median (50th-percentile) in Figure 6.23, one knows that half the values were smaller and half were larger at every frequency. The 2001 ISI had previously been defined based on a week’s data recorded on an autonomous seafloor acoustic recorder (ASAR) during August 2001 (Fig. 7.19 in Blackwell and Greene 2002). Those data showed strong peaks (narrowband components, or tones) at 60, 80 and 150 Hz. The one-third octave band percentiles (Fig. 8.13 in Greene et al. 2002) supported the choice of the corresponding 63, 80 and 160 Hz one-third octave bands to define the ISI for 2001. However, the 5th and 50th percentile levels were relatively strong for one-third octave bands from 31.5 to 800 Hz. The presence of the tones, known to arise from man-made sounds, directed the selection of the three bands cited above. In the present re-consideration of the ISI for 2001 based on 28 Aug.–3 Oct. 2001 data, the initial choice of bands to include in the ISI was guided by the distribution of sound as evident in the percentile narrowband spectral density data (Fig. 6.23). It appears from that Figure that, for most of the time, the most prominent components of sound from the island were in the general range from 30 to 80 Hz.

2 This appendix was prepared by Charles R. Greene Jr., Greeneridge Sciences Inc., Santa Barbara, CA 6-46 Monitoring at Northstar, Open-Water Season

140

130

120

110 /Hz) 2 100 Pa µ 90

SPSD (dB re 1 70

40 101 102 103 Frequency (Hz)

FIGURE 6.23. Percentile sound spectral densities for sounds recorded by near-island Cabled Hydrophone #2, 28 Aug. – 3 Oct. 2001. The hydrophone was ~400 m north of the island’s north edge (at the water line). The sample size was 11,486. Measured levels in each frequency bin were sorted independently to determine the minimum, 5th-percentile, 50th-percentile (median), 95th-percentile, and maximum level for that frequency. (Similar to Fig. 7.19 in Blackwell and Greene 2002, but representing data from a different range of dates.)

Figure 6.24 presents the corresponding percentile distribution of the one-third octave band levels from 10 to 800 Hz. For each one-third octave band, the level is the sum of the spectral densities within the frequency range encompassed by that one-third octave band—for example, the 100 Hz band includes the sound at frequencies from 89 to 112 Hz. Again, the weight of evidence suggests that the one-third octave bands centered at frequencies from 31.5 through 80 Hz represent the important levels of the near-island sounds.

An alternative perspective comes from examining the relationship between the one-third octave band levels and the wind speed. Natural background levels typically are positively related to wind speed; industrial components would not be expected to show this correlation. Wind speed was logged from a gauge on top of the Permanent Living Quarters on Northstar Island. Although this gauge was below the level of the processing units to its north, it provided the best available measure of wind speed at the island. (Wind recorded at mainland facilities was known not to be the same as island wind.) The one- third octave band levels were measured as described above, from one-minute spectrum analyses every 4.37 minutes using sounds from Cabled Hydrophone #2. Chapter 6: Sound Measurements, 2002 Open-Water Season 6-47

140

130

120 Pa) µ

110

100

70 1/3 OctaveBand Level (dB re 1

50 10 16 25 40 63 100 160 250 400 630 1/3rd Octave Band Center Frequency (Hz)

th th th FIGURE 6.24. One-third octave band levels vs. center frequency for the minimum, the 5 -, 50 - and 95 - percentiles, and the maximum observed near Northstar (~400 m from the north edge of Northstar Island) via Cabled Hydrophone #2 during the period 28 Aug. – 3 Oct. 2001. Each statistical spectrum is based on 11,486 one-minute measurements taken every 4.37 minutes. (Identical to Figure 8.13 in Greene et al. 2002.)

It was expected that one-third octave bands most influenced by island sounds would show the least dependence on wind speed. For each of the 20 one-third octave bands centered in the range 10 to 800 Hz, a linear equation of the following form was fitted to the measured band levels and wind speeds: Expected band level = A + B•(wind speed) The fitted values of B were expected to be lowest for those bands least related to the wind speed. The results are presented in Table 6.6. Although all the B-values are low (<0.4) for center-frequencies from 12.5 through 80 Hz, sounds at frequencies below about 30 Hz do not propagate well in the shallow water north of Northstar Island. It is unlikely that these low frequency components would influence bowhead behavior, or even be detected, in the southern part of the whale migration corridor at least several kilometers seaward of Northstar. The coefficients of determination for these regressions were all <0.01 for the 31.5 through 80 Hz bands but were >0.01 for all the other bands. Low values of the coefficient of determination indicate poor correlation. Thus, the regression results are consistent with the choice of the one-third octave bands from 31.5 through 80 Hz for the ISI in 2001. 6-48 Monitoring at Northstar, Open-Water Season

TABLE 6.6. Parameter values for regressions of measured one-third octave band levels vs. wind speed, 28 Aug. – 3 Oct. 2001. Comparable to Table 6.4 in Chapter 6, which provides corresponding 2002 results.

Center P-VALUE Freq. (Hz) AB r2 10 74.713 0.541 <0.0001 0.031 12.5 69.206 0.432 <0.0001 0.031 16 65.411 0.349 <0.0001 0.021 20 69.330 0.385 <0.0001 0.020 25 72.967 0.370 <0.0001 0.022 31.5 82.979 0.015 0.0975 0.003 40 86.087 0.363 0.0042 0.010 50 87.603 0.367 0.0129 0.007 63 89.461 0.319 0.0681 0.004 80 88.383 0.397 0.0151 0.007 100 85.573 0.549 0.0004 0.015 125 86.802 0.477 0.0023 0.011 160 87.351 0.593 0.0001 0.017 200 86.765 0.640 <0.0001 0.021 250 83.898 0.579 0.0002 0.016 315 82.834 0.522 0.0012 0.012 400 83.248 0.626 <0.0001 0.019 500 83.344 0.812 <0.0001 0.036 630 83.729 0.782 <0.0001 0.038 800 83.447 0.649 <0.0001 0.029

Another comparison that might help in defining bands to include in the ISI was to cross-correlate the acoustic measurements for the various one-third octave bands. With certain assumptions, one might expect that the levels in the bands dominated by island sound would not correlate well with those in bands that were dominated by natural sea noise, and that the levels in bands with prominent island sounds might correlate well with those in other bands with island sounds. Table 6.7 presents the correlations, but the expected patterns do not hold particularly well. It is possible that the sources of island sound, which we know will vary with time, do not vary together with time (that is, one band of island sound may be strong while another is weak, and so on). On balance, it appears that the one-third octave bands centered at frequencies from 31.5 to 80 Hz are the best choice in defining the ISI for the whale migration season in 2001. The sound components in these bands appeared to be dominated by Northstar-related industrial sound during the period in question. It is convenient that the five one-third octave bands centered at 31.5 to 80 Hz are also the ones chosen (independently) for inclusion in the ISI for 2002 (see Chapter 6). This enhances the comparability of the data from year to year. Chapter 6: Sound Measurements, 2002 Open-Water Season 6-49 1 Island 1 0.991 1 0.989 0.990 1 0.982 0.963 0.971 1 0.976 0.958 0.941 0.948 1 0.986 0.967 0.955 0.935 0.944 1 0.974 0.962 0.974 0.971 0.959 0.964 1 0.992 0.970 0.963 0.975 0.966 0.950 0.960 1 0.986 0.976 0.964 0.955 0.961 0.940 0.917 0.930 1 0.986 0.961 0.953 0.948 0.934 0.937 0.911 0.882 0.895 Chapter 6, which provides corresponding 2002 results. 2002 corresponding provides 6, which Chapter 1 in 0.965 0.971 0.954 0.941 0.939 0.935 0.929 0.899 0.880 0.883 1 0.955 0.949 0.934 0.894 0.888 0.893 0.883 0.877 0.833 0.805 0.809 ls measured with Cabled Hydrophone #2 ~400 m north of Northstar Northstar m north of #2 ~400 Hydrophone with Cabled measured ls 1 0.920 0.890 0.905 0.866 0.823 0.817 0.830 0.815 0.806 0.763 0.776 0.730 0.794 0.734 0.920 0.873 0.883 0.886 0.870 0.852 0.837 0.839 0.841 0.844 0.805 0.616 0.569 0.592 0.696 0.681 0.589 1 0.667 0.711 0.675 0.648 0.666 0.650 0.643 0.638 0.670 0.569 0.563 0.596 0.602 0.745 1 0.691 0.727 0.7740.634 0.765 0.751 0.593 1 0.619 0.674 0.667 1 692 0.630 0.686 1 0.839 0.809 0.617 0.609 0.634 0.629 0.638 0.646 0.6170.601 0.663 0.621 1 0.909 0.835 0.608 0.636 0.639 0.635 0.634 0.616 0.632 0.635 0.660 0.638 0.917 0.857 0.599 0.584 0.597 0.595 0.591 0.593 0.540 0.555 0.498 0.520 0.558 0.521 0.535 0.580 1016 20 1 25 0.72740 0.729 0.777 5063 0.72580 0.719 0.694 0.720 0.722 0.686 0.676 0.669 0.645 0.632 0.641 0.634 100125160 0.678200 0.680 0.687 0.630250 0.637 0.654 315400 500 0.601630 0.607 0.638 800 12.5 31.5 0.688 0.672 0. 6.7. Cross-correlation of measured 6.7. Cross-correlation one-third octave band leve ABLE 1/3 Octave Bands1/3 Octave 10 12.5 16 20 25 31.5 40 50 63 80 100 125 160 200 250 315 400 500 630 800 T during the period from 28 Aug. through 3 Oct. 2001. Comparable to Table 6.5 to Table 2001. Comparable 3 Oct. 28 Aug. through the period from during CHAPTER 7: ACOUSTIC MONITORING OF BOWHEAD WHALE MIGRATION, AUTUMN 20021

by Charles R. Greene Jr. a, Miles Wm. McLennan a, Susanna B. Blackwell a, Trent L. McDonald b, and W. John Richardson c

aGreeneridge Sciences, Inc. 1411 Firestone Rd., Goleta, CA 93117 (805) 967-7720; [email protected]

bWestern EcoSystems Technology, Inc. 2003 Central Ave., Cheyenne, WY 82001 (307) 634-1756; [email protected]

cLGL Ltd., environmental research associates 22 Fisher St., POB 280, King City, Ont. L7B 1A6 (905) 833-1244; [email protected]

for

BP Exploration (Alaska) Inc. Dept. of Health, Safety & Environment 900 East Benson Blvd, P.O. Box 196612 Anchorage, AK 99519-6612

LGL Report TA2706-2

December 2003

1 Chapter 7 In: W.J. Richardson and M.T. Williams (eds.). 2003. Monitoring of industrial sounds, seals, and bowhead whales near BP’s Northstar oil development, Alaskan Beaufort Sea, 1999–2002. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 7-2 Monitoring at Northstar: Open-Water Season

ABSTRACT

During the bowhead whale migration in September 2002, Greeneridge Sciences, on behalf of BP, implemented an acoustic monitoring program north-northeast of BP’s Northstar oil development at North- star Island (formerly Seal Island). The primary purpose was to assess the effects of Northstar production activities, especially their underwater sounds, on the migration corridor and calling behavior of bowhead whales. This was to be done by comparing the offshore distances of calling bowheads at times with varying levels of industrial sound. If the closest bowhead whales detected by their calls tend to be farther offshore at times when Northstar sounds are stronger than average, this would be an indication some whales are affected by Northstar activities. The geographic scale of any such effect would provide evidence of the magnitude of the effect, if it occurs. This project in 2002 was a follow up to similar efforts in 2000 and 2001. The 2000 work demonstrated that the acoustic localization approach had promise but equipment problems limited the results. The 2001 results provided some evidence that the southern edge of the bowhead migration corridor was deflected offshore at times with high levels of underwater sound near Northstar, but additional data were needed to better characterize the effect and to address remaining uncertainties. The study in September 2002 occurred in much the same way as in 2001. This project required continuous measurements of sound levels near the island over time, and of the locations of calling bowhead whales over the same time. (1) To monitor sounds near the island, two cabled hydrophones and (for backup) two omnidirectional ASARs were installed ~400 m (¼ mi) north of the north edge of the island. These instruments provided a continuous record of underwater sounds near Northstar. (2) To detect and determine the source locations for whale calls in the nearshore part of the bowhead migration corridor, 11 “Directional Autonomous Seafloor Acoustic Recorders (DASARs) were installed in a grid pattern 6.5 to 22 km (4 to 13.7 mi) offshore of Northstar Island. These two types of acoustic data permit analyzing whale call locations in relation to near-island sound levels, the key project objective. Both the near-island hydrophones and the DASARs were installed at the same locations as used in 2001. For 2002, simultaneous data of the two types were available from 31 Aug. to 23 Sept., and cabled hydrophone data were in fact available until 1 Oct. The sounds received by cabled hydrophone in 2002 have been analyzed as broadband signals (10– 1000 Hz) and as one-third octave band levels for the 31 Aug.–1 Oct. period. Compared with 2001, the minimum broadband level ~400 m from the island was about 10 dB higher in 2002, while the maximum levels (generally due to vessel activity near the island) were about 6 dB lower in 2002. The higher minimum level in 2002 may be attributed to a difference in sounds from the drilling and production activities of 2002 as compared with the construction activities on the island during 2001. However, in both years, the times with highest sound levels near Northstar were those when vessels were operating nearby. A total of 10,587 bowhead whale calls was recorded in ~24 days. This is about the same number of calls that were detected during 36 days in 2001. However, in 2002 a higher proportion of the bowhead calls were detected by two or more DASARs, thus providing the intersecting bearings necessary for localization. Also, bearing accuracy was improved in 2002, a higher proportion of the calls were detected by >2 DASARs, and DASARs operated more reliably. (Ten of 11 operated nominally until overturned by a storm on 23 Sept.) Overall, data quality was improved in 2002 compared with 2001. This chapter describes the 2002 procedures and results. Subsequent chapters provide more specifics about the acoustic localization method (Chapter 8), statistical approach (Chapter 9), and results from 2001 and 2002 with regard to the key question of Northstar effects on migrating whales (Chapters 10, 11). Chapter 7: Acoustic Monitoring of 2002 Bowhead Migration 7-3

INTRODUCTION This chapter is a report on acoustic monitoring of bowhead whale migration near the Northstar development during the early autumn of 2002. The primary purpose was to assess the effects of Northstar production activities, as manifested in underwater sounds, on the migration corridor and calling behavior of bowhead whales. This was to be done by comparing the offshore distances of calling bowheads at times with varying levels of industrial sound. If the offshore distances changed with changing sound levels, this would be an indication some whales are affected by Northstar activities. The geographic scale of any such effect would provide evidence of the magnitude of the effect, if it occurs. This chapter describes the basic monitoring activities and results for 2002. In almost all respects the 2002 project paralleled the 2001 effort, with very similar objectives and methods (cf. Greene et al. 2002). Some of the material in Greene et al. (2002) is repeated with appropriate revisions for 2002. However, for detailed descriptions of the basic “Methods”, readers are encouraged to refer to the report for 2001 (Greene et al. 2002) or to Chapters 8 and 9 of this report. Those chapters provide more details on the acoustic and statistical methodologies, respectively. Chapters 10 and 11 of this report use the results of whale migration monitoring for 2001 and 2002 to address the primary objective concerning effects of Northstar on the migrating whales. The following three chapters (Chapters 8–10) have been written in a form suitable for journal publication. The present chapter introduces and supplements Chapters 8–10 by providing additional details concerning project background and objectives, and the specific monitoring activities and results in 2002. This chapter is a much-updated version of the preliminary (90-day) report on this monitoring task (Greene et al. 2003), and it supersedes that report. In 2001 and to a lesser extent in 2002, sounds of vessel traffic to and from Northstar contributed notably to island sound (see Chapter 6). Crewboats and tugs with barges were the primary contributors. With the transition from construction (2000, 2001) to production (2002), vessel traffic diminished. A further reduction in vessel traffic is likely in future when drilling at Northstar is completed. Also, in 2003, BP plans to test a hovercraft for possible future use in lieu of some vessel traffic. The hovercraft is likely to be quieter underwater than are the crewboats. If so, this will further diminish the Northstar-related underwater sounds. BP’s business rationale for this project was driven both by corporate policies and by regulatory requirements. BP corporate policies support studies that objectively assess environmental effects that may result from BP operations. In addition, monitoring the autumn migration of bowhead whales past Northstar was required, during the open-water season of 2002, to satisfy (a) provisions of the North Slope Borough zoning ordinance for Northstar, and (b) the monitoring requirements of the Letter of Authoriza- tion (LoA) issued by NMFS to BP on 14 Dec. 2001. Background

Concern has been expressed that the autumn migration corridor of bowhead whales might be deflected offshore in the Northstar area in response to underwater sounds from construction, operations, and associated vessel and aircraft traffic. Whales, including bowhead whales, are known to avoid various industrial activities when the received sounds are sufficiently strong (Richardson et al. 1995). Previous studies, including work at Seal Island when it was constructed in 1982 by Shell Oil Company (Hickie and Davis 1983) and when equipped for exploratory drilling in 1984 (Davis et al. 1985), showed that most underwater sounds propagating from a gravel island like Northstar are quite weak and usually not detectable beyond a few kilometers offshore of the island. Sounds from impact pile driving are an exception (Moore et al. 1984; Johnson et al. 1986; Blackwell and Greene 2001), but BP has not driven piles at Northstar during the bowhead migration seasons of 2000, 2001, or 2002. Sounds from boat traffic are 7-4 Monitoring at Northstar: Open-Water Season another exception, although vessel (and aircraft) traffic associated with Northstar construction have rarely extended more than 1 km (0.6 mi) north of Northstar toward the bowhead migration corridor. During the planning phase of this project, it was assumed that construction and operational sounds from Northstar would be detectable underwater for only a relatively short distance, typically on the order of a few kilometers. Hence any effects of Northstar construction or operations on the autumn migration corridor were expected to be limited to the small minority of bowheads that move along the southern edge of the migration corridor. For this reason, the effort to monitor Northstar effects on bowhead migration near Northstar needed to concentrate on the southern part of the migration corridor, and had to be designed to detect small-scale effects. In actuality, during the bowhead migration seasons in 2000–2002, sounds from Northstar were often detectable farther offshore, as documented in Blackwell and Greene (2001) for 2000, Blackwell and Greene (2002) for 2001, and Chapter 6 in this volume for 2002. This was primarily attributable to boat operations, including crewboats, tug-and-barge operations, and other vessel operations in the general area. Although many of these vessel movements were in support of Northstar, others had no direct connection with Northstar (see Chapter 1). Specific Objectives

During 1999 and early 2000, BP and its contractors devised a new type of acoustic localization system intended for application during September–October 2000 and future years as necessary to monitor the autumnal migration corridor of bowheads passing Northstar. The main objectives were to document the occurrence of calling bowhead whales in the southern part of their migration corridor in relation to distance offshore of Northstar, and to determine whether their distances from the island varied in direct relation to the sound levels emanating from the island. If so, this would indicate that Northstar affected the distribution and/or the calling behavior of the whales. If there was an effect, the geographic extent of the effect was to be determined, and this information was to be used in estimating the number of whales that were apparently affected. The specific 2002 objectives for this task were listed in a request from BP to NMFS, dated 15 May 2001, for a Letter of Authorization applicable to Northstar operations in 2002. These objectives, closely following those for 2000–2001, were as follows: 1. “to monitor the year 2002 bowhead migration past Northstar by acoustic localization methods, primarily to document the relative numbers of bowhead calls vs. distance offshore in the southern part of the bowhead migration corridor seaward of Northstar; 2. “to document continuously • the characteristics of the sounds propagating away from Northstar, as determined at a close-in monitoring site approximately 300–500 m (1000–1600 ft) offshore from the island, as well as • the characteristics of the background sounds (industrial plus ambient) reaching the recording devices farther offshore in the southern part of the migration corridor; 3. “to assess the effects of Northstar activities on the migration corridor and calling behavior of bowheads by comparing the distances of calling whales from Northstar at times with varying levels of industrial noise; 4. “to track individual bowheads or bowhead groups acoustically when feasible. The tracks will be analyzed to determine whether there is evidence of deflection as the whales approach Northstar and, if so, at what distances and received sound levels; 5. “to compare the acoustic monitoring results with results from MMS or any other available aerial surveys to further assess the strengths and limitations of acoustic and aerial methods in documenting the bowhead migration corridor past Northstar and in detecting any change (probably small- scale, if at all) in the migration corridor near a site like Northstar; Chapter 7: Acoustic Monitoring of 2002 Bowhead Migration 7-5

6. “to estimate the number of bowhead whales whose movements and/or calling behavior were affected by the presence of industrial activities at Northstar in autumn 2002.” The primary goal of this research is to estimate the effect of Northstar sound on migrating bowhead whales, with an intentional emphasis on the southern (nearshore) portion of the distribution of calling whales (that is, those whales most likely to be affected by Northstar sound). Our approach has been to estimate the magnitude in kilometers of the apparent average offshore displacement (if any) as a function of the industrial sound level, thus addressing objective (3) above. General Approach

Our approach is based on the assumption that some effect of Northstar on bowhead whales migrating along the southern edge of the migration corridor was likely, but that this effect would probably be small in spatial scale and intermittent. We expected the effect to be difficult to detect and quantify given those factors, natural variability, and the lack of any true “no Northstar” control data. Therefore, considerable effort was devoted to developing a method that would have high sensitivity, and to development of consensus among stakeholders regarding the approach that should be taken. Two potential monitoring approaches were considered by BP, the Department of Wildlife Manage- ment of the North Slope Borough (NSB), National Marine Fisheries Service (NMFS), and the peer/stakeholder review group: aerial surveys and acoustic localization. BP did not propose to use aerial surveys, on the rationale that even an intensive site-specific aerial survey would provide too few sightings of bowhead whales to detect and characterize a small-scale displacement of the southern edge of their migration corridor. Results of a power analysis supporting that view were presented at the peer/stakeholder meeting in Seattle in May 2000. The power analysis was described in more detail in an Appendix to the final monitoring plan for 2000 (LGL and Greeneridge 2000), repeated as an Appendix to Greene et al. (2001). Instead, BP proposed to develop and use an acoustic localization technique to document the occurrence and locations of calling bowhead whales in the southern part of the migration corridor. In particular, this approach would determine whether the distribution of calling whales in that area is related to short-term variability in sounds emitted by Northstar operations. If so, this would suggest that Northstar sounds have an effect on migrating whales. The geographic scale of this effect should also be determinable, which (in turn) would provide much of the information needed to estimate the numbers of bowheads affected. This approach was discussed and provisionally accepted at the peer/stakeholder review meeting in Seattle in May 2000. This acceptance was subject to completion of a statistical power analysis addressing whether the proposed acoustic approach was likely to detect an effect of the anticipated magnitude if it occurred. That power analysis was completed, and it confirmed that the proposed approach is likely to have the necessary statistical power (see Appendix C in LGL and Greeneridge 2000, also repeated in Greene et al. (2001)). A complication in monitoring effects of Northstar construction (and, in subsequent years, Northstar operations) on bowheads is the fact that the island is present and active continuously. During previous (1996–98) seismic-monitoring projects, effects of the industrial activities on the bowhead migration corridor were assessed by comparing locations of migrating bowheads during times when the airguns were operating vs. shut down (Miller et al. 1999). This approach, involving within-season comparisons between “seismic” and “control” periods, largely avoided the complications associated with known year- to-year variability in the north–south location of the bowhead migration corridor. Within-season comparisons were expected to be less effective in assessing effects of a stationary facility such as Northstar on the migration corridor, i.e., no “control” periods. Nonetheless, at least during the construction phase, appreciable short-term variability in the industrial sounds was anticipated. This variability was expected 7-6 Monitoring at Northstar: Open-Water Season to provide a basis to determine whether, and to what distance, Northstar sounds affect the distribution of calling whales. In fact, short-term variability in industrial sounds continued in 2002, mainly associated with the varying presence or absence of vessels near Northstar (see Chapter 6). Another potential and recognized complication was the presence, in 2002, of the Steel Drilling Caisson (SDC), which was grounded at the McCovey site about 9 km (5 n.mi.) east-southeast of the DASAR array (see Fig. 7.1). It was moved there on 4 Aug. 2002 and was said to be in a quiet “cold- stacked” mode after 14 Aug. and throughout the 2002 bowhead migration. The Concrete Island Drilling System (CIDS), which had been located 1.1 km (0.6 n.mi.) northeast of Northstar during the 2000 bowhead migration and until 31 Aug. in 2001, was not in the region during 2002.

METHODS

Overview of Methods

The methods used in 2002 paralleled the methods of 2001 (Greene et al. 2002). In general, there were two main components: (1) An array of sensors that provided the data needed to detect and determine the locations of bowhead calls occurring in the southern part of the migration corridor offshore of Northstar. (2) Hydrophones near Northstar that provided a continuous record of variability in the underwater sounds (mainly industrial) at that location. (1) The system for detecting and localizing bowhead whale calls consisted of an array of 11 directional autonomous seafloor acoustic recorders (DASARs). The DASARs were installed on 30 and 31 Aug. 2002 at locations 6.5–22 km (4 to 13.7 mi) offshore of Northstar Island (Fig. 7.1). The DASAR locations were within a few meters of the locations used in 2001. As in 2001, for redundancy, two DASARs were installed at the important central location CA/CB. A 12th DASAR was kept as a ready spare. The DASARs began recording underwater sounds immediately after deployment, and recorded continuously at a sampling rate of 1000 samples/second (per channel) from then until retrieval. Retrieval was on 3 Oct., but the units were overturned by a high storm surge late in the morning of 23 Sept. (see high noise and high wind time on 23 Sept. in Fig. 6.15 A,B in Chapter 6), making later data unusable. The DASARs incorporated components from DIFAR (Direc- tional Frequency and Recording) sonobuoys that permitted calculation of bearings to whale calls. When a call was detected by two or (preferably) more DASARs, the location of the calling whale could be estimated by triangulation. Chapter 8 provides details on the characteristics and operation of DASARs in 2002. (2) Underwater sounds near Northstar needed to be recorded continuously. As in 2001, hydrophones were installed about ¼ mile north of the island. Two cabled hydrophones and (as backup) two ASARs (autonomous seafloor acoustic recorders) were installed and operated for the period 31 Aug.– 1 Oct. A map of the installation locations is presented in Chapter 6 (Fig. 6.2). Both cabled hydrophones operated continuously for the entire period, although cabled hydrophone #1 (C.H. #1) was overturned during the period of storm surge on 23 Sept. The data from C.H. #2 were used; the backup data from C.H. #1 and from the two ASARs were redundant and were not analyzed. Chapter 6 provides additional details concerning the field and analysis methods associated with use of C.H. #2 in 2002. Refinements in Methods during 2002

The methods in 2002 were very similar to those in 2001, but a number of improvements, refinements and year-specific adjustments were introduced. Three problems that were encountered in 2001 were addressed and resolved for the 2002 season: Chapter 7: Acoustic Monitoring of 2002 Bowhead Migration 7-7

FIGURE 7.1. Map of DASAR locations with respect to the north edge of Northstar Island, 2002.

• The DASAR disk-writing failed on occasion in 2001, causing the unit to stop recording. When this problem occurred in 2001, the health checks failed and the field crew could retrieve, repair, and replace the faulty units. In 2002, the disk was given more time to spin up, and error handling for disk- writing was improved, including provision to retry if a failure occurred. There were no health check failures in 2002 once the units were installed in working order. • In 2001, some DASAR sensors leaked at the central wire fitting, causing sensor failure. In 2002, a new attachment method incorporated double O-ring seals, and there were no leakage failures. • The “Dogloo” shelter around each DASAR in 2001 appeared to have an effect on the acoustic performance of the DASARs, especially at the high frequencies used by the transponders for health checks. In 2002, a new aluminum tubular frame was built to mount over the sensor and recorder package (Fig. 7.2), and a thin Spandex “sock” was sewn to be stretched over the frame (Fig. 7.3). Consequently, the transponder health checks operated over distances at least twice those of 2001, and the bearing accuracy was improved for 2002. The following four subsections describe four additional types of refinements or year-specific procedures that were applied during the 2002 fieldwork or analyses: 1. The procedures for calibrating the clocks and directional orientations of the individual DASARs were refined, resulting in improvements in calibration. These improvements are described in the next subsection. 7-8 Monitoring at Northstar: Open-Water Season

FIGURE 7.2. DASAR assembly on deck before attaching the “sock” fairing over the frame.

FIGURE 7.3. DASAR on the deck of an ACS “Bay” boat, ready for installation. The small anchor, tag line on a reel stand, and the DASAR are connected. Chapter 7: Acoustic Monitoring of 2002 Bowhead Migration 7-9

2. A new “gain imbalance” correction factor was developed in 2002 and applied to both the 2002 data and to re-analyses of 2001 data that are described in Chapters 9 and 10 of this report. This new procedure corrects for any gain (sensitivity) imbalance in the two directional sensors of each DASAR, reducing the bearing errors and therefore enhancing the location accuracies. Develop- ment of the gain imbalance correction is summarized below and described in detail in Chapter 8, Appendix 8.1. 3. The definition of the “Industrial Sound Index” (ISI) was adjusted relative to the definition used in 2001 (cf. Greene et al. 2002). The spectral characteristics of the sounds emanating from North- star in 2002 differed somewhat between 2001 and 2002, and the definition of the ISI for 2002 needed to be re-examined and, in fact, changed (see Chapter 6 and below). Although not of direct relevance to this chapter, the definition of the ISI for 2001 has also been re-evaluated (see Appen- dix 6.1). The re-analyses of 2001 data described in Chapters 9–11 of this report define the ISI in the same way as for 2002, and in a somewhat different way than in Greene et al. (2002). 4. The maximum distance at which whale calls were detectable at times of high background noise was re-evaluated based on 2002 data. By limiting the statistical analyses (Chapters 9, 10) to calls within this distance, we avoided a bias that would otherwise have occurred (see below). 1. Time and Direction Calibration Acoustic transmissions from known locations at known times allowed us to correct for slight drift in the clock built into each DASAR, and to determine the directional orientation of each DASAR on the sea-floor (Table 7.1). The model J-9 projector was operated near its maximum source level of 150 dB re 1 µPa at 1 m distance (<0.01 watts of acoustic power). These calibration sounds were weaker than those from powered vessels operating in the area, including (when it was underway) the ACS boat used to deploy the J-9 projector. The transmitted sounds were at frequencies above those used in determining the Industrial Sound Index (ISI, see below). However, times when sounds were transmitted, and adjacent times when the ACS boat was near the DASARs for these calibrations, were not included in the DASAR analysis (see Chapters 9, 10). In 2002, calibration transmissions were done on four dates prior to the date (23 Sept.) when the DASARs became non-functional because of the storm: 2, 7, 13 and 21 Sept. On each occasion, these transmissions were emitted at 18 known locations within and around the array as shown in Figure 7.4. (This procedure was used for most of the 2001 calibration transmissions as well.) The same model J-9 projector, amplifier, sound source, and GPS timing were used in 2002 as in 2001, but the projected waveforms were modified. Instead of separate tones and sweeps, a composite sound was used, consisting of a tone at 400 Hz, a down sweep from 400 to 200 Hz, an upsweep from 200 to 400 Hz, and a repeated pattern of the down and up sweeps. Figure 7.5 is an example of the pressure-amplitude vs. time pattern as received by the three component sensors of one DASAR. Figure 7.6 shows a spectrogram of such a calibration transmission as received by the omnidirectional channel of a DASAR about 3 km (1.6 n.mi.) from the projector. Two seconds separated the tone and the sweep transmissions, as can be seen in the Figures. An entire ping transmission required 22 s, and there were 8 s between the two pings shown. The rationale for this choice of waveform was that the tones provided a well-defined start time, necessary for the DASAR clock calibrations, and the bandwidth of the sweeps provided more accurate bearing measurements than a tone. (Tones suffer “phase fluctuations” during propagation because of slight variations in the combination of multiple propagation paths between source and receiver, while broader-bandwidth sounds propagate with more consistency.) 7-10 Monitoring at Northstar: Open-Water Season

TABLE 7.1. DASAR calibration results for 2002, including reference axis orientation (with correction for gain imbalance) as well as clock drift rate. Original Corrected Original Corrected Scatter DASAR Total Ref Bearing Ref Bearing Std. Dev. Std. Dev. Clock Drift after fit Location Pings (Deg) (Deg) Gain Factor (Deg) (Deg) (Sec/Day) (Seconds) NW 126 89.9 89.5 1.0607 1.50 0.87 -0.4836 0.19 CC 198 248.9 248.8 1.0270 1.70 1.60 -1.3582 0.16 WB 138 34.3 34.0 1.0286 1.30 1.20 -1.7558 0.22 CB 194 157.0 157.0 1.0883 2.30 1.50 -1.1756 0.28 CA 192 186.9 187.0 1.1337 2.60 0.99 0.3827 0.21 EA 139 58.6 57.6 1.1594 2.90 0.93 -1.1898 0.34 WA 165 192.9 192.3 0.8661 3.30 1.30 -1.3838 0.19 SE 127 26.7 26.7 1.0038 0.80 0.80 -1.9505 0.54 SW 33 59.0 59.5 1.0781 3.60 3.30 -0.2753 0.30 NE 135 96.0 95.7 0.9294 2.50 0.80 -1.1723 0.21 EB 154 190.5 190.8 1.0529 1.90 1.60 -0.7723 0.19 Avg: 2.22 Avg: 1.35

FIGURE 7.4. Map of the calibration transmission locations, along with the DASAR and island locations.

2. Gain Imbalance Between the Directional Sensors The DIFAR sensor manufacturer employs careful calibration techniques to equalize the sensitivities of the sensors for the two directional channels. However, the manufacturing specification permits bearing errors of ±10° and we desired better performance. There was evidence in our data that the gains (sensitivities) of the DIFAR sensors were slightly unequal, resulting in bearing errors. A gain-correction Chapter 7: Acoustic Monitoring of 2002 Bowhead Migration 7-11

FIGURE 7.5. Sound pressure graphs, in arbitrary units, vs. time for the omni, sine directional channel, and cosine directional channel of a DASAR sensor receiving a calibration ping.

procedure has been developed in 2002, as described in Appendix 8.1 to Chapter 8. The gain correction derived for each individual DASAR in 2002 is shown in Table 7.1. Application of the gain imbalance correction for 2002 reduced the scatter in reference bearings by an average of 39%. This percentage is derived from the ratio of the average corrected standard deviation to the average original standard deviation [100 × (1 – (1.35º/2.22º))], from Table 7.1. From this we can conclude that a good portion of the DIFAR bearing error appears to be a result of gain imbalance in the directional channels. The gain corrections have been measured for all DASARs and were used in routine whale call processing in 2002. More work needs to be done to understand other sources of error. Transducer tilt, which would be small but a possible source of error nonetheless, is an unexplored possibility. To check the usefulness of the gain corrections and the overall accuracy of the localization process, the software used to determine locations of calling whales was used to calculate the locations where calibration pings were emitted. The actual locations were known from GPS readings. (The errors contributed by GPS uncertainty, perhaps about 10 m, have been neglected. Using a 10-m GPS error and worst-case geometry, the bearing error for our 5000 m baseline would be 2•arctan(10/5000) = 0.23°, which is small 7-12 Monitoring at Northstar: Open-Water Season

FIGURE 7.6. Spectrogram of the omni-channel signal for the reception of two calibration pings. Each transmission begins with a constant 400 Hz tone, followed by a 2-s gap, then a downsweep from 400 to 200 Hz, another 2-s gap, an upsweep from 200 to 400 Hz, a final 2-s gap, and a final downsweep from 400 to 200 Hz. compared to the 1.35° average corrected standard deviation of the calibration bearings.) For most pings, the bearings from the nearest DASARs intersect within a small area around the actual ping location. However, bearings to some of the weaker, more-distant pings still include errors of 5–10º. Bearing accuracy increased with increasing signal-to-noise ratio. The gain-correction procedure had not been developed when the 2001 data were described by Greene et al. (2002). However, this new procedure has now been applied retrospectively to the 2001 data, and the re-analyses of 2001 data described in Chapters 9 and 10 incorporate this improvement. The gain correction derived for each individual DASAR in 2001 is shown in Chapter 8, Table 8.2. 3. Industrial Sound Index (ISI) Objective (3) of the monitoring project was “to estimate the effects of Northstar activities on the migration corridor and calling behavior of bowheads by comparing the distances of calling whales from Northstar at times with varying levels of industrial sound.” For this analysis, a continuous measure of industrial sounds emitted by Northstar was needed. Ideally, this measure would represent Northstar sounds only, without any contribution from natural background sounds associated with wind and waves. However, the sounds received at the hydrophone ~400 m from Northstar included a mixture of industrial and natural sounds. Therefore, an “Industrial Sound Index” or ISI was developed that was primarily an index of the industrial components, with as little contribution from natural background sounds as possible. The ISI, as defined for 2001 and 2002, is an important component of the analyses of whale call locations relative to Northstar sound, as described in Chapters 9 and 10. Chapter 7: Acoustic Monitoring of 2002 Bowhead Migration 7-13

Data for the ISI were acquired from a cabled hydrophone located ~400 m north of the north shore of Northstar Island, as described in Chapter 6. The hydrophone was the same sensor located in essentially the same place as in 2001, and the signal analysis was the same (Greene et al. 2002). Signal analysis results for 2002 are presented in Chapter 6, including the variation in broadband level vs. time (Fig. 6.15) and the variability in spectral densities and one-third octave bands (Fig. 6.19). The ISI was constructed by adding together the sound levels in one-third octave bands that appeared to be dominated by industrial components. The ISI did not consider the sound components in additional one-third octave bands that appeared to be substantially influenced by natural sounds. Chapter 6 presents the rationale for the selection of the specific one-third octave bands to be included in the ISI for 2002. The ISI for 2002 was based on the sum of the mean square pressures in the one-third octave bands centered at 31.5, 40, 50, 63, and 80 Hz. This was necessarily a different combination of one-third octaves than was originally used in deriving the ISI for the 2001 season (Greene et al. 2002), although the definition of the ISI for 2001 has recently been re-considered (see Appendix 6.1). Changes in the predominant components of Northstar sound from 2001 to 2002 would not be surprising given that the predominant power source for the island changed from diesel-electric generators in 2001 to gas turbine generators in 2002. In addition, a higher level of barge traffic was present in 2001, and gas turbine compressors were in operation during 2002 that had not been operational in 2001. Total mean- square sound pressure level in the five one-third octave bands considered in 2002 was computed as

dB31.540506380 dB dB dB dB 10 10 10 10 10 ISI = 10log10  10++++ 10 10 10 10 ,  where dB31.5, dB40, dB50, dB63, and dB80 were mean square sound pressure levels in the five relevant one-third octaves (Richardson et al. 1995, p. 30).

4. High Background Noise Cutoff Distance

As in 2001, a potential source of bias existed in the data because whale calls are more difficult to detect during times of high background noise. This difficulty in detecting calls could introduce bias into the analysis because calls far from the DASARs that would be detectable at times of low noise could go un- detected by the hydrophone system during times with high background noise. If this happened, large distances from shore would be over represented in the data at times of low background noise. This could potentially cause us to conclude that offshore displacement was occurring at such times, and conversely could hide a possible offshore displacement during times with high background noise. To assess this potential bias, we calculated distance from a central location within the DASAR array (“CB” in Fig. 7.1) to each call. These distances were summarized (Fig. 7.7) for times with very high, high, and low levels of background noise, as measured in the 5–500 Hz band at the DASAR farthest from North- star (“NE”). "Very high" and "high" noise conditions were defined to be times with broadband noise levels greater than the 90th and 75th percentiles, respectively, of all the noise levels received at the "NE" DASAR (one measure per 15 min). "Low" noise conditions were defined to be times with broadband noise levels less than the 25th percentile. A cutoff distance was established at a distance where the number of calls detected under high-background-noise conditions dropped appreciably. Based on the “shoulder” at 10 km in Figure 7.7, a 10 km cutoff distance was chosen for 2002. For the main analysis addressing Objective (3) for 2002, i.e., quantile regression of offshore distances of bowhead calls vs. ISI, observed calls were eliminated from analysis if their calculated positions were more than 10 km from the “central axis” of the DASAR array. That axis was the straight line from CB past CC and 7-14 Monitoring at Northstar, Open-Water Season 2002: 90-d report

A. Background > 90-th percentile (101.1 dB) Cut-off = 10km Number of Calls 0 10203040 0 1020304050 Distance from Center of DASAR Array (km)

B. Background > 75-th percentile (95.7 dB) Cut-off = 10km Number of Calls 0 50 100 150 0 1020304050 Distance from Center of DASAR Array (km)

C. Background < 25-th percentile (86.8 dB) Cut-off = 10km Number of Calls 0 50 100 150 200 0 1020304050 Distance from Center of DASAR Array (km)

FIGURE 7.7. Distances from estimated locations of calling whales to a central DASAR (CB) during very high (A), high (B), and low (C) background sound conditions in 2002. A cutoff distance of 10 km was established at the “shoulder” evident in (A) and (B). Background sound was measured at the NE DASAR in the 5– 500 Hz band (see text for further details and rationale). Chapter 7: Acoustic Monitoring of 2002 Bowhead Migration 7-15

SW to Northstar (Fig. 7.1).2 Based on these conditions, we included calls detected within an area consisting of a 20-km-wide rectangle (approximately) extending 16.63 km NNE from Northstar to CB, and topped with a half circle of radius 10 km at the offshore end (see Fig. 7.15, later). In our original analyses of the 2001 data, we used a comparable procedure, but with a 12 km rather than a 10 km cutoff distance. The 12 km cutoff distance was appropriate based on the 2001 data (see Fig. 8.28 in Greene et al. 2002). However, to maintain consistency among years, in our recent re-analyses of 2001 data (Chapters 9–11 of this report), we used the more restrictive 10 km cutoff distance for 2001 as well as 2002. Offshore Distances The offshore distances of whale calls detected in 2002 were measured perpendicular to a baseline approximately parallel to the Beaufort Sea coast and through Northstar Island, as in 2001. These distances were analyzed relative to underwater sound level near Northstar in order to determine whether there was evidence of an offshore deflection of calling whales when Northstar (or vessels near Northstar) produced large amounts of underwater sound. The baseline used for these calculations was oriented 108° right of true north (109.4° relative to our UTM grid orientation). Two UTM points on this line were (0, 7975362.82) and (1000000, 7623202.82). The location of Northstar Island was 70.491581º latitude, 148.693517º longitude (NAD 1927), or WGS- 84 UTM location (436747, 7821558). Assuming (x,y) was the estimated location of a whale call and

(x1,y1) and (x2,y2) were two points on the baseline, the perpendicular distance from the whale call to the line was

2 2 D = (x − x p ) + ( y − y p )

where

x p = x1 + u(x2 − x1 ),

y p = y1 + u( y2 − y1 ),

and

(x − x1 )(x2 − x1 ) + ( y − y1 )( y2 − y1 ) u = 2 2 . (x2 − x1 ) + ( y2 − y1 )

The most appropriate choice of baseline orientation depends on how long a section of coastline one considers, whether one considers the barrier islands, and whether one considers only the coastline or also the bathymetric contours (see, for example, Figure 7.15 [later] vs. Figure 1.1 in Chapter 1). We chose a baseline appropriate in relation to the section of coast from longitude 146ºW to 150.5ºW, a distance of about 180 km (97 n.mi.). This was the same region as used in the analyses of Miller et al. (1999), and we used the same 108º orientation for line parallel to the coast as Miller et al. had selected.

2 This criterion retained in the analysis the very small number of calls detected at locations more than 10 km inshore of CB and close to Northstar, on the rationale that these were the most likely to be affected by Northstar and needed to be retained in the analysis. Most of this area was within 10 km of CC, another “central” DASAR. 7-16 Monitoring at Northstar: Open-Water Season

RESULTS

This section is organized with four subsections describing (1) DASAR installation data, (2) ping calibration measurements, (3) underwater sounds near Northstar, and (4) whale call analysis. In the subsection on the whale call analysis, only an overview of the detected calls is given. Analysis of the offshore distances of the calls in relation to the industrial sound index is presented in Chapter 10. DASAR Installation Data

DASARs were installed and retrieved at times and locations summarized in Table 7.2. Although retrieval was on 3 Oct., the unit at location SW tipped over on 4 Sept. at 05:30 am because of storm surge. All the other DASARs tipped over near 10:00 am on 23 Sept. for the same reason. (See Chapter 6, Fig. 6.15B for the Northstar wind data; the coastal current is wind-driven.) Figure 7.8 shows the operating life of each DASAR. Ping Calibration Measurements

Ping calibration transmissions were made on 2, 7, 13, 21 and 29 Sept. However, the transmissions on 29 Sept. occurred after the DASARs tipped over. Table 7.1 presents the results of the reference axis determination for each DASAR, including the correction for gain imbalance. Also shown in Table 7.1 is the correction for clock drift as determined from the calibration data. Time Calibration The clocks in the individual DASARs drifted slightly, and it was important to characterize and correct for this drift in order to synchronize the data from the various DASARs. A least-squares fit of the ping time error versus GPS time, done separately for each DASAR, gave a well defined slope and intercept for each DASAR. Clock drift for the various DASARs ranged from –1.9505 to +0.3827 seconds per day. This enabled a simple calculation of the time correction required in order to estimate the GPS time associated with each whale call detected by a given DASAR. We estimate that, after application of the correction factors for clock drift, the GPS times of most whale calls were estimated within 2 s. This allowed unambiguous matching of calls received by one DASAR with those received at corresponding times by other DASARs. Because each DASAR provides all the data necessary to determine the bearing to a calling whale, there was no requirement for more precise synchronization of the type necessary to localize whale calls based on differences in time-of-arrival at various widely-spaced hydrophones. Reference Axis Calibration The data in Table 7.1 show a maximum standard deviation of 3.3º (unit SW) for the bearings to a source at known locations after applying the correction for gain imbalance. Unit SW tipped over on 4 Sept. so only the 33 pings received on 2 Sept. were useful for bearing calibration. The corresponding standard deviations for the remaining 10 DASARs were from 0.8º to 1.6º, based on much larger numbers of calibration pings on four dates (Table 7.1). Chapter 7: Acoustic Monitoring of 2002 Bowhead Migration 7-17

TABLE 7.2. DASAR locations in 2002, with installation, retrieval and operation timing. See Figure 7.1 for map of DASAR locations. Note that the “Data End” times are shut down times and do not reflect that 10 of 11 DASARs tipped over at about 10:00 am on 23 Sept. SW had tipped over on 4 Sept.

DASAR Installed Recovered Data Start Data End UTM X UTM Y Depth* NW 30 Aug 2002 11:15 03 Oct 2002 12:07 30 Aug 2002 11:25 03 Oct 2002 11:50 443375 7839582 23.8 CC 30 Aug 2002 12:11 03 Oct 2002 14:33 30 Aug 2002 12:25 03 Oct 2002 14:10 443569 7830916 19.2 WB 31 Aug 2002 12:51 03 Oct 2002 15:05 31 Aug 2002 13:00 03 Oct 2002 14:40 438248 7830851 18.6 CB 02 Sep 2002 11:39 03 Oct 2002 14:10 02 Sep 2002 11:50 03 Oct 2002 13:05 446382 7835337 22.9 CA 30 Aug 2002 11:51 03 Oct 2002 13:55 30 Aug 2002 12:06 03 Oct 2002 13:10 446016 7835281 22.9 EA 31 Aug 2002 13:49 03 Oct 2002 11:15 31 Aug 2002 15:05 03 Oct 2002 10:40 450745 7835238 23.5 WA 30 Aug 2002 11:34 03 Oct 2002 13:03 30 Aug 2002 11:45 03 Oct 2002 12:56 441012 7835221 21.6 SE 31 Aug 2002 12:12 03 Oct 2002 09:40 31 Aug 2002 12:33 03 Oct 2002 09:10 445745 7826560 15.8 SW 31 Aug 2002 12:36 03 Oct 2002 15:31 31 Aug 2002 12:49 04 Sep 2002 05:00 440789 7826545 15.8 NE 30 Aug 2002 10:56 03 Oct 2002 11:45 30 Aug 2002 11:16 03 Oct 2002 11:25 448295 7839636 22.9 EB 30 Aug 2002 12:30 03 Oct 2002 10:10 30 Aug 2002 12:46 03 Oct 2002 09:53 448577 7830924 20.0 * Water depth in meters

FIGURE 7.8. The useful life span of each DASAR in 2002.

The omnidirectional channel in unit WA was not useful. A pseudo omni-channel was synthesized from the data in the two directional channels, but the resulting bearings had a 180º ambiguity. For both the calibration pings and the whale calls, bearings from other DASARs permitted us to resolve the ambiguity. When this was done, the standard deviation (after application of the gain correction) was satisfac- tory (1.3º), so bearings from unit WA were used for whale call location analysis. 7-18 Monitoring at Northstar: Open-Water Season

Underwater Sounds near Northstar

Temporal Variation The signals from cabled hydrophone #2 (C.H. #2), located 417 m seaward of the north shore of the island, were analyzed to determine the broadband (10–1000 Hz) level of underwater sound based on a one-minute analysis every 4.37 minutes. The overall results for the season are shown in Figure 6.15A of Chapter 6. Most of the “spikes”, i.e., brief periods with notably higher sound levels, are attributable to movements of the crew boat and tugs coming to and going from the island. The lowest levels in the graph are indicative of the quietest times in the water near the island. Figure 7.9 shows a sample of the same data limited to the two days, 13 and 14 Sept. Those two days included times with high detection rates for bowhead whale calls within the 31 Aug.–23 Sept. period. The nature of the sound spikes is revealed in these graphs. The spikes correspond to the arrival and departure of vessels traveling near Northstar, often those traveling back and forth from West Dock. (Figure 6.16 and associated text in Chapter 6 provides additional evaluation of the specific sources of the “spikes” of vessel sound evident on these dates.) Later in this chapter, the whale call locations will be presented for these same two days as examples of the localization results. Spectral Characteristics To characterize the sounds near Northstar during the study period in 2002, narrowband and one- third octave band levels were calculated for cabled hydrophone #2 on a systematic basis over the operating life of the unit. For the period 31 Aug.–1 Oct. 2002, 10,192 spectrum analyses were calculated. Of these, 7551 were within the 31 Aug.–23 Sept. 2002 period corresponding to times when whale calls were localized. Each spectrum analysis represented the average of sounds over a 1-min period starting 4.37 min after the start of the last such period. For each frequency, the 7551 levels for 31 Aug.–23 Sept. were sorted from smallest to largest in order to determine the minimum, 5th-percentile, median, 95th- percentile, and maximum level. Those values, when plotted, resulted in the five narrowband statistical spectra shown in Figure 6.19A of Chapter 6. In addition, for each 1-min analysis period, one-third octave band levels were computed by integrat- ing over the appropriate frequency cells for each standard center frequency from 10 through 800 Hz. These results were sorted in a manner similar to that applied to the narrowband spectra, resulting in the five one- third octave statistical spectra shown in Figure 6.19B of Chapter 6. The bandwidth of each one-third octave band is 23% of the center frequency, so the one-third octave bandwidths increase with increasing frequency. This accounts for the fact that narrowband spectra tended to diminish with increasing frequency above ~50 Hz, but one-third octave spectra did not. There were no dominant peaks in the one-third octave spectra. There was a rolloff (decrease) in band levels for bands centered at frequencies below 31.5–50 Hz as compared with the bands at somewhat higher frequencies. This rolloff probably comes from a combination of (1) lower levels of low-frequency sound at the sources on and near the island, and (2) waveguide cutoff over the c. 400-m shallow-water path from the island to the hydrophone. The progressively narrower bandwidth of one-third octave bands as center-frequency decreases is also a factor. Industrial Sound Index (ISI) The Industrial Sound Index was described in Methods, and the rationale for its selection and definition are provided in Chapter 6. Some qualities of the ISI are presented here. To review, the ISI for 2002 is the level of sound in the combined one-third octave frequency bands centered at 31.5, 40, 50, 63 and 80 Hz. Figure 7.10A shows, for 20 Sept., the full broadband level (10–1000 Hz) vs. time of day. Chapter 7: Acoustic Monitoring of 2002 Bowhead Migration 7-19

FIGURE 7.9. Broadband (10–1000 Hz) level of underwater sound 417 m north of Northstar vs. local time for (A) 13 Sept. 2002, and (B) 14 Sept. 2002. The spikes correspond to boat arrivals and departures. 7-20 Monitoring at Northstar: Open-Water Season

140

A 130 Pa) µ

120

110

100

10-1000 Hz Band Level (dB re 1 re (dB Hz Band Level 10-1000 90

80 0 4 8 12 16 20 24 Time on 20 Sept. 2002

140

B 130 Pa) µ 120

110

100 One Minute ISI Level (dB re 1 re (dB ISI Level Minute One 90

80 0 4 8 12 16 20 24 Time on 20 Sept. 2002

140

C 130 Pa) µ

120

110

100

Hourly Average ISI Level (dB re 1 90

80 0 4 8 12 16 20 24 Time on 20 Sept. 2002

FIGURE 7.10. Three measures of underwater sound levels 417 m north of Northstar vs. local time for one day, 20 Sept. 2002. All three panels are computed from one-minute averages calculated every 4.37 min. (A) shows the broadband (10-1000 Hz) level, (B) shows the 1-min averaged ISI values computed every 4.37 min, and (C) shows the running average of 14 one-minute averages from the hour preceding the specified time. Chapter 7: Acoustic Monitoring of 2002 Bowhead Migration 7-21

Each value plotted is an average level over a one-minute period, with one such value having been computed for every 4.37 minute interval. Figure 7.10B shows the ISI calculated at the same times, again with one-minute averaging, and Figure 7.10C shows the one-hour moving average of the same one- minute ISI values (from Fig. 7.10B). The smoothing effect of the moving average is manifest in the stretched-out durations of boat operations that account for the spikes in the broadband and one-minute ISI graphs. It is these hourly ISI averages (and other averages over periods shorter and longer than 1 hour) that are used in assessing the influence of Northstar sound on whale call locations. The ISI is a subset of the full 10–1000 Hz band and it never has levels greater than those for the full 10–1000 Hz band. Another comparison of the ISI with broadband sound is by histograms for the entire operating period, 31 Aug.–23 Sept., presented in Figure 7.11. The three histograms compare the distribution of 1-minute and hourly averaged ISI values for comparison with the 10–1000 Hz broadband levels, considering the period from 31 Aug. to 23 Sept. 2002. Again, the ISI should never have levels greater than the full 10–1000 Hz band. For this full 23-day operating period, the average level in the 10–1000 Hz band was 4.33 dB greater than the average of the one-minute 5-band ISI, with a standard deviation of 2.57 dB and a standard error of the mean = 0.03 dB. Whale Call Analysis

Number and Locations of Whale Calls Detected

Figure 7.12 shows the numbers of bowhead calls detected per hour for the period from 31 Aug. to mid-day on 23 Sept. 2002. A whale call detected at multiple DASARs was counted once. Altogether, 10,587 calls were detected during this period. Of these calls, 2022 (19%) were detected by only one DASAR and thus could not be localized, and 643 (6%) of those detected by 2 or more DASARs also could not be localized.3 All of the remaining 7922 calls (75% of 10,587) yielded usable locations. Measured locations of origin for bowhead calls detected on two days, 13 and 14 Sept. 2002, are mapped in Figures 7.13 and 7.14 respectively. The symbol types discriminate calls whose locations were estimated based on bearings from two (+), three (∇), four (), or more (Ο) DASARs. In comparison with the two specific days’ data presented for 2001 (Greene et al. 2002), a higher proportion of the detected calls were localized in 2002. On 13 Sept., 593 of 701 calls were detected by at least two DASARs, the minimum necessary to estimate the whale location. Corresponding figures for 14 Sept. were 350 of 392. For the full 2002 study period, the majority of the located calls (74% = 5837 / 7922) were detected by 3 or more DASARs, with 1429 being detected by 3 DASARs, 1129 by 4 DASARs, and 3279 by 5+ DASARs. This would be expected to provide more accurate locations than obtained in 2001, when the majority of the calls were detected by only one DASAR, and of those detected by 2+ DASARs, the majority were detected by only two units (Greene et al. 2002). (The following subsection summarizes the accuracy of the locations in 2002 relative to distance from the array of DASARs.) Figure 7.15 plots the estimated locations of 4997 bowhead whale calls that were detected in the 31 Aug.–23 Sept. 2002 period. (Overall, 7563 calls were localized, but 2566 of these were estimated to be outside the boundaries of the area mapped in Fig. 7.15.) As in 2001, calling whales were most frequently

3 Of the 643 calls (6%) detected by 2 or more DASARs that could not be localized, 182 (2%) did not yield a location because the maximum likelihood localization procedure did not converge, and 461 (4%) yielded a location but the variance–covariance matrix of the solution was not positive definite. See Chapter 8 for descriptions of the localization process and of these special situations for which the locations could not be determined reliably. 7-22 Monitoring at Northstar: Open-Water Season

A. Broad band (10-1000 Hz) 0 400 1000

B. 1-min ISI (5-bands) 0 400 1000

Number of observations C. 60-min ISI (5-bands) 04080 90 100 110 120 130 Sound level (dB re 1 µ Pa)

FIGURE 7.11. Three histograms of broadband (10-1000 Hz) and industrial sound index (ISI) values for Cabled Hydrophone #2 located 417 m north of Northstar Island. The period spanned is from about noon on 31 August until 10 am on 23 September 2002. The ISI measures the cumulative sound level in five one-third octave bands centered at 31.5, 40, 50, 63, and 80 Hz. Each 1-min ISI value is a 1-minute average every 4.37 minutes. Each 60-min ISI value is an average of the 14 1-minute ISI values occurring during 1 hour between 31 August and 23 September. (This is not a running average like the one in Fig. 7.10C.) Vertical lines show the respective 5th, 50th, and 95th percentiles of each distribution (broadband = 94.8, 103.0 and 117.1 dB re 1 µPa, 1-min ISI = 90.9, 99.1, and 109.2 dB, and the 60-min ISI = 91.8, 99.6, and 106.0 dB).

detected 15 km or more offshore of Northstar, i.e., within, near and beyond the offshore half of the DASAR array. Only a few whale calls were detected within 10 km of Northstar Island in either year. The color coding of the whale-call symbols in Figure 7.15 shows the average Industrial Sound Index (ISI) near Northstar during the 60-minute period preceding detection of that call. Chapter 10 analyzes the relationship between the offshore distances of the whale calls and the then-prevailing ISI. Those analyses considered only the 2057 calls within the “analysis area” that is outlined on the map, defined as described in the Methods. That area consists of a rectangle extending 10 km WNW and 10 km ESE from the “central axis” of the array, and from shore to the CB DASAR (one of the two closely-spaced ones), capped by a semicircle 10 km in radius. Chapter 7: Acoustic Monitoring of 2002 Bowhead Migration 7-23

FIGURE 7.12. Numbers of bowhead whale calls detected per hour over the period 31 Aug.–23 Sept. 2002.

Accuracy of Whale Call Localization The calibration information collected in 2002 (Table 7.1) provided a way to estimate the uncertainty in our estimate of the location of origin of each whale call (see Chapter 8 for procedure). For each call, the uncertainty was expressed in terms of a 90% confidence ellipse: we are 90% confident that the call in question originated inside the bounds of the 90% confidence ellipse. In this subsection, we summarize the average sizes and shapes of the confidence ellipses for bowhead calls detected in various regions within and outside the array of DASARs. It was to be expected that uncertainty would increase with increasing distance from the sensors, given the nature of triangulation. For calls determined to be within the perimeter of the DASAR array, the average 90% confidence ellipse was generally less than 10 ha in size and approximately circular in shape, i.e., with major axis length ≈ minor axis length. (A 10-ha circle has a diameter of 357 m or 1170 ft.). For estimated locations within the perimeter of the array, the length of the major axis of the 90% confidence ellipse averaged 396 m. For estimated locations within 10 km of the center of the array (~5 km outside the perimeter), the corresponding average was 637 m. However, there were small areas within 10 km of the center, notably outside the southeast and northwest sides of the array, where the uncertainty was greater. Beyond about 10 km from the center of the array, localization accuracy deteriorated rapidly (Fig. 7.16). Outside the perimeter of the array, the major axes of the confidence ellipses were generally oriented parallel to rays emanating from the center of the DASAR array, and grew in length rapidly at locations far from the array (Fig. 7.17). 7-24 Monitoring at Northstar: Open-Water Season

FIGURE 7.13. Bowhead whale call locations for the 24 hours of 13 Sept. 2002. Symbol types discriminate calls whose locations were estimated based on bearings from two (+), three (∇), four (), or more (Ο) DASARs. On this date, 149 calls were detected with two DASARs, 117 with three DASARs, 100 with four DASARs, and 227 with five or more DASARs. Locations of those detected by a single DASAR (n = 108) could not be determined.

DISCUSSION

Ping Calibrations

The DIFAR ping processing resulted in bearing measurements with some scatter (measurement error), as expected. However, this variability included an apparent systematic component related to the difference between the transmission azimuth (bearing from DASAR to source location) and the orientation of the DASAR’s reference axis. One possible source of such variation is gain imbalance, a difference in sensitivity between the two azimuthal directional sensors (see Appendix 8.1 in Chapter 8). Calculated corrections for the gain imbalance in each DASAR reduced the spread of DIFAR bearings from the ping calibrations. The uncorrected standard deviations listed in Table 7.1 averaged 2.22º; the corrected standard deviations averaged 1.35º. Chapter 7: Acoustic Monitoring of 2002 Bowhead Migration 7-25

FIGURE 7.14. Bowhead whale call locations for the 24 hours of 14 Sept. 2002. Plotted as in Figure 7.13. On this date, 115 calls were detected with two DASARs, 65 with three DASARs, 55 with four DASARs, and 115 with five or more DASARs. Locations of those detected by a single DASAR (n = 42) could not be determined.

These bearing uncertainties are far less than those measured in 2001, when the standard deviation of the (uncorrected) values averaged 6.7º (from Table 8.1 in Greene et al. 2002). The gain imbalance correction accounts for part of the improvement. A second source of improvement is that only the closest calibration pings, and therefore only those with strongest signal-to-noise ratios, were used in calculating the DASAR reference axis directions in 2002, whereas in 2001 all detected calibration pings were used. In 2001, weaker signal-to-noise ratios resulted in higher bearing variability, especially if the background noise included directional components. Distant vessels and the island itself are potential sources of directional sounds. Third, the reference bearing measurements in 2002 were based on the sweep signals, not tones. Tones are subject to “phase fluctuations” in underwater sound propagation. Such fluctuations are considerably reduced when averaged over a number of frequencies, which is what happens in the reception of the sweep signals. Thus, a sweep is a better signal for bearing calibrations than is a tone. In 2001, both sweeps and tones were used, but in 2002, sweeps were used exclusively. 7-26 Monitoring at Northstar: Open-Water Season

FIGURE 7.15. Map of whale call locations within and near the “analysis area” in 2002. Whale calls are color-coded based on the average ISI ~400 m from Northstar during 1 h preceding the call. Chapter 7: Acoustic Monitoring of 2002 Bowhead Migration 7-27

(A)

(B)

FIGURE 7.16. Mean length (in kilometers) of major axes of 90% confidence ellipses around estimated locations shown as (A) smoothed surface, (B) contour plot. View in (A) is (approximately) from Northstar, looking northeast. Filled circle represents Northstar Island, open inverted triangles represent DASARs, and solid black line is a baseline through Northstar parallel to the coast. 7-28 Monitoring at Northstar: Open-Water Season

x 106 7.87

7.865

7.86

7.855

7.85

7.845 Northing 7.84

7.835

7.83

7.825

7.82 4.3 4.35 4.4 4.45 4.5 4.55 4.6 4.65 4.7 4.75 4.8 Easting x 105

FIGURE 7.17. Mean angle and length of major axis of 90% confidence ellipse around estimated locations. Each arrow is in the direction of the confidence ellipse’s major axis, and its length is proportional to the length of the major axis. Filled circle represents Northstar Island, open inverted triangles represent the DASARs, and solid black line is a baseline through Northstar parallel to the coast.

Underwater Sounds near Northstar

Figure 6.15A (in Chapter 6) and Figure 7.9 present records of broadband (10–1000 Hz) underwater sounds received by cabled hydrophone #2 located about 400 m north of Northstar. The pattern corresponds well with that shown by the corresponding 2001 Figures (Fig. 8.11 and 8.12 in Greene et al. 2002). The spikes in Figure 6.15A, expanded in Figure 7.9, correspond to boat arrivals and departures from the island, primarily the crewboats, just as they did in 2001. Chapter 6 includes a detailed description of the sounds from an arriving and departing crewboat. A notable difference between the 2001 and 2002 broadband levels is that, in 2001, the minimum levels extended down to 80 dB re 1 µPa. In 2002 the minimum levels extended down to 90 dB. The same cabled hydrophone #2 and recorder electronics were used both years, and the hydrophone location in 2002 was within 20 m (66 ft) of its 2001 location—a negligible difference in relation to its 400 m (1300 ft) distance from the island edge. Either the minimum sea noise was higher by about 10 dB in 2002, the minimum sounds from the island were higher by about 10 dB in 2002, or there was an unknown source of additional sound in 2002. The times in 2002 with minimum wind speed were also the times with minimum sound levels. Because minimum levels of sea noise (in the absence of vessels or other man-made sounds) are wind dependent, it is unlikely that the minimum sea noise was higher in 2002. A more likely explanation is that there was a change in the sounds emanating from the island. Drilling, oil production, and gas compression and injection were all occurring in September 2002, whereas none of these activities occurred in September 2001. Also, the power source for the island changed from diesel-powered generators in 2001 to gas-turbine generators in 2002. Some combination Chapter 7: Acoustic Monitoring of 2002 Bowhead Migration 7-29 of these changes in typical activities on the island was probably the reason for the increase in the minimum sound levels from 2001 to 2002. The industrial sound index (ISI) in 2002 was defined by the five one-third octave bands with center frequencies at 31.5 through 80 Hz. In 2001, it seemed appropriate to use three one-third octave bands, those centered at 63, 80 and 160 Hz (Greene et al. 2002). The character of the island sounds appeared to have changed between the two years. Over the 24 days of sound monitoring, the ISI averaged 4.33 dB less than the level in the full 10–1000 Hz band, with a standard deviation of 2.57 dB. In 2001, the average difference was 5.8 dB, understandably larger because the ISI was based on a smaller range of frequencies in 2001 than in 2002. Whale Calls and Locations

In 2001, 10,750 whale calls were detected during a 36-day period (Greene et al. 2002), of which 3153 yielded usable locations. In 2002, 10,587 calls were detected during 24 days of operation, of which 7922 were usable. The higher proportion of whale calls located in 2002 (75%) than in 2001 (29%) represents a major increase in the efficiency and effectiveness of the project in 2002. Also, a much higher proportion of the localized calls were detected by >2 DASARs in 2002, resulting in improved location accuracy. Location uncertainty (as measured by major axis length) was much larger in 2001 (see Fig. 9.5 in Chapter 9) as in 2002. In some locations, error vectors in 2002 were an order of magnitude less in length than they were at the same location in 2001. The improvements in precision are attributable to a combination of factors, including better bearing accuracies and better whale call reception in 2002, perhaps because the DASARs were housed inside nylon socks in 2002 instead of Dogloos as in 2001. Also, ten DASARs (at nine locations) provided usable data at most times during the 2002 field season (Fig. 7.8). In 2001, ten DASARs (at nine locations) were functional for the first few days of operations, but the number of functional DASARs was reduced later in the season (Fig. 8.14 in Greene et al. 2002). Collectively, these conditions resulted in detecting and localizing more calls with higher accuracy in 2002 than in 2001. In 2002, the high background noise cutoff distance was 10 km (Fig. 7.7), while in 2001 the corresponding cutoff distance was 12 km. It is likely that higher background sound levels in 2002 relative to 2001 caused the difference between 10 km and 12 km. However, it is also possible that the cutoff distance was estimated with low precision in both years and that the change from 12 km to 10 km was attributable to random chance. The decrease in the cutoff distance resulted in a smaller “analysis area” in 2002 than in our original 2001 analyses, and this change resulted in a larger proportion (and total number) of calls being excluded from the statistical analysis in 2002 than in the original 2001 analyses. The 2001 data have been re-analyzed for this report using the same procedures as are applied to 2002 (Chapters 9–11). In the re-analysis of the 2001 data, we limited the analysis area to the same size as that used for the 2002 analyses, i.e. based on the 10-km cutoff distance. Overall, the acoustic monitoring effort in 2002 involved various improvements relative to 2001, and the whale call data obtained in 2002 were more numerous and more precise than those obtained in 2001. The increased sample size in 2002 was especially noteworthy in that it occurred despite the fact that the effective field season was shorter in 2002 than in 2001 (23 days vs. 36 days). (The data collected after 23 Sept. in 2002 were not useable because the instruments were disrupted by a storm.) The acoustical monitoring approach continues to be an effective method for documenting the bowhead whale migration past Northstar during late summer and early autumn. It can provide a large sample of data on whale locations, on a continuous basis, over much of the whale migration season. The 2002 data described here are further analyzed in Chapter 10 to 7-30 Monitoring at Northstar: Open-Water Season evaluate whether underwater sounds from activities at Northstar had an influence on bowhead whales in the southern part of their migration corridor during 2002.

ACKNOWLEDGEMENTS

Alaska Clean Seas crews on their Bay-class boats made the ocean work near Northstar Island and at the DASAR array during 2002 safe and successful. Mark Theriault, Gene Pennington, Charles Crabaugh, Gary Seims, Karl Pulliam, Willy Nye and Greg Ronne were particularly helpful. ACS supervisors and coordinators Barron Rutherford, Eric Olsen, Jim Nevels, and Fred McAdams made the scheduling and personnel assignments work. Jonah Leavitt of Barrow helped during field preparations and sea work. Bill Burgess of Greeneridge completed the monitoring tasks at Northstar and at the DASAR array, including retrieval. Bob Norman and Pat Dexter engineered the DASAR re-design and fabrication essential for the success of the instruments in 2002. Kristin Otte, Jane Fleischman, Dorothy Parker, Leslie Lane and Debbie Martinez of Greeneridge performed the whale call localization analysis. Bob Blaylock, Greeneridge, prepared many of the Figures for this chapter. Anne Wright of LGL assisted with report production. Wilson Cullor and Allison Erickson of OASIS Environmental provided helpful logistical connections for the Prudhoe Bay field work. Dr. Ray Jakubczak and Dr. Bill Streever of BP Alaska supported the project, and numerous participants in the peer/stakeholder group convened by NMFS (see specific Acknowledge- ments in Chapter 1) provided guidance and support at earlier stages. Dr. Streever, along with R. Suydam of the North Slope Borough Dept. of Wildlife Management, provided valuable review comments on this report. We thank them all.

LITERATURE CITED

Blackwell, S.B. and C.R. Greene, Jr. 2001. Sound measurements, 2000 break-up and open-water seasons. p. 7-1 to 7-55 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of industrial sounds, seals, and whale calls during construction of BP’s Northstar Oil Development, Alaskan Beaufort Sea, 2000. [April 2001 ed.] LGL Rep. TA2429-2. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 316 p. Blackwell, S.B. and C.R. Greene Jr. 2002. Sound measurements, 2001 open-water season. p. 7-1 to 7-39 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of industrial sounds, seals, and whale calls during construction of BP's Northstar oil development, Alaskan Beaufort Sea, 2001. [Oct. 2002 ed.] LGL Rep. TA2572-2. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 309 p. Davis, R.A., C.R. Greene Jr., and P.L. McLaren. 1985. Studies of the potential for drilling activities on Seal Island to influence fall migration of bowhead whales through Alaskan nearshore waters. Rep. from LGL Ltd., King City, Ont, for Shell Western E&P Inc., Anchorage, AK 99501. 76 p. Greene, C.R., Jr., M.W. McLennan, T.L. McDonald and W.J. Richardson. 2001. Acoustic monitoring of bowhead whale migration, autumn 2000. p. 9-1 to 9-57 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of industrial sounds, seals, and whale calls during construction of BP’s Northstar Oil Development, Alaskan Beaufort Sea, 2000. [April 2001 ed.] LGL Rep. TA2431-2. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 316 p. Greene, C.R., Jr., M.W. McLennan, T.L. McDonald and W.J. Richardson. 2002. Acoustic monitoring of bowhead whale migration, autumn 2001. p. 8-1 to 8-79 In: W.J. Richardson and M.T. Williams (eds.), 2002. Monitoring of industrial sounds, seals, and whale calls during construction of BPs Northstar Oil Development, Chapter 7: Acoustic Monitoring of 2002 Bowhead Migration 7-31

Alaskan Beaufort Sea, 2001. [Oct. 2002 ed.] Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 309 p. Greene, C.R.,Jr., M.W. McLennan and S.B. Blackwell. 2003. Acoustic monitoring of bowhead whale migration, autumn 2002. p. 3-1 to 3-30 In: W.J. Richardson (ed.), Monitoring of industrial sounds and whale calls during operation of BP's Northstar oil development, Alaskan Beaufort Sea, summer and autumn 2002: 90-day report. LGL Rep. TA2706-1. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 66 p. Superseded by the present chapter. Hickie, J. and R.A. Davis. 1983. Distribution and movements of bowhead whales and other marine mammals in the Prudhoe Bay region, Alaska 26 September to 13 October 1982. p. 84-117 In: B.J. Gallaway (ed.), Biological studies and monitoring at Seal Island, Beaufort Sea, Alaska 1982. Rep. from LGL Ecol. Res. Assoc. Inc., Bryan, TX, for Shell Oil Co., Houston, TX. 150 p. Johnson, S.R., C.R. Greene, R.A. Davis and W.J. Richardson. 1986. Bowhead whales and underwater noise near the Sandpiper Island drillsite, Alaskan Beaufort Sea, autumn 1985. Rep. from LGL Ltd., King City, Ont., for Shell Western E & P Inc., Anchorage, AK. 130 p. LGL and Greeneridge. 2000. Technical plan for marine mammal and acoustic monitoring during construction of BP’s Northstar oil development in the Alaskan Beaufort Sea, 2000-2001 (revised 30 Aug. 2000). Rep. from LGL Alaska Res. Assoc. Inc. (Anchorage, AK), LGL Ltd. (King City, Ont.) and Greeneridge Sciences Inc. (Santa Barbara, CA) for BP Exploration (Alaska) Inc., Anchorage, AK. 80 p. Miller, G.W., R.E. Elliott, W.R. Koski, V.D. Moulton and W.J. Richardson. 1999. Whales. p. 5-1 to 5-109 In: W.J. Richardson (ed.), Marine mammal and acoustical monitoring of Western Geophysical’s open-water seismic program in the Alaskan Beaufort Sea, 1998. LGL Rep. TA2230-3. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for Western Geophysical, Houston, TX, and U.S. Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 390 p. Moore, S.E., D.K. Ljungblad and D.R. Schmidt. 1984. Ambient, industrial and biological sounds recorded in the northern Bearing, eastern Chukchi and Alaska Beaufort Seas during the seasonal migrations of the bowhead whale (Balaena mysticetus), 1979-1982. Rep. from SEACO Inc., San Diego, CA, for U.S. Minerals Manage. Serv., Anchorage, AK. 111 p. NTIS PB86-168887. Richardson, W.J., C.R. Greene Jr., C.I. Malme, and D.H. Thomson. 1995. Marine mammals and noise. Academic Press, San Diego, CA. 576 p. CHAPTER 8: USING DIFAR SENSORS TO LOCATE CALLING BOWHEAD WHALES AND MONITOR THEIR MIGRATION1

by Charles R. Greene Jr. a , Miles Wm. McLennan a, Robert G. Norman a, Trent L. McDonald b and W. John Richardson c

aGreeneridge Sciences, Inc. 1411 Firestone Rd., Goleta,, CA 93117 (805) 967-7720; [email protected]

bWestern EcoSystems Technology, Inc. 2003 Central Ave., Cheyenne, WY 82001 (307) 634-1756; [email protected]

cLGL Ltd., environmental research associates 22 Fisher St., POB 280, King City, Ont. L7B 1A6 (905) 833-1244; [email protected]

for

BP Exploration (Alaska) Inc. Dept. of Health, Safety & Environment 900 East Benson Blvd, P.O. Box 196612 Anchorage, AK 99519-6612

LGL Report TA2706-3

December 2003

1 Chapter 8 In: W.J. Richardson and M.T. Williams (eds.). 2003. Monitoring of industrial sounds, seals, and bowhead whales near BP’s Northstar oil development, Alaskan Beaufort Sea, 1999–2002. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 8-2 Monitoring at Northstar: Open-Water Season

ABSTRACT

The purpose of this chapter is to describe the acoustic methods used to determine locations of calling bowhead whales in the southern part of their migration corridor near Northstar. Some results are presented for illustrative purposes, but Chapter 10 is the primary chapter for results. This chapter and Chapter 7 overlap to some extent, the result of writing this chapter to be suitable for journal publication. Bowhead whales, Balaena mysticetus, migrate west during fall in a corridor ~5–40 n.mi. (10– 75 km) off the north coast of Alaska, passing the petroleum developments around Prudhoe Bay during September and early-mid October. In 2000, BP Exploration (Alaska) Inc. constructed an artificial gravel island 5 km (~2.7 n.mi.) offshore from the nearest barrier island for oil production. Operations on the island create sounds that propagate north into the ocean. The southern part of the bowhead migration corridor is exposed to these sounds. BP needed to determine whether these sounds deflected migrating whales farther offshore. Bowheads call during fall migration, and an acoustical approach to localizing calling whales was selected in preference to aerial surveys. The specific technique chosen was to incorporate DIFAR (Directional Frequency and Recording) sonobuoy techniques into autonomous seafloor recorders, designing and building an array of 11 DASARs (Directional Autonomous Seafloor Acoustic Recorders). When two or (preferably) more DASARs detected the same call, the location of the whale was determined from the intersection of the bearings. The purpose of the present chapter is to describe the specialized acoustic methods used to determine the locations of the calling bowhead whales, and to show the types and precision of the data acquired. DIFAR uses an omnidirectional pressure hydrophone and two horizontal, orthogonal particle- velocity hydrophones having dipole patterns. The relative amplitudes of a sound signal at the two dipoles provide a measure of the angle to the source, while the relative phases of the dipole signals with respect to the omni signal resolve the ambiguities. The digital recording system was based on a Persistor Instruments single-board computer and a 30-GByte disk drive. An acoustic transponder provided a way to check the operational health of each deployed DASAR. Batteries were sufficient to record for 45 days. Each unit was installed with a tag line on the bottom between the unit and an anchor, both of whose positions were measured by GPS during deployment. No surface buoys were used because of hazards from drifting ice floes. The array of 11 DASARs was configured as two overlapping hexagons in which the unit-to-unit separation was 5 km (2.7 n.mi.). Calibration transmissions from a J-9 sound projector at GPS-measured locations within and around the array provided measures of the individual DASAR clock drift (for timing whale call arrivals) and reference axis direction (for relating measured whale call bearings to grid north). Standard errors of the bearing measurements to the J-9 projector at distances of 3–4 km (~2 n.mi.) was 2.22º without correcting for gain imbalance in the two directional sensors, and 1.35º with corrections. During 23 days in 2002, 10,587 bowhead calls were detected and 8383 were localized. Of those, 7922 were deemed sufficiently accurate to use in subsequent analyses, with 5837 of those being detected by three or more DASARs. Uncertainty in locations varied with distance and other factors; 90% confidence ellipses were calculated to have an average major axis length of 500–1000 m at distances 10 km from the array center. Chapter 8: DIFAR Sensors for Monitoring Bowhead Migration 8-3

INTRODUCTION

Background

Whales use sound to communicate and to acquire information about their environment. Also, they often react to man-made sounds by changing their behavior and swimming away, or occasionally toward, the sound source (Richardson et al. 1995, 1999). Beginning in late August and continuing past mid- October, bowhead whales (Balaena mysticetus) migrate west across the Beaufort Sea, swimming parallel to the north coast of Alaska (Moore and Reeves 1993). The majority of the whales travel along a corridor 5–40 n.mi. (10–75 km) offshore, with the southern part of that corridor being through water <25 m deep. Alaska Eskimos have traditionally hunted these whales for subsistence. The whalers are concerned about any industrial activity that might result in the bowheads swimming farther offshore and therefore being more difficult to hunt and retrieve. The whalers are also concerned that whales, once deflected, may not readily return to their earlier, nearer-shore paths. In early 2000, BP Exploration (Alaska), Inc., began constructing its Northstar oil development, the first oil production facility seaward of the barrier islands in the Beaufort Sea. The development is centered on Northstar Island, an artificial gravel island 5 km (~2.7 n.mi.) seaward of the coastal barrier islands (Fig. 8.1). Oil production and gas injection began in late 2001, powered by gas-turbine engines, and drilling of additional wells continued through 2002. Northstar is 27 km (15 n.mi.) west of a natural island (Cross Island) from which migrating bowhead whales are hunted in fall. Thus, the possible effects of the industrial operations on the whales, and on their accessibility to subsistence hunters, were of concern.

FIGURE 8.1. Map of the central Alaskan North Slope including Prudhoe Bay and Northstar Island. 8-4 Monitoring at Northstar: Open-Water Season

Because of both BP corporate policy and regulatory requirements, BP decided to study any possible effects of island sound on the path of migrating bowheads. The primary question was whether the whales passing Northstar were farther offshore during times when more sound was emitted from the island and its associated vessels. The approach taken was based on use of a passive acoustic method to determine the locations of calling whales. An acoustical method was selected, in preference to aerial surveys, because acoustics works 24 hours/day, during good weather and bad, night and day, and allows simultaneous monitoring of a large area. Previous experience showed that thousands of bowhead whale calls would be detected during one autumn season of monitoring, providing a far larger sample size than could be obtained from aerial surveys. To compare whale locations with sound emissions, continuous data on sounds from the island were needed. These were obtained with a hydrophone near the island, connected by cable to a digital recording system on the island. Statistical analyses were essential to assess whether any “displacement” effect was evident at times of high sound emission and, if so, to determine the magnitude of displacement. To date, field work has been done during the early autumn of 2000, 2001, and 2002, with equipment development and a pilot study in 2000, and effective data collection in 2001 and 2002. This chapter describes the acoustical methodology developed to detect bowhead whale calls and to locate their source positions. Although the approach was the same in 2001 and 2002, some improvements were made from 2001 to 2002, and the following specific description concerns the 2002 techniques. Some results from 2002 are included to illustrate the approach and the nature and precision of the resulting data. The following chapter (Chapter 9) describes the specialized statistical approach that has been developed for analysis of the whale locations vs. sound emission data. Chapter 10 describes the results from the 2001 and 2002 measurements and analyses concerning Northstar effects on the whale migration corridor. Chapter 11 uses the results to estimate the numbers of bowhead whales apparently affected by Northstar in 2001 and 2002. Additional details concerning the 2001 and 2002 work are given in Greene et al. (2002) and in Chapter 7, respectively.

METHODS

There are two basic methods by which passive acoustic (hydrophone) data can be used to locate the sources of underwater sounds: hyperbolic fixing and triangulation (McLennan and Greene 1996). Hyper- bolic fixing involves measuring differences in the times when a given sound arrives at pairs of hydrophones, usually with cross-correlation. This method has been used extensively in previous studies of bowhead whales (e.g., Greene 1987; Clark et al. 1996). Triangulation involves measuring bearings to sound sources from two or more known locations, and then computing the intersections of the bearings. Hyperbolic fixing with cross-correlation generally performs very well, but it requires synchronized timing of all the hydrophone signals. This is possible if the hydrophone signals are cabled or radioed to a common recording or processing point with a single clock. However, it is difficult to have precise timing with autonomous, asynchronous recorders, each with its own clock, as required here. Surface buoys were undesirable because of hazards from drifting ice, and interconnection by bottom cables was impractical. Triangulation was chosen for this study because synchronized timing of autonomous units would be difficult to achieve, but autonomous bearing measurements could be obtained. Directional Frequency and Recording (DIFAR) sonobuoys (AN/SSQ-53) use a three-channel sensor from which azimuthal bearings to sound sources can be computed (U.S. Navy 1995). Standard DIFAR sonobuoys have been used previously to obtain bearings to calling bowhead whales (Ljungblad 1986; Greeneridge Sciences unpubl. data) and to other baleen whales (e.g., Swartz et al. 2001). We incorporated DIFAR sensors from Chapter 8: DIFAR Sensors for Monitoring Bowhead Migration 8-5

Sparton Electronics (De Leon Springs, FL) into seafloor recorders called DASARs, for Directional Autonomous Seafloor Acoustic Recorder. A DASAR records the three DIFAR channels directly to disk in the seafloor recorder. An array of 11 DASARs deployed in the southern part of the whale migration corridor offshore of Northstar provided the needed intersecting bearings to the calling whales. DIFAR Principles

The concept of DIFAR is to use two horizontal, orthogonal directional sensors and an omnidirectional pressure sensor to sense an acoustic field. The directional sensors are particle velocity hydrophones with dipole patterns. DIFAR sonobuoys use a magnetic compass to provide a geographic reference. However, the DASARs (unlike drifting sonobuoys) sit firmly on the bottom and do not move after installation. Their locations are measured during installation with GPS. Acoustic calibration transmissions from known locations around each DASAR can provide information on their reference directions with respect to grid north. As shown in the “Results” (later), this can provide more precise bearing data than could be obtained by reference to the compass built into a standard DIFAR sensor, especially during operation at high latitudes where magnetic compasses are less accurate and less stable. Figure 8.2 shows the DIFAR sensor pattern. A sound from a given direction results in received levels, from each of the two orthogonal sensors, that are related to their dipole responses. The relative angle

FIGURE 8.2. DIFAR sensor dipole patterns and reference angle. The angle between the Reference Axis and the Source Direction is the angle b used in the section on DIFAR principles. 8-6 Monitoring at Northstar: Open-Water Season to the source is determined from the arctangent of the ratio of the two response magnitudes. The phases of the dipole signals relative to the omnidirectional signal resolve the bearing ambiguities that would otherwise result from the arctangent. The DIFAR sensor functions as if it were a 3-channel analog device, including the following outputs: • Omnidirectional pressure sensitive hydrophone • Two horizontal, orthogonal particle motion sensors whose outputs are proportional to the cosine/sine of the direction to the source. One is the “cosine channel” and the other is the “sine channel”. The DIFAR sensor reference axis is aligned with the cosine channel maximum. Assume the DIFAR sensor is mounted in the water away from any reflecting surfaces and interfering sound sources. Assume there is a source emitting sinusoidal waves so the instantaneous pressure at the omnidirectional sensor is

Po(t) = A sin(ωt) where A is the peak amplitude, ω is the radian frequency, and t is time. If the source direction is at bearing angle b from the reference axis, the instantaneous outputs of the cosine and sine channels will be

Pc(t) = cos(b)•A sin(ωt)

Ps(t) = sin(b)•A sin(ωt) These two channels, plus the omnidirectional pressure channel, are recorded in the DASAR for offline processing. The direction can be found by forming the instantaneous products Qoc(t) = Po(t)•Pc(t) = cos(b)•A2sin2(ωt) Qos(t) = Po(t)•Ps(t) = sin(b)•A2sin2(ωt) Taking the time average of the two products results in the bearing b in degrees as b = 57.2957•arctangent (mean(Qos)/mean(Qoc)) where the arctangent is defined for four quadrants. If other sound sources are audible simultaneously, their signals add to the above expressions and distort the process. Thus it is important to exclude other sources as much as possible. In our processing of the DASAR data, the time and frequency extents of each whale call are first determined by inspection of a spectrogram. Then the signal is time-gated and bandpass filtered tightly before computing the products. These products are derived separately for every 1-ms time sample within the duration of a given bowhead call (typically 1–3 s), and then averaged across the call duration. The resulting averages are used to derive the estimated bearing to that whale call relative to the reference axis. Alternatively, the bearing can be computed from the magnitude and phase of the discrete Fourier transforms of the three sensor channels. For a signal whose frequency changes or spans a range of frequencies, the bearings for every frequency component in the call can be computed and averaged to improve the bearing estimate. In our experience, the time domain and the frequency domain approaches yield very close to the same result. After initially deriving a bearing to each call via both approaches and finding little difference in the results, we now rely on the time-domain method. However, in Results we present an example of the frequency-domain method. Chapter 8: DIFAR Sensors for Monitoring Bowhead Migration 8-7

The DIFAR specification for sonobuoy bearing accuracy calls for angles with respect to magnetic north to be within ±10 degrees. Much better accuracy has been achieved with the DIFAR sensors incorporated into DASARs as described below, with the DASARs firmly on the bottom, calibrated externally, and not reliant on a magnetic compass (see Results). Figure 8.3 graphs the frequency response of DIFAR sonobuoys. It is specified to be from 10 to 2400 Hz, but there is some sensitivity at both higher and lower frequencies. The response is deliberately not flat; it rises with increasing frequency up to ~1000 Hz. This rise is to counter the natural drop in ocean ambient noise with increasing frequency, providing improved dynamic range and noise performance, especially at low frequencies. Most bowhead calls occur in the frequency range from 50 to 400–500 Hz (Clark and Johnson 1984; Würsig and Clark 1993), so the DIFAR coverage is an excellent match. Directional Autonomous Seafloor Acoustic Recorder (DASAR)

The DASAR was conceived to incorporate DIFAR technology in an autonomous recorder that would operate for at least 30 days in the water north-northeast of Northstar Island, at depths up to 30 m. To achieve frequency coverage from 50 to almost 500 Hz, a sample frequency of 1000 samples/second was selected. For anti-aliasing, the three acoustic channels were lowpass filtered at 400 Hz. With three acoustic channels, 3000 samples/second had to be written to a computer disk storage unit.

FIGURE 8.3. DIFAR frequency response envelope (from U.S. Navy 1995). The response is specified to be within ±3 dB at 100 Hz. 8-8 Monitoring at Northstar: Open-Water Season

Although DASAR orientation on the bottom was determined by calibration (see below), a compass was still desired in the DASAR as a check on any possible changes in orientation during a deployment. The compass in the standard DIFAR sensor used excessive battery power because of the high effective sample rate needed for the acoustic data. However, we did not require frequent compass readings. A substitute flux-gate compass, small in size and low in power drain (Model V2X from PNI Corp., Santa Rosa, CA), was installed in the sensor. Its output was recorded once for every disk write, which occurred at intervals of 46.6 min. Compass bearings were stored in the compact flash memory in the DASAR, independent of the disk storing the acoustic data. Digital Recording Figure 8.4 is a block diagram of a DASAR. A Persistor Instruments Inc. (Bourne, MA) single-board computer, model CF1, serves as the central processor unit and controls the sampling, buffer storage, and disk recording from the buffer. The instrument uses a 196 MByte compact flash memory for double buffer- ing so no data are lost during the disk write operation. However, in practice, the sounds of the disk drive increased the local acoustic noise for about 30 s each time disk writing occurred, once every 46.6 minutes. A disk of 30 GByte capacity (IBM Travelstar 2.5”) provides continuous storage for over 45 days’ recording. The disk is controlled by Persistor Instruments’ “BigIdea” IDE controller. A custom circuit card handles signal conditioning, analog-to-digital conversion for the three DIFAR channels, compass I/O, and various switching and linear regulators to power the CF1, compass, and sensor head. Transponder for Health Checks An acoustic transponder (Benthos model UAT-376, North Falmouth, MA) was built into each DASAR to permit checks on operational health during a deployment. This was essentially a check on the digital recording. If a sensor quit working, it was not likely to be reported as “bad health”. The transponder was powered by its own 9-V cells so it would continue operating if the main DASAR batteries failed. Upon interrogation at 26 kHz by a topside unit, each transponder transmitted on one of seven frequencies: 25, 27, 28, 29, 30, 31 or 32 kHz. It responded with different codes for “OK” or “failed”. The topside unit was a Benthos model DRI-267A, built as a diver-operated transponder/interrogator and deployed on a hand-held pole over the side of a boat. Upon receipt of transponder signals, it read out distance as well as the operational code. It operated over distances of at least 200 m, but GPS always brought the boat to within about 40 m of the DASAR being interrogated. In this project, the “health” of each deployed DASAR was checked about once per week. No failures were detected in 2002, but in 2001 some reports of “bad health” were received, and in these cases it was possible to retrieve and replace the failed DASAR. Physical Construction and Batteries The pressure housing is a cylinder 12” in diameter and 14” high (30 × 36 cm). The recording electronics and the transponder are mounted on the underside of the lid, which incorporates double O-ring seals. Three packs (Nexergy model SMP-7S4P, Escondido, CA) of 28 D-cell alkaline batteries are wired in series and parallel to provide 10.5 volts. They are secured by a plate and threaded rods anchored in the cylinder bottom. Diodes protect the packs from internal shorts. Figure 8.5A shows a DASAR as mounted on its frame and within its cage. On the top of the DASAR are two plastic posts 8.75” (22 cm) tall, extending upward from opposite sides of the top plate. Between them is a rubber band 1” (2.5 cm) wide from which the DIFAR sensor is suspended. The top of the sensor is secured to a plastic block with a double O-ring seal. The wires from the compass and three acoustic sensors pass from the sensor through the block and into a plastic tube and Chapter 8: DIFAR Sensors for Monitoring Bowhead Migration 8-9

SONOBUOY HEAD (MODIFIED DIFAR -53E) INCL V2X COMPASS TRANSPONDER (DATASONICS UAT-376)

SIG COND, STATUS ADC, VREG (CUSTOM)

CF MEMORY CPU CCA CARD (PERSISTOR BACKPLANE (196 MB) CF1)

IDE CONTROLLER CCA (PERSISTOR)

2.5" IDE HARD DRIVE (30 GB) 10.5VDC

BATTERY PUCK (Alkaline 'D' cells)

10.5VDC

FIGURE 8.4. Block diagram of the DASAR recording system. 8-10 Monitoring at Northstar: Open-Water Season

FIGURE 8.5. (A) Photograph of a DASAR assembly. (B) DASAR on deck with anchor and tag line, ready for deployment. Chapter 8: DIFAR Sensors for Monitoring Bowhead Migration 8-11 thence to a waterproof access in the main body of the DASAR (Fig. 8.5A). The rubber suspension permits the sensor to “float” relatively freely in the water for sensing the particle motion essential to the directional sensing, while at the same time minimizing twisting. As deployed in 2002, a latex “sock” secured over an aluminum cage shielded the sensor from motion induced by water currents. Figure 8.5B portrays a DASAR in its sock on the deck ready for deployment with the small anchor and tag line. (In 2001, a different frame and a self-supported thin plastic dome were used.) Each DASAR was installed on the bottom with a tag line between the DASAR and an anchor, both of whose positions were measured by GPS during deployment. No surface buoys were used because of hazards from drifting ice floes. The DASARs were retrieved from depths of 16–24 m by grappling for the tag line based on the GPS coordinates. The DASAR Array

The basic array configuration is a series of equilateral triangles. Bearing measurements from pairs of elements in a triangle provide good resolution of source location for sources at any angle around the triangle. Six triangles sharing a center unit formed a hexagon (Fig. 8.6). Two overlapping hexagons completed the array of 10 locations oriented roughly north-northeast from Northstar Island, toward the path of migrating bowheads. The basic spacing between adjacent DASARs was 5 km (2.7 n.mi.), selected based on

FIGURE 8.6. Map of DASAR locations with respect to Northstar Island. 8-12 Monitoring at Northstar: Open-Water Season known bowhead call detection distances of 15–20 km and experience with 2 km spacing in a 1986 application of passive localization (Greene 1987). The closest and farthest units were, respectively, 7 and 22 km (4 and 12 n.mi.) northeast of Northstar, with the most offshore unit being about 1/3rd of the distance from Northstar to the shelf-break. Twelve DASARs were constructed and tested. Two were intended to be spares, but it made sense to deploy one of the spares as a redundant element at a central location in the 10-position array. Thus, two DASARs were used at the center of the northern hexagon, spaced about 200 m apart at locations CA and CB (Fig. 8.6). Time and Reference Axis Calibration

Calibrations were very important to the success of the triangulation approach to whale call localization. The times had to be known within 1–2 s to permit associating calls received on two or more independent DASARs, as necessary to localize based on bearings from two or more DASARs to the same call. With 5 km between adjacent DASARs, the maximum difference in acoustic travel time to an adjacent pair of DASARs is 3.3 s. The travel time across the longest separation of DASARs (units NE and SW in Fig. 8.6) is 10 s. Reference axis calibration was very important to minimizing localization errors. Clearly, the smaller the uncertainty in a bearing measurement, the smaller the uncertainty in the whale call location. A detailed description of localization errors is contained in Appendix 8.2. The approach taken to calibration for both time and reference axis direction was to transmit a signal from a boat at known locations around and within the DASAR array. The equipment used included a U.S. Navy model J-9 underwater sound projector (U.S. Navy Sound Reference Laboratory, Newport, RI), a 400-W Mackie FR series M-800 amplifier (Woodinville, WA), a notebook computer, and a Trimble (Sunnyvale, CA) GPS receiver with a pulse time signal output precisely at every UTC (Universal Time Coordinated) second. The notebook computer contained the waveform file and repro- duced it for amplification and transmission on receipt of the timing signal from the GPS. Although the J-9 is not rated for source levels greater than ~150 dB re 1 µPa-m, its efficiency is very low and a high- powered amplifier is required to achieve such a source level. The waveform of the calibration signal required a clear start time for clock timing and wide bandwidth for determining the reference axis bearing. The sound waveform used in 2002 (Fig. 8.7) began with a 400-Hz tone for 5 s, followed by a 2-s gap, a downsweep from 400 to 200 Hz, another 2-s gap, an upsweep from 200 to 400 Hz, a 2-s gap, and a final downsweep from 400 to 200 Hz. (In future applications, the 2-s gaps will be omitted.) The total duration of one transmission was 22 s. For each calibration location, two full transmissions occurred at sound projector depth 9 m followed by two more at depth 6 m. All transmissions were analyzed as received at the closest two or three DASARs, generally within distance 3–4 km. Calibration stations were selected in and around the array such that each DASAR received calibration signals from at least four (and often six) nearby stations on different bearings. Figure 8.8 shows the 18 calibration stations used in 2002 in relation to the 11 DASARs. Obtaining bearings to several known sites around each DASAR permitted searching for possible systematic bearing errors. One such error was found and attributed to “gain imbalance” between the two directional channels. Chapter 8: DIFAR Sensors for Monitoring Bowhead Migration 8-13

FIGURE 8.7. A one-minute spectrogram of two calibration transmissions.

FIGURE 8.8. Map of the calibration signal transmission stations (+) around the array of DASARs (▲). 8-14 Monitoring at Northstar: Open-Water Season

Gain Imbalance Between the Directional Sensors

The DIFAR sensor manufacturer, Sparton Electronics, employs careful calibration techniques to equalize the sensitivities of the sensors for the two directional channels. However, the manufacturing specification is for bearing accuracies of ±10° and we desired better performance. There was evidence in our data that, as installed in the Beaufort Sea, the gains (sensitivities) of the DIFAR sensors were slightly unbalanced, resulting in systematic bearing errors. A gain-correction procedure has been developed. It is described in Appendix 8.1. The results are presented in the Results section of this Chapter. Localization Process and Its Accuracy

The locations of the calling whales were estimated based on triangulation. However, when more than two bearings to a given call were available, the 2+ intersections of bearings usually did not coincide exactly. A maximum-likelihood approach was developed to estimate the most likely location, taking account of the variability in bearings from each DASAR as determined during calibration. Also, an estimate of the precision of each whale call location was needed for use as a weighting factor in subsequent analyses. Appendix 8.2 describes the localization process and the derivation of the localization accuracy measures. High Background Noise Cutoff Distance

The objective of the project was to evaluate whether offshore distances of whales were related to the level of industrial sound measured near the island (400 m away). Early data indicated that, during high background noise conditions at the DASAR array, fewer whale calls were detected. Further analysis confirmed that, at such times, proportionally fewer calls were detected at longer distances. To minimize the influence of variations in background noise on the observed distribution of whale call locations, it was necessary to limit the study area to the region within which whale calls were likely to be detected even during high noise conditions. The appropriate distance was determined by comparing the distribution of distances from the array-center to the detected whale calls during times with high and low background noise (see Results).

RESULTS

Examples of the results of the calibration process and the localization of whale calls for 2002 are presented here. Chapter 9 shows how these results can be analyzed statistically to assess whether bowhead whales are displaced offshore at times when much industrial sound emanates from the Northstar area, and Chapter 10 describes the results from the 2001 and 2002 seasons. In 2002, the DASARs were deployed on 30–31 Aug. and recovered on 3 Oct. One DASAR apparently physically tipped over on 4 Sept., but the other 10 functioned well until 23 Sept. when all apparently tipped over during an unusually strong storm. Thus, the effective study period was from 31 Aug. to 23 Sept. Health checks were done about once per week, and there were no reports of “bad health”. (In 2001, some such reports were received, and this allowed us to replace failed units with functional ones.) Calibration of DASAR Clock Times and Reference Axis Bearings

The calibration results for each DASAR are presented in Table 8.1. Calibration data were collected on 2, 7, 13 and 21 Sept. 2002. For 10 of 11 DASARs, the calibration was based on 126–198 calibration pings. For the DASAR at location “SW”, only 33 pings (all from 2 Sept.) were useable; SW was the unit that upset in high storm surge on 4 Sept. The lack of additional measurements at SW may account for a higher standard deviation in the bearing measurements at unit SW (Table 8.1). Chapter 8: DIFAR Sensors for Monitoring Bowhead Migration 8-15

TABLE 8.1. DASAR calibration results for 2002, including corrections for gain imbalance in the two directional channels and clock drift coefficients.

Original Corrected Original Corrected Scatter DASAR Total Ref Bearing Ref Bearing Std. Dev. Std. Dev. Clock Drift after fit Location Pings (Deg) (Deg) Gain Factor (Deg) (Deg) (Sec/Day) (Seconds) NW 126 89.9 89.5 1.0607 1.50 0.87 -0.4836 0.19 CC 198 248.9 248.8 1.0270 1.70 1.60 -1.3582 0.16 WB 138 34.3 34.0 1.0286 1.30 1.20 -1.7558 0.22 CB 194 157.0 157.0 1.0883 2.30 1.50 -1.1756 0.28 CA 192 186.9 187.0 1.1337 2.60 0.99 0.3827 0.21 EA 139 58.6 57.6 1.1594 2.90 0.93 -1.1898 0.34 WA 165 192.9 192.3 0.8661 3.30 1.30 -1.3838 0.19 SE 127 26.7 26.7 1.0038 0.80 0.80 -1.9505 0.54 SW 33 59.0 59.5 1.0781 3.60 3.30 -0.2753 0.30 NE 135 96.0 95.7 0.9294 2.50 0.80 -1.1723 0.21 EB 154 190.5 190.8 1.0529 1.90 1.60 -0.7723 0.19 Avg: 2.22 Avg: 1.35

The benefit of correcting for the gain imbalance between the two directional sensors of each DASAR is evident by comparing the standard deviations of the “original” (uncorrected) bearings, which averaged 2.22°, with those of the “corrected” bearings (average 1.35°). Corresponding average values for 2001 were, respectively, 6.60° and 6.28° (see Table 8.2). The higher values in 2001 resulted primarily from using tones as calibration sounds, in contrast to the frequency sweeps used in 2002. The differences in the DASAR housings may have been a minor factor.

TABLE 8.2. DASAR calibration results for 2001, including corrections for gain imbalance in the two directional channels.

Original Corrected Original Corrected DASAR Total Ref. Bearing Ref. Bearing Std. Dev. Std. Dev. Location Pings (Deg) (Deg) Gain Factor (Deg) (Deg) CAa 204 211.8 213.0 1.0439 7.52 7.50 CBa 111 246.2 248.4 0.9924 5.41 5.37 CBb 70 226.3 226.4 0.9559 11.25 11.28 CBc 214 114.7 114.7 1.0540 3.53 3.36 CCa 91 168.5 172.5 0.7762 7.92 7.43 EAa 253 183.0 182.6 1.0607 6.23 6.19 EAb 515 248.3 248.3 0.9897 2.85 2.85 EBa 452 208.0 208.1 0.6574 12.90 10.94 NEa 871 179.5 181.2 0.8446 6.30 5.60 SEa 264 30.2 30.3 1.0237 9.16 9.09 SWa 184 219.4 220.9 0.8138 5.62 5.27 WAa 216 189.5 189.5 1.0041 5.60 5.61 WAb 220 124.5 124.5 0.9560 3.76 3.63 WBa 536 37.8 38.2 0.8823 4.41 3.79 Avg: 6.60 Avg: 6.28 8-16 Monitoring at Northstar: Open-Water Season

Based on the DIFAR sonobuoy specification that bearing errors should be within ±10º, standard deviations on the order of 3º had been expected. The improvement over the specification (in 2002) is partly from the broadband calibration signals used, partly from the gain-correction procedure, and partly because the orientations of the bottom-mounted DASARs were fixed and externally calibrated. In contrast, a DIFAR sonobuoy is designed to drift, with bearings being determined in relation to an internal magnetic compass. The time-drift calibrations showed that the clocks in the DASARs drifted by as much as 2 seconds per day (Table 8.1). However, the drift-rate was quite stable over time. Figure 8.9 presents, for one DASAR, an example of the measurement data and the least-squares fit of a linear model for clock drift. After compensation for the measured drift, the times of events recorded between calibrations generally could be determined to an accuracy of 1 or 2 s. This was adequate for assessing whether a whale call received at several DASARs represented a single call from one whale. The residual measurement scatter is mostly due to imprecision in determining the time of the received signal during data analysis. This precision can be improved, as described in the Discussion. Whale Call Localization

During the period from mid-day on 31 Aug. until mid-day on 23 Sept. 2002, 10,587 whale calls were detected. Of these, 2022 (19%) were detected by only one DASAR and thus could not be localized. Also, 182 (2%) of the remaining 8565 calls did not yield a location because the maximum likelihood procedure (Appendix 8.2) did not converge, and 461 (5%) of the remaining 8383 calls yielded a location but the variance-covariance matrix of the solution was not positive definite. Overall, 7922 (75%) of the 10,587 calls yielded useable locations in 2002.

FIGURE 8.9. Example of a DASAR clock drift measurement from calibration transmissions. From the DASAR at location CC in 2002. Chapter 8: DIFAR Sensors for Monitoring Bowhead Migration 8-17

Figure 8.10 presents an example of a whale call bearing measurement, in this case from DASAR unit EB. The bearings of all the 1-ms time samples spanning the duration of the call are plotted as vectors from the DASAR location. The bearing estimate comes from the arctangent of the average of the north- south components divided by the average of the east-west components. Figure 8.11 shows the call waveform. The same call as received at the same DASAR was also processed by the frequency domain method. Figure 8.12 presents the magnitude of the sound pressure density spectrum computed over the duration of the call as delimited by the analyst. The relatively high spectrum levels at 190–320 Hz represent the whale call. The rising slope of the background noise with increasing frequency results from the corresponding emphasis of higher frequencies by the DIFAR sensor (see Fig. 8.3.). Figure 8.13 presents the bearing–frequency result for the same call, with the bearing selected from the time-domain solution superimposed across the frequencies selected by the analyst. The specific call illustrated above was detected via six DASARs, and Figure 8.14 shows the intersections of those bearings, including the bearing from Figure 8.10 (unit EB). Of the 7922 useable call locations during the 23-day study period in 2002, the numbers based on 2, 3, 4 and 5+ bearings were 2085, 1429, 1129, and 3279, respectively.

FIGURE 8.10. Example of a time-domain bearing display illustrating the spread of the bearings computed for each 1-ms time sample. From the DASAR at location EB in 2002. 8-18 Monitoring at Northstar: Open-Water Season

FIGURE 8.11. Whale call waveform corresponding to the bearing data in Figure 8.10.

FIGURE 8.12. Sound pressure spectral density for the whale call in Figures 8.10 and 8.11. The peak at about 190–320 Hz represents the whale call. The increasing slope in the background noise is a result of the sloped sensitivity of the DIFAR sensor as shown in Figure 8.3. Chapter 8: DIFAR Sensors for Monitoring Bowhead Migration 8-19

FIGURE 8.13. The directional spectrum for the whale call in Figures 8.10–8.12. The bearing from the time domain analysis (Fig. 8.10) is drawn for the frequency band selected by the analyst.

A B

7831 EB

7830

7829 UTM Northing (km) UTM Northing

7828

447 448 449 UTM Easting (km)

FIGURE 8.14. Map of an example of intersecting bearings to a whale call, including the bearing from unit EB shown in Figures 8.10–8.13. (A) is the large-area view, and (B) is an exploded view of the intersection area showing the 90% confidence ellipse and the best estimate of the calling whale’s location. 8-20 Monitoring at Northstar: Open-Water Season

The numbers of bowhead whale calls detected per hour varied widely through the operational period (Fig. 8.15). Call detection rates depended not only on the number of whale calls, but also on the background noise level, as high background noise tended to mask the more distant calls (see below). The best estimates of the locations of the 350 localizable bowhead whale calls detected on 14 Sept. 2002 are mapped in Figure 8.16. Localization Accuracy

Figure 8.17 maps one measure (error ellipse long axis length) of the accuracy of whale call localizations for locations around the DASAR array, based on all 7922 of the useable call locations from 2002. The calculations were based on the estimated accuracy of individual locations, which in part depended upon error in the calibration bearings (see Appendix 8.2). As expected, accuracy tended to deteriorate with increasing distance from the center of the array of DASARs. For most locations within the perimeter of the array, the length of the major axis of the 90% confidence ellipse averaged between 0.25 and 1 km. For most locations 10 km from the center of the array (~6 km outside the perimeter), the corresponding average was ~1 km. Average localization accuracy was apparently poorer in an area southwest of the DASARs and northwest of Northstar, but very few calling whales were detected there (see Fig. 7.15 in Chapter 7). Beyond about 10 km from the center of the array, localization accuracy deteriorated rapidly (Fig. 8.17). High Background Noise Cutoff Distance

Distances from center DASAR CB to all whale calls detected during times of very high, high, and low background noise were determined (Fig. 8.18). “Very high” and “high” background levels were defin-

FIGURE 8.15. Numbers of bowhead whale calls detected per hour over the period 31 Aug.–23 Sept. 2002. A total of 10,587 calls were received at one or more of the DASARs. Chapter 8: DIFAR Sensors for Monitoring Bowhead Migration 8-21

FIGURE 8.16. Whale call localizations for 14 Sept 2002. Filled upright triangles represent DASAR locations. Symbol types discriminate calls whose locations were estimated based on bearings from two (+), or more (Ο) DASARs. On this date, 42 calls (not localizable) were detected on only ,(ٱ) three (∇), four one DASAR, 115 calls on two DASARs, 65 with three DASARs, 55 with four DASARs, and 115 with five or more DASARs. The WNW–ESE line through Northstar Island shows the general trend of the coastline; the autumn bowhead migration is to the WNW, parallel to shore. ed to be those at times when broadband (5–500 Hz) sound levels were greater than the 90th and 75th percentiles, respectively, of all noise levels received at DASAR NE, the one farthest from Northstar, at 15-min intervals. “Low” background levels were defined to be those at times when broadband levels at “NE” were less than the 25th percentile levels there. The proportion of the detected calls that were at distances >10 km was lower during the very high and high background noise conditions. To minimize potential biases, calls >10 km from the “central axis” of the DASAR array were not used in statistical analyses of call locations relative to Northstar noise (see Chapters 9 and 10). That central axis extended from CB to Northstar. Thus, the study area for those statistical analyses was a “silo-shaped” area (Fig. 8.17B) that was 20 km wide extending offshore to DASAR CB, capped by a semi-circle of radius 10 km centered at CB. 8-22 Monitoring at Northstar: Open-Water Season

FIGURE 8.17. Mean length (in kilometers) of major axis of 90% confidence ellipses around estimated locations shown as (A) smoothed surface, (B) contour plot. View in (A) is (approximately) from Northstar, looking northeast. Filled circle represents Northstar Island, open triangles represent DASARs, and solid black line is a baseline through Northstar parallel to the coast. Chapter 8: DIFAR Sensors for Monitoring Bowhead Migration 8-23

A. Background > 90-th percentile (101.1 dB) Cut-off = 10km Number of Calls 0 10203040 0 1020304050 Distance from Center of DASAR Array (km)

B. Background > 75-th percentile (95.7 dB) Cut-off = 10km Number of Calls 0 50 100 150 0 1020304050 Distance from Center of DASAR Array (km)

C. Background < 25-th percentile (86.8 dB) Cut-off = 10km Number of Calls 0 50 100 150 200 0 1020304050 Distance from Center of DASAR Array (km)

FIGURE 8.18. Distances from estimated locations of calling whales to the center DASAR (CB) during very high (A), high (B), and low (C) background sound conditions in 2002. A cutoff distance of 10 km was established at the “shoulder” evident in A and B. Background sound was measured at the NE DASAR in the 5–500 Hz band (see text for further details and rationale). 8-24 Monitoring at Northstar: Open-Water Season

DISCUSSION AND CONCLUSIONS

Background noise has a strong influence on the detectability of whale calls. At times with high background noise, calls with low source levels, and other calls whose received levels at the DASARs are low because of propagation from long distances, are less likely to be detected. The variability in background noise is one reason why simple counts of whale calls (e.g., calls detected per hour) are not a reliable index of the number of whales in the area. (The potentially variable calling rate of individual whales is another reason.) In using the data from this project, it has been important to limit the size of the study area to a region where calling bowhead whales are generally detectable even when the background noise is high. In 2002, the “high background cutoff distance” was determined to be 10 km (Fig. 8.18). A corresponding analysis for 2001 showed it to be 12 km in that year (Greene et al. 2002). Potential Improvements to DASAR Performance

The controlling software will be revised to permit entering a sleep mode after initialization, with startup programmed for a time after the units are on the bottom and stable. In the past, when DASARs have been writing to disk while being handled, write failures have sometimes occurred. In future applications, the sound transmitted for calibration of times and reference bearings will be frequency sweeps or a pseudo random noise waveform of at least 200 Hz bandwidth. Wideband sounds have been shown to produce less bearing scatter than tones. There will be no gaps in the sound transmissions (cf. Fig. 8.7). Also, eight more calibration stations will be added to the 18 used in 2002, providing better azimuthal calibration data for the peripheral DASARs and hence better gain imbalance corrections. Data analysis will be enhanced to permit more accurate measurements of the arrival times of the calibration transmissions. Also, more automation in detecting whale calls will be investigated. At present, whale calls are detected based on simultaneous visual inspection of spectrograms of the data from all active DASARs, one minute at a time. The computer might be programmed to detect whale calls by using their frequency characteristics and the fact that they are stronger than the background sounds (Mellinger and Clark 1997, 2000). Utility of DASARs in Localizing Sound Sources

This application of DIFAR techniques, as incorporated into specially designed bottom-mounted recording units, has been successful in monitoring the locations of calling bowhead whales over extended periods in a remote region. This approach allows distributed data acquisition, and requires only moderately accurate synchronization (e.g., ±1 or 2 s) of the time bases of the various recording units. This approach is suitable in situations where it would be difficult or impossible to acquire all the data with a single recording system and time base, as needed for localization by time-of-arrival methods. Experience from two years’ use of DASARs to localize bowhead whale calls reinforces the concept of applying DIFAR sensor technology to sound source localization. The frequency range of 10–500 Hz, as presently recorded by the DASARs, suits the frequency ranges of most bowhead calls, and may also be suitable for most other baleen whales (Richardson et al. 1995). However, the DIFAR sensors are useful at frequencies >2 kHz and could conceivably be used to detect and localize clicks from sperm whales and other marine life. A higher frequency range would require a higher sampling rate, which would affect recording life depending on disk space. Chapter 8: DIFAR Sensors for Monitoring Bowhead Migration 8-25

The present DASAR units are designed for deployment in relatively shallow water, and retrieval by a simple grappling technique. The physical configuration of the units would need to be revised if deeper deployments were required. However, the DIFAR-based approach should be useable in deeper water.

ACKNOWLEDGEMENTS

Dave Iddings, Quy Nguyen, Bob Early and Patrick Dexter designed the DASAR pressure housing, sensor suspension, and sensor wire routing (including waterproofing). Mark Cavalier, Nautronix-MariPro, fabricated the aluminum frames and cages. Kent Reinhardt of Sparton Electronics was very helpful with engineering details of the DIFAR sensors, permitting us to disable the built-in compass and extract the sensor signals before they were multiplexed. David Christian made the cable harness from the sensor to the recording electronics, and he wired the DASAR electronics and batteries, making battery installation simple and foolproof. During deployment, calibrations, and retrieval, crews of Alaska Clean Seas provided excellent navigation and seamanship consistent with safe operations. Jonah Leavitt of Barrow, Alaska, helped with mobilization at Deadhorse, Alaska, as well as deployment and retrieval at sea. Mike Williams of LGL helped in various ways. Wilson Cullor and Allison Erickson of OASIS Environmental (on behalf of BP) helped with logistics planning for the field work. Dr. Bill Streever of BP and Dave Trudgen of OASIS (on behalf of BP) guided the project. Dr. Streever reviewed this chapter, as did R. Suydam of the North Slope Borough Dept. of Wildlife Management. BP Exploration (Alaska) Inc. supported the work. We thank them all.

LITERATURE CITED

Bailey, T.C. and A.C. Gatrell. 1995. Interactive spatial data analysis. Longman Scientific & Technical, Essex, U.K. Clark, C.W. and J.H. Johnson. 1984. The sounds of the bowhead whale, Balaena mysticetus, during the spring migrations of 1979 and 1980. Can. J. Zool. 62(7):1436-1441. Clark, C.W., R. Charif, S. Mitchell and J. Colby. 1996. Distribution and behavior of the bowhead whale, Balaena mysticetus, based on analysis of acoustic data collected during the 1993 spring migration off Point Barrow, Alaska. Rep. Int. Whal. Comm. 46:541-552. Greene, C.R. 1987. Acoustic studies of underwater noise and localization of whale calls. (Chap. 2, 128 p.) In: LGL and Greeneridge, Responses of bowhead whales to an offshore drilling operation in the Alaskan Beaufort Sea, autumn 1986. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for Shell Western E & P Inc., Anchorage, AK. 371 p. Greene, C.R., Jr., M.W. McLennan, T.L. McDonald and W.J. Richardson. 2002. Acoustic monitoring of bowhead whale migration, autumn 2001. p. 8-1 to 8-79 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of industrial sounds, seals, and whale calls during construction of BP's Northstar oil development, Alaskan Beaufort Sea, 2001, Draft, Oct. 2002 ed. LGL Rep. TA2573-2. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. Lenth, R.V. 1981. On finding the source of a signal. Technometrics 23:149-154. Mardia, KV. 1972. Statistics of direction data. Academic Press, New York, NY. 357 p. Mardia, K.V., J.T. Kent and J.M. Bibby. 1979. Multivariate analysis. Academic Press, London, U.K. McLennan, M.W. and C.R. Greene. 1996. Passive acoustic localization and tracking of vocalizing marine mammals using buoy and line arrays. Appendix In: Report of the cetacean acoustic assessment workshop. Rep. by 8-26 Monitoring at Northstar: Open-Water Season

Australian Nature Conservation Agency (now Biodiversity Group of Environment Australia), Hobart, Tas. var. pag. Ljungblad, D.K. 1986. Endangered whale aerial surveys in the Navarin Basin and St. Matthew Hall planning areas, Alaska. In: Aerial surveys of endangered whales in the northern Bering, eastern Chukchi, and Alaskan Beaufort Seas, 1985: with a seven year review, 1979-85. Appendix E in NOSC Tech. Rep. 1111; OCS Study MMS 86- 0002. Rep. from Naval Ocean Systems Center, San Diego, CA, for U.S. Minerals Manage. Serv., Anchorage, AK. NTIS AD-A172 753/6. Mellinger, D.K. and C.W. Clark. 1997. Methods for automatic detection of mysticete sounds. Mar. Freshwat. Behav. Physiol. 29(1-4):163-181. Mellinger, D.K. and C.W. Clark. 2000. Recognizing transient low-frequency whale sounds by spectrogram correlation. J. Acoust. Soc. Am. 107(6):3518-3529. Moore, S.E. and R.R. Reeves. 1993. Distribution and movement. p. 313-386 In: J.J. Burns, J.J. Montague and C.J. Cowles (eds.), The bowhead whale. Spec. Publ. 2. Soc. Mar. Mammal., Lawrence, KS. 787 p. Richardson, W.J., C.R. Greene Jr., C.A. Malme and D.H. Thomson. 1995. Marine mammals and noise. Academic Press, San Diego, CA. 576 p. Richardson, W.J., G.W. Miller and C.R. Greene Jr. 1999. Displacement of migrating bowhead whales by sounds from seismic surveys in shallow waters of the Beaufort Sea. J. Acoust. Soc. Am. 106(4, Pt. 2):2281. Swartz, S.L., A. Martinez, T. Cole, P.J. Clapham, J.A. Hildebrand, E.M. Oleson, C. Burks and J. Barlow. 2001. Visual and acoustic survey of humpback whales (Megaptera novaeangliae) in the eastern and southern Caribbean Sea: preliminary findings. NOAA Tech. Memo. NMFS-SEFSC-456. NOAA Fisheries, Southeast Fish. Sci. Cent., Miami, FL. 37 p. U.S. Navy. 1995. Production sonobuoy specification for bathythermograph transmitting set AN/SSQ-36B and sonobuoy AN/SSQ-53D, 57B, 62C, 77B, 86 and 110A. Rev. 3, 12 June 1995. Progr. Exec. Office, Air Antisub- marine Warfare, Assault, and Special Missions (ASWASM), U.S. Navy, Arlington, VA. 79 p. Wand, M.P. and M.C. Jones. 1995. Kernel smoothing. Chapman & Hall/CRC, Boca Raton, Fl. White, G.C. and R.A. Garrott. 1990. Analysis of wildlife radio-tracking data. Academic Press, New York, NY. 383 p. Würsig, B. and C. Clark. 1993. Behavior. p. 157-199 In: J.J. Burns, J.J. Montague and C.J. Cowles (eds.), The bowhead whale. Spec. Publ. 2. Soc. Mar. Mammal., Lawrence, KS. 787 p. Chapter 8: DIFAR Sensors for Monitoring Bowhead Migration 8-27

APPENDIX 8.1: GAIN IMBALANCE BETWEEN DIRECTIONAL SENSORS 2 AND ITS EFFECT ON BEARING ACCURACY

Summary. This Appendix investigates the question of gain imbalance between the two DIFAR directional sensors. A method is developed to quantify and correct for this factor, thereby improving the accuracy of measured bearings to whale calls. An understanding of DIFAR (Directional Frequency and Recording) principles is assumed. The approach is based on calibration ping data taken for time and bearing calibration. The DIFAR hydrophones were the acoustic sensors in the DASARs. These small units were installed in water 20–25 meters deep with an unknown orientation, and left to record sounds in the 10–500 Hz band. In order to measure the orientation on the bottom and calibrate the clock time drift, acoustic pings were transmitted under the water from known locations. By measuring the bearing of each ping with the DIFAR logic and comparing it to the known bearing from the DASAR to the projector position, it was possible to measure the reference axis direction on the bottom. (A magnetic compass is included in the DIFAR transducer; however, the DASARs were installed at a high geomagnetic latitude, and due to the extreme dip of the magnetic field, an additional check was needed.) Figure 8.19 presents the measured reference bearing for DASAR CBa as determined via each of the nearest calibration pings received by that DASAR. The scatter is larger than expected. Figure 8.19 shows a typical case; similar data exist for each of the other DASARs. The reference axis direction is given by

bref = btrue – bd (1) where btrue is the known true bearing of the ping, and bd is the measured DIFAR bearing. The scatter is about 10 degrees peak-to-peak. This meets the bearing accuracy specified for DIFAR, but is larger than desired. Figure 8.19 suggests that the bearing scatter may be systematic, not random. Ping sequence number is related to time of transmission, and the true bearing of the transmissions varied with time. The data suggest that the measured reference axis direction might be azimuth dependent and not statistically random. Figure 8.20 presents the same data plotted versus true bearing. At any given true bearing, the scatter is on the order of 3º–5º, but the variation across bearing appears to follow an orderly pattern, suggesting that improvement would be possible if the cause of the variation could be determined and modeled. The first effort to explain the scatter centered on acoustic propagation causes. Reflecting objects near the DASAR could explain some bearing error; however the installation area is uniform and relatively flat. Vertical multipaths (which do exist) would not introduce azimuthal bearing errors. We were led to investigate the effects of a gain imbalance in the DIFAR directional channels and to find a means of correcting for it. The Effects of a Gain Imbalance

The DIFAR hydrophone includes three channels, one omni and two directional. The bearing measurement process requires that the acoustic sensitivity of the two directional channels be equal. The DIFAR hydrophone is complex and includes many adjustments by the manufacturer to match the outputs, including temperature compensation. However, the manufacturing specification permits bearing errors of ±10º and

2 This appendix was prepared by Miles W. McLennan, Greeneridge Sciences Inc., Santa Barbara, CA 8-28 Monitoring at Northstar: Open-Water Season

FIGURE 8.19. The calculated reference bearings for DASAR CBa based on calibration pings in 2002. The horizontal axis is sample number (ping sequence number) in order of time. From each location of the ping projector, 4-8 pings were transmitted. For each DASAR, bearings were determined to the pings projected from 4-6 positions at different azimuths. The vertical axis is the calculated bearing of the reference axis of the DIFAR hydrophone.

FIGURE 8.20. Measured reference axis direction vs. true grid bearing for DASAR CBa. (These are the same measurement data presented in Fig. 8.19.) Chapter 8: DIFAR Sensors for Monitoring Bowhead Migration 8-29 we desired better performance. Gain imbalance seemed a possibility. We first investigate what the effect would be of such a gain imbalance. We introduce three new variables: g gain factor of sine channel relative to cosine channel (g=1 means no imbalance)

berr error angle introduced by the gain imbalance

bmeas measured error including the error, such that

bmeas = bd + berr (2)

Assume there is an acoustic signal of amplitude A arriving from bearing bd relative to the DIFAR reference axis. The x and y components of that signal will be

vx = A sin bd

vy = A cos bd (3) However if there is a relative gain error, these will be measured as

vxmeas = g · A sin bd

vymeas = A cos bd (4) and the measured bearing will be

bmeas = arctan2( vxmeas/vymeas) = arctan2(g · vx/vy) (5) where arctan2 is a form of the arc tangent that accounts for all four quadrants. This measured bearing can

be converted to the error berr by subtracting the known bd. Figure 8.21 below shows the bearing error resulting from values of g ranging from 0.7 to 1.5. When the error is plotted versus the true azimuth, it very closely resembles a sine wave with two cycles in the 360 degrees of azimuth, with amplitudes ranging up to about 10º for the extreme values of gain shown. In fact, it is NOT exactly a sine wave. In such a 2-cycle sine wave the maximum value would be at exactly 45º (northeast) off the main axis. In this function, the maximum is lower or higher, depending on the value of g. Manipulation of (5) shows the DIFAR bearing of the maximum is

 1  b = arcsin  (6) d max    1+ g  This is graphed in Figure 8.22 below. Note that this is just the principal bearing value. There will be similar maxima and minima near the corresponding points on the (almost) sine wave; a positive peak at 180º plus b d max , and there will be negative peaks at minus b d max and 180 minus b d max . The magnitude of the error at that maximum is

  1  b = arctan.5 g −  (7) err max      g  This is plotted in Figure 8.23. This Figure presents the magnitude of the maxima and minima of the near- sinusoidal waves. 8-30 Monitoring at Northstar: Open-Water Season

FIGURE 8.21. Azimuthal bearing error due to DIFAR gain imbalance.

FIGURE 8.22. Bearing of maximum bearing error vs. gain imbalance. Chapter 8: DIFAR Sensors for Monitoring Bowhead Migration 8-31

FIGURE 8.23. Maximum magnitude of bearing error vs. gain imbalance.

Measuring the Actual Error and Gain Factor

The relations above give a feeling for the error behavior as a function of possible gain factors, and the errors observed in Figures 8.19 and 8.20 are in the same range. The actual data include random errors from other sources, so we must use a statistical approach to characterize the actual errors, while measuring the reference bearing. We do so by fitting a curve similar to Figure 8.21 to the distribution in Figure 8.20. The basic equation we use is

btrue = bref +bmeas – be (8) where, to review,

btrue is the absolute bearing measured by GPS or other means

bref is the true bearing of the reference (cosine) channel axis

bmeas is the measured bearing of the target from the reference axis

bmeas = bd + be

bd bearing of sound source relative to DASAR reference

be is the small error due to gain imbalance 8-32 Monitoring at Northstar: Open-Water Season

We assume be can be calculated as

becalc = a · f(g,bd) = a · (atan2(g · sin(bd),cos(bd)) – bd)(9) where a is an unknown amplitude and f is a function giving the shape of the error as shown in Figure 8.21.

The maximum value of this will be a · fmax, where fmax is given by equation (7). We also have a body of measurements allowing us to measure the error for a single trial. From (8) we say

bemeas = bref +bmeas –btrue (10)

We wish to compare the calculated form and the measured data, and solve for the values of bref and gain factor g. We do this by a least-squares fit of the function f to the measured data, then solving for the gain and bref. Because of the non-linearity of the equations, it is necessary to use a starting guess for g (for instance 1.2), then solve iteratively for new values.

We start with the i-th measurement giving btrue(i) and bmeas(i). We assume a starting value for g and calculate bd(i) as

bd(i) = atan2( sin(bmeas(i))/g, cos(bmeas(i))) (11)

This can inserted into (9) to give a calculated value of the error bcalc(i). The difference of these is the measurement error.

error(i) = bemeas(i) –becalc(i)

= bref(i) +bmeas(i) –btrue(i) –a*f(g,bd(i)) (12)

One measure of the goodness of a fit is the sum of the squares of all measurements: S = ∑error(i)2 (13) i

We wish to select the parameters bref and a such that S is a minimum. There will be two simultaneous equations. One is ∂S ∂error(i) = ∑2error(i) = 0 (14) ∂bref ∂bref Observing from (12) that ∂error(i) = 1, (15) ∂bref equation (14) may be manipulated to the form

bref *∑1− a *∑ f (i) = ∑[]btrue (i) − bmeas (i) (16) i i i The second of the two equations is ∂S ∂error(i) = ∑2error(i) = 0 (17) ∂a i ∂a Chapter 8: DIFAR Sensors for Monitoring Bowhead Migration 8-33

From (12) ∂error(i) = − f (i) (18) ∂a Thus equation (17) may be manipulated to give

2 bref *∑ f (i) − a*∑ f (i) = ∑ f (i)*(btrue (i) − bmeas (i)) (19) i i i

Equations (16) and (19) are two linear equations for the unknowns bref and a. In matrix form they are

 nm − ∑ f bref   ∑(btrue − bmeas )   2   =   (20) ∑ f − ∑ f  a  ∑ f *(btrue − bmeas ) where nm is the total number of measurements and all summations are over all measurements. This system

can then be solved for bref and a. bref is used as is. The value of a is used to improve the estimate of the gain factor as follows: the maximum value of the calculated error is found as a · fmax where fmax is found from equation (8) using the old value of g. Then a new value of a is computed by inverting (8):

tbe = tan(a · fmax) (21)

rg = tbe ± 1+ tbe2 (22)

g = rg2 (23) where tbe and rg are just intermediate variables for computing g. Equation (22) gives two possible values of g that are inverse to each other. If tbe is greater than zero, use the smaller value; if tbe is positive, use the larger. The entire calculation is then repeated from (11) above, until a approaches a value of 1.000. This implies that g has attained its final value. In typical cases this took around 5 iterations. Results

Figure 8.24 shows the measured and calculated values from one such set of measurements. It appears that a substantial proportion of the deviations in Figure 8.21 are due to some mechanism such as the gain imbalance. However, there are also other measurement errors. Once the gain correction factor and reference bearing have been determined for a given DASAR, the correction may be applied in either of two ways: 1. by calculating the correction factor in degrees for each measurement, or 2. by multiplying the digitized output of the sine channel by the inverse of g before measuring bearings. The first of these methods is used if the digitized time series has already been used to calculate bearings. The correction is applied to the calculated bearings. The second method is applied if the digitized time series is available but has not already been used for the bearing measurement.

The data plotted in Figure 8.19 were processed according to the previous section. bref was measured as 156.97º with g = 1.0883. The error (the difference between measured and true bearings) when plotted against true bearing was (approximately) 2 cycles of a sine wave with amplitude 2.45º. The data were re- measured after dividing the sine channel by g. The results are in Figure 8.25. 8-34 Monitoring at Northstar: Open-Water Season

FIGURE 8.24. Measured and computed errors for the ping calibrations for DASAR CAa.

FIGURE 8.25. Reference bearing measurements for DASAR CBa after correcting for gain imbalance. Chapter 8: DIFAR Sensors for Monitoring Bowhead Migration 8-35

The gain imbalance correction reduced the scatter (±s.d.) for DASAR CBa from 2.3º to 1.3º. How- ever, there is clearly some additional non-random source of error in this case. This second data set measured with gain correction was run through the same analysis. The results showed that there was no 2-cycle sine wave contribution remaining. Out of curiosity, the data were run again looking for 3- and 4-cycle sine waves. These yielded peak errors of –0.88º and +0.18º, respectively. However, we cannot at present imagine any physical explanation for this apparent 3-cycle dependence. 8-36 Monitoring at Northstar: Open-Water Season

APPENDIX 8.2: LOCALIZATION ACCURACY

The precision of whale call locations, in the form of error (or, more accurately, “confidence”) ellipses, was estimated following the methods of Lenth (1981) under which individual bearings from a DASAR to a received whale call were assumed to follow a Von Mises distribution (Mardia 1972). The Von Mises distribution is analogous to the normal distribution, but applies to directional data in which angles x and x + 360º are the same. The Von Mises distribution has probability density function

−1 f (θ;µ,κ ) = [2πI 0 (κ )] exp[κ cos(θ − µ)]

where µ is the mean bearing, κ is the concentration parameter, and I0 is a modified Bessel function. Assuming s is the standard deviation of bearings calculated in the usual way from a random sample of bearings, the concentration parameter can be calculated approximately as (1− A) 2 (0.48794 − 0.82905A −1.3915A2 ) κˆ −1 ≈ 2(1− A) + A 2 −1 sπ   A = exp   .  2 180   Estimates of s for each DASAR were computed from the gain-corrected calibration data (presented in the Results section). Histograms of the differences between gain-corrected bearings and true bearings (known during calibration) confirmed that it was reasonable to assume bearings followed a Von Mises distribution. When two or more DASARs received the same call, the (x,y) location with maximum statistical likelihood, according to the Von Mises distribution, was used as the estimate of the call’s location. Random errors in bearings (observed bearing – true bearing) from each DASAR to a given call were assumed to be independent. The statistical likelihood of the bearings was constructed and maximized as described in footnote (3). An estimate of the variance-covariance matrix of (x,y),  Var(x) Cov(x, y) Q =   , Cov(x, y) Var( y)  was obtained using formulas given in Lenth (1981) and White and Garrott (1990). Relying on the fact that maximum likelihood estimates are asymptotically normally distributed, the area of a 90% confidence ellipse for the true call position was calculated as

3 The statistical likelihood was a function of the unknown parameters κ and µi, where i = 1, 2, …, indexed the various DASARs receiving the call. The statistical likelihood represented the probability of observing the entire set

of bearings given hypothetical values of µi and κ. To assure that all µi intersected at a common point, the statistical th likelihood was re-parameterized by setting µi = arctan((y-yi)/(x-xi)), where (xi,yi) was the location of the i DASAR

and (x,y) was the point where all µi intersected. After re-parameterization, the statistical likelihood was a function of the unknown parameters x, y, and κ. Given estimates of κ from the calibration data, estimates of x and y (the call’s location) were obtained by repeatedly evaluating the statistical likelihood over a range of values until the probability of the observed bearings was maximized. The (x,y) point that maximized the statistical likelihood was used as the estimate of the call’s location. Chapter 8: DIFAR Sensors for Monitoring Bowhead Migration 8-37

1/ 2 2 a = π Q χ 2,0.10 ,

2 th where χ 2,0.10 was the 10 quantile of a Chi-square distribution having 2 degrees of freedom and |Q| = Var(x)Var(y) – Cov(x,y)2 (White and Garrott 1990). The length of the major axis of the error ellipse was calculated as

2 l1 = 2(λ1χ 2,0.10 )

where λ1 was the largest eigenvalue of Q. Similarly, the angle of the major axis was obtained from the eigenvector corresponding to the largest eigenvalue as

−1 φ1 = cos (ν1,1 )

where ν1,1 was the first element of the first eigenvector of Q (Mardia et al. 1979). Not all calls received by 2+ DASARs were localizable or suitable for inclusion in subsequent analyses. Calls were discarded for one of three reasons. (1) A location could not always be computed because the maximum likelihood procedure occasionally failed to converge on a maximum. Generally, this happened when the geometry of the estimated bearings was unfavorable, i.e., when bearings did not intersect, or when multiple bearings intersected at widely separated locations. Convergence failure meant that neither the location nor its covariance could be calculated. (2) In some cases, the maximum likelihood procedure converged on a location estimate, but the variance-covariance matrix, Q, was not positive definite. Positive definite variance-covariance matrices were required for the corresponding error ellipse to have positive area. Non-positive definite Q tended to occur when bearings were nearly parallel. This typically occurred at locations far from the DASAR array or near lines connecting pairs of DASARs. (3) In a few cases, the maximum likelihood solution was “behind” one of the DASARs. This occurred if the angle reported by the DASAR and the angle to the estimated location differed by 180°. For example, if a DASAR reported a bearing to a call of 45°, and the maximum likelihood procedure converged on a location 225° from the DASAR, the solution was “behind” the DASAR and was discarded. This situation generally occurred when only two bearings were available and when those two bearings were divergent and did not intersect (except if one or both bearings were changed by 180°). To map the typical precision of the estimated call locations near the DASAR array, values of a

(error ellipse size), l1 (long axis length), and φ1 (angle of long axis) were calculated for each call and smoothed using a kernel density estimator modified for continuous data. The usual Gaussian kernel density estimator for point location data (Wand and Jones 1995) was modified for continuous data by multiplying values of the continuous variable by the usual kernel weights and computing a weighted average, i.e.,

n ∑wxX(,ii ,) hy i=1 µˆ ()yx| = n ∑wxX(,i ,) h i=1 where µˆ ()yx| was the estimated mean of variable y at location x (e.g., y = area of 90% confidence th ellipse), Xi was the coordinates of the i observed location, yi was the value of the continuous variable at the ith observed location, h was the smoothing parameter, and 8-38 Monitoring at Northstar: Open-Water Season

 1 T  wxX(,iii ,) h=−− exp ()() x X x X 2h2 

(Bailey and Gatrell 1995). The smoothing parameter h dictated the degree of smoothing and for this analysis h was set to 1000 meters. The weighted average, µˆ ( yx| ), was computed at every intersection of a 500 × 500 meter grid. The resulting surface of µˆ(y | x) was then plotted for inspection. CHAPTER 9: A STATISTICAL APPROACH FOR ASSESSING DISPLACEMENT OF BOWHEAD WHALE LOCATIONS ASSOCIATED WITH ELEVATED SOUND LEVELS 1

by Trent L. McDonald a and W. John Richardson b

a Western EcoSystems Technology, Inc. 2003 Central Ave., Cheyenne, WY 82001 (307) 634-1756; [email protected]

b LGL Ltd., environmental research associates 22 Fisher St., POB 280, King City, Ont. L7B 1A6 (905) 833-1244; [email protected]

for

BP Exploration (Alaska) Inc. Dept. of Health, Safety & Environment 900 East Benson Blvd, P.O. Box 196612 Anchorage, AK 99519-6612

LGL Report TA2706-4

December 2003

1 Chapter 9 In: W.J. Richardson and M.T. Williams (eds.). 2003. Monitoring of industrial sounds, seals, and bowhead whales near BP’s Northstar oil development, Alaskan Beaufort Sea, 1999–2002. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 9-2 Monitoring at Northstar, Open-Water Season

ABSTRACT

In many environmental monitoring and impact studies, the animals closest to the source of a stimulus are more likely to be affected by the stimulus than are animals far from the source. Also, effects on the closer animals are likely to be stronger. The primary interest in many situations is estimating the magnitude of stimulus levels that cause change in the lower tail of a distance measurement’s distribution, where maximal effects might occur. Here, a statistical method is proposed to estimate the deflection of animal locations (or alternative variables that are subject to disturbance effects) in relation to levels of a time- varying stimulus that diminishes in intensity through space (such as sound). The proposed method focuses on univariate distance distributions and estimates the relationship between a distribution’s lower quantiles and levels of the stimulus. To do this, quantile regression is applied. Quantile regression requires that animal locations and levels of the stimulus be recorded over a period of time, and that stimulus levels vary over this interval. Complications include overcoming lack of independence in responses, non-linearity of quantile relationships if there is no effect, and the fact that researchers may not know (a priori) how long levels of the stimulus need to be elevated before a response occurs and becomes large enough to be detectable. A detailed example is given describing application of the method to a study of bowhead whales (Balaena mysticetus) migrating past an oil and gas production island that is a source of time-varying levels of underwater sound.

INTRODUCTION

A key question in many environmental monitoring situations is whether or not animals are displaced or deflected by some sort of man-made disturbance. In many situations, the potential displacement may be caused by a man-made stimulus that varies over time and diminishes with distance away from the human activity. Sound is the most obvious type of stimulus that varies through time, diminishes with increased distance form the source, and commonly elicits disturbance effects. However, other stimuli with these characteristics exist, such as air or water borne pollutants and odors, and perhaps certain visual stimuli whose conspicuousness varies both with time and distance. For example, displacement of ungulates might be related to time-varying noise from logging, vehicle traffic, or other industrial activities (e.g., Rowland et al. 2000; Dyer et al. 2001). Some migrating whales are known to deflect off-course in response to underwater noise produced by offshore oil exploration (Richardson et al. 1995b). Most human developments are unlikely to disturb all members of an animal population. In these cases, it would be desirable to use a data analysis technique that estimated the proportion of the population affected by the disturbance, and the extent of disturbance to that proportion of the population. When the development is outside or at one edge of the population’s range, it is most likely to affect animals occurring close to the development, and is likely to affect those individuals most strongly. It is possible in such cases for the disturbance effect to be confined to animals on the proximal edge of the range, and for the average animal’s location to be unaffected. To detect and estimate the magnitude of such an effect, it would be desirable to use an analysis technique that focuses on the fringes or tails of some distribution associated with the population where the effect is expected to be most pronounced. For example, the average offshore distance of migrating bowhead whales (Balaena mysticetus) near an oil production island in nearshore waters may not be affected by underwater sounds produced by the island and its associated vessels if the industrial sounds diminish to low or undetectable levels at the middle of the migration corridor. However, the whales closest to the island may deflect farther offshore when large amounts of underwater sound are being produced. It would be of interest to have a method that would address whether there is evidence of Chapter 9: Statistical Approach for Assessing Displacement 9-3 displacement of the closer whales at times with high stimulus levels, and if so, how the displacement effect varies with stimulus level. Such a technique should be designed to detect effects even if they are limited to one edge of the distribution, allowing for the likelihood that there will be either a lesser effect or no effect on the average animal. This paper outlines a statistical analysis designed to estimate changes in the tail of a distance distribution that is associated with a time-varying and distance-varying stimulus (like sound). The analysis makes use of an established but little-known statistical technique called quantile regression (Koenker and Basset 1978) to associate changes in the tails of a distribution with increased (or decreased) levels of a stimulus produced by a human activity. The analysis allows estimation of a wide variety of functional relationships between various quantiles of an animal population’s distribution and stimulus levels, including linear, polynomial, non-parametric (such as spline), and threshold relationships. The threshold relationship may be particularly useful because it is flexible, and can model a progressively increasing quantile above some threshold stimulus level, and either zero slope or a different functional relationship at lower stimulus levels. This technique is useful in situations where no true reference or control sites exist for the human activity whose effects are under scrutiny. Reference sites are not required by the technique because low and high levels of the disturbance stimulus effectively function as the “reference” and “treatment” situations in a regression setting. As such, this technique will only be useful if the disturbance stimulus varies in a meaningful manner over time. Further, if the disturbance stimulus varies faster than animals can react and change locations, and if stimulus levels cannot be quantified over longer time periods, the method might be of little practical use. The underlying sound avoidance model assumed by the method presupposes that animals in the general area of the disturbance source increase their distance from the source when the stimulus emanating from the source increases. Under this model, maximum displacement potential exists for animals nearest the source during periods of high stimulus production. Conversely, when little stimulus is being produced, animals are expected to naturally occur closer to the disturbance source. Thus, displacement is expected to diminish as the strength of the stimulus (e.g., source level of man-made sound) decreases or as initial distances from the source increase. Animals far from the source are expected to show little or no displacement even at high-stimulus times due to the decay of received stimulus levels with increasing distance. As a result, displacement of the lower tail of the animal’s distance distribution should provide evidence for displacement if it occurs and if this avoidance model is correct. The project that prompted development of the technique was designed to monitor the southern edge of the bowhead whale migration corridor near an artificial island constructed 5 km offshore from barrier islands in the Beaufort Sea near Prudhoe Bay, Alaska. The artificial island, named Northstar, was built and is being operated by British Petroleum (BP) as an oil production platform. In autumn, bowhead whales migrate past Northstar, traveling parallel to shore at widely varying distances offshore of the island. The closest whales pass within several kilometers of the island during at least some years, within the zone where they would be exposed to underwater sound from the island at times with high sound emissions (Chapter 6). Monitoring of the whales was required to address regulatory requirements under the Marine Mammal Protection Act, and concerns by subsistence hunters that their hunt would be disrupted by offshore displacement of the whales migrating closest to shore. The key objective of the monitoring was to estimate the offshore displacement of the southern edge of the bowhead migration corridor, if any, at times when higher-than-average levels of underwater sound were being emitted from Northstar (Chapters 7, 10). The possibility of broader scale effects on portions of the migration corridor distant from Northstar, either >26 km offshore or >10 km alongshore, were not investigated. It is assumed that underwater sound effects, if any, are maximal for whales <26 km offshore of Northstar, and 9-4 Monitoring at Northstar, Open-Water Season if no effects were found near Northstar, no sound effects would be expected to occur in regions farther offshore. As part of the monitoring, underwater sounds were recorded continuously via hydrophones located about 400 m north of the island throughout much of the bowhead migration season (Chapter 6). In addition, an array of 11 Directional Autonomous Seafloor Acoustic Recorders (DASARs) was deployed 6–21 km offshore of Northstar Island to record and locate the origins of bowhead whale calls. The DASARs provided a large number of whale call locations within the southern part of the migration corridor between 29 Aug. 2001 and 2 Oct. 2001 (Chapters 7, 8). The next section describes the methods designed to estimate a change in the tails of a distance distribution associated with increased levels of a stimulus. In that section, the data necessary to carry out the analysis and the basic quantile regression procedure are described. Included is a permutation-based procedure to estimate the shape and variation of the quantile regression line if no effect were occurring. The section after that contains an example of the technique as applied to the bowhead monitoring data collected near Northstar Island. A brief discussion and summary follow. Chapter 10, following this one, describes application of the present technique to the whale monitoring data from both 2001 and 2002.

METHODS

Response Distribution

The analysis of disturbance described here involves detecting and characterizing changes in the statistical distribution of a response variable that is expected to change as levels of a stimulus increase or decrease. In many situations, distance to the source of the stimulus is a natural choice for the response variable because distance quantifies an important aspect of the spatial distribution of animals surrounding the source. For example, if fish are potentially disturbed by warm effluent from a power plant during periods of high power production, it makes sense to measure the radial or in-stream distances of fish from the plant’s outfall during periods of high and low power production. If elk are disturbed by vehicle traffic along a road, it makes sense to measure the distances of radio-collared elk to the nearest point on the road during periods of high and low vehicle traffic. In this case, it is probably most appropriate to view the stimulus (noise and traffic) as emanating from the entire length of the source (road). In other cases, it may make sense to measure the distance of animals from a landscape feature other than the position of the stimulus source. For example, if whales typically migrate roughly parallel to shore, and an activity on or near shore might deflect them farther offshore, it may make sense to measure distance of the whales from the coast (or from a line parallel to the coast), not radial distance from the source point. An unaffected whale migrating parallel to the coast past a sound source would first show large radial distances, then shorter radial distances as it passed the source, and then longer radial distances again. In contrast, the same whale would display a constant offshore distance. The latter measure would provide a simpler basis for determining whether varying levels of the source had an effect on distance. Other response variables aside from distance are possible. For example, one could measure a behavioral variable in relation to stimulus level, and assess whether the distribution of that variable changed in relation to stimulus level. Alternatively, it might, at least in theory, be possible to measure the levels of a short-lived hormone or enzyme in the blood of animals exposed to varying levels of a stimulus that produces stress. In any case, a univariate measure of the potential response must be defined, and the statistical distribution of that response must be sensitive to the type of disturbance postulated. If the disturbance stimulus is produced continually by the human activity, but at varying levels, it may be necessary to accumulate or summarize values of the stimulus for some time period immediately Chapter 9: Statistical Approach for Assessing Displacement 9-5 preceding measurement of the response variable. For example, if the stimulus is sound, a strong sound of short duration may not induce an animal to react, but exposure to strong sounds (or perhaps weaker sounds) for extended periods may cause a reaction. Thus, sound levels should be summarized or averaged for a period of time immediately preceding measurement of the response. The exposure time that induces an animal to react will often be unknown a priori. If so, it will be necessary to estimate the relationship between stimulus levels and responses separately for several different exposure times. The exposure time producing the largest response can then be identified. If this is done, common sense and caution should be employed to avoid violating scientific principles and inflating experiment-wide significance levels. For example, the separate exposure times to consider can be specified a priori, rather than investigating ever-increasing exposure times until a significant effect is found. Quantile Regression

This method for detecting disturbance is based on the relationship between quantiles of the response distribution and varying levels of the stimulus, as compared with the relationship expected if stimulus levels do not affect responses. For example, the 5th quantile of a response distribution may be similar to that expected under the assumption of no effect at low or moderate stimulus levels, but higher than expected at high stimulus levels (Fig. 9.1). The main goal of this approach is to estimate the relationship between a quantile (or quantiles) of the response distribution and levels of the stimulus, and compare that to the range of relationships expected under the assumption of “no effect” (i.e., statistical independence). If the observed relationship falls outside the range expected by chance, the magnitude of displacement as a function of stimulus level can be described. For example, the magnitude of change in a certain quantile when stimulus levels increase from x to y can be computed. This, in conjunction with data on the distribution of stimulus levels and (ideally) estimates of other quantiles, can be used to estimate the proportion of the population affected when stimulus levels increase from x to y. The number of individuals affected can also be estimated under this scenario if an estimate of population size is available. It should be noted that if the rate at which responses are acquired is related to stimulus levels, the analysis cannot differentiate between shifts in quantiles and changes in measurement acquisition rates. The basic analysis utilized in this method is quantile regression (Koenker and Bassett 1978). A quantile of a distribution is a number below (or above) which a certain percent of the distribution lies (also called a percentile). For example, if the 5th quantile of a distribution is x, 5 percent of the observations in the distribution fall below x. Contrary to regular regression methods that estimate a relationship between the mean of a response variable’s distribution and one or more predictor variables, quantile regression estimates a relationship between a quantile (e.g., the median = 50th quantile, or the 5th quantile) of a response’s distribution and predictor variables. Quantile regression is used here to estimate the relationship between a quantile of the disturbance distribution and stimulus levels. The general form of the quantile regression equation is

Qyτ [ ]= β011+++β x ... β ppx where Qτ[y] denotes the τ-th quantile of the random variable y, x1 through xp denote predictor variables, and

β0 through βp denote unknown coefficients to be estimated. ˆ Coefficients of the quantile regression, βi , are estimated by minimizing a sum of weighted residuals, where each residual’s weight depends on whether the observation was above or below the th th estimated quantile line. Let yi be the i observed response and Xi be the i vector of predictor variables associated with yi. For example, the vector of predictor variables, Xi, could equal [ 1, stimulus level, 9-6 Monitoring at Northstar, Open-Water Season

Observed 5th quantile at 105 = 18 km Distribution at 105 Expected = 17 km

45 Observed 40

35 Relative Frequency Relative 0455 10152025303540 Distance from baseline (km) 30

Observed 20 Distribution at 120 15

10 Distance from baseline (km)

5 Expected

0 85 90 95 100 105 110 115 120 125 Relative Frequency 0455 10152025303540 Stimuli produced by source Distance from baseline (km) 5th quantile at 120 = 28 km Expected = 22 km

FIGURE 9.1. Pictorial representation of the method for detecting disturbance based on comparing an observed and expected relationship between a specific quantile of a response distribution and varying levels of a stimulus. In this example, the response is “Distance from baseline”, similar to the bowhead whale example given later. Two response distributions are shown, one at 105 units of stimulus (upper left) and one at 120 units of stimulus (lower left), with their respective 5th quantiles. If the 5th quantile for each stimulus level were plotted, the observed (solid) line in the right hand graph would result. The expected 5th quantile of distance for each stimulus level (dashed line), assuming independence of the stimulus and response, does not plot as a straight line with zero slope, but is established by a permutation method (see Methods).

2 (stimulus level) ]. The model estimated by this Xi vector would include an intercept (quantified by the 1) plus linear and quadratic effects of the stimulus level. Throughout this paper, it is assumed that the Xi vector only contains variables that are functions of the stimulus levels, and not functions of other variables like age, sex, season, etc. The “Discussion” has a brief note on why this is necessary and on potential extension of this technique to include non-stimulus covariates. To estimate the coefficients, an iterative updating process is used to minimize the weighted sum of ˆ residuals. During the initial iteration, initial estimates of all βi values are made and residuals are calculated ˆ as ri = (yi - Xi β ). The initial estimates have no effect on the final results except in cases where the initial estimates are several orders of magnitude from the correct values and yield pathologically incorrect models. th Weights associated with the i residual are calculated as w1i=τ-I(ri < 0), where I( ri < 0) is an indicator function taking the value 1 if ri < 0, and 0 otherwise (recall, τ is the quantile of interest, e.g., τ=0.05 for the 5th quantile, τ=0.25 for the 25th quantile). The sum of weighted residuals to be minimized is Chapter 9: Statistical Approach for Assessing Displacement 9-7

nn swr11==−<∑∑ rii w r iτ I() r i 0  , ii==11 and final estimates of coefficients satisfy

n ˆ   βτβ==min{}swr1 min β∑ rii w1  . i=1

Direct minimization of swr1 can be carried out using any one of a multitude of nonlinear minimization routines such as nlminb() available in S-Plus (MathSoft 1999) and Proc Nlin in SAS (SAS Institute 1999). General minimization routines usually implement a variant of the Newton-Raphson search algorithm. In some settings, particularly when locations of animals are being analyzed, a measure of the precision of the responses may be available. In this case, it makes sense to further weight residuals of the quantile regression by the inverse of estimated precision of the response. Supposing that the estimate of precision for response i is ai, weights associated with the precision of the i-th response could be w2i = n -1 -1 naij/ ∑ a , where n is total sample size. In this case, the minimized sum of weighted residuals is j=1

n nn -1 -1 (1) swr212jiii==∑∑∑ riii w w n / araτ-I() r<0  i=1 j=1 i=1 This effectively down-weights those observations with poor precision. Adoption of a specific form of weighting function, such as the inverse of estimated precision (1/a), as suggested here, will sometimes be based on theory or precedence. More often it will be based on a subjective a priori judgment regarding the utility of the derived weights in a specific setting. Form of Relationship

The method for detecting disturbance first establishes a plausible model for the relationship between a particular quantile of the response and stimulus level. Establishing a plausible model should follow accepted model selection practices, and may include choosing a model from among a set of models based on significance levels of coefficients, forward or backward or stepwise variable selection, and professional judgment. Once a plausible form of the quantile regression model is chosen, the method computes (1-α) % bounds for the model under the assumption that responses and stimulus levels are statistically independent. If the observed quantile regression relationship goes outside these bounds at some stimulus levels, significant disturbance can be inferred to occur at those stimulus levels. Establishing bounds for the regression if response and stimulus are independent is covered in the next section. The remainder of this section discusses some useful regression models that could be applied in particular situations. The key feature of the estimated quantile regression equation is that it fit a particular quantile of the data’s scatter-cloud reasonably well. In some applications, particularly when the distribution of the stimulus is more-or-less uniform, it may be possible for the linear model

Qy=τ []β01+ β x , where x is the stimulus level, to fit adequately. In other applications, it may be necessary to use higher order polynomial functions of x, such as quadratic or cubic, i.e., 9-8 Monitoring at Northstar, Open-Water Season

2 Qy=τ []β01+ β x+β 2x or

23 Qy=τ []β01+ β x+β 2x+β 3x . In some applications, it may be useful to fit a threshold model for biological reasons, or to approximate a complex function with a series of straight lines. Threshold models, called “piecewise linear” by Neter et al. (1983, p. 346) or “multiphase regression” by Seber and Wild (1989, p. 433), fit two or more linear line segments in different ranges of a predictor variable. If a threshold at x = T1 has been provisionally identified, a 1-threshold model estimates two line segments that connect at x = T1 and that have separate slopes above and below x = T1. This type of single threshold model fits the equation

Qy=τ []β01+ β x + β 2Ix>Tx-T ( 1 )( 1 ),(2) where I(x>T1) is again an indicator function for the event x > T1 and β1, β2, and β3 are coefficients to be estimated. Slope of the estimated line for x < T1 is β1, while slope for x > T1 is (β1+β2). A 2-threshold model estimates one slope for x < T1, no slope for T1 < x < T2, and a separate slope for x > T2 (T1 ≤ T2). This model is given by

Qy=τ []β01++β IxTx-T ( 2 )( 2 ).(3)

If T1 = T2, the two-threshold model collapses to the single threshold model in Equation (2).

In practice, the thresholds T1 and, optionally, T2 will not be known and must be estimated. Estimation of T1 (and T2) is accomplished either by use of a grid search over the range of possible values for T1 (and T2 subject to T1 < T2) or by specifying the threshold(s) as additional parameters in the nonlinear minimization methods required to estimate the β coefficients. When thresholds are fitted in the model, derivatives of the parameters are not continuous and nonlinear minimization routines that rely on these derivatives being continuous may have trouble finding the minimum sum of weighted residuals. In this case, a manual (or automatic) search for the minimum sum of weighted residuals over a grid of possible threshold values will assure that a (global) minimum of the sum of weighted residuals is achieved. Another type of model that should be applicable in a wide range of situations employs spline functions (de Boor 1978; Green and Silverman 1994; Venables and Ripley 1999). The use of spline functions accomplishes scatter plot smoothing. This results in a fitted relationship that is quite general in shape, fitting a point cloud with just about any shape. However, the fitted coefficients may be difficult (or impossible) to interpret biologically. Regression splines split the range of the stimulus (the x range) at a list of values known as knots. Between each pair of knots, the regression relationship is a cubic polynomial that is continuous at the knots and has continuous first and second derivatives (Venables and Ripley 1999, p. 282). The form of the quantile spline regression equation is

Qy=τ []β00bx ()+ β 11b () x +...+β k+4k+4bx () where the bi(x) are the components of the spline (which are in turn functions of the stimulus level x) and k is the number of specified knots. The coefficients β0 through βk+4 will rarely, if ever, be interpretable biologically, and it may be useful to think of the spline regression as

Qy=xτ []fk () Chapter 9: Statistical Approach for Assessing Displacement 9-9 where f(x) is simply a smooth function of the stimulus levels x. Large values of k (many knots) result in more complex functions. While cross-validation and other data-driven techniques can attempt to choose an optimum k, in practice k will generally be chosen based on a judgment that the fitted line adequately fits the data scatter cloud While splines can be difficult to visualize and even more difficult to compute by hand, their implementation is straightforward provided a suitable computer routine is used. The computer package S-Plus implements a particular parameterization of splines, known as B-splines, in its function named bs. Expected Quantiles Under No Effect

Before assessing the degree to which the response (as summarized by a fitted quantile regression) depends on the stimulus, it is necessary to understand the expected relationship if there is no effect. Under the assumption that the response distribution is unrelated to (i.e., statistically independent of) the stimulus distribution, the relationship between a particular quantile of the response and stimulus can, but will not always, be a horizontal line. This complicates detection and estimation of a disturbance effect. Even in the absence of any effect, the relationship may be linear, quadratic, higher order polynomial, logarithmic, threshold, or more complex (Fig 9.2). It appears that the slope and shape of a quantile line depend on the shapes of the response and stimulus distributions, and in particular the amount of skewness in each. Thus, unless assumptions can be made about the shape of the response and stimulus distributions, no assumption can be made regarding the shape of the quantile regression in the absence of an effect. A permutation or stratified permutation method (Manly 1997) is proposed to estimate what the quantile regression model would look like under the assumption of no relationship between responses and stimulus. The proposed method takes account of the actual distributions of response and stimulus values. Assuming Responses are Independent Assuming, temporarily, that all responses are statistically independent of one another, and that responses and stimulus are independent, any observed response could have occurred with any observed stimulus level with equal probability. Consequently, any random permutation of the response values would be equally likely to have occurred under these assumptions. Again, temporarily assuming responses are independent, the suggested permutation method is as follows: 1. Construct a random permutation of the responses. A random permutation is a random shuffling of the order of the response values. 2. Pair the random permutation of responses with the original (un-permuted) stimulus values. This step randomly associates each stimulus value with one response value. 3. Fit the quantile regression using the randomly permuted data and store the fitted line. 4. Repeat steps 1 – 3 a large number of times (say 500). For a given stimulus level, the mean of the 500 response values generated by the permutation method is the best estimate of the true quantile assuming no relationship between responses and stimulus. The best estimate of the relationship between a given quantile and stimulus level in the absence of any stimulus effect is derived by connecting the mean response values at sequential stimulus levels (e.g., solid lines in Fig. 9.2). Confidence intervals for the 5th quantile line, assuming no relationship, are constructed based on the distribution of the quantile regression lines fitted to the permuted data. For example, the values of the 500 5th quantile regression lines at a particular stimulus level are 500 estimates of the 5th quantile at that stimulus level 9-10 Monitoring at Northstar, Open-Water Season 5 101520 5 101520 Normal random variable Normal random variable

90 95 100 105 110 80 90 100 110 120 Normal random variable Uniform random variable 10 15 20 25 Left-skewed random variable 5 10152025 Left-skewed random variable

95 100 105 110 80 90 100 110 120 130 Right-skewed random variable Normal mixture random variable

th FIGURE 9.2. Four plots of statistically independent random variables showing the 5 quantile of y (solid line) and its 90% bounds (dashed lines). Under repeated generation of the x and y random variables, 90% of the observed 5th quantiles are within the bounds shown as dashed lines on each plot. Note that the 5th quantile (solid line) is appreciably curved in one graph, and not entirely straight in two of the other three graphs. if responses are not related to stimulus levels. The 5th and 95th percentiles of the 500 values across all stimulus levels form a 90% bound on the quantile regression line when responses are not related to stimulus levels. If a 95% confidence bound is preferred, then the 2.5th and 97.5th percentiles of the 500 values can be used. If the direction of a possible disturbance effect can be predicted, as is typical, one-sided confidence intervals can be constructed from the results of the permutation exercise. A 95% one-sided confidence bound can be constructed by considering only the upper 95th percentile of the 500 permutation quantiles. If the observed quantile line falls above this one-sided confidence bound, we can infer that the quantile was displaced farther from the source than expected under no effect. Interdependence of Responses In most situations where animal locations are analyzed, some amount of temporal correlation will exist among locations. Furthermore, in some situations it may not be possible to determine whether two or more animal locations represent the same individual. The latter problem results in a lack of statistical independence and can occur, for example, during aerial surveys, or with remote autonomous cameras, or with acoustic detection of animal calls. If responses are correlated in time, proper permutation of responses on stimulus levels, as described in the previous subsection, is more complicated. If correlation exists among the responses, the permutation method Chapter 9: Statistical Approach for Assessing Displacement 9-11 repeatedly shuffles groups of responses that are assumed to be independent (or uncorrelated) onto stimulus levels. The size and structure of response groups that are independent must be estimated prior to implementing the permutation method. For example, if whale locations are being determined from recorded whale calls, individual whales cannot be identified and the amount of time that must elapse between observations before they become independent needs to be estimated. In many cases, Moran’s I statistic (Moran 1950; Cressie 1993, p. 427) can be used to estimate the time (or distance) required for responses to become uncorrelated. When responses are correlated, the permutation procedure is a moving block method similar to that of Fitzenberger (1997). The permutation method that should be used to estimate bounds on the quantile regression, under the assumption that responses are not related to stimulus levels, is as follows: 1. Assume that the amount of time that must elapse between responses in order for them to be uncor-

related is t. Assume the earliest response was recorded at time t0, and that the latest response was

recorded at time tn. Let ∆t = tn − t0. Effectively, t0 is the beginning of the study, tn is the end of the study, ∆t is the duration of the study, and t is the time interval that will be permuted.

2. Construct a grid of time intervals by first choosing a random start between t0 – t and t0. Let this random start be denoted by m, where t0 – t < m < t0. Then define the grid of time intervals as (m, m+t], (m+t, m+2t], …, (m+(k-1)t, m+kt], where k = the smallest integer greater than ∆t/t (i.e., k = ∆t/t), and the notation (x,y] implies an interval where x < t ≤ y. Within each time interval, responses are potentially correlated, but between intervals, responses are uncorrelated. For example, if t = 2

hours, t0 = 00:00:00 on 1 Sept. 2001, tn = 00:00:00 on 2 Sept. 2001 (thus, ∆t = 24 hours and k = 13), and the random number m = 22:37:00 on 31 Aug. 2001, the grid of points constructed in this step (all in 2001) would be (22:37:00 31 Aug, 00:37:00 1 Sept.], (00:37:00 1 Sept., 02:37:00 1 Sept.], … , (22:37:00 1 Sept., 00:37:00 2 Sept.] 3. Randomly permute the time intervals. Continuing the example in step 2, there were 13 intervals of length 2 hours constructed between 22:37:00 31 Aug. and 00:37:00 2 Sept. Label these intervals from 1 to 13 and then randomly permute the numbers 1, 2, … , 13. If the random permutation resulted in the ordering {5, 3, 6, 2, 7, 9, 10, 1, 8, 12, 13, 4, 11}, the first interval would be (06:37:00 1 Sept., 08:37:00 1 Sept.], the second would be (02:37:00 1 Sept., 04:37:00 1 Sept.], and so on. All response values should remain associated with their original time intervals. Stimulus levels should not. 4. Reorder the response values according to the random permutation of intervals, but do not reorder responses within intervals. For example, if interval (06:37:00 1 Sept., 08:37:00 1 Sept.] contained 5 responses and interval (02:37:00 1 Sept., 04:37:00 1 Sept.] contained 10 responses, the 5 values from (06:37:0 1 Sept., 08:37:00 1 Sept.] would be the first 5 values in a new “permuted” data set, the 10 values from (02:37:00 1 Sept., 04:37:00 1 Sept.] would be the next 10 values in the new “permuted” data set, and so on. 5. Associate the reordered responses with the original ordering of stimulus levels. Assuming that there were n responses and associated stimulus levels to begin with, step 4 results in a restricted random ordering of the n responses, but not of the n stimulus levels. Here, the reordered responses are matched with the non-reordered stimulus levels so that responses measured at one time are paired with stimulus levels measured at other times. 6. Re-fit the quantile regression using the randomly permuted data resulting from step 5, and store the results (i.e., the fitted line). 7. Repeat steps 2 – 6 a large number of times (say 500). 9-12 Monitoring at Northstar, Open-Water Season

This permutation method is valid because it assures that responses are independent of predictor variables in each permuted data set (permuted data set results from step 5) and the quantile regression fit to the permuted data (step 6) represents a random realization of the regression under the hypothesis of independence. Furthermore, the permutation method preserves the correlations among observations within time intervals, and treats the time intervals (of length t) as the basic unit of replication. After 500 realizations of the quantile regression are generated, confidence intervals for the regression relationship, assuming no stimulus–response relationship, are constructed in the same manner as in the independent case. In some applications, it will be obvious that the total number of responses declined or increased through time as a result of emigration or change in behavior. In these cases, it may be necessary to implement a stratified permutation method wherein time intervals are only permuted within strata. For example, if the total number of responses per unit time after ts was different from that before ts, one could define two strata as (t0, ts ] and (ts, tn]. The intervals defined in step 2 would then be restricted to each stratum, and the permutations of steps 3 and 4 would only permute intervals within strata. In step 6, stratum boundaries are ignored and the model is fitted to all data. If the observed quantile regression line goes outside the upper or lower bounds that were constructed assuming no relationship between responses and stimulus, we can infer that the observed quantile regression estimate at a particular stimulus level was unlikely (probability <10%) to have occurred by chance if in fact responses were unrelated to stimulus levels. Therefore, we can conclude with 90% confidence that the assumption of no relationship is false and that responses and stimulus levels are related. Furthermore, using this technique we can identify the stimulus levels where displacement is significantly different from that expected by chance alone. A point estimate of displacement at those levels is the value of the original quantile regression at a particular level of stimulus minus the mean quantile regression obtained by randomly permuting the responses. A confidence interval on the estimated displacement at a particular stimulus level is simply the difference between the original quantile regression estimate and the lower and upper endpoints of the confidence bounds constructed by permuting the responses.

EXAMPLE: NORTHSTAR AND BOWHEAD WHALES, 2001

In 1999 and 2000, BP proposed to develop and apply an acoustic localization technique to document the occurrence and locations of calling bowhead whales in the southern part of their migration corridor near BP’s Northstar artificial island. The Northstar facilities, including the island, were then being planned and built in preparation for oil production offshore from Prudhoe Bay, Alaska. Design of a study to assess the effects of Northstar on bowhead whales was challenging given the difficult study environment (Barnes et al. 1984), high inherent variation in whale distribution (Moore 2000; Treacy 2000), and the fact that migrating bowheads are below the surface and invisible ~90% of the time (Richardson et al. 1995a). Also, no nearby location seemed very appropriate as a reference area, given what was known about whale activities and other human activities in the region, and the island was expected to produce underwater sound more or less continuously, with no truly inactive times. Although underwater industrial sounds from Northstar were expected to be more or less continuous, relatively large temporal fluctuations in sound levels were anticipated due to differing types of operations at Northstar. Utilizing this fact, the monitoring approach was designed to determine whether the offshore distribution of calling whales was related to temporal variability in sounds emitted by Northstar operations. If so, this would suggest that Northstar sounds affect the movements (or calling behavior) of migrating whales. To implement the acoustical monitoring approach near Northstar, it was essential to determine (1) the locations of calling whales in the southern part of the whale migration corridor, and (2) levels of Chapter 9: Statistical Approach for Assessing Displacement 9-13 underwater industrial sound near the island. Because any displacement effect was expected to be relatively small in geographic scale, and evident primarily (or solely) in the southern (inshore) part of the whale migration corridor, only the inshore part of the migration corridor needed to be monitored. If island sounds caused whales near the island to swim farther offshore, and all other aspects of whale behavior were the same, it was hypothesized that calling whales would tend to be farther offshore during times with high sound emission than during times with less Northstar sound. This method was intended only to address questions about the occurrence and geographic scale of displacement effects more or less directly offshore of Northstar. Questions about the distance east of Northstar at which any displacement effect might begin, and the distance west of Northstar to which displacement might persist, are of considerable interest but were outside the scope of this study. Methods

Whale calls were localized from 29 Aug. 2001 to 2 Oct. 2001 using 11 Directional Autonomous Sea- floor Acoustic Recorders (DASARs) deployed in two overlapping hexagonal arrays within the southern part of the whale migration corridor (Fig. 9.3). When a call occurred, each DASAR that received the call provided a bearing (Chapter 8). Standard triangulation techniques (Lenth 1981; White and Garrott 1990; Chapter 8) were used to estimate the location of the call near the intersection(s) of bearings from multiple DASARs. For each call location, a 90% confidence ellipse was calculated based on the number of DASARs that received the call, disparity of the estimated bearings, and estimated inherent variation in bearings from each DASAR, assuming bearings followed the von Mises distribution (Mardia 1972; Chapter 8). The perpendicular distance from the best estimate of each call’s location to a line oriented parallel to the general trend of the coast and through Northstar was computed. This measure of “offshore distance” was used as the response variable. Northstar was potentially affecting whales in the southern edge of the migration corridor if the lower tail of the distribution of offshore distances increased as industrial sounds from Northstar increased. Underwater sounds produced by the island and associated vessels were measured from continuous recordings of sounds received by a hydrophone positioned near the island and just above the seafloor. Water depth around Northstar is 12 m. The island hydrophone was placed 400 m seaward (north) of the north edge of the island, and was connected by cables to digital sound recording apparatus on the island. The sampling rate of the digital recording system was 2000 samples/s, providing acoustic data up to ~1000 Hz. Sound received 400 m from Northstar was partly of industrial origin, but also partly of natural origin due to the sounds generated by wind and wave action. Analysis of the 2001 acoustic data showed that sounds in five one-third octave frequency bands, centered at 31.5, 40, 50, 63, and 80 Hz, was predominantly of industrial origin. The total sound pressure in these one-third octaves, expressed in dB re 1 µPa, was used as an Industrial Sound Index (ISI; see Chapter 6 and Appendix 6.1). Details regarding construction, deployment, operation, calibration, and retrieval of the gear, and processing of data both to locate whales and to measure island sounds, can be found in Chapter 8 and, specifically for 2001, in Greene et al. (2002). The analysis was limited to calls within the “analysis zone” outlined in Figure 9.3. Calls at locations >10 km from the center of the DASAR array (unit CB) or >10 km from a line connecting CB to Northstar were excluded. Although many calls were detected at greater distances, the DASAR array could not reliably detect calls at distances greater than 10–12 km during times with high background noise (Greene et al. 2002; Chapter 8). Offshore distances of calls in the “analysis zone” were analyzed for a disturbance effect by estimating the relationship between the 5th quantile of offshore distance and ISI using the two-threshold model in Equation (3). The two-threshold model was sufficiently general to fit the data adequately. The square 9-14 Monitoring at Northstar, Open-Water Season

FIGURE 9.3. DASAR locations, analysis area, and estimated locations of calling whales detected near Northstar Island from 29 Aug. to 2 Oct. 2001. Localized calls are color-coded by average industrial sound index (ISI) as received 400 m north of Northstar during the 60-min preceding detection of that call. Chapter 9: Statistical Approach for Assessing Displacement 9-15 root of the area of each call’s 90% confidence ellipse was used to weight the quantile regression minimizing function as in Equation (1). Actual weights used in the quantile regression were proportional to a-1/2, where a = area of the 90% confidence ellipse, and scaled to sum to the number of received calls. This generally resulted in observations far from the center of the DASAR array and calls with poor precision being down-weighted during the estimation process. The duration of sustained high sound levels necessary to induce a calling whale to displace (or to call less often) was unknown. Consequently, Northstar ISI levels associated with each whale call were averaged over several different time periods immediately preceding the call and separate quantile regressions (one for each averaging time) were used to obtain separate estimates of displacement in the 5th quantile. The time periods considered covered the anticipated range of times necessary to manifest an effect in whales. Twenty-one time periods were considered, ranging from 5 to 160 minutes (i.e., 5, 10, 15, 20, 30, 40, 45, 50, 60, 70, 75, 80, 90, 100, 105, 110, 120, 130, 140, 150, and 160 minutes), and were established before estimation began. For each averaging time, the two-threshold quantile regression of Equation (3) was fitted. To estimate the apparent displacements, the difference between the estimated 5th quantile of offshore distance and the mean quantile expected if there was no effect was computed for each of ISI = 73, 75, 77, …, 129 dB re 1 µPa. For example, suppose that the estimated 5th quantile of offshore distances was 15 km when ISI averaged over the 60 minutes preceding detection of the call was 120 dB, and that under the assumption of no effect, we expect that the 5th quantile would be 10 km offshore. The estimated offshore displacement of the 5th quantile at a stimulus level of 120 dB (measured 400 m from Northstar) would then be calculated as 15 km − 10 km = 5 km. Confidence intervals for the parameters of the quantile regression and derived estimates were constructed using the stratified permutation method (Manly 1997). From collected data, it was impossible to know whether two calls detected at similar times were from separate or independent whales. As a result, it was necessary to estimate the amount of elapsed time necessary between two calls in order for their associated offshore distances to be uncorrelated. This elapsed time was estimated using Moran’s I statistic (Moran 1950; Cressie 1993) and the associated autocorrelation function (acf). Individual Moran’s I statistics at a particular lag time were deemed significantly different from zero if a Bonferroni- corrected 95% confidence interval did not contain 0. Once the length of time within which offshore distances were potentially correlated was determined, the permutation technique randomly shuffled time intervals of this length during each iteration of the process. Due to a general reduction in the number of whale calls detected during the latter half of the 2001 study (Fig. 9.4), the set of time intervals occurring after 14 Sept. was viewed as different from the set of time intervals occurring up to 14 Sept. Because of this, a stratified bootstrap approach was adopted wherein time intervals up to 14 Sept. were placed in a separate stratum and permuted separately from the later time intervals. Results

The 90% confidence ellipses around estimated call locations generally had major axes less than 2.5 km for calls within the perimeter of the DASAR array during 2001 (Fig. 9.5). Within the larger analysis zone extending out to 10 km from the central axis of that array, major axes of 90% confidence ellipse varied in length between 0.5 km and 12 km, with 50% of the ellipses having major axes <5 km (Fig. 9.5). (As noted in Chapters 7 and 8, localization accuracy was much improved in 2002.) In all, 1121 calls were available for the analysis during the 29 Aug. to 2 Oct. period in 2001, and their locations ranged from 3.6 km seaward of the baseline through Northstar (and parallel to shore) to ~26 km offshore. The latter is 9-16 Monitoring at Northstar, Open-Water Season Offshore Distance (km) 5 10152025

29-Aug-2001 10-Sep-2001 22-Sep-2001 Date

FIGURE 9.4. Offshore distances to whale call locations (n = 1121) obtained within the analysis zone from 29 Aug. to 2 Oct. 2001. Dashed vertical line occurs at 00:00 on 15 Sept. 2001 (midnight between 14 and 15 Sept.). an artificial cutoff, representing the distance between the northeastern edge of the analysis zone and Northstar (Fig. 9.3). These 1121 calls exclude 2032 additional calls whose estimated locations were outside the analysis zone, along with 7597 calls that could not be localized because they were detected by only one DASAR, or for other reasons (see Chapter 8). Based on Moran’s I statistic, offshore distances of whale calls were significantly and positively correlated when the calls were <6 h apart, negatively correlated when calls were 6–16 h apart, positively correlated again when calls were 16–24 h apart, and both positive and negatively correlated when calls were >24 h apart (Fig. 9.6). The correlations beyond 24 h were assumed to be spurious in the absence of any known biologically plausible explanation. (Feeding aggregations might persist for extended periods, but in that case one would expect more consistently positive correlations at intervals shorter than 24 h.) Although no specific explanation was evident for the negative correlations at intervals of 6–16 h, the permutation interval was set to t = 24 h to guard against potential adverse effects of the positive correlations at 16–24 h intervals. If the negative and positive correlations after 6 h are not the result of real correlation among whale calls or whales, but instead (in some unknown manner) are artifacts of the data collection and analysis, a permutation interval of t = 6 h would have been appropriate. If correlations beyond 6 h are indeed spurious, our use of t = 24 under-represents the true sample size and unnecessarily broadens the reported confidence intervals. Figures 9.7 through 9.12 show scatter plots of distance offshore versus the underwater sound index (ISI) 400 m from Northstar, averaged over the 5, 15, 30, 60, 70, and 90 minutes preceding detection of the individual calls. As anticipated, industrial sounds produced by Northstar varied substantially throughout Chapter 9: Statistical Approach for Assessing Displacement 9-17

7850 10 0 2 20 15 7845 10

5 5 7840 .5 10 2 7835 5 CB 2. a 5 re A is 7830 s ly UTM Northing (km) 10 a n A 7825

10 5 1 7820 Northstar (Seal Island)

425 430 435 440 445 450 455 460 UTM Easting (km)

FIGURE 9.5. Average major axis length (kilometers) of 90% confidence ellipses around estimated call locations in 2001, shown as a smoothed contour plot. Filled circle represents Northstar Island, open triangles represent DASARs, and solid black line is the baseline parallel to the coast. Calls detected within the outlined “analysis zone” (see Fig. 9.3) were used in the analysis. -0.5 0.0 0.5 Correlation (Moran's I)

0 1020304050 Hours between locations

FIGURE 9.6. Moran’s I statistics and estimated autocorrelation function for offshore distances of calls detected in the analysis zone in 2001. Vertical bars are Bonferroni-corrected 95% confidence intervals on Moran’s I. 9-18 Monitoring at Northstar, Open-Water Season 5 Minute Sound Window

6 Offshore Distance (miles)

6 Offshore distance (km)

4 5 10152025

80 90 100 110 120 130 Industrial Sound Index at 400 m (dB re 1 µ Pa)

FIGURE 9.7. Offshore distances of bowhead calls vs. average industrial sound index (ISI) 400 m north of Northstar in the 5 minutes previous to the call, 2001. Quantile regression (solid line) relates 5th quantile of offshore distances for bowhead calls to 5-min ISI. 90% confidence interval (dashed lines) for the quantile regression if there is no effect was estimated via a stratified permutation method (see th “Example: Methods” section). Average 5 quantile assuming no effect is also plotted (– ▲ –). ISI was total sound level in five 1/3rd-octave bands dominated by industrial components. Offshore distances were measured perpendicular from a baseline through Northstar Island to the call location. Only the 1121 calls within an analysis zone extending out to 10 km from the “central axis” of the DASAR array are considered (Fig. 9.3). Symbol diameter is approximately proportional to the weight given each observation in the quantile regression (see text). the study period, providing offshore distance information at both low and high sound levels. Quantile regressions for 16 other intervals in the 5–160 min range, as listed earlier, were also estimated but are not plotted here. Included in each Figure are the average 5th quantile line assuming no effect, and its 90% bounds. The quantile regression lines estimated in the 15, 30, 60, and 70 minute Figures all indicate that the two thresholds estimated by the model are similar and in the range of 98 to 105 dB re 1 µPa. th The 5 quantiles of offshore distance expected if there was no effect (– ▲ – lines in Fig. 9.7–9.12) were not constant across varying sound (ISI) levels. For all averaging times plotted here except 90 min, the 5th quantile under the assumption of no effect declined between low ISI values (~78 dB) and moderate ISI (~98–105 dB), but then increased from moderate ISI to high ISI (~125 dB). This “cupped” shape was due to the nonsymmetrical offshore-distance and ISI distributions, and the particular configuration of weights used in the analysis. Chapter 9: Statistical Approach for Assessing Displacement 9-19 15 Minute Sound Window

6 Offshore Distance (miles)

6 Offshore distance (km)

4 5 10152025

80 90 100 110 120 130 Industrial Sound Index at 400 m (dB re 1 µ Pa)

FIGURE 9.8. Offshore distances of bowhead calls vs. average industrial sound index (ISI) 400 m north of Northstar in the 15 minutes previous to the call, 2001. Otherwise as in Figure 9.7. 30 Minute Sound Window

6 Offshore Distance (miles)

6 Offshore distance (km)

4 5 10152025

80 90 100 110 120 130 Industrial Sound Index at 400 m (dB re 1 µ Pa)

FIGURE 9.9. Offshore distances of bowhead calls vs. average industrial sound index (ISI) 400 m north of Northstar in the 30 minutes previous to the call, 2001. Otherwise as in Figure 9.7. 9-20 Monitoring at Northstar, Open-Water Season 60 Minute Sound Window

6 Offshore Distance (miles)

6 Offshore distance (km)

4 5 10152025

80 90 100 110 120 130 Industrial Sound Index at 400 m (dB re 1 µ Pa)

FIGURE 9.10. Offshore distances of bowhead calls vs. average industrial sound index (ISI) 400 m north of Northstar in the 60 minutes previous to the call, 2001. Otherwise as in Figure 9.7. 70 Minute Sound Window

6 Offshore Distance (miles)

6 Offshore distance (km)

4 5 10152025

80 90 100 110 120 130 Industrial Sound Index at 400 m (dB re 1 µ Pa)

FIGURE 9.11. Offshore distances of bowhead calls detected in the analysis area vs. average industrial sound index (ISI) 400 m north of Northstar in the 70 minutes previous to the call, 2001. Otherwise as in Figure 9.7. Chapter 9: Statistical Approach for Assessing Displacement 9-21 90 Minute Sound Window

6 Offshore Distance (miles)

6 Offshore distance (km)

4 5 10152025

80 90 100 110 120 130 Industrial Sound Index at 400 m (dB re 1 µ Pa)

FIGURE 9.12. Offshore distances of bowhead calls detected in the analysis area vs. average industrial sound index (ISI) 400 m north of Northstar in the 90 minutes previous to the call, 2001. Otherwise as in Figure 9.7.

When sound levels in the 5 ISI bands were comparatively high (>112 dB re 1 µPa as measured 400 m from Northstar), the estimated 5th quantile line fell outside what could be expected by chance only for the 70- and 80-minute ISI averaging times, and even then just barely. (The 80-min results were inspected but are not presented here.) When ISI values were averaged for 70 or 80 min, the estimated 5th quantile was 15 km offshore at 120 dB. For ISI levels from 113 dB to 122 dB re 1 µPa, the 5th quantile of offshore distance was farther offshore than expected given the upper bound of the 90% confidence region if offshore distance and ISI levels were unrelated (Fig. 9.11). This result effectively has 95% confidence if the problem is viewed from a one-tailed testing perspective, as is generally appropriate in a study of disturbance effects. Estimated displacement for ISI > 105 dB and varying averaging times appears in Figure 9.13. The 5th quantile was estimated to displace farther offshore by 3.2 km (90% CI = 0.2 km to 5.2 km) when ISI values averaged during the 70 min preceding the call were 119 dB re 1 µPa (Fig. 9.14). When sound levels in the 5 ISI bands were low (<95 dB), the estimated 5th quantile line was below what could be expected by chance for most of the sound integration times from 15 to 90 min. Although close to the boundary of the confidence region for some of these averaging times, the estimated 5th quantile of offshore distance was closer to Northstar than expected at ISI levels between 85 dB and 95 dB, by ~0.75–1.0 km (Fig. 9.8–9.12). 9-22 Monitoring at Northstar, Open-Water Season Offshore Displacement (miles)

ISI Level

121 dB

119 dB 5 117 dB 115 dB

113 dB 1

0 111 dB

109 dB

5 107 dB Offshore Displacement (km)

0 0123

0 50 100 150 Sound Integration time (minutes)

th FIGURE 9.13. Estimated displacement of the 5 quantile of offshore distances as a function of ISI level and sound averaging (integration) time, 2001. Lower 90% confidence limits on displacement were >0 only for ISI levels from 113 dB to 122 dB re 1 µPa and averaging times from 70 to 80 min. Offshore Displacement (miles)

0 ISI Level

. 119 dB 5

5 Offshore Displacement (km)

0 012345

0 50 100 150 Sound Integration time (minutes)

th FIGURE 9.14. Estimated displacement of the 5 quantile of offshore distances in 2001 when ISI levels averaged over varying lengths of time reached 119 dB. Dashed lines indicate lower and upper 90% limits for displacement at each integration time. 90% confidence limits on displacement of the 5th quantile are > 0 only for integration times from ~70 to 80 min. Chapter 9: Statistical Approach for Assessing Displacement 9-23

DISCUSSION

This paper describes a new statistical approach for assessing the possible displacement effects of anthropogenic stimuli that diminish with increasing distance and vary over time. One of the unusual features of the proposed method is that (in our application) it concentrates on one tail of the distribution of the response variable, not on a measure of central tendency. This is an appropriate approach when the effect is expected to be largely confined to one extreme of the distribution. In the case of disturbance, the effect would normally be expected to be most evident among the closest animals, and focusing attention on the closest animals maximizes the potential to detect an effect if it occurs. It should be kept in mind, however, that there are (by definition) fewer observations in the extreme of a distribution, and precision for estimating extreme quantiles (say the 5th) is lower than when estimating a measure of central tendency. In the Northstar example, the reduced precision inherent in estimating a quantity in the tail of a distribution was apparently offset by the large number of recorded whale calls . We have arbitrarily selected the 5th quantile of the response distribution as our measure of effect in the whale example. For a given distribution, quantiles that are close to one another, e.g. the 3rd, 5th and 10th quantiles, will necessarily be closely correlated. Thus, for purposes of determining if there is a disturbance effect on the closest animals, it should generally make little difference which quantile, of a group of similar low quantiles, one uses in the analysis. However, the magnitude of the estimated displacement (if any) will depend on the quantile considered. There may be value in repeating the analysis with different quantiles in order to better measure the effect size at different levels of the stimulus variable (see Chapter 11). In this application, the quantile regression equation contains only one explanatory variable, stimulus level. This allows estimation of models that are linear, quadratic, threshold, spline, etc., functions of stimulus levels, but not models that include factors such as year, age, sex, season, habitat, elevation, etc. The only difficulty in extending this method to allow inclusion of non-stimulus covariates is establishing confidence bounds on the no-effect model. The permutation method used here to establish those bounds was effective because the model was only a function of stimulus levels. It is unknown whether the same permutation method will be useful if non-stimulus covariates must be considered, or whether a different permutation (or bootstrap) method will need to be employed. Ongoing statistical research is being conducted to identify permutation and bootstrap methods useful for regression analyses involving multiple x variables, and analysis of variance (ANOVA) involving more than one factor (Manly 1997). From initial results of that research, it appears that permuting or bootstrapping residuals of a multi-variable model provides a useful test of a single model coefficient or ANOVA factor in the usual normal theory setting where regression is through the mean. However, these techniques have not been investigated when quantile regression is applied, and their performance in this case is unknown. In the example, our “analysis zone” extended seaward to 10 km beyond the “CB” DASAR, or about 26 km seaward of the Northstar offshore oil development (Fig. 9.3), and spanned only the inshore half (approximately) of the bowhead migration corridor. The distribution of offshore distances varies from year to year, but in an average year only about half of the population of bowheads migrates within 26 km of Northstar (Treacy 2000, 2002). Thus, the 5th quantile of the offshore distances within our analysis zone actually represents the (approx.) 2.5th quantile of the overall distribution. Provided that the quantile being estimated is well inside the analysis zone even at high stimulus levels (as here), results would be unchanged if one monitored the full width of the migration corridor and estimated the 2.5th quantile. However, it would have been much more difficult and costly to obtain data across the full width of the migration corridor, and this would have provided little improvement in characterizing any disturbance effect of a nearshore oil 9-24 Monitoring at Northstar, Open-Water Season development. Limited resources were better spent on characterizing the (inshore) part of the response distribution where maximal effects were expected. Similar situations may occur in other applications as well. One advantage of the proposed analysis approach is that autonomous data collection equipment can be used in many cases to record daytime and nighttime observations, and in any weather conditions. This increases sample size and provides a continuous record of animal behavior or responses. However, such equipment may not allow the researcher to determine reliably whether different observations come from the same or different individual animals. The present approach provides a way to deal with this problem. Not- withstanding its advantages, autonomous data collection equipment often is costly and difficult to operate and maintain. Consequently, the costs and benefits of such approaches must always be compared to the cost and benefits of traditional approaches. Development of the present statistical method was prompted by the need for a method that could quantify possible displacement of migrating bowhead whales exposed to varying levels of underwater sound from an offshore oil development. Our example comes from the first season (2001) of successful monitoring of bowheads migrating past that oil development. Although the 2001 results suggest that whales are farther than expected from Northstar when ISI values were high, and closer than expected to Northstar when ISI values were moderate, the data from this one season were limited in various ways (see Chapter 10). Also, there are other non-statistical complications that need to be explored before drawing firm conclusions (see Greene et al. 2002 and Chapter 10). However, the same monitoring and analysis approach has been applied again in 2002, and this has provided an additional and in several respects improved dataset in 2002. Results from 2002 are described in Chapters 7 and 10.

ACKNOWLEDGEMENTS

Greeneridge Sciences Inc., Santa Barbara, CA, was responsible for the development and implementation of the overall acoustical monitoring approach, with the sponsorship and strong encouragement of BP Exploration (Alaska) Inc. Thanks to C.R. Greene Jr., M.W. McLennan, R.G. Norman, S.B. Black- well, W.C. Burgess, Q. Nguyen, J. Seitz and others associated with Greeneridge for developing and/or implementing the acoustic technique. Alaska Clean Seas (ACS) crews on their Bay-class boats made the 2001 ocean work safe and successful. M. Theriault, B. Venable, K. Pulliam, W. Nye and G. Ronne, along with B. Rutherford, E. Olsen, J. Nevels, F. McAdams and W. Henry, were particularly helpful. J. Leavitt of Barrow helped during field preparations and sea work. K. Otte, J. Fleischman, S. Love, R. Fortune, J. Lynch, J. Seitz, J. Leavitt and D. Martinez of Greeneridge performed the whale call localization analysis for 2001, and R. Blaylock of Greeneridge helped with data summarization. Dr. B. Streever of BP provided much support and advice during development and implementation of the analytical method, including numerous useful comments on the approach, its applicability to Northstar and other situations, and this manuscript. Dr. R. Jakubczak of BP was instrumental in the initial development of the approach; he and D. Trudgen of OASIS Environmental (for BP), along with B. Streever, strongly supported the project, and provided reviews of earlier reports plus helpful suggestions on direction. M. Williams and A. Wright of LGL, and W. Cullor and A. Erickson of OASIS (on behalf of BP), helped in various ways. At WEST, R. Nielson and C. Nations assisted with the statistical analysis, and B.F.J. Manly commented on the statistical methodology and the draft manuscript. Numerous participants in a peer/stakeholder group convened by NMFS provided guidance and support, and R. Suydam of the North Slope Borough Dept. of Wildlife Management provided helpful comments. Thanks to all. Chapter 9: Statistical Approach for Assessing Displacement 9-25

LITERATURE CITED

Barnes, P.W., D.M. Schell and E. Reimnitz (eds.). 1984. The Alaskan Beaufort Sea ecosystems and environments. Academic Press, Orlando, FL. 466 p. Cressie, N.A.C. 1993. Statistics for spatial data, revised edition. John Wiley & Sons, New York, NY. de Boor, C. 1978. A practical guide to splines. Springer-Verlag, Berlin. Dyer, S.J., J.P. O'Neill, S.M. Wasel and S. Boutin. 2001. Avoidance of industrial development by woodland caribou. J. Wildl. Manage. 65(3):531-542. Efron, B. 1979. Bootstrap methods: another look at the jackknife. Ann. Stat. 7:1-26. Fitzenberger, B. 1997. The moving blocks bootstrap and robust inference for linear least squares and quantile regressions. Journal of Econometrics 82:235-2887. Green, P.J. and B.W. Silverman. 1994. Nonparametric regression and generalized linear models, a roughness penalty approach. Chapman & Hall, London, U.K. Greene, C.R., Jr., M.W. McLennan, T.L. McDonald and W.J. Richardson. 2002. Acoustic monitoring of bowhead whale migration, autumn 2001. p. 8-1 to 8-79 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of industrial sounds, seals, and whale calls during construction of BP's Northstar oil development, Alaskan Beaufort Sea, 2001, Draft, Oct. 2002 ed. LGL Rep. TA2573-2. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. Johnson, D.H. 1999. The insignificance of statistical significance testing. J. Wildl. Manage. 63(3):763-772. Koenker, R. and G. Bassett, Jr. 1978. Regression quantiles. Econometrica 46:33-50. Lenth, R.V. 1981. On finding the source of a signal. Technometrics 23:149-154. Manly, B.F.J. 1997. Randomization, bootstrap and Monte Carlo methods in biology, 2nd ed. Chapman and Hall, London, U.K. 281 p. Mardia, K.V. 1972. Statistics of direction data. Academic Press, New York, NY. 357 p. MathSoft Inc. 1999. S-Plus 2000 guide to statistics, vol.1. Data Analysis Products Division, MathSoft, Seattle, WA. Moore, S.E. 2000. Variability of cetacean distribution and habitat selection in the Alaskan Arctic, autumn 1982-91. Arctic 53(4):448-460. Moran, P.A.P. 1950. Notes on continuous stochastic phenomena. Biometrika 37:17-23. Neter, J., W. Wasserman and M. H. Kutner. 1983. Applied Linear Regression Models. Richard Irwin Inc., Home- wood, IL. page 346. Richardson, W.J., K.J. Finley, G.W. Miller, R.A. Davis and W.R. Koski. 1995. Feeding, social and migration behavior of bowhead whales, Balaena mysticetus, in Baffin Bay vs. the Beaufort Sea—regions with different amounts of human activity. Mar. Mamm. Sci. 11(1):1-45. Richardson, W.J., C.R. Greene Jr., C.I. Malme and D.H. Thomson. 1995. Marine mammals and noise. Academic Press, San Diego, CA. 576 p. Rowland, M.M., M.J. Wisdom, B.K. Johnson and J.G. Kie. 2000. Elk distribution and modeling in relation to roads. J. Wildl. Manage. 64(3):672-684. SAS Institute Inc. 1999. S AS/STAT user’s guide, release 8 edition. SAS Institute Inc. Cary, NC. Seber, G.A.F. and C.J. Wild. 1989. Nonlinear regression. John Wiley & Sons, New York, NY. 9-26 Monitoring at Northstar, Open-Water Season

Treacy, S.D. 2000. Aerial surveys of endangered whales in the Beaufort Sea, fall 1998-1999. OCS Study MMS 2000- 066. U.S. Minerals Manage. Serv., Anchorage, AK. 135 p. Treacy, S.D. 2002. Aerial surveys of endangered whales in the Beaufort Sea, fall 2001. OCS Study MMS 2002-061. U.S. Minerals Manage. Serv., Anchorage, AK. 117 p. Venables, W.N. and B.D. Ripley. 1999. Modern applied statistics with S-plus, 3rd edit. Springer-Verlag, New York, NY. White, G.C. and R.A. Garrott. 1990. Analysis of wildlife radio-tracking data. Academic Press, New York, NY. 383 p. CHAPTER 10: ACOUSTIC LOCALIZATION OF BOWHEAD WHALES NEAR A BEAUFORT SEA OIL DEVELOPMENT, 2001–2002: EVIDENCE OF DEFLECTION AT HIGH-NOISE TIMES? 1

by W. John Richardson a , Charles R. Greene Jr. b, and Trent L. McDonald c

a LGL Ltd., environmental research associates 22 Fisher St., POB 280, King City, Ont. L7B 1A6 (905) 833-1244; [email protected]

b Greeneridge Sciences, Inc. 4512 Via Huerto, Santa Barbara, CA 93110 (805) 967-7720; [email protected]

c Western EcoSystems Technology, Inc. 2003 Central Ave., Cheyenne, WY 82001 (307) 634-1756; [email protected]

for

BP Exploration (Alaska) Inc. Dept. of Health, Safety & Environment 900 East Benson Blvd, P.O. Box 196612 Anchorage, AK 99519-6612

LGL Report TA2706-5

December 2003

1 Chapter 10 In: W.J. Richardson and M.T. Williams (eds.). 2003. Monitoring of industrial sounds, seals, and bowhead whales near BP’s Northstar oil development, Alaskan Beaufort Sea, 1999–2002. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 10-2 Monitoring at Northstar, Open-Water Season

ABSTRACT

Bowhead whales, Balaena mysticetus, migrate west parallel to the north coast of Alaska during fall. In 2000, BP Exploration (Alaska) Inc. began constructing the Northstar Development 5 km (3 mi.) seaward of the barrier islands near Prudhoe Bay. Construction at the Northstar artificial island continued during the autumn of 2001, and oil production was underway in autumn 2002. The objective of this study was to determine whether underwater sound propagating seaward from Northstar deflected the southern part of the bowhead migration corridor farther offshore. An acoustical approach was used to determine whether the closest of the calling whales tended to be farther offshore at times when levels of underwater noise from Northstar and its support vessels were above average. An array of 11 Directional Autonomous Seafloor Acoustic Recorders (DASARs) offshore of Northstar provided bearings to calling whales. When two or more DASARs detected the same call, the location of the whale was estimated from the intersections of the bearings. Calling whales were localized for 36 days in 2001 and 23 days in 2002. In those periods, totals of 1121 and 2057 calls (respectively) were localized within a 20 × 26 km (12 × 16 mi.) area seaward of North- star where comparable data could be obtained at all times. For those calls, quantile regression was used to determine the relationship between the 5th quantile distance from shore and an index of the level of underwater sound coming from Northstar Island and its support vessels. In both 2001 and 2002, the apparent southern edge of the migration corridor was slightly farther offshore at the noisiest times as compared with typical times, with ≥95% confidence (one-sided). In both years, the maximum displacement effect was evident when industrial sound index (ISI) was averaged over ~70 min. When the 70-min ISI was high, the 5th quantile of the offshore distances was displaced offshore by ~2.3–3.3 km (1.4–2.1 mi) in 2001, and by ~2.2–4.7 km (1.4–2.9 mi) in 2002, as compared with that expected with no displacement. However, there was considerable uncertainty in the estimated displacement distances. The high sound levels occurring at times with evidence of displacement were mainly from vessels, not Northstar island itself.

INTRODUCTION

Whales often react to man-made sounds by changing their behavior and swimming away, or occasionally toward, the sound source (Richardson et al. 1995b, 1999). Beginning in late August and continuing past mid-October, bowhead whales (Balaena mysticetus) migrate west-northwest across the southern part of the Beaufort Sea, swimming parallel to the north coast of Alaska (Moore and Reeves 1993; Treacy 2002a,b; Fig. 10.1). The majority of the whales travel along a corridor 10–75 km (6–47 mi.) offshore, over the con- tinental shelf. Each autumn, Alaskan Eskimos hunt these whales for subsistence. Bowhead whales are known to show disturbance reactions to various types of underwater industrial sounds (Richardson and Malme 1993). The hunters are concerned about any industrial activity that might result in the bowheads swimming farther offshore and therefore being more difficult to hunt and retrieve. The hunters are also concerned that whales, once deflected from nearshore paths, may not readily return to their original paths. In early 2000, BP Exploration (Alaska), Inc., began constructing its Northstar oil development, the first oil production facility offshore of the barrier islands in the Beaufort Sea. The development is centered on Northstar Island, an artificial gravel island 5 km (3 mi.) seaward of the natural barrier islands, and a few kilometers inshore of the southern edge of the main migration corridor of the bowhead whales (Fig. 10.1). The artificial island was built in early 2000, and facilities were installed on the island during the summers of 2000 and 2001. Oil production and gas injection began in late 2001, powered by gas-turbine engines, and drilling of additional wells continued through 2002. Chapter 10: Deflection of Bowhead Migration? 10-3

FIGURE 10.1. Northern Alaska, showing the main autumn migration corridor of bowhead whales (arrows) and the location of the study area near the Northstar oil development (rectangle). Dashed line is the 100 m depth contour, near the shelf-break.

Northstar is 27 km (17 mi.) west of a natural island (Cross Island) from which migrating bowhead whales are hunted in fall. Thus, the possible effects of the industrial operations on the whales, and on their accessibility to subsistence hunters, were of concern. Because of both BP corporate policy and regulatory requirements, BP decided to study any possible effects of island sound on the paths of bowheads migrating along the southern edge of the migration corridor. Design of a study to assess the effects of Northstar on bowhead whales was challenging for a variety of reasons. Levels of underwater sound propagating from oil industry facilities on gravel islands tend to be low (Richardson et al. 1995b, p. 127–130). Thus, any displacement of whales was expected to be small in geographic scale, limited to the southern extremity of the migration corridor, and difficult to detect and quantify. In the absence of industrial activities, there is already high within- and among-year natural variation in whale distribution, related in part to variable ice conditions (Moore 2000; Treacy 2000, 2002a,b). Migrating bowhead whales are below the surface ~90% of the time, resulting in low sighting probability during aerial surveys (Richardson et al. 1995a). Also, it was difficult to identify any type of data that would serve as a “no Northstar” control. No other location(s) near Northstar seemed very appropriate as reference areas, given what was known about whale activities and other human activities in the region. The high year-to-year variability in the migration corridor would limit the value of comparisons with pre-Northstar aerial survey data. Once construction began in early 2000, the island was present continuously and was expected to produce underwater sound more or less continuously, with no truly inactive times. The arctic environment, with the possibility of drifting ice at any time, and the likelihood of freeze-up before the end of the autumn migration season, imposed additional constraints. The key objective of the monitoring was to estimate the offshore displacement of the southern edge of the bowhead migration corridor, if any, at times when higher-than-average levels of underwater sound were being emitted from Northstar island and its associated vessels. From the early planning stages, this question was viewed as being one-sided, given that any displacement effect that might occur at high sound levels was anticipated to be an avoidance effect. This formulation of the key objective does not require pre- Northstar “control” data, or data from a separate “reference” area. A passive acoustic method was selected to determine the locations of calling whales. An acoustical method was selected, in preference to aerial surveys, because acoustics works 24 hours/day (night and day), during good weather and bad, and allows 10-4 Monitoring at Northstar, Open-Water Season simultaneous monitoring of a large area. Previous experience showed that thousands of bowhead whale calls would be detected during one autumn season of monitoring, providing a far larger sample size than could be obtained from aerial surveys. To compare whale locations with sound emissions, continuous data on sounds from the island were needed. These were obtained with a hydrophone near the island, connected by cable to a digital recording system on the island. Statistical analyses were essential to assess whether any “displacement” effect was evident at times of high sound emission and, if so, to determine the magnitude of displacement. To date, field work has been done during the early autumn of 2000, 2001, and 2002, with equipment development and a pilot study in 2000, and effective data collection in 2001 and 2002. This project required development of new acoustic monitoring equipment and a new statistical approach. In Chapter 8, Greene et al. describe the methods developed to detect bowhead whale calls, to locate their source positions, and to estimate the accuracy of those positions. Some results from 2002 are included in Chapter 8 to illustrate the approach. In Chapter 9, McDonald and Richardson describe the specialized statistical approach developed for analysis of the whale locations vs. sound emission data. The most novel feature of the analysis is the need to assess effects on the edge of an observed distribution (here the southern edge of the whale migration corridor), rather than effects on a measure of central tendency. For this, the chosen method is quantile regression (Koenker and Basset 1978); we consider primarily the 5th quantile of the distribution of distances-from-shore. Other features include allowance for the potential lack of independence of whale calls that are close in time and space, and allowance for the possibility that the 5th quantile distance-from-shore may vary even if there is no effect of sound on these distances. The method provides an ability to fit a variety of types of statistical models to the offshore distances vs. sound levels. In this case, a threshold or “piecewise linear” model is used (Neter et al. 1983). Chapter 9 uses results from 2001 to illustrate the statistical approach. The present chapter uses the results from these acoustical and statistical techniques, as applied in both 2001 and 2002, to evaluate Northstar effects on the southern part of the whale migration corridor in those years. Additional details concerning the specific 2001 and 2002 work are given in Greene et al. (2002) and in Chapter 7, respectively. The conclusions of this chapter are subject to caveats and provisos created by the data collection and analysis methods. The main issues are addressed in the “Discussion” section. Also, this chapter is based on results from only two field seasons. The bowhead migration corridor is known to be variable from year to year (Treacy 2002), and specific relationships between the migration corridor and Northstar sound may vary among years. The acoustical monitoring approach has been applied again in 2003, apparently successfully, so additional relevant data are expected to become available when analysis of the 2003 data is completed.

METHODS

In order to evaluate the relationship (if any) between the location of the southern part of the bowhead migration corridor and the fluctuating sounds from Northstar, there were three requirements. (1) Determine the locations of whales in the southern part of the migration corridor. (2) Measure the sounds emanating from Northstar. (3) Use statistical methods to evaluate the relationship of (1) and (2). Data of types (1) and (2) were obtained essentially continuously from 29 Aug. to 2 Oct. 2001, and from 31 Aug. to 23 Sept. 2002. The start of the main bowhead migration season is around 1 Sept., and migration continues into mid-October (Moore and Reeves 1993). However, vessel operations near North- star ended around 7 Oct. in 2001 and 2 Oct. in 2002 as sea ice formed. Thus, in both years it was necessary to retrieve the equipment in early October, before the end of the bowhead migration. In 2002, although the equipment was deployed until early October, storm-surge upset much of the gear on 23 Sept., and useful data collection ended then (see Chapter 7 for details). Chapter 10: Deflection of Bowhead Migration? 10-5

Whale Call Localization

A triangulation method was used to locate whale calls. Bearings to calling whales were determined by an array of 11 independently-operating “Directional Autonomous Seafloor Acoustic Recorders” (DASARs) designed specifically for this study. When two or more DASARs detected a given call, the location of the calling whale was determined from the intersection(s) of the bearings. The following is a summary of the DASAR design and array configuration; additional details are provided in Greene et al. (Chapter 8). Each DASAR incorporated the directional sensors from a DIFAR (Directional Frequency and Recording) sonobuoy, type AN/SSQ-53, from Sparton Electronics (De Leon Springs, FL). A DIFAR sensor provides three channels of acoustic data, based on one omnidirectional pressure sensor and two orthogonal particle-velocity sensors (U.S. Navy 1995). The DASARs were designed to record the three DIFAR channels directly to a disk in the seafloor recorder. We sampled each channel at 1 ms intervals, providing acoustic data up to 500 Hz—adequate for most bowhead whale calls (Clark and Johnson 1984). A 30 Gbyte disk in each DASAR provided for 45 days of recording. A standard DIFAR sonobuoy includes a magnetic compass to provide the necessary directional reference. In our case, each DASAR was at a fixed (but initially unknown) orientation on the bottom. We determined that orientation by projecting calibration sounds from known (via GPS) locations around each DASAR. This was done at (approximately) weekly intervals during the 2001 and 2002 field seasons. Pro- cessing of these signals identified the orientation of each DASAR, and confirmed that it was stable over the field season (see Greene et al., Chapter 8). The calibration sounds were projected at precisely known times, which allowed us to quantify and correct for drift in the clock within each DASAR. After this compensation, times of events could be determined to an accuracy of 1 or 2 s, adequate to assess whether a whale call received at several DASARs represented a single call from one whale. The DASARs were deployed at the vertices of a series of equilateral triangles with sides 5 km (3.1 mi.) long, configured as two overlapping hexagons (Fig. 10.2). The array of DASARs extended roughly north-northeast from Northstar Island into the southern part of the bowhead migration corridor. The closest and farthest units were, respectively, 7 and 22 km (4 and 14 mi.) northeast of Northstar. The most offshore unit was about 1/3rd of the distance from Northstar to the shelf-break, which is near the outer edge of the main migration corridor (Fig. 10.1). Standard triangulation techniques (Lenth 1981; White and Garrott 1990) were used to estimate the location of each call based on the intersection(s) of bearings from multiple DASARs. For each call location, a 90% confidence ellipse was calculated based on the number of DASARs receiving the call, disparity of the intersections, and inherent variation estimated from calibration data for each DASAR (see Appendix 8.2 in Chapter 8). The square-root of the area of that ellipse was a measure of precision, which was used as a weighting factor in the statistical analyses. Analysis of the call location data showed that the number of calls detected >10–12 km from the center of the array tended to be proportionally lower at times when background noise was high (Greene et al. 2002 and Chapter 8). This partial “cutoff” at 10 km (2002) or 12 km (2001) during high-noise times apparently represented masking of some of the more distant calls at those times. To minimize the potential for bias associated with variable background noise, statistical analyses for 2001 and 2002 were limited to calls detected within an “analysis area” whose boundaries were 10 km from the “central axis” of the DASAR array (Fig. 10.2). The central axis was a straight line from DASAR “CB” to Northstar. 10-6 Monitoring at Northstar, Open-Water Season

FIGURE 10.2. Locations where 11 DASARs were deployed northeast of the North- star oil development in 2001 and 2002. There were two nearby DASARs at the CA and CB locations. Bowhead whale calls detected within the "analysis area" were used in the main statistical analysis. Dashed WNW–ESE line through Northstar is the baseline from which the perpendicular “offshore distance” was measured.

Sounds from Northstar

Underwater sounds produced by the island and associated vessels were measured from continuous recordings of signals received by a hydrophone positioned just above the seafloor (depth 12 m) near the island. An ITC model 8212 hydrophone was placed ~400 m seaward (north) of the north edge of the island. It was connected by double-armored well-logging cable to digital sound recording apparatus on the island, similar to that used in a DASAR. Sampling rate was 2000 samples/s, with 16 bit quantization, providing acoustic data up to ~1000 Hz. Sound spectral densities were determined for a 1-min period every 4.37 min, or ~330 times per 24-h day, throughout each field season. These data were used to determine broadband and one-third octave band levels in the 10–1000 Hz frequency range for each sampling time. There were 11,846 mea- Chapter 10: Deflection of Bowhead Migration? 10-7 surements for the 28 Aug.–3 Oct. 2001 period, and 7551 measurements for the 31 Aug.–23 Sept. 2002 period. Sound received ~400 m from Northstar was partly of industrial origin, but also partly of natural origin, associated with wind and wave action. For purposes of relating whale call locations to sounds from Northstar, it was desirable to have a measure of sound that (insofar as possible) represented the industrial but not the natural components. Separate analyses of the 400 m acoustic data from 28 Aug.–3 Oct. 2001 and 31 Aug.–23 Sept. 2002 showed that, for both years, sound in five 1/3-octave frequency bands, centered at 31.5, 40, 50, 63 and 80 Hz, was predominantly of industrial origin (see Chapter 6). In both years, there were prominent and recurrent tonal sounds in most of those bands. The total sound pressure in these five 1/3rd octaves, expressed in dB re 1 µPa, was used as an Industrial Sound Index (ISI). As compared with the overall broadband level in the 10–1000 Hz band, the “5-band ISI” averaged 5.5 ± s.d. 3.02 dB less in 2001, and 4.3 ± s.d. 2.57 dB less in 2002. Figure 10.3A vs. B shows the relationship between the 10–1000 Hz values as compared with the corresponding 5-band ISI values for one sample day in 2002.

140 140

A B 130 130 Pa) Pa) µ µ

120 120

110 110

100 100 One Minute ISI Level (dB re 1 re (dB Level ISI Minute One

10-1000 Hz Band Level (dB re 1 90 90

80 80 0 4 8 12 16 20 24 0 4 8 12 16 20 24 Time on 20 Sept. 2002 Time on 20 Sept. 2002 140

C FIGURE 10.3. Sound levels recorded by cab- 130 led hydrophone ~420 m north of Northstar at Pa) µ various local times during 20 Sept. 2002. 120 (A) 10–1000 Hz broadband levels averaged over 1 min and measured every 4.37 min. rd 110 (B) Corresponding levels in the five 1/3 octave bands with center frequencies 31.5–80 100 Hz; this is the Industrial Sound Index (ISI). (C) Corresponding 60-min moving-average

Hourly Average ISI Level (dB re 1 90 ISI, where each 60-min value is the average of 13 or 14 1-min ISI values.

80 0 4 8 12 16 20 24 Time on 20 Sept. 2002 10-8 Monitoring at Northstar, Open-Water Season

Statistical Analyses

The response variable was a measure of “offshore distance”, specifically the perpendicular distance from the best estimate of each call’s location to a line oriented parallel to the general trend of the coast and through Northstar (Fig. 10.2). Offshore distances of calls in the “analysis zone” (Fig. 10.2) were analyzed for a disturbance effect by estimating the relationship between the 5th quantile of offshore distance and ISI (Koenker and Basset 1978), and comparing that with the expected relationship if there were no effect of Northstar sound. The following is a summary of the procedure; for details, see McDonald and Richardson (Chapter 9). Exclusion of Non-Northstar Vessels in 2002 Vessels not associated with routine Northstar operations occasionally traveled offshore of Northstar toward or into the whale migration corridor. When vessels approach bowhead whales, the whales often show avoidance at distances as much as 2–4 km (Richardson et al. 1985; Richardson and Malme 1993). The one vessel for which we had specific records of movements seaward of Northstar during 2002 was the 13-m (42 ft) boat that we used to calibrate and service the DASARs weekly (approx.). As the objective of this study was to evaluate effects of routine Northstar operations on the southern edge of the whale migration corridor, we excluded whale calls localized when our vessel was operating >2 km offshore of Northstar, and during the 2-h periods following such times. The additional 2-h exclusion provided for any carry-over effect of previous boat disturbance to subside. Two hours was chosen based on how long bowhead avoidance reactions in response to boats typically continue after the boat moves away (usually ½– 1 h, Richardson et al. 1985; Richardson and Malme 1993), plus a 1–1½ h allowance for displacement effects to subside. This procedure was applied only in 2002, not 2001. The excluded periods in 2002 totaled 55.5 h, equivalent to 10% of the field season. Of the 2237 calls localized within the analysis zone (Fig. 10.2) in 2002, 8.0% (180 calls) were excluded as being potentially influenced by our vessel. By excluding times when the monitoring vessel was operating offshore of Northstar during 2002, the analysis focused more directly on the question of displacement by industrial and vessel activities at Northstar. If times when the monitoring boat was operating offshore had been included, this could have confounded the analysis at least slightly. Some whales in the analysis area near the vessel might be disturbed by the vessel even though the ISI (measured near Northstar) would not be elevated appreciably. Inclusion of data from these times would be expected to reduce the strength of the apparent relationship between the location of the southernmost whales and Northstar sounds. Also, use of the monitoring vessel is likely to be reduced in future seasons. It is appropriate to seek to understand the effect of Northstar industrial activities per se on the bowhead migration corridor, without including confounding influences by the monitoring boat. Noise from other vessels not associated with Northstar was also a contributor to the background sound levels offshore of Northstar, and a potential confounding influence on bowhead whales. However, noise from other vessels could not easily be distinguished from noise associated with other sources contributing to background sound, and in particular could not be associated with any specific vessel. Therefore, “other vessel” noise is a contributor to the (unexplained) residual variability in the relationship between offshore distances and sound levels. Although it would be desirable to take account of this source of variability, an apparent displacement effect associated with Northstar was evident for both 2001 and 2002 even in the presence of this residual variability (see “Results”, below). The concern about confounding by “other vessels” would be of greater importance if an apparent displacement effect had not been found. Chapter 10: Deflection of Bowhead Migration? 10-9

Quantile Regression of Offshore Distances vs. ISI A two-threshold model was fitted to the remaining offshore distances vs. ISI. Threshold or “piecewise linear” models (Neter et al. 1983) are appropriate when there is a particular stimulus level (threshold) above which an effect is suspected, and below which either there is no effect, or the relationship is different. In this case, we fitted a model with two thresholds. This model estimates one slope for ISI < T1 (T1 is the lower threshold), no slope for T1 < ISI < T2 (T2 is the upper threshold), and a separate slope for ISI > T2 . This model is given by

Qy=τ []β01++β IxTx-T ( 2 )( 2 ).(1) where Qτ[y] denotes the τ-th quantile of the random variable y (here “offshore distance”), x is the predictor variable (here ISI), and β0 through β2 are unknown coefficients to be estimated. I is a dichotomous indicator function that is either 0 or 1 depending on whether its argument is true or false

(e.g., I (x

To estimate the regression coefficients β0 through β2, an iterative nonlinear minimization routine was used to minimize the weighted sum of residuals; we used nlminb() available in S-Plus (MathSoft 1999). The square root of the area of each call’s 90% confidence ellipse was used as a weight in the quantile regression minimizing function. This resulted in observations far from the center of the DASAR array and calls with poor precision being down-weighted during the estimation process. Averaging Time for ISI We suspected that any response by bowhead whales to Northstar sound might be a function of the sound over a period much longer than the 4.37-min interval between our successive 1-min sound measurements. Typical swimming speed for a bowhead during autumn migration is 4–5 km/h (Koski et al. 2002), so whales would take a few hours to travel across our “analysis area”, which was 20 km wide in the ESW–WNW direction (Fig. 10.2). However, the effective integration or averaging time for a bowhead was not known a priori. For 2001, Northstar ISI levels associated with each whale call were averaged over a variety of different time periods immediately preceding the call, and separate quantile regression analyses (one for each averaging time) were done. The 21 averaging-times considered covered the anticipated range of times necessary to manifest an effect in whales, ranging between 5 and 160 min. Based on the 2001 results, the primary analysis for 2002 averaged the ISI levels over the 70 min preceding each call. However, other averaging times were subsequently investigated for 2002 as well (see “Results”). Figure 10.3C shows examples of 60-min average ISI values in relation to the original ISI values at 4.37-min intervals from which the 60-min values were derived. Expected Quantiles if No Effect Confidence intervals for the quantile regression and derived estimates were constructed using a stratified permutation method (Manly 1997). From collected data, it was impossible to know whether two calls detected at similar times were from independent whales, two whales moving in coordination, or a single whale. We estimated the elapsed time necessary between two calls in order for their offshore distances to be uncorrelated. This elapsed time was estimated using Moran’s I statistic (Moran 1950) and the associated autocorrelation function (acf). An individual Moran’s I statistic at a particular lag time was deemed significantly different from zero if a Bonferroni-corrected 95% confidence interval did not contain zero. Once the length of time within which offshore distances were potentially correlated was determined, the permutation technique randomly shuffled time intervals of this length during each of 10-10 Monitoring at Northstar, Open-Water Season

500 iterations of the process. In both years, the number of whale calls detected during the first and last halves of the study period differed. Calls were more abundant early than later in the 2001 season (Fig. 9.4 in Chapter 9). Conversely, calls were less abundant early than later in the 2002 season (see Fig. 10.5, later). Because of this, a stratified permutation approach was adopted wherein time intervals before and after 00:00 on 15 Sept. in 2001, and before and after 12:00 on 13 Sept. in 2002, were permuted separately. These breakpoints were chosen subjectively based on inspection of call rates as evident in Figures 9.4 and 10.5. Simulations based on bivariate randomized data show that, even when Y (offshore distance) is independent of X (ISI), the 5th quantile of Y may not be constant across the range of X (see Fig. 9.2 and associated text in Chapter 9). To determine confidence intervals if there were no effect of ISI on offshore distances, the quantile regression was fitted to the randomly permuted data and the process was repeated 500 times. For a given stimulus level, the 5th and 95th percentiles of these fitted quantile regressions were 90% confidence bounds assuming no relationship of offshore distances to ISI. If the observed quantile regression line went outside these bounds, we inferred that offshore distances at corresponding ISI levels were probably affected by Northstar sound. Estimating Offshore Displacement vs. ISI To estimate the “offshore displacement” of the 5th quantile for any given ISI value, we found the difference between (a) the actual fitted quantile regression at a particular ISI level and (b) the mean quantile regression obtained by randomly permuting the offshore distances. We estimated displacement in this manner for 2-dB steps of ISI from ISImin to ISImax. A confidence interval on the estimated displacement at a particular ISI level is simply the difference between the actual quantile regression estimate and the lower and upper endpoints of the confidence bounds constructed by permuting the offshore distances.

RESULTS Whale Call Locations

For 2002, locations of origin were estimated for 7563 bowhead whale calls detected during the 31 Aug.–23 Sept. period. (This excludes 359 calls detected during times potentially affected by the boat used to service the DASARs.) Some of these estimated locations were 20 km or more away from the DASAR array, and were quite uncertain (Greene et al., Chapter 8). However, 4997 of the 7563 calls were estimated to be within the area mapped in Figure 10.4, and 2057 of these were within the “analysis area” extending out to 10 km from the central axis of the DASAR array. In 2002, the highest concentration of localized calls was in a band parallel to shore centered ~20 km (12 mi) offshore of Northstar, near the seaward edge of the DASAR array (Fig. 10.4). The migration corridor extended offshore far beyond the “analysis area”, but the densest band of detected calls crossed the northern part of the analysis area. East of the analysis area, the southern edge of the migration corridor was apparently near Cross Island. Within the analysis area, the closest calls were ~7 km offshore of the baseline through Northstar (Fig. 10.4, 10.5). The number of whale calls localized within the analysis area tended to be higher during the latter half of the 2002 field season, after 13 Sept. (Fig. 10.5). Corresponding graphs for 2001 are included as Figures 9.3 and 9.4 in McDonald and Richardson (Chapter 9). In 2001, locations were estimated for 3153 bowhead whale calls detected during the 29 Aug.– 2 Oct. period. Of these, 2217 were within the area mapped in Figure 10.4, and 1121 of these were within Chapter 10: Deflection of Bowhead Migration? 10-11

FIGURE 10.4. Map of whale call locations within and near the “analysis area” in 2002. Calls are color-coded based on average industrial sound index 400 m from Northstar during the 1 h preceding the call. 10-12 Monitoring at Northstar, Open-Water Season Offshore Distance (km) 10 15 20 25

31-Aug-2002 08-Sep-2002 16-Sep-2002 Date FIGURE 10.5. Offshore distances to whale call locations (n = 2057) obtained within the analysis area from 31 Aug. to 23 Sept. 2001 (excluding times potentially affected by our boat). Dashed vertical line occurs at 12:00 on 13 Sept. 2002. See Figure 9.4 in Chapter 9 for corresponding 2001 graph. the analysis area. Calls were numerous from about 15 km offshore of Northstar seaward to well beyond the analysis area, as in 2002. The southern edge of the migration corridor east of the study area was again close to Cross Island (Fig. 9.3 in Chapter 9). A few calls may have occurred slightly closer to the baseline through Northstar in 2001 than in 2002, with the closest of the estimated locations being ~3 km seaward of the island. However, those calls were well inshore of the DASAR array, and these estimated locations are less accurate than those closer to the array (see Fig. 9.5 in Chapter 9 for estimates of location accuracy in 2001). In 2001, more whale calls were localized in the first half of the study period than in the second half, in contrast to 2002. This may have been partly an artifact resulting from a decline in the number of functioning DASARs late in the 2001 season (usually 8–10 before 15 Sept.; diminishing to 7–8 in mid-Sept. and to 4 on 1–2 Oct.). In 2002, 10 or (briefly) 11 DASARs operated throughout the season. Autocorrelation analysis of the 2002 offshore distances indicated that the distances became uncorrelated with one another after an elapsed time of about 12 hours (Fig. 10.6). A corresponding analysis of the 2001 data showed that significant correlations (positive and negative) persisted for at least 24 hours (Fig. 9.6 in Chapter 9). Therefore, the duration of the time periods in the permutation procedure used to determine confidence intervals was set to t = 12 hours for 2002 but t = 24 hours for 2001. Sounds 400 m from Northstar

The characteristics and variability in the sounds received ~400 m offshore of Northstar throughout the 2001 and 2002 field seasons have been described for 2001 by Greene et al. (2002), and for 2002 by Blackwell and Greene (Chapter 6). The distributions of Industrial Sound Index values (as defined in the Chapter 10: Deflection of Bowhead Migration? 10-13 Correlation (Moran's I) -0.4 0.0 0.4

0 1020304050 Hours between locations

FIGURE 10.6. Moran’s I statistics and estimated autocorrelation function for offshore distances of calls detected in the analysis zone in 2002. Vertical bars are Bonferroni-corrected 95% confidence intervals on Moran’s I. See Figure 9.6 in Chapter 9 for corresponding 2001 graph.

“Methods”), measured ~400 m from Northstar, are shown in Figure 10.7. The median ISI values were 96.3 and 99.1 dB re 1 µPa for 2001 and 2002, respectively, based on five 1/3rd octave bands (Fig. 10.7A,B). These values were slightly lower than the corresponding medians for the overall 10–1000 Hz band, 102.6 and 103.0 dB for 2001 and 2002, respectively. The lower ISI values were expected, given the additional frequencies included in the 10–1000 Hz band as compared with the ISI. Mean 1-min ISI values for 2001 and 2002 were 5.5 and 4.3 dB (respectively) below the mean 10–1000 Hz broadband values. The ISI values were more variable in 2001 than in 2002 (Fig. 10.7A vs. B). The 5th and 95th percentiles in 2001 were 83.4 and 118.7 dB re 1 µPa (a span of 35.3 dB), whereas those in 2002 were 90.9 and 109.2 dB (a span of only 18.3 dB). This is important in relation to potential effects of variable Northstar sound on the offshore distribution of whale calls. Effects are more readily detected when the predictor variable (here ISI) varies across a wider range. Thus, one might expect any potential effect of variable sound levels on whale call distribution to be more difficult to detect in 2002 than in 2001, if other factors were equal. The ISI values ~400 m from Northstar averaged over the 70-min periods preceding detections of whale calls within the analysis area are shown in Figure 10.7C,D. In 2001, the values ranged from 78.4 to 125.5 dB re 1 µPa, with the 5th and 95th percentiles 82.5 and 107.7 dB (median 88.8 dB). In 2002, the 70- min ISI values ranged from 88.2 to 126.1 dB re 1 µPa, with 5th and 95th percentiles 93.0 and 106.2 dB (median 99.2 dB). Variability was again notably less in 2002 than in 2001 (Fig. 10.7C,D). 10-14 Monitoring at Northstar, Open-Water Season

2001 2002

A B Number of 1-minute ISI levels Number of 1-minute ISI levels 1-minute intervals 0 200 400 600 800 1000 0 200 400 600 800 1200

70 80 90 100 110 120 130 70 80 90 100 110 120 130 5-band ISI level (dB re 1 µ Pa) 5-band ISI level (dB re 1 µ Pa)

C D Number of 70-minute ISI levels Number of 70-minute ISI levels 70-minute intervals 0 50 100 150 200 0 100 200 300 400

70 80 90 100 110 120 130 70 80 90 100 110 120 130 5-band ISI level (dB re 1 µ Pa) 5-band ISI level (dB re 1 µ Pa)

FIGURE 10.7. Industrial Sound Index (ISI) 400 m north of Northstar averaged over (A,B) 1-min periods at 4.37-min intervals, and (C,D) 70-min periods preceding the times of the whale calls in the analysis zone. A and C apply to 29 Aug.–3 Oct. 2001 (11,846 1-min samples and 1121 whale-call times). B and D apply to 31 Aug.–23 Sept. 2002 (7551 1-min samples and 2057 whale-call times). 5th, 50th and 95th percentile values are also shown by vertical lines. The ISI included five 1/3rd octave bands (see “Methods”).

Call Locations vs. Sound Level near Northstar

2001 Results Offshore distances of calls in the “analysis zone” were analyzed for a disturbance effect by estimating the relationship between the 5th quantile of offshore distance and ISI, weighting each case by the square root of the area of each call’s 90% confidence ellipse. Northstar ISI levels associated with each whale call were averaged over several different time periods immediately preceding the call, and separate two-threshold quantile regressions (one for each averaging time) were fitted. Figure 10.8 shows the results for 70-min averaging time. McDonald and Richardson (Chapter 9) show corresponding graphs for 5, 15, 30, 60 and 90 min averaging times. Included in each Figure are the average 5th quantile line assuming no effect, and its 90% confidence bounds. The quantile regression lines estimated based on the 15–70 min averaging times indicate that the two thresholds estimated by the model are similar, and in the range 98 dB to 105 dB. Chapter 10: Deflection of Bowhead Migration? 10-15

The 5th quantiles of offshore distance expected if there was no effect were not constant across varying sound (ISI) levels. For the 70-min averaging time (Fig. 10.8) and most other averaging times plotted in Chapter 9, the 5th quantile under the assumption of no effect declined between low ISI values (~78 dB) and moderate ISI (~98–105 dB), but then increased from moderate ISI to high ISI (~125 dB). When sound levels in the 5 ISI bands were comparatively high (>112 dB re 1 µPa as measured 400 m from Northstar), the estimated 5th quantile line fell outside what could be expected by chance, though only for the 70- and 80-minute ISI averaging times, and even then just barely. (The 80-min results were inspected but are not presented here.) For ISI levels from 113 to 122 dB re 1 µPa, the 5th quantile of offshore distance was farther offshore than expected given the upper bound of the 90% confidence region if offshore distance and ISI levels were unrelated (Fig. 10.8). This result effectively has 95% confidence if the problem is viewed from a one-tailed testing perspective. A one-sided approach is generally appropriate in a study of disturbance effects, and was proposed during planning of this study. The 5th quantile was estimated to displace 3.2 km offshore (90% CI = 0.2 km to 5.2 km) when ISI values averaged during the 70 min preceding the call were 119 dB re 1 µPa (Fig. 10.8). Estimates of displacement were positive for ISI values ≥107 dB re 1 µPa and most averaging times up to 80 min. However, the 90% confidence intervals showed that displacement was statistically larger than zero only with ISI levels 113–122 dB and averaging times 70–80 min (Fig. 9.13 and 9.14 in Chapter 9). At such times, the best estimates of the offshore displacement of the 5th quantile distance ranged from 2.3 to 3.3 km (Appendix 10.2A). When sound levels in the 5 ISI bands were moderately low (~85–95 dB), the estimated 5th quantile line was below what could be expected by chance for most of the sound integration times between 15 and 90 min (e.g., Fig. 10.8; see also Fig. 9.8 – 9.12 in Chapter 9). This result indicated that, in 2001, the 5th quantile of offshore distance was closer to Northstar than expected at ISI levels between 85 dB and 95 dB, by ~0.75–1.0 km. This effect, although slight, could indicate an attraction effect at moderate as compared with low sound levels, transitioning to an avoidance effect at high as compared with moderate sound levels. 2002 Results The 2001 results were suggestive of slight displacement of the closest whales at times with the highest sound levels, and perhaps a slight attraction effect at moderate sound levels. However, there was some concern that, in 2001, this effect was found by exploratory analysis of numerous averaging times, and the effect was evident only for a few of those averaging times—specifically 70 and 80 min. For 2002, we decided (a priori) to base our primary analysis on the averaging time that had showed the strongest apparent effect in 2001, which was 70 min. Only after evaluating the effect based on that averaging time would we check to see if the strongest apparent effect was, again in 2002, with averaging times near 70 min. Based on the 70-min average ISI, the quantile regression lines again show that the two thresholds estimated by the model are very close, and near 98 dB re 1 µPa (Fig. 10.9). The model suggests that the 5th quantile of offshore distances became closer to shore as ISI values increased from ~88 dB to 95−100 dB, and then became farther offshore from 95−100 dB to ~126 dB (Fig. 10.9). However, the expected 5th quantile distance in the absence of any Northstar effect also became closer to shore as ISI increased up to 95–100 dB, and then farther offshore as ISI increased further (see dashed lines in Fig. 10.9). For 70-min ISI values below 109 dB, the fitted 5th quantile was within the 90% confidence bounds assuming no Northstar effect. However, for 70-min ISI above 109 dB, the fitted line was outside the 90% confidence 10-16 Monitoring at Northstar, Open-Water Season

70 Minute Sound Window

6 Offshore Distance (miles)

6 Offshore distance (km)

4 5 10152025

80 90 100 110 120 130 Industrial Sound Index at 400 m (dB re 1 µ Pa)

FIGURE 10.8. Offshore distances of bowhead calls localized within the “analysis area” vs. average industrial sound index (ISI) ~400 m north of Northstar in the 70 min previous to the call, 2001. Quantile regression (solid line) relates 5th percentile of offshore distances for bowhead calls to 70-min ISI. 90% confidence interval (dashed lines) for the quantile regression if there is no effect was estimated via a stratified permutation method (see “Methods”). Average 5th percentile assuming no effect is also plotted

(– ▲ –). Symbol diameter is proportional to the weight assigned to that case, based on the area of its 90% confidence ellipse. region, indicative of an offshore displacement. As noted for 2001, this result effectively has 95% confidence given that a one-tailed approach is appropriate and was planned a priori. After finding evidence of a displacement effect at high 70-min ISI levels in 2002, we repeated the 2002 analysis with 18 averaging times from 5 min to 130 min. Detailed results for seven additional averaging times (5, 15, 30, 60, 80, 90, and 120 min) are shown in Appendix 10.1. Figure 10.10 summarizes results for all averaging times. The strongest displacement effect was for averaging times of 60–75 min, with maximum apparent effect at ~70 min averaging time (Fig. 10.10). However, at high ISI values, the fitted line extended outside the 90% confidence bounds around the no-effect line for all averaging times from 5 to 75 min (but not for longer averaging times). Thus, the 2002 results showed a probable displacement effect no matter which of a wide range of averaging times was considered (Fig. 10.11). The apparent displacement was positive at ISI levels from ~100 dB re 1 µPa upward, and was beyond the 90% confidence region for ISI values higher than 109–113 dB when ISI was averaged over periods ranging from 5 to 75 min (Fig. 10.8; Appendix 10.1). For 70-min ISI values of 109, 119 and 126 dB, the best estimates of the displacement were 2.2 km, 3.8 km and 4.7 km, respectively. For 70-min ISI = 119 dB, the 90% confidence interval for the displacement was 0.37 to 7.32 km (Fig. 10.11). In Chapter 10: Deflection of Bowhead Migration? 10-17

70 Minute Sound Window

6 Offshore Distance (miles)

6 10 15 20 25 Offshore distance (km)

90 100 110 120 Industrial Sound Index at 400 m (dB re 1 µ Pa)

FIGURE 10.9. Offshore distances of bowhead calls vs. average industrial sound index (ISI) ~400 m north of Northstar in the 70 min previous to the call, 2002. The analysis was based on 2057 calls occurring within the analysis zone. Otherwise as in Figure 10.8.

0 Offshore Displacement (miles) ISI Level

127 dB . 5 125 dB 123 dB

121 dB 2

. 119 dB 0 117 dB 115 dB

. 113 dB 5 111 dB

109 dB 1

0 107 dB

5 Offshore Displacement (km)

0 012345

0 20406080100120 Sound Integration time (minutes)

th FIGURE 10.10. Estimated displacement of the 5 percentile of offshore distances as a function of ISI level and sound averaging (integration) time, 2002. Lower 90% confidence limits on displacement were >0 only for ISI levels >109–113 dB re 1 µPa and averaging times 5–75 min. See Appendix 10.2A for 2001. 10-18 Monitoring at Northstar, Open-Water Season Offshore Displacement (miles)

ISI Level 119 dB 2

1 Offshore Displacement (km)

0 0246

0 20 40 60 80 100 120 Sound Integration time (minutes)

th FIGURE 10.11. Estimated displacement of the 5 percentile of offshore distances when ISI levels averaged over varying lengths of time reached 119 dB re 1 µPa, 2002. Dashed lines indicate lower and upper limits of the 90% confidence region under the assumption of no displacement effect. See Appendix 10.2B for 2001. general, the 2002 results show that there probably was an offshore displacement effect at times when the industrial sound levels near Northstar were highest, but the estimated size of this displacement effect was quite imprecise. In 2002, contrary to 2001, there was no indication of an “attraction” effect at low-to-moderate ISI values. The 5th quantile offshore distance became closer to shore as ISI increased from low to moderate values, but this trend was expected even on the assumption of no disturbance effect. The fitted 5th quantile at low-to-moderate ISI was well within the 90% confidence region for no effect.

DISCUSSION

This study has demonstrated that the acoustical localization method is a practical method for assessing the extent to which the Northstar oil development influences bowhead whales migrating west in the southern (inshore) part of their migration corridor. Evidence consistent with slight offshore displacement of the southern edge of the bowhead migration corridor at times with high levels of industrial sound has been found in both 2001 and 2002: • The 2001 data show evidence of offshore displacement of the southern edge of the migration corridor at the times with strongest Northstar sound, but only when that sound is averaged over a specific interval (~70 min) preceding each whale call. For that averaging time, the data indicate that we can be 95% confident (though barely so) that there was offshore displacement when sound levels were high. The evidence of displacement was not convincing (confidence <95%) when sound levels were averaged over other intervals. For 2001, there was no a priori basis on which to predict the most appropriate averaging time. Chapter 10: Deflection of Bowhead Migration? 10-19

• In 2002, more whale calls were localized, and the precision of localization was improved. The 2002 data, when analyzed based on the 70-minute averaging time suggested by the 2001 results, again showed evidence of offshore displacement of the southern edge of the migration corridor at the times with strongest Northstar sound. In 2002, in contrast to 2001, the displacement effect was evident (with ≥95% confidence) for a wide range of averaging times (5 to 75 min). The effect evident in 2002 and apparently also 2001 might be different or absent in another year with differing ice conditions, migration corridor, or industrial activities. The data pertain to only a five-week segment of the bowhead migration season in 2001, and a 3.5-week segment in 2002. Both of those years were light ice years. There is much year-to-year variation in the migration corridor of bowheads, in ice conditions, and in other aspects of the environment. The “Northstar effect” evident in the southern part of the bowhead migration corridor at high-noise times in 2001 and (more strongly) in 2002 might not occur in a year when the overall migration corridor in the central Alaskan Beaufort Sea is farther offshore. The effect might be different, or absent, in a year when ice has a strong effect on the migration corridor. It also may not be evident in future years if the underwater sounds created by Northstar during oil production in later years are weaker and/or less variable than those during 2001–02. Boat traffic, which was responsible for most of the high ISI levels measured in 2001–02 (see Chapter 6), is expected to diminish in future as drilling is completed. Also, a hovercraft may replace some boat (and helicopter) traffic in future, and the hovercraft may produce less underwater sound than is produced by vessels. BP applied the same acoustical monitoring approach during 2003 as in 2001–02, and the additional 2003 data are expected to become available by mid- 2004. The 2001 data are consistent with the possibility of an attraction effect at times with moderate levels of Northstar sounds as compared with times having low levels. The closest whale calls tended to be closer at times with moderate Northstar sounds than with weaker Northstar sounds. However, no corresponding “attraction” effect was evident in the larger 2002 dataset. Effect of Northstar-Related Sound on Call Locations

Based on 2001 and especially 2002 results, we are reasonably confident (≥95%) that, at times with high levels of underwater sound near Northstar Island averaged over 70 min, the spatial distribution of bowhead calls near Northstar was affected. For several reasons, the confidence level is higher for 2002 than for 2001: • The 2002 call localization data were more numerous and had greater precision given various technical improvements made in 2002, and given the continued operation of 10 DASARs throughout the 2002 field season. (In 2001, there had been gradual attrition in the number of functioning DASARs.) • During times in 2002 with high levels of industrial sound, the fitted 5th-quantile of offshore distance extended well outside the range expected if there were no Northstar effect. In 2001, the fitted line extended barely outside that range. • In 2002, unlike 2001, the primary averaging time for sound levels was selected a priori (based on 2001 results). Furthermore, in 2002 the displacement effect was evident not only for that averaging time, but also for a wide range of alternative averaging times. Considering the 2001 and 2002 data together, it is apparent that—at least in those years—as the industrial sound level near Northstar increased above a certain threshold value, the southernmost of the whale calls tended to be increasingly far offshore. This result addressed one of the key objectives: “to 10-20 Monitoring at Northstar, Open-Water Season estimate the effects of Northstar activities on the migration corridor and calling behavior of bowheads by comparing distances of calling whales from Northstar at times with varying levels of industrial sound”. In 2002, the magnitude of offshore displacement, based on the 5th quantile of the offshore distances, is estimated to have reached a maximum of about 4–5 km at the rare times with the highest levels of Northstar-related sound (Fig. 10.10, 10.11). However, this estimated displacement has broad 90% confidence limits, ranging from <1 km to 7+ km (Fig. 10.11). If Northstar-related vessel activity is reduced in future, as expected, sound levels high enough to cause this displacement will become even less frequent. The analysis of the distribution of bowhead calls relative to Northstar sound depended on two main types of data―sound measurements near Northstar and localization of calling bowheads―, and on several decisions as to how these data should be analyzed. These topics are considered briefly here: 1. During the monitoring periods in 2001 and 2002, continuous data on low-frequency (up to 1000 Hz) underwater sounds were available from several instruments deployed within a few hundred meters offshore of Northstar. For the present analysis, we used data from a single hydrophone (C.H.2) that operated continuously throughout the two seasons at a location ~400 m seaward of Northstar Island. We restricted the acoustic data to five 1/3-octave bands in which the sounds seemed to be dominated by industrial sound. This resulted in a sound index that best represented the variability in industrial sound emissions. However, the numerical value of this index was inevitably lower than the overall sound pressure level in the 10–1000 Hz band—a commonly-used measure of industrial sounds. The mean difference between our industrial sound index and the 10–1000 Hz level was 5.5 ± s.d. 3.02 dB in 2001, and 4.3 dB ± s.d. 2.57 dB in 2002. 2. The displacement effect became conspicuous at sound index values (averaged over 70 min) above ~113 dB in 2001 and ~109 dB in 2002. By “conspicuous”, we mean “large enough such that the fitted 5th-quantile offshore distance was outside the 90% confidence region if there were no Northstar effect”. These ISI values correspond (respectively) to ~118 and 113 dB re 1 µPa on a 10–1000 Hz basis. The sound effect on offshore distances may have started to become evident at a sound index (averaged for 70 min) as low as 105 dB re 1 µPa in 2001, and 100 dB in 2002. These are the levels above which the fitted 5th-quantile of offshore distances exceeded (barely) the most likely value in the absence of any effect. These values correspond to 10–1000 Hz values of ~110 and 104 dB re 1 µPa. 3. The sound levels mentioned thus far in this chapter were all at a location only ~400 m from the north edge of Northstar Island. The whale calls that were localized were all farther away than that. Sound levels tend to decrease with increasing distance from the source (see Chapter 6). Thus, some bowheads were apparently reacting to sounds at received levels lower than those that were directly measured 400 m from Northstar. The estimated received levels of the industrial sound index (ISI) at whale call locations, averaged over the 60-min period preceding detection of each call, are shown in Figures 10.12 and 10.13. Received levels were estimated assuming that sound attenuated at the rate of 15 log (range) from 400 m to the distance of the whale. Of relevance here are the whales within the analysis zone; those are the cases considered in Figures 10.8–10.11. The 60-min average ISI at call locations within the analysis zone was only rarely in the range 100–110 dB re 1 µPa (yellow symbols in Fig. 10.12–10.13), and only infrequently in the range 90–100 dB (green symbols). For the vast majority of the calling whales in the analysis zone, the estimated received level was <90 dB (blue and purple symbols). Broadband (e.g., 10–1000 Hz) received levels at the whale locations would have averaged 4–5 dB higher, as previously described. Even so, received broadband levels at the locations of the calling whales rarely if ever exceeded 115 dB re 1 µPa. Thus, the displacement effect evident at times with comparatively high levels of Northstar sound was occurring in situations where Chapter 10: Deflection of Bowhead Migration? 10-21

the received sound levels (averaged over 60 min) were not very high—less than 115 dB re 1 µPa on a broadband basis. 4. The accuracy of the call locations determined in this study has been discussed in Chapters 7–9. In general, the bearings determined by the DASARs in 2002 were more accurate than those for 2001. Also, in 2002, whale calls tended to be detected by more DASARs than in 2001, resulting in more bearings being available for the average call in 2002 than in 2001. For both reasons, locations tended to be more accurate in 2002. Nonetheless, there was some uncertainty in the estimated location of each calling whale in both years, especially for whales that were well outside the perimeter of the array of DASARs. The potential negative effects of this inaccuracy on the overall statistical analysis of call locations vs. sound were minimized by two methods. (1) After the uncertainty in each call location was estimated using available calibration data, uncertainty was used as a weighting factor in the analysis, with imprecise locations being down-weighted. (2) For other reasons (avoidance of a possible bias), it was necessary to limit the study area to the zone within 10 km east, west and offshore of the central axis of the DASAR array. This automatically excluded calls at locations distant from the instruments, which constituted most of the calls whose positions were especially uncertain. 5. We analyzed the distribution of whale calls based on their perpendicular distances offshore from a line drawn through Northstar Island and parallel to the coast. Thus, a call that originated 10 km offshore of this line was considered to be at the same offshore distance whether it was directly offshore of Northstar or as much as 10 km to the east or west. (Calls more than 10 km east or west were excluded, as noted above.) This approach has some advantages and some disadvantages over the simplest alternative, analyzing radial distances from Northstar. As described in Chapter 9, it seemed most appropriate to determine the extent to which, for calling whales in the general Northstar area, distance-from-shore was related to the sound level. However, in follow-up analyses it might also be useful to test for an effect of sound on radial distances. 6. The displacement effect did not become conspicuous until the industrial sound index at 400 m, averaged over ~70 min, exceeded 113 dB (2001) or ~109 dB (2002) re 1 µPa, i.e., overall 10–1000 Hz band level above ~118 or 113 dB. The 70-min sound index 400 m from Northstar exceeded these levels for less than 5% of the whale calls during each of the years (Fig. 10.7). Thus it can be assumed that only a small fraction of the total bowhead population passed at times with any appreciable Northstar effect. Furthermore, at those times, the Northstar effect was expected to be restricted to only the southern part of the migration corridor, and the observed distributions of call locations in 2001–02 (Fig. 10.8, 10.9) were consistent with that expectation. Further analyses based on fitting quantiles other than the 5th also indicated that the displacement effect waned with increasing distance offshore (see Fig. 11.3 in Chapter 11). The number of bowheads potentially affected is discussed in Chapter 11, but it is obvious that this must have been only a small fraction of those passing through the monitoring zone, and an even smaller fraction of the population. The analysis zone extended offshore 26 km from Northstar, to the 30-m depth contour (e.g., Fig. 10.2). This is near, or perhaps slightly inshore of, the middle of the migration corridor in an average year (LGL and Greeneridge 1996; Miller et al. 1999; Treacy 2000; see Chapter 11). Thus, it is probable that no more than half of the whales were close enough to shore to pass through the analysis zone. Although the DASARs do not detect or localize bowheads beyond the analysis zone as reliably as they do for whales within that zone, the large number of whales occurring >26 km seaward of Northstar is evident in a general way from Figures 10.12 and 10.13. 10-22 Monitoring at Northstar, Open-Water Season

FIGURE 10.12. Map of whale call locations within and near the “analysis area” in 2001. Calls are color- coded based on estimated average industrial sound index at whale location during 1 h preceding the call. Chapter 10: Deflection of Bowhead Migration? 10-23

FIGURE 10.13. Map of whale call locations within and near the “analysis area” in 2002. Calls are color- coded based on estimated average industrial sound index at whale location during 1 h preceding the call. 10-24 Monitoring at Northstar, Open-Water Season

7. A further consideration is that the bowhead migration past Northstar continues until mid or late October. In both 2001 and 2002, routine vessel operations near Northstar ceased in early October, when freeze-up started to occur. High values of the industrial sound index were largely if not entirely attributable to vessels operating near Northstar (see Chapter 6). We do not have data on sound levels near Northstar after freeze-up. However, after that date, there probably were few if any occasions when industrial sound levels were high enough to cause any “Northstar effect” on migrating bowhead whales. This would further reduce the percentage of the population that might have been affected. Chapter 11 provides additional details concerning numbers of bowhead whales potentially affected. Statistical Analysis Issues

Many of the passing whales are expected to emit a number of calls in succession, and this raises concerns about lack of independence of the individual calls. In a preliminary analysis of the 2001 data described in Greene et al. (2002), we tested for autocorrelation in the distance-offshore data, and then applied a bootstrapping (re-sampling) approach. In the present re-analysis of the 2001 data and initial analysis of 2002 data, we use a closely related but further refined approach. After applying autocorrelation analysis to determine the interval within which distances are interrelated, we now use a permutation method to estimate the expected relationship between offshore distances and sound level. This allows for the interdependence of data collected within short time periods, and avoids overestimating the strength of the apparent relationship between sound levels and call distribution. The above approach also allows for the possibility that the quantile distance might vary with sound level even in the absence of a Northstar effect. McDonald and Richardson (in Chapter 9) showed that such an effect can occur when the distributions of the two variables under consideration are non-uniform, as they typically will be. The specific mathematical reason(s) for this effect have not been investigated. However, it appears that the slope and shape of a quantile line in the absence of any dependence of the response variable on the stimulus depend on the distributions of the two variables, and in particular the amount of skewness in each (Chapter 9). In Figures 10.8 and 10.9, the line showing the “average 5th percentile assuming no effect” (– ▲ –) is in fact curved. This illustrates that, with the actual distributions of offshore distances and 70-minute sound levels as measured in 2001 and 2002, the expected relationship if there is no Northstar effect was not a straight horizontal line. The preliminary analysis of 2001 data given in Greene et al. (2002) did not allow for that possibility, and as a result, tended to overestimate both the displacement at high sound levels and the possible attraction effect at low-to-moderate sound levels. The revised approach used in the present analyses avoids this problem. The outcome of the analyses (especially in 2001) depended on the averaging time that was used when determining the Northstar sound level. Apparent offshore displacement at times with relatively high sound emissions was evident in both years when sound was averaged for ~70 minutes. However, for 2001 in particular, displacement was less evident (or not evident) based on shorter or longer averaging times. Prior to collecting the 2001 data, there was no specific basis on which to predict the averaging time most relevant to bowhead whales. It was reasonable to assume that sounds received by a whale over some substantial period preceding detection of a given whale call would influence the location of the whale. If the apparent offshore displacement at high-sound times was at least partly the result of an actual offshore displacement of whales, and not solely attributable to a reduced tendency to call when exposed to industrial sound, it is not surprising that displacement would be most evident when sound is averaged over a relatively long (e.g., 70-min) period. Migrating bowhead whales travel at an average speed of only ~4–5 km/h, and net speeds are lower for whales that linger to feed or socialize (Richardson et al. 1995a; Koski et al. 2002). Any substantial displacement in response to strong sounds would require some time to become evident. Chapter 10: Deflection of Bowhead Migration? 10-25

Prior to our initial analysis of the 2001 data (Greene et al. 2002), we decided to consider 5, 15, 30 and 60 min averaging times. In hindsight, it was possible that the apparent displacement observed at times with strongest sounds might have been stronger with an averaging time longer than 60 min. As a result, we investigated a wider range of averaging times in the present re-analysis of the 2001 data, and in our analysis of 2002 data. For both years, the displacement effect was most evident when sound was averaged over ~70 min preceding the time when a given whale call was heard. When the 2001 data were re-analyzed, the effect was not very convincing for any averaging other than 70–80 min. However, for 2002, the effect was evident across a wide range of averaging times, from 5 to 75 min. A reviewer has asked whether the analytical approach assumes that bowhead whales return to their original distance offshore after noise is reduced, and how the analysis would be affected if whales do not return to their original distance offshore after passing Northstar. The longer that migrating whales remain displaced after passing Northstar, the farther west the displacement effect would extend. It was outside the scope of this study to address how long and how far any displacement effect might extend. However, uncertainty about the duration and westward extent of displacement should not affect our analysis, which deals with the occurrence and geographic extent of offshore displacement within a relatively narrow (20 km) east–west zone near Northstar. For 2001, there may be some reason for concern about the effects of conducting multiple analyses with varying averaging times on the confidence bounds for the 5th-quantile distance offshore. Some statisticians would recommend correcting the probability statements that we have made to account for the fact that we conducted multiple analyses on the same 2001 data set. Any such correction would involve a reduction in the stated confidence level. Other statisticians would recommend no correction of probability statements based on multiple analyses. They might argue that, because the analyses we conducted were planned in advance, we do not need to correct for multiple analyses. Still other statisticians have expressed the opinion that no study ever needs to correct for multiple analyses of the same data, whether analyses were planned or not. We have not applied any such adjustment for 2001, primarily to avoid further complications in presentation. Instead, our confidence bounds for 2001 should be considered “approximate”. For 2002, there is no such concern about multiple analyses. There were two reasons for this improvement: (a) In 2002, unlike 2001, there was an a priori basis (i.e., the 2001 results) for selecting a 70-min analysis time for the primary analysis. (b) In 2002, the results were similar for a wide range of averaging times, and not strongly dependent on the specific averaging time used. Given the 2001 and 2002 results, we will again use 70 minutes as the a priori averaging time during analyses of the corresponding data that have been collected during 2003. In 2000, before collecting any data, we decided on a 1-tailed analysis approach. We assumed that any relationship of the “distances offshore” to Northstar sound would involve increasing distances with increasing sound level. The appropriateness of a 1-tailed approach was discussed at the peer/stakeholder review meeting prior to the 2000 field season, and was subsequently mentioned in the monitoring plans for 2000 and 2001–02 (submitted Aug. 2000 and May 2001), and in previous reports on the 2000 and 2001 monitoring. The results from 2001 (e.g., Fig. 10.8) are consistent with the possibility that the closest bowhead calls tended to become somewhat closer to Northstar as ISI levels increased from low to moderate. However, the 2002 data (which are more numerous and more precise) did not show any evidence of an “attraction” effect. Whether or not any slight “attraction” effect ever occurs at low-to- moderate sound levels (see below), it is still appropriate to adopt a 1-sided approach when considering the 10-26 Monitoring at Northstar, Open-Water Season possibility of a displacement effect at moderate-to-high sound levels, i.e., for the portion of the fitted threshold model above the threshold(s). Possible Attraction to Weak Industrial Sounds

It would not be entirely surprising if some whales in the southern part of the migration corridor were somewhat attracted toward the relatively weak received levels that they might perceive during times when sound emissions were at moderate levels. Bowheads and other whales have often been shown to exhibit tolerance of industrial sounds that are only slightly above ambient noise levels provided those sounds are not associated with any obvious threat (Chapter 9 in Richardson et al. 1995b). Minke whales often approach and even swim under stationary or slow-moving vessels (Winn and Perkins 1976; I.W.C. 1982; Leatherwood et al. 1982; Tillman and Donovan 1986). Gray whales, especially in the calving lagoons in winter, are often attracted to quiet, idling, or slow-moving boats, especially in late winter (Swartz and Cummings 1978; Norris et al. 1983; Withrow 1983; Bryant et al. 1984; Dahlheim et al. 1984; Jones and Swartz 1984, 1986). Attraction to weak, steady sounds has not been demonstrated for bowheads, but the possibility should not be discounted without further consideration of the data. In doing this, it would be very desirable to take account of the industrial-to-ambient ratios at the locations of the calling whales. At times when ISI at 400 m is moderate, i.e., when there might have been a slight attraction effect during 2001, received levels at the locations of even the most inshore whales will often be below the ambient noise and presumably undetectable. It is difficult to see how there could be attraction to Northstar sound in these circumstances. Unfortunately, we do not have precise data on ambient noise levels near each calling whale, so the industrial-to-ambient ratios to which each whale is exposed are not specifically determinable. Advantages and Limitations of the Acoustical Monitoring Approach

The acoustical method used in this study has several major advantages. As compared with aerial surveys, the main advantages are the large number of detections and localizations that it can provide, and its ability to operate autonomously, day or night, and in any weather conditions. Statistical power analyses were conducted prior to the start of this study to evaluate the anticipated ability of this acoustical approach vs. intensive site-specific aerial surveys. In both cases, it was assumed that the primary objective was to detect small-scale deflection of bowheads in the southern part of their migration corridor at high-noise times. The results of the power analyses were described in Appendices to the final monitoring plan for 2000 (LGL and Greeneridge 2000), and were repeated in Appendices to Greene et al. (2001). Those analyses showed that the most intensive aerial surveys that might have been practical would have very little ability to detect an effect of the anticipated geographic scale. In contrast, the acoustical approach was predicted to have the necessary statistical power. The actual sample sizes achieved with the acoustical method in 2001 and (especially) 2002, and the geographic scale of the apparent displacement effect during high-noise times, have proven to be generally consistent with those assumed in the a priori power analyses. Although those have not been updated based on actual results from 2001 and 2002, it is clear that the decision to adopt an acoustical approach rather than aerial surveys was appropriate. An effect of the magnitude evident from the acoustical monitoring during 2001 (barely) and 2002 (more clearly) would not have been detectable by aerial surveys. However, the acoustical approach also has limitations that need to be recognized when evaluating its suitability for any particular purpose. The acoustical approach depends on the calling behavior of Chapter 10: Deflection of Bowhead Migration? 10-27 bowhead whales, and on the ability of the instrumentation to detect and localize the calls. Both calling behavior and the capabilities of the localization system are complex, and several factors are known or suspected to influence each of them. The following list is the same one provided in previous reports on the acoustic localization efforts at Northstar in 2000 and 2001 (Greene et al. 2001, 2002), updated to take account of new information. 1. Only whales that call can be detected and localized. Any silent whales will be missed. If the proportion of whales that pass by without calling varies from time to time, this could result in biases. 2. The number of calls emitted by whales is not necessarily in direct proportion to the number of whales present. Whales may call at a higher rate on one occasion than another. 3. The rate and intensity (source level) at which whales call may be affected by the sounds (anthropogenic or natural ambient) that they hear. 4. There are limits on the distance to which a given call can be detected and localized. With the DIFAR method, a bearing can be obtained to any call that is detected, but localization accuracy diminishes as the number of DASARs detecting a given call decreases, and as the distance of the calling whale from the DASARs increases. Considerable improvements in localization accuracy were achieved in 2001 vs. 2000, and in 2002 vs. 2001. Also, a weighting procedure was applied during analysis to reduce the influence of the less reliable locations without discarding them altogether. 5. The source level of a whale call affects the ability of a DASAR (or any hydrophone) to detect the call. The higher the source level, the farther away the call will be detected, other factors being equal. As noted above, source levels might be affected by exposure to anthropogenic or natural background noise. 6. Background noise (anthropogenic plus natural ambient) at the location of a DASAR (or any hydrophone) affects its ability to detect a given whale call. In this study, all DASARs were close enough to Northstar such that, at times of high industrial sound output by or near Northstar, weak calls that would otherwise be detectable by the DASARs might be masked by industrial sounds. Also, natural ambient noise levels vary widely. Results from 2000 showed that there was a strong negative correlation between number of calls detected and background noise level (Greene et al. 2001). Results from 2001 and 2002 showed that the distances of the localized whale calls from a reference point near the center of the DASAR array depended strongly on background noise level. The number of calls detected dropped off sharply at distances beyond 10–12 km at times with high background noise, whereas many calls were detected at longer distances at the quiet times (see Chapter 8). To avoid biases associated with masking of distant whale calls at high- noise times, we limited the analysis area to a zone within 10 km of the central axis of the DASAR array (Fig. 10.2). This required exclusion of a large number of whale calls detected at greater distances (see Fig. 10.12, 10.13), thus reducing sample size. However, most of the excluded calls were from whales located well offshore, where Northstar sounds would be undetectable or weak, and at distances where Northstar was likely to have little if any effect on whale distribution. Exclusion of these calls probably had little effect on statistical power, and avoided biases that might otherwise have resulted. 7. Given these considerations, one cannot directly determine whether a change in the call detection rate represents a change in (a) the number whales present, (b) their calling rate, or (c) the system’s ability to detect the calls. However, considerable information about the effective monitor- 10-28 Monitoring at Northstar, Open-Water Season

ing radius can be derived, when the detected calls are localized, based on two types of data: the distance at which the number of detected calls starts to decline (e.g., Fig. 8.18 in Chapter 8), and the signal-to-noise ratios of the calls received from whales at various distances. Based on the 2000 study, we noted that it is important to ensure that the array of DASARs extends far enough south to monitor the southernmost portions of the migration corridor even at times when background noise levels are high. In 2001 and 2002, the DASARs were placed 1.85 km (1.15 mi) south of their 2000 locations. Also, in 2001 and 2002, another DASAR location was added to the southwest part of the array (WB in Fig. 10.2). As noted in (6) above, our analyses considered only the calls within the radius where most calls were detectable even at times with high noise levels, thus avoiding a potential bias. 8. To assess whether the distribution of calling whales detected north of Northstar is (or is not) related to the level of industrial sound from Northstar, it is necessary that there be substantial temporal variation in underwater sound emitted by Northstar activities. We expected that sound emissions would be quite variable during construction and during subsequent times with vessel traffic and/or drilling. (a) Results from 2001 confirmed that there was much variability in underwater sound levels near Northstar during the monitoring period from 28 Aug. to 3 Oct. 2001. Nonetheless, the great majority of the calls were at times when the sound level near Northstar was below that at which offshore deflection of the closest bowhead whales became evident (Fig. 10.8). (b) During the monitoring period in 2002, there was less temporal variability in sound levels than in 2001 (Fig. 10.7). This was probably a result of the transition from the late phases of construction in 2001 to initial production (plus drilling) in 2002. The reduced variability in Northstar sounds would be expected to make it more difficult to detect any effect of variable Northstar sound on the southern edge of the migration corridor. Nonetheless, this effect was more clearly evident in 2002 than in 2001, probably because of the larger sample size in 2002 and the greater precision of the call locations in 2002. (c) Our data did not extend into the latter part of the 2001 or 2002 bowhead migration periods, as it was necessary to retrieve the acoustic instrumentation before vessel operations were curtailed by ice. (Also, the 2002 season was curtailed when a storm upset the DASARs on 23 Sept. 2002.) Vessel operations near Northstar ceased by early October in both 2001 and 2002. Thereafter, underwater sound levels were unmeasured but most if not all the time were probably below the level where deflection of bowheads would occur. (d) There may be even less variability in underwater sound emissions, and generally lower levels of underwater sound, during “routine” production operations in future years, after drilling ends and/or if vessel traffic to Northstar is reduced. Thus, deflection of bowheads may become less common, smaller in magnitude, and more difficult to document in future. 9. If there is an offshore deflection or change in calling behavior but it is limited to a small radius around Northstar, the number of call localizations within the impact area might be too small to allow detection of a difference during the high-sound vs. low-sound periods. This would be most likely in a year when the general bowhead migration corridor is farther offshore than average, or a year when sound emissions are relatively low most of the time. In 2001 and especially 2002, there was sufficient variability in sound, and a sufficient effect on the bowhead migration, that the statistical technique was able to identify an effect (though barely so in 2001). If no similar effect were found in another year, or in some future application of this methodology elsewhere, the precision of the statistical estimators should be determined for various sizes of hypothesized effects. Chapter 10: Deflection of Bowhead Migration? 10-29

This would be an update of the a priori power analysis given in Appendix D of Greene et al. (2001), taking account of actual sample size, call locations, and analysis methodology. 10. There could be a change in the relative numbers of calls detected at different distances offshore even if there were no corresponding change in the distribution of whales. An effect such as that detected in 2001 (barely, Fig. 10.8) or 2002 (more clearly, Fig. 10.9) indicates that some change in bowhead behavior occurred in response to industrial sound. However, it is not immediately obvious whether this industry effect involved a change in whale distribution, calling rate, source levels of the calls, or some combination of these parameters. In our description of the results, we often state or imply that the effect represents an offshore deflection upon exposure to high sound levels. This type of effect was predicted based on reactions of bowheads to industrial activities in other situations. However, we acknowledge that the sound effect on locations where calls are detected could be attributable, at least in part, to a sound effect on some other aspect of bowhead behavior. To some degree, it may ultimately be possible to discriminate these potential types of industry effects based on localization data, sequences of calls, and received levels (and perhaps types) of calls. This level of refinement has not been attempted here. However, in 2001 and 2002, the received levels and types of calls were logged for possible future analysis if warranted. 11. Another complication in assessing the disturbance effects of any one activity, in this case North- star, is the occurrence of other potentially disturbing activities in the area: • 2001: The Concrete Island Drilling Structure (CIDS) had been cold-stacked 1.1 km (0.7 mi) northeast of Northstar since August 2000. In August 2001, there was vessel activity around CIDS as it was prepared for its eventual westward departure on 31 Aug. 2001. The activities at and around CIDS were a confounding factor during the first three days of our 2001 monitoring period (28–31 Aug.; see Chapter 2). • 2002: The Steel Drilling Caisson (SDC) arrived in the Prudhoe Bay area from the west on 4 Aug. 2002. By 14 Aug., it had been emplaced 18 km (11 mi) east of Northstar (Fig. 10.13), and cold-stacking was completed. The SDC was present just east of our analysis area, but inoperative, for the remainder of the 2002 open-water season. • Bowhead hunting near Cross Island, located 28 km (17.5 mi) east of Northstar, also might influence whale distribution, calling behavior, or both. The eastern edge portion of our analysis area was 12 km (7.5 mi) west of Cross Island (Fig. 10.12, 10.13). However, it is not known how far west of Cross Island any hunting effects might extend. The Nuiq- sut hunters were at Cross Island from 3 to 26 Sept. in 2001, and from 30 Aug. to 21 Sept. in 2002. In 2001, they hunted intermittently from 4 or 5 Sept. to 22 Sept., and landed three whales, on 5, 10 and 22 Sept. (M. Galginaitis, pers. comm.). In 2002, they hunted intermittently until 18 Sept., and struck five whales: two on 5 Sept. (one lost; one landed); one on 12 Sept. (sank; retrieved 19 Sept.); one on 13 Sept. (landed that day); and one on 15 Sept. (sank; retrieved 17 Sept.) (Galginaitis 2003). • Boat traffic aside from Northstar support: Activities of the boat we used to deploy, service, calibrate, and retrieve seafloor recorders may have caused intermittent disturbance to some bowheads close to the vessel. We conducted such operations for several hours once or twice per week during the 2001 and 2002 monitoring periods. The 2002 analyses in this chapter exclude the times potentially influenced by this boat traffic (see “Methods”). However, this was not done for 2001, and this boat traffic might be a minor 10-30 Monitoring at Northstar, Open-Water Season

confounding factor in 2001. Also, records of other vessel traffic unrelated to Northstar were lacking or limited in detail for both years, but there were tens of vessel trips in waters west and east of Northstar during our study period in 2002 (Table 2.1 in Chapter 2), and no doubt also in 2001. Most of this traffic was probably fairly close to shore and distant from the whale migration corridor. However, some of these trips may have influenced some whales and added unexplained variance to the analysis of Northstar effects, as noted in the “Methods”. Notwithstanding these complications, the acoustical monitoring approach has the great advantage of being able to detect the locations of large numbers of bowhead whales as they pass the Northstar area, and to do so by both day and night, independent of darkness, visibility, or weather. Although this method has limitations and biases, so do other methods that have been suggested for use. Aerial surveys, in particular, also involve a variety of biases in detectability of bowhead whales (Thomas et al. 2002). In addition, sample sizes achievable with aerial surveys are 1–2 orders of magnitude smaller, and inadequate to address the primary project objective. Ability of Approach to Meet Objectives

The specific objectives of the acoustic monitoring project are listed in Chapter 7 of this report, and are discussed one-by-one here. Objectives of the 2000, 2001 and 2002 work were essentially the same. The 2000 work was valuable in developing the technique, but problems with equipment severely limited the results. These problems were partially overcome in 2001, and largely overcome in 2002. This section concentrates on the results from 2001 and 2002. It is an update of discussion included in Greene et al. (2001, 2002), now taking account of the successful monitoring effort in 2002. Objective 1 “to monitor the year 2002 bowhead migration past Northstar by acoustic localization methods, primarily to document the relative numbers of bowhead calls vs. distance offshore in the southern part of the bowhead migration corridor seaward of Northstar.” This objective was partially met in 2000, and was met in 2001 and 2002, although in each case for only a fraction of the bowhead migration period. In 2000, bowhead whale calls were recorded successfully, in large numbers. The 2000 results provided data on call detection rates at several locations varying distances from Northstar during about 20 days of the fall migration. Directions to calling whales were measured for all call detections, but in 2000 these bearings were not sufficiently accurate for reliable localization of the individual calling whales. The bearings (and hence localizations) obtained in 2000 were substantially less precise than had been anticipated. In 2001 and 2002, all necessary types of data, i.e., call locations and industrial sound levels near Northstar, were collected continuously over 36-day and 23-day periods, respectively, including the early and middle portion of the bowhead migration season. Also, the DASARs (and the call data stored in those instruments) were all recovered successfully just prior to freeze up. Localization accuracy was improved in 2001 and further improved in 2002 based on improved design and reliability of the DASARs, and a more systematic and comprehensive set of field calibrations of the DASAR clocks and orientations (see Chapters 7, 8). Objective 2 “to document continuously • the characteristics of the sounds propagating away from North- star, as determined at a close-in monitoring site approximately 300–500 m (1000–1600 ft) Chapter 10: Deflection of Bowhead Migration? 10-31

offshore from the island, as well as • the characteristics of the background sounds (industrial plus ambient) reaching the recording devices in the southern part of the migration corridor;” This objective was not met in 2000 except over the period 25–30 Sept, when a backup ASAR provided continuous hydrophone data from a site ¼ n.mi. (~455 m) NNE of Northstar. This objective was met in 2001 and 2002. The problems with the cabled hydrophone system encountered in 2000 were avoided in 2001 and 2002 by deploying redundant hydrophones and recording systems near Northstar, and by confirming frequently, throughout the field season, that useful data were being acquired. Results from the DASARs indicated that sounds from vessels operating near Northstar were sometimes detectable even at the DASARs situated farthest offshore. (The DASAR most distant from Northstar was at the NE location, 22 km (13.7 mi) northeast of Northstar—see Fig. 10.2.) One remaining instrumentation issue is the fact that the electronic noise floor of the DASARs is sufficiently high that electronic noise masks background sounds at times when they are weak. This had little if any negative effect on the detection of whale calls within the primary study area (within 10 km of the central axis of the DASAR array). However, it did reduce the ability of the DASARs to provide data on the full range of variability in background sounds. Objective 3 “to assess the effects of Northstar activities on the migration corridor and calling behavior of bowheads by comparing the distances of calling whales from Northstar at times with varying levels of industrial sound;” This objective was not met in 2000 because of the imprecision of the directional data and the lack of continuous data on sounds emitted by Northstar. Simultaneous data of both types were obtained for substantial parts of the 2001 and 2002 bowhead migration seasons, and this objective was met in 2001 and 2002. Instrumentation problems that were encountered in 2000 were much reduced in 2001, and further reduced in 2002, allowing us to obtain the needed types of data continuously. In particular, localization accuracy and DASAR reliability were progressively improved, allowing locations of calls within and near the DASAR array to be determined with sufficient accuracy in 2001 and (especially) 2002 for use in the analysis. The statistical approach that had been proposed for estimating the effects of Northstar sound on bowhead whales in the southern part of the bowhead migration corridor requires continuous data on sound emissions from Northstar and on locations of whale calls. Data acquisition problems in 2000 prevented application of the statistical approach in that year. Both types of data were obtained in 2001 and 2002. That allowed us to estimate the occurrence and magnitude of an apparent offshore displacement of some bowheads during the 2001 and 2002 autumn migrations, as related to industrial sound associated with Northstar. The results have confirmed that the number of bowhead calls that can be localized is sufficient for a meaningful analysis. This was especially so in 2002, when DASAR performance was best. Procedures have been developed • to avoid the bias that might otherwise occur as a result of variations in background noise level and detection-radius (Chapter 8); • to allow for the fact that calls detected close to one another in time may not be independent (Chapter 9); and • to estimate the expected relationship of the 5th quantile of offshore distances to sound levels in the absence of a Northstar effect. This relationship is more complex than had originally been assumed, 10-32 Monitoring at Northstar, Open-Water Season

and it was necessary to reanalyze the 2001 data in the present report in order to allow for recently recognized complications (Chapter 9). The analysis of 2001 data in Chapter 9, also summarized in the present chapter, supersedes that in Greene et al. (2002). Although we expected the Northstar effects to diminish with increasing distance offshore, the techniques used in this study were not originally designed to determine how far offshore the effects extended. However, the data nonetheless show that the geographic extent of the displacement effect was progressively less for the 10th – 30th quantiles of offshore distance, as compared with the 5th quantile (see Fig. 11.3 in Chapter 11). Thus, as expected, the displacement effect diminished with increasing distance offshore. Neither Objective 3 nor any of the other objectives dealt with the question, “if displaced offshore, how long do bowheads remain displaced, and how far west does the displacement effect persist”. Although this question if of interest, especially to bowhead hunters, it was outside the scope of the present study as defined during the planning phase. Objective (3) was “to assess the effects of Northstar on the migration corridor and the calling behavior of bowheads…”. The acoustic localization technique that has been applied cannot directly determine whether a change in the call detection rate close to shore at high noise times represents an offshore displacement, a change in calling rate, or a combination of the two. Either one would represent a change in the behavior of the migrating whales. This is discussed above under “Advantages and Limitations…”, items (7) and (10). Given this inability to distinguish Northstar effects on distribution from those on calling behavior, we have viewed objective (3) as requiring us “to assess the effects of Northstar on the migration corridor and/or the calling behavior…”. However, “/or” does not appear in the objective as stated. Little specific analysis of calling behavior has been done based on the 2001 or 2002 data. It would not be useful to do simple plots or simple analyses of number of calls vs. ISI. The number of calls detected is strongly and inversely affected by background noise level. That was shown directly from the 2000 results (Fig. 9.14 in Greene et al. 2001), and indirectly for 2001–2002 from Figure 8.18 in this report and Figure 8.28 in Greene et al. (2002). Thus, any analysis of numbers of calls detected would need to allow for variation in background noise levels, distribution of the calls, etc. The lack of independence of calls that are close in space and time would also be a complication. Although potentially of interest, a meaningful analysis of possible Northstar effects on call numbers (and types) would not be simple. Objective 4 “to track individual bowheads or bowhead groups acoustically when feasible. The tracks will be analyzed to determine whether there is evidence of deflection as the whales approach Northstar and, if so, at what distances and received sound levels;” This objective was originally worded in terms of the feasibility of tracking bowheads acoustically, and then reworded as above for 2002. The feasibility objective was met to a limited extent with the 2000 and 2001 data. In 2000, the lack of precise, reliable localizations precluded detailed analysis of sequences of calls. However, the observed sequences of call bearings vs. time (Greene et al. 2001) showed several sets of successive points that appeared to represent “tracks in bearing”. These data suggested that, when the localization problems encountered in 2000 were solved, it would likely be possible to track some calling bowheads (or closely spaced groups) via the planned acoustical method. Results from 2001 showed apparent temporal sequences of calls progressing WNW toward and through the DASAR array, or passing just offshore of the DASARs (Greene et al. 2002). Specific analysis of these call sequences was not pursued in 2001 or 2002 because the statistical analysis based on individual call locations was able to characterize a Chapter 10: Deflection of Bowhead Migration? 10-33

Northstar effect on locations of detectable whale calls. Any analysis of call sequences would be complicated by the fact that it is impossible to be certain whether successive calls from locations close to one another are created by the same bowhead (or the same group of bowheads). Objective 5 “to compare the acoustic monitoring results with results from MMS or any other available aerial surveys to further assess the strengths and limitations of acoustic and aerial methods in documenting the bowhead migration corridor past Northstar and in detecting any change (probably small-scale, if at all) in the migration corridor near a site like Northstar;” Review and analysis of historical aerial survey data before the 2000 field season showed conclusively that the number of aerial survey sightings that could be acquired near Northstar, even with a dedicated site- specific aerial survey project, would be far from adequate to detect an offshore displacement of the anticipated magnitude if it occurred. Conversely, acoustic localization held promise as being able to meet this objective (LGL et al. 2000; see also Appendices C and D in Greene et al. 2001). Based on that assessment, BP proposed and the peer/stakeholder group agreed (at its May 2000 meeting) that the acoustical approach to monitoring should be adopted, and that site-specific aerial surveys would not be effective. The same group, at its June 2001 meeting, re-affirmed that the acoustical approach should be applied again during the autumn of 2001. As implemented in 2001 and especially 2002, the acoustic method was able to identify and characterize a small-scale change in the pattern of call detections during the minority of the time when much underwater sound was emitted near Northstar. Results from 2000, 2001 and 2002 showed that bowheads moved past Northstar in approximately their normal migration corridor during the portions of those migration seasons when monitoring was possible. As in years previous to construction of Northstar Island in early 2000, the main corridor in 2000–02 was well offshore from Northstar, but a small minority of the individuals passed within a few kilometers of Northstar (e.g., Fig. 10.12, 10.13; see also Greene et al. 2001). Nonetheless, during parts of the 2001 and especially 2002 seasons when sound emissions near Northstar were well above average, the proportion of the detected calls that originated from locations along the southern edge of the migration corridor was lower than expected if there were no Northstar effect (Fig. 10.8, 10.9). Figure 9.23 in Greene et al. (2001) mapped sightings of bowhead whales in the central part of the Alaskan Beaufort Sea during broad-scale aerial surveys conducted by MMS during the late summer and autumn of 2000 (see also Treacy 2002a). The number of sightings within that area in 2000 (14 sightings) was lower than average. This was probably attributable, in large part, to difficult weather conditions for aerial surveys during much of the 2000 bowhead migration season. Similarly, there were very few sightings in that area during MMS’s aerial surveys in 2001 (Treacy et al. 2002b), again because of difficult weather conditions for aerial surveys. Results from MMS’s 2002 aerial surveys are not yet available. Given the low number of MMS sightings in this region in 2000 and 2001, even a dedicated daily site- specific aerial survey program in the Northstar area (had it been done during those years) probably would have provided <100 bowhead sightings per year. Furthermore, the 2001 and 2002 acoustic localization results indicate that there was a Northstar effect on the call distribution only during the small percentage of time when sound levels were highest. Only a small percentage of any aerial survey results would represent whale distribution at times when a Northstar effect was occurring. Thus, the conclusion from the a priori statistical power analysis regarding the inability of aerial survey methods (either broad-scale MMS or site- specific) to detect a small-scale displacement undoubtedly applied to 2000, 2001 and 2002. 10-34 Monitoring at Northstar, Open-Water Season

Objective 6 “to estimate the number of bowhead whales whose movements and/or calling behavior were affected by the presence of industrial activities at Northstar in autumn 2002.” This objective was to be addressed primarily by using the acoustical localization data to determine the scale of any deflection of the bowhead migration corridor. This was not possible in 2000 given the data limitations that year. In 2001 and 2002, we were able to demonstrate that an effect occurred, and that this effect only became evident when sound levels near Northstar exceeded a specific value. For the minority of the times when the underwater sound levels near Northstar exceeded that level, the analysis technique provided an estimate of the extent (in kilometers) of the effect, and the dependence of the effect size on the sound level. The analysis also provided approximate confidence bounds on the estimated displacement effect at various sound levels. These results indicate that the general acoustic-localization approach provides a basis for estimating how often an effect occurs, and thus an upper limit on the fraction of the bowhead population that is potentially affected. Also, the approach provides a basis for estimating how far offshore the effect extends at times when it does occur. However, the variability in 2001 results, depending on the averaging time that is assumed, means that a decision must be made as to the most appropriate averaging time. (This concern was reduced in 2002, when averaging time had much less effect on the outcome.) Also, there are complications in converting the estimated offshore displacement of the “5th percentile whale” under specified sound conditions into an estimate of the number of whales potentially affected. To address those complications, Chapter 11 of the present report, concerning “Estimated Numbers of Seals and Whales Potentially Affected”, includes refinements relative to the estimation procedure used in 2001. As discussed earlier, the acoustical method does not reveal directly whether the demonstrated effect represents displacement, a change in calling behavior, or some combination of the two. The acoustic monitoring approach has provided an empirical basis for estimating numbers of whales that might have been affected by Northstar in 2001 and especially 2002, despite some limitations discussed in Chapter 11. We are not aware of any alternative monitoring method that could have provided this information. In particular, aerial surveys would not have provided sufficient coverage and sightings of bowheads to detect any change in whale distribution during the minority of the time when Northstar-related sounds were strong enough to have a local effect on bowhead distribution. The acoustic localization method was able to document the frequency and scale of such an effect on the distribution of bowhead calls.

ACKNOWLEDGEMENTS

Numerous people associated with Greeneridge Sciences Inc. were involved in the development and field application of the DASARs, and in analysis of the acoustical data. We thank Bob Norman, Susanna Blackwell, Bill Burgess, Bill McLennan, and the many other Greeneridge personnel listed in the Acknow- ledgements sections of Chapters 7 and 8. During DASAR deployment, calibrations, and retrieval, crews of Alaska Clean Seas provided excellent navigation and seamanship consistent with safe operations. Jonah Leavitt of Barrow, Alaska, helped with the fieldwork. At WEST, Chris Nations and Ryan Nielson assisted with statistical analysis, and Bryan F.J. Manly commented on the statistical approach. Ted Elliott, Mike Williams, and Anne Wright of LGL helped in various ways. Wilson Cullor and Allison Erickson of OASIS Environmental (on behalf of BP) helped with logistics planning for the field work. Bill Streever and Ray Jakubczak of BP, and Dave Trudgen of OASIS (on behalf of BP), guided and supported the project, and Bill Streever provided valuable comments on a draft of this chapter. BP Exploration (Alaska) Inc. supported the Chapter 10: Deflection of Bowhead Migration? 10-35 work. A peer/stakeholder group convened by NMFS in May 2000, June 2001 and June 2003 provided helpful advice. R. Suydam of the North Slope Borough Dept of Wildlife Management provided useful comments on the draft report. We thank them all.

LITERATURE CITED

Bryant, P.J., C.M. Lafferty and S.K. Lafferty. 1984. Reoccupation of Laguna Guerrero Negro, Baja California, Mexico, by gray whales. p. 375-387 In: M.L. Jones, S.L. Swartz and S. Leatherwood (eds.), The gray whale Eschrichtius robustus. Academic Press, Orlando, FL. 600 p. Clark, C.W. and J.H. Johnson. 1984. The sounds of the bowhead whale, Balaena mysticetus, during the spring migrations of 1979 and 1980. Can. J. Zool. 62(7):1436-1441. Dahlheim, M.E., H.D. Fisher and J.D. Schempp. 1984. Sound production by the gray whale and ambient noise levels in Laguna San Ignacio, Baja California Sur, Mexico. p. 511-541 In: M.L. Jones, S.L. Swartz and S. Leather- wood (eds.), The gray whale Eschrichtius robustus. Academic Press, Orlando, FL. 600 p. Galginaitis, M. 2003. Annual assessment of subsistence bowhead whaling near Cross Island: ANIMIDA Task Order 004 – Annual report for 2002. (Draft). Rep. from Applied Sociocultural Res., Anchorage, AK, for LGL Alaska Res. Assoc. Inc., and U.S. Minerals Manage. Serv., Anchorage, AK. Greene, C.R., Jr., M.W. McLennan, T.L. McDonald and W.J. Richardson. 2001. Acoustic monitoring of bowhead whale migration, autumn 2000. p. 9-1 to 9-57 In: W.J. Richardson and M.T. Williams (eds), Monitoring of industrial sounds, seals, and whale calls during construction of BP’s Northstar Oil Development, Alaskan Beaufort Sea, 2000. [Draft, April 2001.] LGL Rep. TA2431-2. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 316 p. Greene, C.R., Jr., M.W. McLennan, T.L. McDonald and W.J. Richardson. 2002. Acoustic monitoring of bowhead whale migration, autumn 2001. p. 8-1 to 8-79 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of industrial sounds, seals, and whale calls during construction of BP's Northstar oil development, Alaskan Beaufort Sea, 2001, Draft, Oct. 2002 ed. LGL Rep. TA2573-2. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. IWC. 1982. Report of the special meeting on Southern Hemisphere minke whales, Cambridge, 22-26 June 1981. Rep. Int. Whal. Comm. 32:697-745. Jones, M.L. and S.L. Swartz. 1984. Demography and phenology of gray whales and evaluation of whale-watching activities in Laguna San Ignacio, Baja California Sur, Mexico. p. 309-374 In: M.L. Jones, S.L. Swartz and S. Leatherwood (eds.), The gray whale Eschrichtius robustus. Academic Press, Orlando, FL. 600 p. Jones, M.L. and S.L. Swartz. 1986. Demography and phenology of gray whales and evaluation of human activities in Laguna San Ignacio, Baja California Sur, Mexico: 1978-1982. Rep. from Cetacean Res. Assoc., San Diego, CA, for U.S. Mar. Mamm. Comm., Washington, DC. 69 p. NTIS PB86-219078. Koenker, R. and G. Bassett, Jr. 1978. Regression quantiles. Econometrica 46:33-50. Koski, W.R., T.A. Thomas, G.W. Miller, R.E. Elliott, R.A. Davis and W.J. Richardson. 2002. Rates of movement and residence times of bowhead whales in the Beaufort Sea and Amundsen Gulf during summer and autumn. p. 11- 1 to 11-41 (Chap. 11) In: W.J. Richardson and D.H. Thomson (eds.), Bowhead whale feeding in the eastern Alaskan Beaufort Sea: update of scientific and traditional information, vol. 1. OCS Study MMS 2002-012; LGL Rep. TA2196-7. Rep. from LGL Ltd., King City, Ont., for U.S. Minerals Manage. Serv., Anchorage, AK, and Herndon, VA. 420 p. 10-36 Monitoring at Northstar, Open-Water Season

Leatherwood, S., F.T. Awbrey and J.A. Thomas. 1982. Minke whale response to a transiting survey vessel. Rep. Int. Whal. Comm. 32:795-802. Lenth, R.V. 1981. On finding the source of a signal. Technometrics 23:149-154. LGL and Greeneridge. 1996. Northstar marine mammal monitoring program, 1995: baseline surveys and retrospective analyses of marine mammal and ambient noise data from the central Alaskan Beaufort Sea. LGL Rep. TA2101-2. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK. 104 p. LGL and Greeneridge. 2000. Technical plan for marine mammal and acoustic monitoring during construction of BP’s Northstar oil development in the Alaskan Beaufort Sea, 2000-2001 (revised 30 Aug. 2000). Rep. from LGL Alaska Res. Assoc. Inc. (Anchorage, AK), LGL Ltd. (King City, Ont.) and Greeneridge Sciences Inc. (Santa Barbara, CA) for BP Exploration (Alaska) Inc., Anchorage, AK. 80 p. Manly, B.F.J. 1997. Randomization, bootstrap and Monte Carlo methods in biology, 2nd ed. Chapman and Hall, London, U.K. 281 p. MathSoft Inc. 1999. S-Plus 2000 guide to statistics, vol. 1. Data Analysis Products Division, MathSoft, Seattle, WA. Miller, G.W., R.E. Elliott, W.R. Koski, V.D. Moulton and W.J. Richardson. 1999. Whales. p. 5-1 to 5-109 In: W.J. Richardson (ed.), Marine mammal and acoustical monitoring of Western Geophysical’s open-water seismic program in the Alaskan Beaufort Sea, 1998. LGL Rep. TA2230-3. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for Western Geophysical, Houston, TX, and U.S. Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 390 p. Moore, S.E. 2000. Variability of cetacean distribution and habitat selection in the Alaskan Arctic, autumn 1982-91. Arctic 53(4):448-460. Moore, S.E. and R.R. Reeves. 1993. Distribution and movement. p. 313-386 In: J.J. Burns, J.J. Montague and C.J. Cowles (eds.), The bowhead whale. Spec. Publ. 2. Soc. Mar. Mammal., Lawrence, KS. 787 p. Moran, P.A.P. 1950. Notes on continuous stochastic phenomena. Biometrika 37:17-23. Neter, J., W. Wasserman and M. H. Kutner. 1983. Applied Linear Regression Models. Richard Irwin Inc., Home- wood, IL. page 346. Norris, K.S., B. Villa-Ramirez, G. Nichols, B. Würsig and K. Miller. 1983. Lagoon entrance and other aggregations of gray whales (Eschrichtius robustus). p. 259-293 In: R. Payne (ed.), Communication and behavior of whales. AAAS Selected Symp. 76. Westview Press, Boulder, CO. 643 p. Richardson, W.J. and C.I. Malme. 1993. Man-made noise and behavioral responses. p. 631-700 In: J.J. Burns, J.J. Montague and C.J. Cowles (eds.), The bowhead whale. Spec. Publ. 2. Soc. Mar. Mammal., Lawrence, KS. 787 p. Richardson, W.J., M.A. Fraker, B. Würsig and R.S. Wells. 1985. Behaviour of bowhead whales Balaena mysticetus summering in the Beaufort Sea: reactions to industrial activities. Biol. Conserv. 32(3):195-230. Richardson, W.J., K.J. Finley, G.W. Miller, R.A. Davis and W.R. Koski. 1995a. Feeding, social and migration behavior of bowhead whales, Balaena mysticetus, in Baffin Bay vs. the Beaufort Sea—regions with different amounts of human activity. Mar. Mamm. Sci. 11(1):1-45. Richardson, W.J., C.R. Greene Jr., C.A. Malme and D.H. Thomson. 1995b. Marine mammals and noise. Academic Press, San Diego, CA. 576 p. Richardson, W.J., G.W. Miller and C.R. Greene Jr. 1999. Displacement of migrating bowhead whales by sounds from seismic surveys in shallow waters of the Beaufort Sea. J. Acoust. Soc. Am. 106(4, Pt. 2):2281. Chapter 10: Deflection of Bowhead Migration? 10-37

Swartz, S.L. and W.C. Cummings. 1978. Gray whales, Eschrichtius robustus, in Laguna San Ignacio, Baja California, Mexico. MMC-77/04. Rep. from San Diego Nat. Hist. Mus. for U.S. Mar. Mamm. Comm., Washington, DC. 38 p. NTIS PB-276319. Thomas, T.A., W.R. Koski and W.J. Richardson. 2002. Correction factors to calculate bowhead whale numbers from aerial surveys of the Beaufort Sea. p. 15-1 to 15-28 In: W.J. Richardson and D.H. Thomson (eds.), Bowhead whale feeding in the eastern Alaskan Beaufort Sea, vol. 1. LGL Rep. TA2196-7; OCS Study MMS 2002-012. Rep. from LGL Ltd., King City, Ont., for U.S. Minerals Manage. Serv., Anchorage, AK, and Herndon, VA. 420 p. Tillman, M.F. and G.P. Donovan. 1986. Behaviour of whales in relation to management: report of the workshop. Rep. Int. Whal. Comm., Spec. Iss. 8:1-56. Treacy, S.D. 2000. Aerial surveys of endangered whales in the Beaufort Sea, fall 1998-1999. OCS Study MMS 2000- 066. U.S. Minerals Manage. Serv., Anchorage, AK. 135 p. Treacy, S.D. 2002a. Aerial surveys of endangered whales in the Beaufort Sea, fall 2000. OCS Study MMS 2002-014. U.S. Minerals Manage. Serv., Anchorage, AK. 111 p. Treacy, S.D. 2002b. Aerial surveys of endangered whales in the Beaufort Sea, fall 2001. OCS Study MMS 2002-061. U.S. Minerals Manage. Serv., Anchorage, AK. 117 p. U.S. Navy. 1995. Production sonobuoy specification for bathythermograph transmitting set AN/SSQ-36B and sonobuoy AN/SSQ-53D, 57B, 62C, 77B, 86 and 110A. Rev. 3, 12 June 1995. Progr. Exec. Office, Air Antisub- marine Warfare, Assault, and Special Missions (ASWASM), U.S. Navy, Arlington, VA. 79 p. White, G.C. and R.A. Garrott. 1990. Analysis of wildlife radio-tracking data. Academic Press, New York, NY. 383 p. Winn, H.E. and P.J. Perkins. 1976. Distribution and sounds of the minke whale, with a review of mysticete sounds. Cetology 19:1-12. Withrow, D.E. 1983. Gray whale research in Scammon's Lagoon (Laguna Ojo de Liebre). Cetus 5(1):8-13. 10-38 Monitoring at Northstar, Open-Water Season

APPENDIX 10.1: 2002 RESULTS WITH DIFFERENT ISI AVERAGING TIMES

The following graphs show the offshore distances of bowhead calls localized within the “analysis area” in 2002 vs. average industrial sound index (ISI) ~400 m north of Northstar in various intervals (“sound windows”) previous to the call. Quantile regression (solid line) relates 5th percentile of offshore distances for bowhead calls to ISI averaged over the specified interval. 90% confidence interval (dashed lines) for the quantile regression if there is no effect was estimated via a stratified permutation method (see “Methods” and th Chapter 9). Average 5 percentile assuming no effect is also plotted (– ▲ –). ISI was total sound level in five 1/3rd-octave bands dominated by industrial components. Offshore distances were measured perpendicular from a baseline through Northstar Island to the call location. Only the 2057 calls within an analysis zone extending out to 10 km from the “central axis” of the DASAR array are considered. Symbol diameter is proportional to the weight assigned to that case, based on the area of its 90% confidence ellipse. See McDonald and Richardson (Chapter 9) for corresponding graphs for 2001.

5 Minute Sound Window

6 Offshore Distance (miles)

6 10 15 20 25 Offshore distance (km)

90 100 110 120 Industrial Sound Index at 400 m (dB re 1 µ Pa) Chapter 10: Deflection of Bowhead Migration? 10-39

15 Minute Sound Window

6 Offshore Distance (miles)

6 10 15 20 25 Offshore distance (km)

90 100 110 120 Industrial Sound Index at 400 m (dB re 1 µ Pa)

30 Minute Sound Window

6 Offshore Distance (miles)

6 10 15 20 25 Offshore distance (km)

90 100 110 120 Industrial Sound Index at 400 m (dB re 1 µ Pa) 10-40 Monitoring at Northstar, Open-Water Season

60 Minute Sound Window

6 Offshore Distance (miles)

6 10 15 20 25 Offshore distance (km)

90 100 110 120 Industrial Sound Index at 400 m (dB re 1 µ Pa)

80 Minute Sound Window

6 Offshore Distance (miles)

6 10 15 20 25 Offshore distance (km)

90 100 110 120 Industrial Sound Index at 400 m (dB re 1 µ Pa) Chapter 10: Deflection of Bowhead Migration? 10-41

90 Minute Sound Window

6 Offshore Distance (miles)

6 10 15 20 25 Offshore distance (km)

90 100 110 120 Industrial Sound Index at 400 m (dB re 1 µ Pa)

120 Minute Sound Window

6 Offshore Distance (miles)

6 10 15 20 25 Offshore distance (km)

90 100 110 120 Industrial Sound Index at 400 m (dB re 1 µ Pa) 10-42 Monitoring at Northstar, Open-Water Season

APPENDIX 10.2: ESTIMATED DISPLACEMENT IN 2001 Offshore Displacement (miles)

ISI Level

121 dB

119 dB 5 117 dB 115 dB

113 dB 1

0 111 dB

109 dB

5 107 dB Offshore Displacement (km)

0 0123

050100150 Sound Integration time (minutes) th APPENDIX 10.2A. Estimated displacement of the 5 quantile of offshore distances as a function of ISI level and sound averaging (integration) time, 2001. Lower 90% confidence limits on displacement were >0 only for ISI levels 113–122 dB re 1 µPa and averaging times 70–80 min. From Chapter 9, Figure 9.13. Offshore Displacement (miles)

0 ISI Level

. 119 dB 5

5 Offshore Displacement (km)

0 012345

0 50 100 150 Sound Integration time (minutes)

th APPENDIX 10.2B. Estimated displacement of the 5 quantile of offshore distances in 2001 when ISI levels averaged over varying lengths of time reached 119 dB. Dashed lines indicate lower and upper 90% limits for displacement at each integration time. 90% confidence limits on displacement of the 5th quantile are > 0 only for integration times between ~70 and 80 min. From Chapter 9, Figure 9.14. CHAPTER 11:

ESTIMATED NUMBERS OF SEALS AND WHALES POTENTIALLY AFFECTED BY NORTHSTAR ACTIVITIES, NOV. 2001 – OCT. 2002 1 by

Valerie D. Moulton a, Michael T. Williams b, W. John Richardson c and Trent L. McDonald d , 2

a LGL Ltd., environmental research associates 388 Kenmount Rd., POB 13248 Stn. A, St. John’s, Nfld. A1B 4A5 (709) 754-1992; [email protected]

b LGL Alaska Research Associates Inc. 1101 E. 76th Ave., Suite B, Anchorage, AK 99518 (907) 562-3339; [email protected]

c LGL Ltd., environmental research associates 22 Fisher St., POB 280, King City, Ont. L7B 1A6 (905) 833-1244; [email protected]

d Western EcoSystems Technology, Inc. 2003 Central Ave., Cheyenne, WY 82001 (307) 634-1756; [email protected]

for

BP Exploration (Alaska) Inc. Dept of Health, Safety & Environment 900 East Benson Blvd. P.O. 196612 Anchorage, AK 99519-6612

LGL Report TA2707-6 December 2003

1 Chapter 11 In: W.J. Richardson and M.T. Williams (eds.). 2003. Monitoring of industrial sounds, seals, and bowhead whales near BP’s Northstar Oil Development, Alaskan Beaufort Sea, 1999–2002. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD.

2 Primary authors: VDM and MTW for seal sections; WJR and TLM for whale sections. 11-2 Monitoring at Northstar, 1999–2002

ABSTRACT

A Letter of Authorization (LoA) issued by NMFS to BP authorized the “taking” of small numbers of seals and whales incidental to Northstar production activities. The numbers potentially affected during the ice- covered, break-up, and open-water periods in November 2001 through October 2002 have been estimated based on monitoring results. An estimated 110 ringed seals, 1 bearded seal, and probably no spotted seals were present near BP’s Northstar activities during the combined ice-covered, break-up, and open-water seasons of 2001–02. These figures are less than the corresponding “takes” authorized by the LoA issued by NMFS to BP, and most of these seals do not seem to have been negatively affected by Northstar activities. These totals exclude ~143 seals that dived in response to aircraft overflights during the spring aerial surveys; those incidents occurred over a much larger area and were separately authorized via a Scientific Research Permit issued by NMFS to LGL. The overall results suggest that any effects of Northstar production activities on seals were minor, short-term, and localized, with no consequences for the seal populations. There was no indication of any effects of the Northstar development on availability of seals to subsistence hunters during the 2002 ice-covered, break-up, or open-water periods. Acoustic localization data indicated that, during late summer and early autumn of 2001 and 2002, a small number of bowhead whales in the southern part of the migration corridor (closest to Northstar) were affected by vessel or Northstar operations. This occurred during 4–5% of the time when levels of underwater sound from these activities were highest. At these times, most “Northstar sound” was from maneuvering vessels, not the island itself. The best (though quite imprecise) estimates of the number of bowheads that were apparently deflected offshore by ≥2 km (1.2 mi) were 13 bowheads in 2001 and 19 bowheads in 2002, i.e., <0.2% of the population. Actual numbers would vary relative to these imprecise estimates depending on the accuracy of assumptions and sampling variability. It is possible that the apparent deflection effect was, at least in part, attributable to a change in calling behavior rather than actual deflection, but in either case there was a change in the behavior of a small number of whales. The occurrence of a small “Northstar effect” on a low number of bowheads in each of 2001 and 2002 would be well within the provisions of the LoA. Based on the acoustic monitoring results, it is unlikely that the apparent displacement extended far enough east and northeast to affect whales being hunted near Cross Island. There was no specific information on numbers of gray or beluga whales (if any) that may have been close enough to Northstar to be disturbed by Northstar production and drilling operations in 2002. Given that gray whale sightings have been infrequent this far to the east in previous years, it is most likely that no gray whales were affected by Northstar in 2002. For belugas, the estimated numbers that might approach within an assumed 1–2 km (0.6–1.2 mi) disturbance radius are 10–20 in an average year.

INTRODUCTION

BP began producing oil from the Northstar Development at the end of the previous reporting period (31 Oct. 2001), and continued production activities throughout the November 2001 through October 2002 period considered here. Chapter 2 describes BP’s activities during the ice-covered and open-water periods from Nov. 2001 through Oct. 2002. The Letter of Authorization (LoA) issued by NMFS for production activities at Northstar required monitoring of those activities to estimate the nature, amount, and geographic extent of any effects on marine mammals or subsistence hunting (NMFS 2001). One of the primary purposes of monitoring was to assess the numbers of seals and whales potentially affected by BP’s Northstar activities, and to compare those estimates with the numbers authorized in the LoA. Chapter 11: Seals & Whales Potentially Affected 11-3

Estimates of the number of seals and whales that might be affected by Northstar production and maintenance activities were discussed by NMFS (2000a) in the preamble to the regulations issued in May 2000. The LoA issued by NMFS on 14 December 2001 under those regulations set limits on the types and levels of “taking” that were authorized (NMFS 2001). Specifically, the LoA stated that (i) During the ice-covered and ice-breakup periods, up to 164 ringed seals and 5 bearded seals may be incidentally harassed annually; (ii) During the ice-covered period, up to 5 ringed seals may be incidentally killed annually; (iii) During the open-water period, up to 27 ringed seals, 5 spotted seals, 5 bearded seals, 5 gray whales, and 91 beluga whales may be incidentally harassed annually; and (iv) During the open-water period, 765 bowhead whales (maximum of 1533 bowheads in 2 out of 5 seasons, or 3585 in 5 years) may be incidentally harassed. The monitoring work planned for the period November 2001 through October 2002 was done, and the results have been described earlier in this report. Chapters 3–5 describe the two types of monitoring conducted during the 2001–02 ice-covered season, and Chapters 6–10 describe the monitoring during the 2002 open-water season. Monitoring during the ice-covered season of 2001–02 included • on-ice measurements of sounds (underwater and in-air) and vibrations in the ice during February– March 2002 (Chapter 3), and • fixed-wing aerial surveys of seals during May–June 2002 (Chapters 4 and 5). There were no on-ice dog-assisted searches for seal structures during the ice-covered season of 2001–02. BP’s activities at Northstar from 1 March 2002 until ice break-up did not extend into previously undisturbed areas. Therefore, no searches for seal structures were required (as per NMFS 2001).

Monitoring during the open-water season of 2002 included • underwater acoustic measurements during September (Chapter 6), and • acoustic monitoring of calling bowhead whales during the autumn migration (Chapters 7–10). The next section of this chapter estimates the number of seals potentially affected by activities associated with BP’s Northstar development during the ice-covered period from November 2001 through mid-June 2002. A subsequent section estimates numbers of seals and whales potentially affected during the break-up and open-water periods in 2002 (mid-June through October). In addition, this chapter discusses potential impacts of Northstar activities on subsistence hunting for seals and whales near the development area. When appropriate, we use more than one procedure to estimate the numbers of marine mammals potentially affected. This approach identifies a range of potential numbers, as is appropriate given that the criteria for disturbance are not well defined. To date, the MMPA and its implementing regulations have not had a clear operational definition of “take by harassment”. As a result, there has been much debate concerning how substantial and prolonged a change in behavior must be before it constitutes a “take by harassment”.

2001–2002 ICE-COVERED SEASON

During the ice-covered season, the only species of seal or whale that occurs regularly in the area of landfast ice surrounding Northstar is the ringed seal. An occasional bearded seal can occur in the landfast 11-4 Monitoring at Northstar, 1999–2002 ice in some years. Bowhead and beluga whales are absent from the Beaufort Sea in winter, and in spring their eastward migrations are through offshore areas north of the landfast ice, which excludes whales from areas close to Northstar. Possible Displacement from Northstar Development Area

It was not expected that there would be any injuries or mortality to seals as a result of the activities at and near Northstar on the sea ice. However, operational activities at Northstar might have displaced some seals from the area, or at least caused some seals to cease using one or more of their breathing holes or lairs. For purposes of estimating the numbers of seals potentially affected by Northstar activities, we defined two zones of progressively increasing size around Northstar: (1) The “Northstar development zone”, as it existed in the winter–spring of 2001–02, was defined as the artificially thickened ice road and Northstar Island (see linear area shaded dark gray in Fig. 11.1). This narrow strip totaled ~1.2 km2. We did not include, as part of the “development zone”, a small area near Northstar where emergency equipment was occasionally tested (Fig. 2.16), or the area along the pipeline route where monitoring for oil spills was conducted intermittently (Chapter 2). These “excluded” activities were limited in scope, geographic extent, and frequency, i.e., five or fewer operations during the reporting period. It was considered unlikely that they would have any biologically important effects on seals, given the very limited evidence of such effects even after the much more intensive industrial activities on the sea ice in 2000–02 (Moulton et al. 2001, 2002, Chapter 5; Williams et al. 2001a, 2002b). (2) The “potential impact zone” was larger than the “Northstar development zone”, totaling 29.5 km2. For the 2001–02 season, we assumed that seals within 0.64 km of the primary ice road and 1.85 km of the artificial island could potentially be affected (BPXA 1999). The 1.85 km zone around the island was also assumed to apply during the latter phases of construction in 2001 (Williams et al. 2002a), whereas a 3.7 km zone was assumed during initial construction in 2000 (Williams et al. 2001b). BP’s activities during the latter stages of construction and production were less intense than the initial construction activities. It is estimated that ringed seals, when underwater, can sometimes hear production noise up to 1.5 km from Northstar (see Chapter 3). The maximum distance at which airborne sound from Northstar is audible to seals may, at times, be at least 5 km (Chapter 3). However, detection at those distances would be intermittent, and the received sound would be faint. (Sounds from Northstar activities were at times detectable by instruments at greater distances, but seals are less sensitive than instruments at low frequencies—see Chapter 3.) NMFS (2001, p. 65930) states that “NMFS does not consider an animal to be ‘taken’ if it simply hears a noise, but does not make a biologically significant response to avoid that noise.” Given the lack of evidence for reduced seal density near Northstar (Chapter 5), the 1.85 km buffer is likely conservative, i.e., an overestimate of the area within which seals are affected by Northstar sounds. The “potential impact zone” included the areas within 0.64 km of the primary ice road and 1.85 km of the artificial island, which are shaded medium-gray in Figure 11.1, plus the “Northstar development zone”. Seals near the island might also be affected during helicopter landings and takeoffs at the island. The low-altitude portion of the helicopter route, as depicted in Figure 2.4 (Chapter 2), was largely included in the potential impact zone and was therefore not treated separately. Helicopter altitude was normally 457 m (1500 ft) while en route. The “potential impact zone” also included the areas within a few hundred meters west and southwest of the island where occasional equipment testing and spill drills were conducted (see Fig. 2.16 in Chapter 2), and the northernmost 1.5 km of the pipeline route where oil- spill inspections were conducted on five occasions. Chapter 11: Seals & Whales Potentially Affected 11-5

FIGURE 11.1. Distribution of ringed seal sightings near Northstar during two aerial survey replicates in 2002 (30 May–2 June and 3–7 June). The “Northstar development zone” consists of the ice road and artificial island (dark gray linear area). The “potential impact zone” (see text) is indicated with medium gray shading. The 4 and 10-km bins are measured from the edge of the former area. The 4-10 km “reference” zone was used to calculate expected number of seals within the potential impact zone in the absence of industrial activity at Northstar. 1 km = 0.62 mile = 0.54 n.mi. 11-6 Monitoring at Northstar, 1999–2002

It was assumed, based on previous studies, that seals would tolerate some level of ongoing noise and disturbance from distant industrial activities on a permanent production island. Also, results from aerial monitoring in 2000–02, along with the on-ice studies in 2000–01, indicate that the distance categories applied in defining the “potential impact zone” are overestimates of actual impact distances. As a result, our estimates of numbers of seals within the “potential impact zone” no doubt overestimate the numbers of seals significantly affected by Northstar. In order to evaluate how seals may have been affected we estimate the number of seals that might have occurred within the potential impact zone if the Northstar Development had not been constructed. This estimate is compared to observed densities of ringed seals in that area. The difference is an estimate of the number of ringed seals that may have been displaced. The expected number in the absence of industrial activities was based on surveys of otherwise-similar non-industrial areas, or surveys of the Northstar area in non-industrial years. In these comparisons, we used average densities from BP/LGL aerial surveys in various zones in 1997–2002 (Miller et al. 1998; Link et al. 1999; Moulton et al. 2000, 2001, 2002, Chapter 4 of this study). When calculating the number of ringed seals expected within the potential impact zone in the absence of industrial activity, we considered the area at a distance of 4–10 km from the Northstar Develop- ment Zone as our “reference” area. We limited the reference area to these nearby locations in case there were undocumented characteristics (independent of industry) of the Northstar area that made it either more or less “favorable” to ringed seals. We also controlled for water depth, because ringed seal density in our overall study area is highest in waters 5–20 m deep. We would expect higher densities in some areas 4–10 km from Northstar than within the potential impact zone even in the absence of Northstar activity; the 4–10 km zone includes some areas deeper than any areas in the potential impact zone (Fig. 11.1). To calculate the number of ringed seals expected to occur within the potential impact zone in the absence of avoidance, we applied densities from four water depth categories (0–3 m, 3–5 m, 5–10 m, 10– 15 m) within the 4–10 km “reference” zone. These four densities (in seals/km2) were multiplied by the corresponding four stratum areas within the potential impact zone (totaling 29.5 km2) to give the expected numbers of ringed seals within that zone, i.e., within 0.64 km of the ice road to the island and/or within 1.85 km from the artificial island. This was done separately based on 1997, 1998, 1999, 2000, 2001, 2002, and average 1997–2002 data from the BP/LGL surveys. Based on the 1997–2002 surveys of the 4– 10 km “reference” zone, the expected “uncorrected” numbers of seals in the potential impact zone in the absence of any avoidance ranged from 11 to 22, depending on the year. The estimate based on 2002 survey data was 22 seals (Table 11.1). The estimates quoted above are the estimates prior to correction for the proportions of the seals that would be missed by aerial surveyors because of “detection bias” or “availability bias”. (“Detection bias” refers to the fact that aerial surveyors do not see every seal that is on the ice and potentially sightable. “Availability bias” refers to the fact that seals are not always hauled out above the ice and snow, and thus “available” to be seen by aerial surveyors.) Instead, the estimates quoted above represent the average numbers of seals that one would expect to see during a single aerial survey of the entire potential impact zone, i.e., assuming 100% on-transect coverage. We adjusted the above estimates to allow for seals hauled out but not sighted by observers (detection bias; ×1.22, based on Frost et al. 1988) and for the proportion of ringed seals not hauled out during the survey coverage (availability bias; ×2.33, based on Kelly and Quakenbush 1990). These adjustments increased the estimated numbers of seals present within the potential impact zone by a Chapter 11: Seals & Whales Potentially Affected 11-7

TABLE 11.1. Numbers of ringed seals expected to occur in spring 1997–2002 within the “Potential Impact Zone” (as defined for 2002) in the absence of any Northstar impact, based on observed seal densities in a reference area 4–10 km away from Northstar. The potential impact zone included areas within 0.64 km of the ice road and within 1.85 km of Northstar/Seal Island.

Expected Number of Seals Within Observed Density Potential Impact Zone Within Potential Expected Densitya Impact Zone BP/LGL Survey (seals/km2) Uncorrected Corrected Totalb (seals/km2) 1997 0.47 14 39 0.45 1998 0.38 11 32 0.48 1999 0.39 12 33 0.23 2000 0.52 16 44 0.45 2001 0.53 16 45 0.67 2002 0.75 22 63 0.51 Average 1997-2002 0.51 15 43 0.47

a These average uncorrected densities are based on data from the zone 4-10 km away from the 2002 development zone, controlling for water depth by weighting density based on the proportions of the potential impact zone within the various depth strata. b "Uncorrected" multiplied by the 1.22 correction factor for seals hauled out but not seen by observers (Frost et al. 1988), and by the 2.33 correction factor for seals not hauled out (Kelly and Quakenbush 1990). combined factor of ×2.84, within the limits of rounding error (Table 11.1). For example, based on the 2002 survey data, the actual numbers of seals expected to occur within the potential impact zone in the absence of any avoidance effect increased from the uncorrected estimates of 22 seals to a corrected estimate of 63 seals. The adjustments for “availability” bias used above and in our previous related reports are based on a study of five ringed seals whose haul-out behavior was monitored until 4 June during two springs (1982 and 1983) near Reindeer Island, within our study area (Kelly and Quakenbush 1990). More recently, Kelly et al. (2002), in a preliminary report, have estimated daily availability correction factors ranging from 1.63 to >11 in 1999 and 2000. Those authors have continued their study in 2001 and 2002. We have not used correction factors from Kelly et al. (2002) given the preliminary status of the results. Many factors confound the use of correction factors based on a small number of tagged seals in a limited area (Frost et al. 2002); perhaps most noteworthy is the large variability in individual seal basking behavior between and within years. Depending on the actual correction factor applicable to the surveys conducted in a given year, the total number of seals in the area could be over- or underestimated when the generic ×2.33 factor of Kelly and Quakenbush (1990) is used. However, the relative densities and numbers at different distances from Northstar in a given year should be unaffected. No ringed seals were observed in the small (1.2 km2) linear area defined as the “Northstar development zone” during late spring in 2002 (Fig. 11.1). Given the expected seal densities (Table 11.1) and small area of the development zone, we would not have expected more than 1–2 sightings there even in the absence of an industry effect. The observed density of ringed seals within the potential impact zone in 2002 was 0.51 seals/km2, based on 26 seals observed there during the surveys. That density was lower than expected based on surveys 4–10 km away from Northstar in 2002 (0.75 seals/km2; Table 11.1). The estimated density in the potential impact zone, when adjusted by the two correction factors, is 1.45 seals/km2, as compared with an estimated of 2.13 seals/km2 based on data from the “reference area” 4–10 11-8 Monitoring at Northstar, 1999–2002 km away. The 1.45 seals/km2 figure corresponds to an estimated 43 seals actually present within the 29.5 km2 area of the potential impact zone―20 seals less than the 63 expected based on survey results 4–10 km away. The 20-seal difference derived above did not necessarily mean that 20 seals had moved away from the potential impact area in response to Northstar activities: After allowance for the effect of environmental factors on observed seal density, there was no evidence of a “Northstar effect” on seal density in 2002 or any other year (Chapter 5). The seemingly lower densities near Northstar in 2002 may have been related to the more southerly location of the ice edge in 2002 than in other survey years (Chapter 4). In 2002, the edge of the landfast ice was only 15.8 km north of Northstar Island, whereas in 2000 and 2001 it was 27.4 and 25.7 km away, respectively. In most years, seal densities are higher near the ice edge (Frost et al. 2002), and this was the case in 4 of the 6 years from 1997 to 2002 (Moulton et al. 2002; Chapters 4 and 5). In 2002, the overall trend was for diminishing seal densities with increasing distance from the ice edge (Fig. 5.2), although raw densities 0–5 km inshore from the ice edge were in fact lower than those 5– 10 km inshore of the ice edge (Fig. 4.9). In any event, our comparison of seal densities in the potential impact zone in 2002 with those 4–10 km away is likely confounded by the fact that, in that year, one would expect higher densities 4–10 km away (and nearer the ice edge) regardless of distance from Northstar. If there was some displacement of ringed seals away from Northstar in the winter of 2002, we would have expected fewer seals within the potential impact zone during 2002 than in the pre- construction years―1997, 1998, and 1999. That was not observed, although inter-year comparisons should be treated cautiously given the possibility of year-to-year differences in sightability of seals during aerial surveys. Also, there was no evidence of reduced densities near Northstar in 2000, a winter of intensive construction activity, or in 2001 (Chapter 5). The presence of numerous seals within the potential impact zone during late spring of 2000, 2001 and 2002 indicates that any displacement effect was localized and, if it occurred at all, involved only a small fraction of the seals that would otherwise have been present. A complication in our analysis is that we do not know whether the individual seals seen near Northstar during aerial surveys in late May and early June 2002 had been resident at the same locations during the winter/early spring period. Results from monitoring of seal structures periodically throughout the winter and spring of 2000–2001 showed that seals in the Northstar area are quite mobile (Williams et al. 2002b), and they are likely to be especially mobile during spring at the onset of breakup. Acoustic Monitoring

Measurements of sounds and vibrations produced by Northstar-related industrial activities during the winter of 2002 were recorded at a variety of distances from the island (see Chapter 3). These measurements were made on two successive days along a “dogleg” transect that extended to 5 km north and then to a maximum of 9.4 km northwest from Northstar on the landfast ice (Fig. 3.1). On 28 February 2002, recordings were acquired over a 4-hour period at five sites ranging from 470 m to 7620 m from Northstar Island. On 1 March 2002, recordings were acquired over a 5.5-hour period at seven sites ranging from 220 m to 9400 m from Northstar Island. Two rolligon-type vehicles were used to travel along the transect. Chapter 3 contains the details of the equipment and procedures used during acoustic monitoring. It is possible that some seal structures were encountered during travel along the acoustic transect, and that some seals may have been exposed to sounds and vibrations from activities associated with the acoustic measurements. However, no seals or seal structures were observed during acoustic monitoring Chapter 11: Seals & Whales Potentially Affected 11-9 activities. We consider acoustic monitoring to be similar in scope to the intermittent activities associated with oil spill detection and equipment testing, as described above. Although there are no specific data, we assume that any seals along and near the acoustic transect during monitoring did not make “biologically significant responses”, as discussed by NMFS (2000b, p. 60409) and NMFS (2001, p. 65930). Fixed-wing Aerial Surveys

In order to evaluate potential changes in ringed seal density and distribution relative to Northstar activities, we conducted aerial surveys over the potential impact zone near Northstar as well as other reference areas to the east and west of Northstar (see Chapters 4 and 5 for details). Overflights by the survey aircraft at an altitude of 91 m (300 ft) ASL were expected to elicit behavioral responses by a minority of the seals on the ice (Richardson et al. 1995; Born et al. 1999). This was observed consistently in 2002 and previous years; the majority of seals showed no apparent reaction to the aircraft (see Table 4.2 in Chapter 4). Behavioral responses from the seals that apparently reacted to the aircraft ranged from very slight (looking up at the aircraft) to diving into the water. None of these behavioral reactions are expected to have serious consequences for the individual seals involved. A Scientific Research Permit (SRP 481-1623-01) issued by NMFS to LGL authorized the disturbance of any number of ringed, bearded or spotted seals that might be affected by these aerial surveys. Any such disturbance does not fall under the provisions of the LoA issued by NMFS to BP. This section applies several methods to estimate the numbers of seals that may have reacted to aircraft overflights during the 2002 aerial surveys, and is presented as required by SRP 481-1623-01. Overall, 55 ringed seals were observed diving down holes or cracks in the ice during the 2002 aerial surveys. This represents 1.9% of the 2963 ringed seals sighted in both fast and pack ice whose behavior was recorded (see Table 4.2). If we apply this percentage to the ringed seals with “unknown” behavior (724 seals), then there were an additional 14 suspected dives, for a total of 69. These 69 ringed seals represent a minimum estimate of the number of dives that may have occurred in response to the aircraft. Some of the unseen seals in the relatively narrow (271 m wide) unsurveyed area directly under the aircraft presumably dived in response to the aircraft. The unsurveyed area underneath the aircraft totaled 1563 km2 (0.271 km × 5766 km). If we apply the overall observed density for ringed seals on the landfast ice during the 2002 surveys (0.72 seals/km2), we estimate that 1125 seals may have been located underneath the aircraft. If we arbitrarily assume that seals under the aircraft were twice as likely (3.8%) to dive as compared with those to the side (1.9%), the estimated number of seals below the aircraft that dived is 42 seals. Combining the previous estimate of 69 seals with the suspected 42 seals that dived below the aircraft, we estimate that at least 111 seals dove in response to the survey aircraft. Additional seals were encountered but not recorded as the aircraft flew over the ice while in transit between successive transects. Approximately 90% of the over-ice distance flown consisted of flights along transect lines. The remaining 10% consisted of brief between-transect flight segments, averaging ~4 km long as compared with an average transect length of 44 km. Half of these between-transect segments were over land or grounded ice near the south edge of the survey area, and were unsuitable as ringed seal habitat. Thus, the 111 instances when seals along transects dived should be increased by a factor of about 5% to allow for unrecorded seals overflown during the between-transect segments at the northern end of transects. This results in a total of about 117 dives observed or presumed to have occurred in response to the aircraft. Furthermore, observers probably missed some seals that were hauled out (detection bias). Frost et al. (1988) estimated that a factor of ×1.22 should be applied to correct for detection bias. This results in a 11-10 Monitoring at Northstar, 1999–2002 total estimate of about 143 dives. We do not consider the additional seals that simply “looked” or “moved” as the aircraft passed (see Table 4.2) to have been disturbed. The best estimate of ringed seal dives in response to the aircraft is 143. Factors contributing to the uncertainty in this estimate include the following: • A few seals may have dived at a distance far enough ahead of the aircraft that the observers did not sight them. • A few seals may have dived in response to the aircraft in the few seconds after the aircraft passed, in some cases behind the field of view of the observer • It is possible that a few unseen seals beyond the transect-width (more than 546 m to the side of the aircraft flight path) dived in response to the aircraft. • Each transect was surveyed twice, and some of the between-transect segments may have been covered more than two times. Thus, some of the same seals may have reacted to the aircraft on two different days or possibly, in a very few cases, more than two days. An unknown but probably small proportion of the observed dives might have occurred for some reason other than the aircraft overflight. However, for purposes of this analysis, we assume that all observed dives were in response to the aircraft. In total, nine bearded seals were seen in 2002; none of these were observed diving. Four of the bearded seals were recorded as having “looked” in response to the aircraft, two bearded seals did not respond overtly, and the behavior of the remaining three bearded seals was recorded as unknown. Combined Impact of Northstar Activities during Ice-Covered Season

BP’s updated Petition for Regulations (BPXA 1999) estimated that, on average, about 91 ringed seals might be expected to occur within the potential impact zone around Northstar construction activities during the ice-covered season in the absence of any avoidance. An upper limit of 125 seals was predicted during construction. That Petition estimated that, after production began, 77 ringed seals would be in the (smaller) impact zone during the ice-covered season, with an upper limit of 105 seals. • BP’s request for an LoA for the initial period of production (BPXA 2001) revised the latter figures to 53 and 139, respectively. In the resultant “production LoA”, NMFS authorized the “taking by harassment” of a maximum of 164 ringed seals during the ice-covered and break-up seasons combined: the aforementioned 139 seals plus an additional 25 during the break-up period (NMFS 2001). • The production LoA also states that “During the ice-covered period, up to 5 ringed seals may be incidentally killed annually”. • It was expected that one to five bearded seals might also occur within 2–4 km of the Northstar facilities during any one winter/spring season. The production LoA authorizes the incidental harassment of up to 5 bearded seals during the ice-covered and break-up periods. In actuality, there was no evidence, and no reason to suspect, that any seals were killed or seriously injured by Northstar-related activities during the winter of 2001–02. During this period there was no impact pile driving or similarly noisy activity that could have exposed any marine mammals to received levels of 180 or 190 dB re 1 µPa, above which there is concern about the possibility of physical effects (NMFS 2000a, 2001). Oil spills would be a concern with regard to the potential for injuries or deaths. Chapter 11: Seals & Whales Potentially Affected 11-11

There were no oil spills in the usual sense onto the ice or into the water during the 2001–02 ice covered season. However, a total of ~45 L of hydrocarbon condensates released via the gas flare was deposited onto the sea ice within a 334 m2 (i.e., 0.0003 km2) area adjacent to and downwind of the flare (see Chap- ter 2). This area was so small (and so close to Northstar) that it is highly unlikely that any seal would have been affected. The flare has subsequently been modified to avoid the “flare sneezes”, and BP believes that the problem has been rectified (see p. 2-25 in Chapter 2). Based on observed ringed seal densities in 2002 in a reference area 4–10 km from the Northstar development zone (corrected for observer and availability bias), we calculated that 63 ringed seals would have been expected in the “potential impact zone” in the absence of Northstar construction. This number corresponds to a raw survey density of 0.75 seals/km2. In fact, the observed raw density in the potential impact zone was 0.51 seals/km2 (Table 11.1), corresponding to a corrected estimate of 43 ringed seals within the potential impact zone. The difference between these two figures (20 seals) could be taken as an estimate of the number of seals displaced, or whose behavior was changed sufficiently to make them undetectable to aerial surveyors. However, that figure would have wide confidence limits. Also, as previously discussed, it is probable that the 20-seal difference was at least partly attributable to factors unrelated to Northstar, given the lack of evidence for a “Northstar effect” on seal density after allowance for other variables (Chapter 5). The highest estimate of the number of seals potentially affected by Northstar during the 2002 ice-covered season is 63; this is the number of seals expected within the potential impact zone in the absence of Northstar. However, of this estimated 63 seals, about 43 tolerated the proximity of Northstar and remained in the potential impact zone. Another estimate of the number potentially affected is the 20 seals that might have moved out of that zone. The estimated number of ringed seals within the potential impact zone in spring 2002 (43) and the expected number if there had been no Northstar effect on numbers present (63) were both similar to the latest “best estimate” by BPXA (2001) of the numbers that might be present and potentially affected (53). Both figures were considerably less than the estimated maximum number that might be present and potentially affected (139—see BPXA 2001), which was the number of ringed seals authorized by NMFS to be “taken” during the ice-covered period. The fact that ~43 seals remained in the potential impact area within 0.64–1.85 km of the Northstar infrastructure, despite the production activities, drilling, traffic on the ice road, and other operational activities, suggests that these seals were not disturbed seriously. These ringed seals were widely distributed in the “potential impact zone” (Fig. 11.1). No bearded seals were sighted within 10 km of Northstar during the 2002 aerial surveys (see Chapter 4). There is no indication that any bearded seals were affected by Northstar activities during the ice-covered season of 2002. Approximately 143 ringed seals, but (probably) no bearded seals, dived in response to aircraft overflights during the spring seal survey in 2002. The great majority of these seals were distant from Northstar and thus different individuals than those potentially affected by industrial or monitoring activities near Northstar. This behavioral reaction is not expected to have any long-term deleterious impact on seals. The 2002 aerial surveys were conducted under Scientific Research Permit 481-1623-01 issued by the NMFS to LGL Ltd. Disturbance to seals during the aerial surveys was authorized under that Permit.

2002 BREAK-UP AND OPEN-WATER SEASONS

Potential sources of disturbance to marine mammals from the Northstar project during the open- water season of 2002 consisted primarily of vessel and helicopter traffic, plus the ongoing production and 11-12 Monitoring at Northstar, 1999–2002 drilling operations on the island (see Chapter 2). Helicopter flights to and from Northstar were especially frequent during break-up and freeze-up, and less so during the open-water period. Vessel traffic included crew vessel operations between West Dock and Northstar throughout the open-water period; and tug and barge trips during July to early October. As described in Chapter 6, vessel traffic near Northstar was responsible for most of the peaks in underwater sound associated with Northstar, whereas production and drilling operations produced relative steady sound at lower levels. There was no major sealift of equipment to Northstar in 2002. Estimated Number of Seals Present

Break-Up Period The available evidence indicates that no more than a few seals near Northstar were affected by Northstar activities during winter or spring, and there is no reason to believe that effects were greater during the subsequent break-up period. If any seals were affected during break-up in some subtle way, it is probable that some of these would be some of the same individuals already counted as present and potentially affected during the latter stages of the ice covered season. Even so, our estimates of numbers of seals present and potentially affected during break-up assume complete weekly turnover. BP’s Petition for Regulations (BPXA 1999) assumed that seals within 1 km (0.6 mi) of the island might be affected by construction and other activities on the island. The area of water within 1 km of the island is 3.11 km2. During aerial surveys in the 30 May–7 June 2002 period, the observed average density of seals hauled out on the landfast ice depended on the specific area considered, but was 0.51 seals/km2 within the “Potential Impact Zone” (Table 11.1). When correction factors are applied to allow for seals present on the ice but not seen (×1.22) and below the surface of the ice (×2.33), the estimated actual density of seals was 1.45 seals/km2. Thus, an estimated 5 seals would be present within 1 km of the island (1.45 seals/km2 × 3.11 km2). The Petition for Regulations assumed that, during the break-up and open-water seasons, there could be complete turnover in the seals present on a weekly basis. If so, and assuming that break-up lasted six weeks (early June to mid-July), the total number of seals potentially present might be about 30 (5 seals × 6 weeks). Open-Water Period For the 2002 season, we estimated the number of seals exposed to and potentially affected by Northstar activities assuming that • seals within 1 km of the island might be affected; • the density of seals within that area would be no more than 2× the density observed during boat- based surveys for seals within the general Prudhoe Bay area in 1996–2000 (0.19 seals/km2 × 2 = 0.38 seals/km2); and • seals within the affected area are replaced once for each of fifteen 7-day intervals during the open-water period. The first of these points assumes that seals in open water are not significantly affected by passing vessels (or helicopters) that they could occasionally encounter in areas >1 km from Northstar. Passing boats and helicopters might cause startle reactions and other short-term effects. However, NMFS has indicated that short-term behavioral effects having no negative consequences for biologically important activities are not relevant in estimating the number of ringed seals potentially affected (NMFS 2000b, p. 60409). Chapter 11: Seals & Whales Potentially Affected 11-13

Based on the above assumptions, an estimated 18 seals might be present and potentially affected, i.e., 3.11 km2 × 0.38 seals/km2 × 15 weeks. Ringed seals constituted 94% of the seals identified in the area during the open-water seasons of 1996–2000, with 4.3% being bearded seals and 1.5% spotted seals. Thus, of the estimated 18 seals, about 17 would be ringed seals, one might be a bearded seal, and probably none would be spotted seals. These estimates are subject to wide uncertainty (in either direction) given the uncertainties in each of the three assumptions listed above. There is no specific evidence that any of the seals occurring near Northstar during the 2002 open- water season were disturbed or otherwise affected by BP’s activities, but the possibility of such effects was not specifically studied in 2002. Following NMFS (2000b, 2001), we assume that simple exposure to sound, or brief reactions that do not disrupt behavioral patterns in a biologically significant manner, do not constitute harassment or “taking”. Combined Break-Up and Open-Water Periods Overall, on the order of 48 seals (30 + 18) might have been present within 1 km of Northstar Island at some time during the break-up and open-water periods during 2002, given the assumptions listed above. This total assumes complete turnover of the seals each week for 21 weeks. The actual number of different individuals present would probably be lower, and there is no indication that these seals were deleteriously affected. The LoA issued by NMFS indicated that incidental harassment was authorized for 25 ringed seals during the break-up period and 27 during the open-water period (total of 52), plus 5 bearded and 5 spotted seals during the open-water period (NMFS 2001). The actual estimates derived above for 2002 are 30 ringed seals present during break-up and 17 during the open-water period (total of 47), plus one bearded seal and probably no spotted seals. The total estimate for ringed seals during the break-up plus open-water periods (47) was less than the authorized total (52). Also, there is no reason to believe that most of the 47 seals estimated to be present were in fact affected by the presence of Northstar. Estimated Number of Whales Potentially Affected

Bowhead Whales The acoustic monitoring of the bowhead whale migration, as described in Chapters 7–10, was designed to determine (with the limitations discussed in Chapter 10) whether bowhead whales were displaced offshore by Northstar construction and operational activities and, if so, to what geographical extent. If the southernmost calling bowheads detected by the acoustic monitoring system tended to be farther offshore when Northstar operations were noisy than when they were quieter, this was to be taken as evidence of an effect. The geographic scale of any documented effect, as a function of Northstar sound level, would provide a basis for estimating the number of whales affected. In fact, the monitoring results provided evidence of a displacement effect in the southern part of the migration corridor during a minority of the time when the industrial noise level in the water near Northstar was highest. This evidence was stronger in 2002 than in 2001 (Chapter 10). Previous Estimates for 2000 and 2001.—In 2000, calling bowheads were monitored acoustically during 2–28 Sept., but technical problems reduced the accuracy of the call localizations and prevented measuring underwater sounds near Northstar until late in the season (Greene et al. 2001). In the absence of these data, the number of bowheads that might have been affected by Northstar activities in 2000 was estimated by an alternative method. This was based on (1) the distance from Northstar within which received levels of underwater sounds might exceed 115 dB re 1 µPa, estimated to be 4 km; and (2) historical data concerning the proportion of the bowhead population expected to approach within 4 km of 11-14 Monitoring at Northstar, 1999–2002

Northstar. With that approach, we estimated that as many as 202 bowheads might be affected by Northstar in an average year. The estimate was ~727 bowheads if the migration corridor was as close to shore as in 1997, the year when bowheads were closer to shore than in any other year in the 1979–2000 period (Williams et al. 2001b). The 202 estimate, and especially the 727 estimate, were probably overestimates of the actual number of bowheads that, in 2000, would have received a sound level of ≥115 dB re 1 µPa if they had not deflected away from Northstar. In 2000, the received sound levels dropped below 115 dB well within 4 km of Northstar much of the time (Blackwell and Greene 2001). Also, the limited available aerial survey data from 2000 suggested that the bowhead migration corridor in the general Prudhoe Bay/Northstar region was not unusually close to shore in 2000 (Treacy 2002a).3 In 2001, the acoustic monitoring approach was attempted again, this time with improved instrumentation and procedures. The required types of data were obtained continuously from 28 Aug. to 2 Oct. 2001 (see Greene et al. 2002 and Chapter 9). For 2001, Williams et al. (2002a) used a preliminary implementation of the originally planned approach to estimate the number of bowheads that were apparently affected by Northstar. That analysis indicated that whales in the southern part of the migration corridor (closest to Northstar) may have been affected by vessel or Northstar operations during the 10.8% of the time when levels of underwater sound from these activities were highest. At these times, the main components of “Northstar sound” were from maneuvering vessels, not from the island itself. An estimated 1140 bowheads passed Northstar at such times, and an estimated maximum of 200 of these whales (≤2% of the population) were potentially affected by vessel or Northstar operations. This effect consisted of either an offshore displacement of whales traveling near the southern edge of the migration corridor or some change in their calling behavior or both. Further analysis of the 2001 data based on improved and more quantitative procedures (see below) suggests that the preliminary ≤200 estimate for 2001 was an overestimate of the number of whales that were substantially displaced. We arbitrarily consider an apparent displacement by >2 km to be “substantial”. Refinements in Analysis Procedures for 2001 and 2002.—This chapter develops updated estimates of the number of bowheads potentially affected in 2001, along with corresponding estimates for 2002. A new and less subjective procedure is proposed and used for this purpose. The uncertainties in the new estimates have not been estimated, and the new estimates could either over- or under-estimate the actual numbers of bowhead whales displaced or otherwise affected. However, the new approach is more objective and data-based than the preliminary approach used in our previous analysis of the 2001 data. As compared with the previous estimates for 2001, the present procedures for 2001 and 2002 incorporate the following refinements: 1. The measure of industrial sound used in the revised 2001 and the 2002 analyses is based on five adjacent 1/3rd octave bands. Reasons for use of the five adjacent bands in both years are explained in Chapter 6. In brief, this procedure appears to provide the best available measure of the components of underwater sound attributable to industrial activities associated with Northstar. This measure is called the Industrial Sound Index or ISI.

3 For the “West” portion of the Alaskan Beaufort as a whole, including Northstar, bowhead sightings in 2000 tended to be closer to shore than in most other years (see Table 7 in Treacy 2002a). However, the low mean and median distance-from-shore values in 2000 were caused by numerous sightings close to shore in the extreme western part of the Beaufort Sea, west of Harrison Bay. The few sightings in 2000 in the Prudhoe Bay/Northstar area were at relatively “normal” distances offshore, as were the larger number of sightings in the “East” region east of Prudhoe Bay (see Figures 7–10 in Treacy 2002a). Chapter 11: Seals & Whales Potentially Affected 11-15

2. The present analyses for 2001 and 2002 are based on the average ISI value for the 70-min period preceding the time of a given whale call. For both years, the offshore displacement effect at times with relatively high industrial sound was most evident when the ISI was averaged over about 70 min (Chapters 9, 10). The size of the displacement effect was not very sensitive to the specific averaging time in 2002, but was sensitive to that time in 2001. To ensure that the number of whales affected is not underestimated, we base this analysis on the averaging time for which the apparent displacement effect was a maximum (i.e., 70 min). 3. The estimated locations of the whale calls detected in 2001, as well as those detected in 2002, have been refined based on use of a correction factor for “gain imbalance” between the two directional sensors in each DASAR (see Chapter 8). Thus, the present estimates of the locations (and offshore distances) of the calling whales during 2001 are slightly more precise than were the original estimates used by Williams et al. (2002a). 4. The “analysis area” considered in the present report (Chapters 9–11) extends 10 km to the west, north and east from the “central axis” of the array of DASARs. The seaward (northerly) edge of the analysis area is now 26.6 km offshore of Northstar. In earlier analyses of 2001 data, the analysis area extended 12 km from the central axis, and to 28.6 km offshore of Northstar. Reasons for the adjustment in the size of the analysis area are explained in Chapter 8. 5. The statistical analysis procedure used to determine the size of the offshore displacement effect now allows for the fact that the expected 5th (or other) quantile of offshore distance can vary somewhat with ISI even if Northstar sound has no effect on offshore distances. The existence of this important complication is shown in Chapter 9. That chapter also describes a refinement of the statistical procedures that allows for this complication. 6. Previous estimates of numbers of bowhead whales potentially displaced in 2001 included a subjective assessment as to how far beyond the 5th quantile distance-from-shore the displacement effect may have extended. That was necessary because it was likely that, at times, displacement would extend beyond the 5th quantile distance. Whales slightly farther offshore than the 5th percentile whale call would be exposed to somewhat lower sound levels than those at the 5th percentile distance-from-Northstar. These whales presumably would be less affected, i.e., deflected by lesser amounts, assuming that the observed “Northstar effect” is a result of offshore deflection. However, in circumstances where there is a substantial displacement of the 5th quantile distance-from-shore, it is reasonable to expect that lesser but still notable displacements could occur for whales somewhat farther offshore. Williams et al. (2002a) noted that, “in future, it might be useful to fit curves to the 10th, 15th, etc., percentile distances-from-Northstar as well as to the 5th percentile, especially when a larger sample size is available. This could show how much farther offshore, beyond the 5th percentile distance, the apparent displacement effect extends. That information, in turn, would help in estimating the number of whales affected.” We have applied this refinement to the 2001 as well as the 2002 data. Specific Estimation Procedure.—In estimating the number of bowheads potentially displaced by Northstar in 2001 and 2002, our approach is based on the quantile regression methods described in Chapters 9 and 10, with the refinements listed in the previous subsection. In general, the procedure was as follows: We used the underwater sound measurements near Northstar to determine the proportion of the whale calls (and, we assume, the proportion of the whales) occurring when the industrial sound index was in each of several categories (5 dB bins). For each of those categories, we estimated the percentage of the passing whales that was apparently displaced by 2, 3 and 5 km. (The specific procedure is 11-16 Monitoring at Northstar, 1999–2002 described below.) We allowed for the fact that only about one-half of the whale population would be expected to pass through the analysis area (i.e., through waters ≤26.6 km seaward of Northstar). By combining these percentages and fractions, and applying them to the overall population size (approx. 10,000 bowheads), we estimated how many bowheads were apparently displaced offshore by at least 2, 3 and 5 km. These are three alternative estimates of the number of bowheads potentially affected. The estimated numbers displaced by 2, 3 and 5 km diminished progressively, as expected. As noted previously, it is possible that the apparent displacement effect is partly or perhaps even entirely attributable to a change in the calling behavior of the whales close to shore at high noise times rather than an actual offshore deflection. In either case, there was a change in the behavior of some whales as a result of exposure to Northstar activities, and the geographic scale of this effect on behavior can be estimated from the call localization data. The following are the specific steps applied in this chapter to implement, separately for each of 2001 and 2002, the estimation procedure summarized above: 1. Based on the underwater sounds recorded continuously ~400 m from Northstar (see Greene et al. 2002 and Chapter 6), determine the proportion of the whale calls occurring in the analysis area when the Industrial Sound Index (ISI), averaged over 70-min periods immediately preceding the calls, was in each of five 5-dB bins: 102.5–107.5, 107.5–112.5, … , 122.5–127.5. (All dB values are dB re 1 µPa, considering the five 1/3rd octave bands centered at 31.5–80 Hz.) Lower ISI values were not relevant because there was no apparent displacement of bowheads at times when ISI was in the range 102.5–107.5 dB or in lower ISI categories. Results of this step are shown in the 3rd column of Table 11.2. 2. Fit additional quantile regression models to the data on offshore distances vs. ISI. Chapters 9 and 10 describe the fitting of the 5th quantile. Here we used the same procedures to fit the 2.5th (2002 only), 7.5th, 10th, 12.5th (2001 only), 15th, 20th, 25th, 30th and 35th quantiles.4 As examples, the continuous lines in Figure 11.2 show the 10th and 20th as well as the 5th quantiles as fitted to the 2002 data. 3. For each fitted quantile regression, apply the permutation method described in Chapter 9 to determine the offshore distance expected at each ISI level in the absence of any effect of Northstar th sound on offshore distance. The “– ▲ –” lines” in Figure 11.2 show (for 2002) these expected 5 , 10th and 20th quantile offshore distances in the absence of a Northstar effect. 4. For the mid-points of the five 5-dB bins considered in (1) above (i.e., for ISI = 105, 110, 115, 120, and 125 dB re 1 µPa), plot the differences between the fitted quantile regressions and the corresponding expected values. Figure 11.3 is an example of this, considering the case of 120 dB in 2002. 5. From plots such as Figure 11.3, determine what percentage of the bowheads passing through the analysis zone at times with a given ISI would be expected to be displaced by 2, 3 and 5 km (columns 4, 5 and 6 of Table 11.2).

4 For 2001, sample size was insufficient to allow us to fit the 2.5th quantile meaningfully. For 2002, the 12.5th quantile was anomalous relative to the 10th and 15th quantiles because of one heavily-weighted call location. Excluding the 12.5th quantile in 2002 resulted in higher estimates of the numbers of whales displaced. Chapter 11: Seals & Whales Potentially Affected 11-17

TABLE 11.2. Estimated number of bowhead whales apparently displaced offshore by at least 2 km, 3 km, and 5 km during 2001 and 2002. The Table considers times when 70-minute ISI levels at the time of the call were moderate and high (> 102.5 dB re 1 µPa). Each ISI bin was 5 dB wide, centered on the reported level. Midpoint of Percentage Percentage of whales within analysis Estimated number of 70-min ISI of calls in ISI zone displaced by at least whales displaced Year bin (dB) bin 2 km 3 km 5 km 2 km 3 km 5 km 105 3.84% 0.00% 0.00% 0.00% 0.0 0.0 0.0 110 1.43% 0.00% 0.00% 0.00% 0.0 0.0 0.0 115 1.96% 6.24% 0.00% 0.00% 6.1 0.0 0.0 2001 120 1.25% 6.70% 5.40% 0.00% 4.2 3.4 0.0 125 0.54% 6.81% 5.70% 0.00% 1.8 1.5 0.0 2001 Totals: 12-13 ~5 0

105 12.59% 0.00% 0.00% 0.00% 0.0 0.0 0.0 110 2.72% 6.54% 0.00% 0.00% 8.9 0.0 0.0 115 0.78% 15.96% 5.71% 0.00% 6.2 2.2 0.0 2002 120 0.10% 23.26% 9.94% 0.97% 1.1 0.5 0.0 125 0.15% 35.32% * 20.86% 2.59% 2.6 1.5 0.2 2002 Totals: ~19 4-5 0-1

* This value obtained by extrapolating displacement of the 25th and 30th quantile regressions to 2 km.

Number of whales in migration corridor = 10000 Proportion of population passing through the 0.5 analysis zone =

6. Assume that the total number of bowhead whales that travel west in the waters off Northstar during late summer and autumn is 10,000 (George et al. 2002). 7. Assume that 50% of these whales travel through the analysis area (extending seaward to 26.6 km NNE of Northstar); the remaining 50% travel west farther offshore, seaward of the analysis area. The 50% figure is based on late summer/autumn aerial surveys (MMS and industry) in the central Alaskan Beaufort Sea (Flaxman Island to Thetis Island) during 1979–2000, compiled by Miller et al. (1999) for 1979–1998 and subsequently updated to included 1999 and 2000. The 50% figure takes account of differences in survey effort at different distances offshore. The actual value derived from this process was 49%. This percentage would vary considerably from year to year, as there is substantial annual variability in the average distance of bowhead sightings from shore (Treacy 2002b). Given this variability, rounding from 49% to 50% is reasonable. 8. For each 5-dB category of ISI, estimate the number of whales apparently displaced by a specific amount (say x km) by multiplying the following four factors: • proportion of whale calls when ISI is in that 5-dB category (Table 11.2, column 3); • proportion of bowhead whales passing during times with that ISI level that are displaced by x km (Table 11.2, columns 4–6); 11-18 Monitoring at Northstar, 1999–2002 70 Minute Sound Window

6 Offshore Distance (miles)

6 10 15 20 25

Offshore distance (km) Observed Expected Quantile 20th 10th 5th 4

90 100 110 120 Industrial Sound Index at 400 m (dB re 1µ Pa) th th th FIGURE 11.2. Estimated 5 , 10 , and 20 quantile regressions of offshore distances of bowhead whale calls localized within the “analysis area” during 2002 vs. average industrial sound index (ISI) ~400 m north of Northstar in the 70 min previous to the call. Expected quantile lines assuming no effect of Northstar on offshore distances (– ▲ –), as derived by the permutation method described in Chapter 9, are also shown. Displacement of a given quantile at a particular ISI value is the difference between the estimated and expected values at that ISI level.

• proportion of bowhead whales that pass through the analysis area (assumed to be 0.5); • bowhead population size (assumed to be 10,000). Then add the results for the various 5-dB categories. The results of this process are shown in the three columns on the right side of Table 11.2. Estimated Number of Bowheads Apparently Displaced.—Based on the above procedure, the estimated numbers of bowheads apparently displaced by at least 2, 3 and 5 km during 2001 were, respectively, 13, 5 and 0 whales, rounded up (Table 11.2). Corresponding estimates for 2002 were 19, 5 and 1 whales. Decimals have been rounded up to the next integer in each case. These numbers are not additive. For example, in 2002, the 19 whales estimated as deflected by at least 2 km includes the 5 whales deflected by at least 3 km. Thus, the total number estimated to be deflected by at least 2 km in 2002 is 19, not 24 or 25. As previously discussed, the apparent displacement effect could in fact be partly indicative of a Northstar effect on calling behavior rather than an actual displacement. Chapter 11: Seals & Whales Potentially Affected 11-19 120 dB Displacement (km) 2345

0% 5% 10% 15% 20% 25% 30% Regression quantile

FIGURE 11.3. Estimated displacement of offshore distance quantiles in 2002 at ISI = 120 dB re 1 µPa. Filled circles were estimated from quantile regressions. Open circle was estimated by linear extension of the curve below the 2.5% quantile. Dashed lines illustrate that, for ISI = 120 dB, the estimated percentages of the offshore distances that displaced by 5 km, 3 km and 2 km were 0.97%, 9.94%, and 23.26%, respectively. Displacement of the 12.5% quantile is plotted, but was excluded from the curve. For that quantile, a single whale-call with high weighting resulted in an uncharacteristically low estimate of displacement; however, inclusion of the 12.5% quantile would not affect the estimated percentages displaced by 2 km, 3 km, and 5 km when ISI was 120 dB.

We have not attempted to estimate how many whales were apparently deflected by 1 km or other distances <2 km. However, the smaller the criterion of displacement (e.g., 1 km vs. 2 km), the larger the number of animals that would be affected by at least that amount. The estimated numbers “displaced” by ≥2 km are very low, in relation to the total population size, for three interrelated reasons: (1) For the great majority of the whale calls, the measured 70-min industrial sound index ~400 m from Northstar was <107.5 dB re 1 µPa. At those times, there was no evidence that any bowheads were displaced offshore—even the whales along the southern edge of the migration corridor closest to Northstar. Only 5.2% of the calls localized in the analysis area during 2001, and 3.8% during 2002, occurred when average ISI during the preceding 70 min was ≥107.5 dB re 1 µPa (Table 11.2). Thus, one would expect ~95–96% of the 10,000 bowheads to pass Northstar at times when sound levels were too weak for any whales to be displaced. (2) At the times with somewhat higher ISI values, the displacement effect was limited to small distances, and only a small percentage of the few whale calls heard at those times showed evidence of displacement. (3) At the times with highest ISI values (e.g., ≥117.5 dB re 1 µPa), the displacement effect was 11-20 Monitoring at Northstar, 1999–2002 somewhat larger, affecting a somewhat larger fraction of the whales passing at those times, but the effect was still confined to the southern part of the migration corridor. Also, only a small fraction of the whale calls were localized during these highest ISI times: 1.8% in 2001, and 0.25% in 2002. Thus, only a few whales were estimated to be displaced at the high-ISI times. For 2001, the present estimates of numbers potentially displaced offshore are much lower than the preliminary estimate of ≤200 bowheads derived by Williams et al. (2002a). The preliminary estimate was overestimated for two main reasons: (1) It was based on precautionary (i.e., high) and more subjective estimates of the distance offshore to which displacement might extend, given that only the 5th quantile of offshore distances had been fitted at that time. We now consider other fitted quantiles as well, and these allow a more objective assessment. (2) The preliminary calculations were based on an assumption that the 5th quantile of offshore distance would be a constant if there were no Northstar effect. However, sub- th sequent investigation has shown that the 5 quantile expected with no effect (– ▲ – lines in Fig. 10.8, 10.9 and 11.2) tends to be farther offshore at high than at moderate ISI values. This results in a considerably smaller difference between the fitted and expected 5th-quantile offshore distances at high ISI times (i.e., a smaller displacement effect) than had been expected when the preliminary analysis was done. The estimated numbers of whales displaced offshore are based on the fitted quantile regressions relative to the best estimate of the corresponding quantile distance if there were no Northstar effect. However, these fitted and expected values are subject to uncertainty. As shown in Chapters 9 and 10 for the 5th quantile, the 90% confidence bounds for the estimated displacement are generally wide. The lower edge of the 90% confidence bound around the estimated displacement is often ≤0 km, and the upper edge often about twice the best estimate of displacement. Confidence bounds on the estimated numbers of whales displaced by x km have not been estimated, but it is clear that they would also be fairly wide. To a first approximation, one can say that there is roughly a 50% chance that the true value was higher than the best estimate, and a 50% chance that it was lower. For 2001, in particular, the evidence of offshore displacement was somewhat weak, and depended on the duration over which sound level was averaged (Chapter 9). The largest apparent effect was found with 70-min averaging, and that is the averaging time we have considered in this chapter. Alternative analysis approaches involving other averaging times, when applied to the 2001 data, showed little or no evidence of an offshore displacement at the times of highest noise. The 2001 data (but not the 2002 data) were also consistent with the possibility of an attraction effect at times of moderate Northstar sound relative to times with weaker sound. In general, there is much uncertainty about the strength and geographic extent of the apparent displacement effect in 2001. Thus, there is much uncertainty in the estimated numbers of whales displaced in 2001. For 2002, the estimates of numbers displaced are more robust but still imprecise. This is partly a result of the larger sample size in 2002 than in 2001 (i.e., more whales localized within the analysis zone) and the improved precision of location estimates. In addition, in 2002 the results were less dependent on the interval over which sound level was averaged, there was a higher degree of confidence that the displacement effect was real, and the scale of the displacement effect was estimated with greater precision. The above estimation process tacitly assumes that whales traveling past Northstar after 2 Oct. 2001 and 23 Sept. 2002 (when monitoring ended) encountered the same range of sounds as those passing up to those dates. • However, vessel activity near Northstar had essentially ended by early October in both years, as freeze-up occurred. The apparent displacement effects prior to freeze-up were at high-noise times Chapter 11: Seals & Whales Potentially Affected 11-21

when “Northstar sounds” were dominated by vessel sounds, not sounds from the island itself (Chapter 10). After freeze-up, strong vessel noise would not occur often if at all. • Helicopter traffic increased when vessel traffic ended (see Chapter 2), but helicopter flights in support of Northstar did not extend appreciably north of Northstar and thus rarely if ever approached the bowhead migration corridor. Underwater sound from helicopters (unlike that from vessels) does not propagate far from the flight path (Urick 1972; Patenaude et al. 2002). Use of helicopters after freeze-up would not cause appreciably elevated noise levels in waters north of Northstar. Thus, there probably were few if any occasions after early October when sound levels were high enough to cause appreciable deflection. The proportions of the bowhead migration that occurred after early October in 2001 and 2002 are unknown. [The last bowhead sightings during MMS’s 2001 aerial surveys were on 2 Oct., but aerial survey coverage thereafter was limited (Treacy 2002b). MMS’s 2002 aerial survey data are not yet available at the time of writing.] Bowhead migration through the Prudhoe Bay area commonly continues until mid- to late October. Based on this consideration, plus (for 2001) our use of the sound-averaging-time associated with maximum apparent deflection, our estimates of the numbers of bowhead whales displaced are more likely to overestimate than to underestimate the actual numbers. There is no established criterion for determining how large the displacement would need to be before a bowhead whale would be considered “taken by harassment”. However, it seems improbable that the apparent displacements evident in the southern part of the migration corridor at high-noise times during 2001 and 2002 would have negative consequences for the few individual whales involved let alone their population. There is considerable natural variation in the distances of bowheads from shore both within and between years (Treacy 2002a,b). Thus, the displacement would need to be by many kilometers before the whales could be said to following a migration route outside the normal range of routes. Offshore displacement of the migration route of a given whale by 2 or 3 km, or even 5 km, is well within the natural range of variability, and is unlikely to have negative consequences for the individual whale. An exception could occur if the whales were displaced from a localized area of particular significance to bowheads. However, bowheads did not show any special tendency to congregate or feed near Northstar prior to the construction of Northstar Island in 2000 (e.g., Miller et al. 1996; Treacy 2002b). Given those considerations, plus the small numbers of whales apparently displaced appreciably by Northstar, no biologically significant effects on individuals or their population are expected as a result of the types of displacements evident to date. Likewise, given the apparent rarity of displacements by ≥2 km, there is no evidence that impacts from Northstar are changing the migration corridor in a significant manner even in the area immediately north of Northstar. Although we often describe the apparent “Northstar effect” on the distribution of whale calls as being indicative of deflection offshore, it is possible that the effect represents a change in the calling behavior of the whales close to shore rather than an actual offshore deflection. These possibilities cannot be distinguished with confidence. In either case, i.e., deflection or altered calling behavior, there was apparently a change in the behavior of some whales as a result of exposure to Northstar activities. Our estimates of the number of whales involved are only small proportions of the bowhead population. For the small number of whales affected, it is not known how long the effect (whether slight offshore displacement or altered calling behavior) persisted. However, even if this effect were prolonged, it is unlikely that a slight offshore displacement or a change in calling behavior among whales traveling through largely open water would result in long-term negative effects on biologically important activities 11-22 Monitoring at Northstar, 1999–2002 of those individuals. In the absence of important negative effects on the small percentage of individuals subject to apparent displacement, no significant effects on their overall population would be expected. Potential implications for subsistence hunting are addressed below. Gray Whales The LoA indicated (NMFS 2001) that no more than five gray whales could be incidentally harassed by Northstar production activities in the open-water season. There are no specific data on the numbers of gray whales (if any) that were present near Northstar in 2002. Gray whales are uncommon in the Prudhoe Bay area, with no more than a few sightings in any one year, and usually no sightings (Miller et al. 1999; Treacy 2000, 2002a,b). There were no sightings of gray whales during the MMS aerial surveys of the Beaufort Sea in 2002 (S. Treacy, MMS, pers. comm.). Gray whales do not call very often when on their summer feeding grounds, and the infrequent calls are not very strong (M. Dahlheim and S. Moore, NMFS, pers. comm.). No gray whale calls were recognized in the data from the acoustic monitoring system near Northstar in 2002 (or in 2000–01). It is most likely that no gray whales were affected by activities at Northstar in 2002. Beluga Whales There are no specific data on the numbers of beluga whales (if any) disturbed or otherwise affected by Northstar production activities in 2002. During MMS aerial surveys in 2002, the majority of beluga sightings were far offshore, near or beyond the shelf-break, and the few nearshore sightings were east of Flaxman Island (S. Treacy, MMS, pers. comm.). Specific monitoring of the beluga migration was not identified as a requirement, given the predominantly offshore nature of beluga migration across the Alaskan Beaufort Sea during late summer and autumn. The DASARs were not designed to monitor the beluga migration; the DASARs did not record sounds with frequencies above ~500 Hz, and beluga calls are at frequencies higher than that. Williams et al. (2001b), following procedures similar to those of Miller et al. (1999), used historical aerial survey data to estimate the number of belugas that might approach the Northstar site in the absence of any disturbance: • Aerial survey data from 1979 to 2000, including both MMS and LGL surveys, were used to estimate the proportion of belugas migrating through waters ≤4 km seaward of Northstar. Of the belugas traveling through the surveyed waters (generally inshore of the 100-m contour), the overall percentage seen ≤4 km offshore of Northstar during 1979-2000 was 0.62% (8 of 1289 belugas). The maximum percentage for any one year was for 1996, when 6 of 153 (3.9%) were ≤4 km offshore of Northstar. These figures are based on beluga sightings within the area 147°00’ to 150°30’W. • Most beluga whales migrate far offshore; the proportion migrating through the surveyed area is unknown but was assumed by Miller et al. (1999) to be ≤20%, which is probably an overestimate. • The disturbance radius for belugas exposed to construction and operational activities in the Beaufort Sea, although not well defined, is apparently considerably less than that for bowheads (Richardson et al. 1995). BPXA (1999) assumed that the potential radius of disturbance was ~1 km around the island. (There are no Northstar-specific data that could be used to obtain a better estimate than this ~1 km figure.) Based on the assumed 1 km radius, we would expect Chapter 11: Seals & Whales Potentially Affected 11-23

that no more than 20% of the belugas migrating ≤4 km seaward of Northstar would approach within 1 km of the Northstar island in the absence of any industrial activity there. • The current size of the Beaufort Sea population of beluga whales is ~39,258 animals (Angliss and Lodge 2002). • Satellite-tagging data show that some members of the Chukchi Sea stock of belugas could also occur in the Beaufort Sea generally near Northstar during late summer and autumn (Suydam et al. 2001, 2003). However, they (like the Beaufort belugas) tend to remain at or beyond the shelf break when in the Alaskan Beaufort Sea during that season. That, combined with the small size of the Chukchi stock, means that consideration of Chukchi belugas would not appreciably change the estimated numbers of belugas that might occur near Northstar. From these values, the number of belugas that might approach within 1 km of Northstar (in the absence of industrial activities) during a given open water season is ~10 belugas based on the average distribution: 0.0062 × 0.2 × 0.2 × 39,258. If the disturbance effects extended to a radius of 2 km, then the estimated number of belugas potentially involved during a typical year would be ~20.

EFFECT ON ACCESSIBILITY TO HUNTERS

Residents of the village of Nuiqsut are the primary subsistence users in the Northstar project area. The subsistence harvest during winter and spring is mainly ringed seals. During the open-water period both ringed and bearded seals are taken commonly. Bowhead whales are hunted from camps on Cross Island, east of Northstar, during September. Beluga whales are rarely taken by Nuiqsut hunters. Seals

Nuiqsut hunters may hunt seals year-round; however during recent years, most of the seal harvest has been in the early summer in open water (Thomas Napageak, pers. comm.). The most important seal hunting area for Nuiqsut hunters is off the Colville Delta, extending as far west as Fish Creek and as far east as Pingok Island (149º40'W). Seal hunting occurs in this area by snow machine before ice break-up, and by boat during summer. Pingok Island, the closest edge of the main hunting area, is ~27 km (17 mi) west of Northstar. Disturbance from production activities at Northstar during the winter of 2001–02 did not displace seals to the extent that their availability to hunters from Nuiqsut would be reduced. However, some concern has been expressed that seals near Northstar may be avoided by hunters given the perception that these seals would be at greater risk of exposure to contaminants, including oil from Northstar. No hunters were observed near the Northstar development during LGL’s intermittent monitoring work in November 2001 through early June 2002. In winter and spring, a small number of ringed seals may have been displaced from the immediate location of the development area. However, if this occurred, the effect was sufficiently limited and localized as to be undetectable by aerial surveys (Chapter 5). Given the aerial survey results, any net displacement of ringed seals that might have occurred during the ice-covered period was small and would not have affected their availability to subsistence hunters at that time of year. Similarly, any disturbance caused by flying over seals during the aerial surveys described in this chapter and (in more detail) in Chapters 4 and 5 had little potential to influence seal hunting activities by residents of Nuiqsut. The western edge of the survey area is 13.7 km (8.5 mi) east of Pingok Island, the eastern edge of the preferred hunting area. No hunters or signs of subsistence hunting were observed on 11-24 Monitoring at Northstar, 1999–2002 the ice during any of the flights during the present 2002 surveys or during the 1997–2001 surveys (G. Miller, LGL, pers. comm.; M. Williams and V. Moulton, pers. obs.). Bowhead Whales

Nuiqsut hunters establish hunting camps on Cross Island during early September each year in order to hunt for bowhead whales. Cross Island is located 28 km (17.5 mi) east of Northstar. Most bowheads that are taken by Nuiqsut hunters are struck east or north of Cross Island (Long 1996). In recent years, Nuiqsut hunters have had a quota of four whales. The number of whales landed has ranged from two to four per year during 1995 to 2002. Three bowhead hunting seasons have occurred at Cross Island subsequent to the start of construction at Northstar: • In 2000, four bowheads were landed, on dates ranging from 2 to 9 Sept. The 2000 hunt was completed successfully at an earlier date than normal despite the ongoing Northstar operations west of Cross Island. • In 2001, three bowheads were landed, on 5, 10 and 22 Sept. The Nuiqsut hunters indicated that the 2001 hunt was more difficult than the 2000 hunt (M. Galginaitis, Applied Sociocultural Research, pers. comm., April 2002). This was partly because of more difficult ice and weather conditions in 2001. However, the hunters reported that the whales also were more difficult to find, more skittish, and generally more difficult to hunt in 2001. They suggested that this could be attributable to industrial activities, vessel traffic, or perhaps killer whales to the east of Cross Island (M. Galginaitis, pers. comm.). • In 2002, four bowheads were landed; they were struck on 5, 12, 13 and 15 Sept. The 2nd and 4th of these whales initially sank, and were not landed until 7 and 2 days after they were struck, respectively (Gal- ginaitis 2003). The hunters did not make any specific comments about possible Northstar effects on the 2002 hunt during interactions with the sociocultural researcher who documented that hunt (M. Galgin- aitis, pers. comm., May 2003). The acoustic localization system deployed offshore of Northstar was not designed to monitor bowhead calls reliably as far away as the Cross Island area. However, it did provide much information on the positions and distances-from-shore for bowhead calls detected west of Cross Island but east of Northstar (see 10.12 and 10.13 in Chapter 10). Most calling bowheads were at least 10 km offshore of a line parallel to the coast and through Northstar, and the densest concentration of calls was 15 km or more offshore of that line (Fig. 10.8, 10.9). The historical aerial survey data show that most bowheads sighted in the general Prudhoe Bay area prior to 2000 (when Northstar was constructed) were also 15 km or more offshore of a similar baseline (e.g., LGL and Greeneridge 1996). There was no indication of any large scale change in the migration corridor in 2001 or 2002 as compared with years before 2000, notwithstanding the evidence of a slight offshore displacement of some whales at the few times when industrial sounds were strongest. The acoustic localization data revealed a “Northstar effect” on the distribution of whale calls during the 5% (2001) or 3% (2001) of the time when Northstar sounds (averaged over 70 min) were strongest. This effect was wholly or largely attributable to sounds from maneuvering vessels rather than Northstar island itself (Chapter 10). The effect was most evident during the 1.8% (2001) or 0.25% (2002) of the time when sound levels in the “industrial bands”, averaged over 70 min, were >117.5 dB re 1 µPa at a location 400 m from Northstar. That level is equivalent, on average, to >121–122 dB in the overall 10–1000 Hz band, given that levels in the 10–1000 Hz band average 4–5 dB higher than levels in the “industrial bands”. Received levels of industrial sound at the locations of the whales (which were much more than 400 m from Northstar) would be much lower, as shown in Chapter 10 (Fig. 10.12 and 10.13). Chapter 11: Seals & Whales Potentially Affected 11-25

Specific industrial-to-ambient ratios at the locations of bowheads that deflected are uncertain, but probably low. Even during the infrequent times of highest noise, the spatial scale of the effect was small, estimated as ≤5 km but with considerable uncertainty. The sample size under these relatively high-noise conditions was low, resulting in broad confidence limits around the estimate of average offshore displacement. The results at these noisiest times are consistent with the possibility that, on these infrequent occasions, effects on bowheads, either displacement or effects on calling patterns, might have extended as much as 7–10 km (4.3–6.2 mi; Fig. 10.9, 10.11). This is based on a comparison of the fitted 5th quantile regression line to the lower edge of the 90% confidence region. It is equally likely that the displacement effect with high ISI was <1 km, based on the difference between the fitted 5th quantile regression line and the upper edge of the 90% confidence region (Fig. 10.9, 10.11). The most likely size of the displacement effect was intermediate between these two extremes, as represented by the difference between the fitted 5th quantile regression line vs. best estimate of the 5th quantile without any Northstar effect (Fig. 10.9, 10.11). Because the sound levels associated with appreciable displacement occurred infrequently (<1% of the time), this alone could not account for reports that “skittish” behavior was common among bowheads off Cross Island in 2001. Indeed, the best estimates for 2001 and 2002 are that only ~13 and ~19 bowheads, respectively (<0.2% of 10,000), were deflected by as much as 2 km (1.2 mi) as a result of Northstar activities. Those estimates are imprecise, and the actual numbers displaced by ≥2 km might have been somewhat larger. However, given the low proportion of time with sounds strong enough to elicit apparent displacement, and given the limited geographic extent of displacement even at those times, the number of bowheads displaced by ≥2 km must have been small even if somewhat larger than our current “best estimates”. Because the effects are limited to the bowheads closest to Northstar, it is unlikely that displacement effects would extend far enough to the east and northeast to affect whales being hunted near Cross Isl. There is concern among the Barrow hunters that offshore displacement of bowhead whales at industrial sites east of Barrow would affect the accessibility of whales to Barrow hunters. Point Barrow is ~300 km (190 mi) west-northwest of Northstar, and the eastern edge of the Barrow hunting area is about 50 km (30 mi) closer to Northstar. This study was not designed to assess how far west of Northstar the apparent displacement effect extended. It can be argued that, given the natural day-to-day and year-to- year variability in the distances of bowheads from shore, and in their directions of movement, an offshore displacement by a few kilometers is unlikely to persist after a 250 km westward migration. However, there is no specific evidence on that point. Even if displacement or other disturbance effects did persist far to the west, the evidence indicates that they would involve only the small percentage of the whales that passed close to Northstar at high-noise times.

SUMMARY A Letter of Authorization (LoA) issued by NMFS to BP authorized the “taking” of small numbers of seals and whales incidental to Northstar production activities (Table 11.3). The numbers of seals and whales present and potentially affected by Northstar production activities during the ice-covered, break- up, and open-water periods in November 2001 through October 2002 have been estimated based on information in previous chapters of this report (Table 11.3). The ice-covered and open-water seasons are not discrete, and the intervening “break-up” period has been combined with the open-water season in estimating the “numbers potentially affected”. The following paragraphs summarize the basis for the estimates in Table 11.3. 11-26 Monitoring at Northstar, 1999–2002

TABLE 11.3. Authorized annual takes incidental to Northstar production activities, and estimated numbers of seals and whales present and potentially affected during November 2001 – October 2002.

Authorized Annual Harassment Takes # Present &/or Potentially Affected_ Ice- Break-up & Ice- Break-up & Covered Open-Water Covered Open-Water Season Seasons Total Season Seasons Total Ringed Seal 139 a 52 191 63b,c 47b 110b Bearded Seal 5 5 10 - 1 1 Spotted Seal - 5 5 - - 0 Bowhead Whale - 765 d 765 d -~19e ~19e Gray Whale - 5 5 - ~0 ~0 Beluga Whale - 91 91 - ~10–20 ~10–20 a In addition, the LoA authorized up to 5 ringed seals to be incidentally killed annually during the ice-covered period. In fact, there was no evidence of any seal deaths as a result of Northstar activities. b Northstar production activities probably had little or no effect on most of these seals, as discussed in text. c Excludes an estimated 143 seals that dove in response to the passing aircraft during spring seal surveys; possible disturbance to seals during those surveys was authorized separately by NMFS under provisions of the Scientific Research Permit (SRP 481-1623-01) issued to LGL. d Up to 765 bowheads annually, with maxima of 1533 bowheads in 2 out of 5 seasons, and 3585 in 5 years. e Bowhead estimates represent the best estimates of the numbers displaced offshore by 2 km or more, from Table 11.2. The 2001 estimate (~13) supersedes the preliminary estimate given in Williams et al. (2002a).

Seals

During the 2001–02 ice-covered season, about 63 ringed seals were expected to occur within the area potentially impacted by BP’s Northstar development if there had been no industrial activities there during 2002. This estimate is based on the density of seals 4–10 km (2.5–6.2 mi) from the potential impact zone in comparable water depths, corrected for availability and detection biases. The potential impact zone to which these estimates apply was defined as extending from 0.64 km from the ice road and 1.85 km (0.4 to 1.15 mi) from Northstar Island. The observed density and the estimated number of seals within the potential impact zone in 2002 were lower than the expected density based on the reference area 4–10 km away. This difference was probably related to the more southerly location of the ice edge in 2002 vs. other survey years, but could be interpreted as indicating that ~20 seals had been displaced by Northstar activities. During aerial surveys, 26 seals were observed in the potential impact zone. Also, the results of the multivariate analysis showed there was no evidence of a “Northstar effect” on seal density (Chapter 5). Overall, the results indicate that, if there were any effects of industrial activities at Northstar on seal densities, these effects were likely very minor and localized. There was no evidence that any seals were killed or seriously injured by Northstar-related activities. We estimated that there were ~143 occasions when ringed seals (and no occasions when bearded seals) dived into the water in response to aircraft overflights during the aerial surveys in spring 2002. Only a small minority of the seals reacted by diving; most seals showed no apparent response to the aircraft. Seals that dived into the water once or twice in response to aircraft overflights are unlikely to have suffered deleterious effects. Most diving responses occurred far from Northstar, and they were authorized by a Scientific Research Permit issued by NMFS to LGL, separate from the LoA issued to BP. Chapter 11: Seals & Whales Potentially Affected 11-27

During the break-up and open-water seasons, no evidence of seal injuries or fatalities was evident, nor was it expected. Approximately 48 different seals (47 ringed seals and 1 bearded seal) are estimated to have occurred within a 1 km (0.62 mi) radius around Northstar Island at some time during the break-up and open-water seasons. This estimate is based on the densities of seals observed during the 2002 aerial surveys (applied to the break-up season) and during vessel-based surveys during 1996–2000 (applied to the open-water season). The estimate takes account of correction factors for missed animals, and assumes weekly turnover of individuals during the break-up and open-water periods. It is unlikely that these seals were affected deleteriously by Northstar activities. For the year as a whole (ice-covered, break-up, and open-water seasons combined), an estimated 110 ringed seals, 1 bearded seal, and probably no spotted seals were present near BP’s Northstar activities. These figures are less than the corresponding “takes” authorized by the LoA issued by NMFS to BP (Table 11.3). Furthermore, most of these seals do not seem to have been negatively affected by Northstar activities, and it would appear that very few of them should be counted as “takes”. These totals exclude the ~143 seals that dived in response to aircraft overflights during the spring aerial surveys; those incidents were separately authorized via a Scientific Research Permit issued by NFMS to LGL. The overall results suggest that any effects of Northstar production activities on seals were minor, short-term, and localized, with no consequences for the seal populations. Whales

For bowhead whales, acoustic localization data indicated that a small number of whales in the southern part of the migration corridor (closest to Northstar) were affected by vessel or Northstar operations during the 4–5% of the time when levels of underwater sound from these activities were highest. At these times, the main components of “Northstar sound” were from maneuvering vessels, not from the island itself. The best (though quite imprecise) estimates of the number of bowheads that were apparently deflected offshore by ≥2 km (1.2 mi) were 13 bowheads in 2001 and 19 bowheads in 2002 (i.e., <0.2% of the population). Actual numbers would vary relative to these imprecise estimates, depending on the accuracy of assumptions and sampling variability. It is possible that the apparent deflection effect was, at least in part, attributable to a change in calling behavior rather than actual deflection, but in either case there was a change in the behavior of a small number of whales. Most of the affected whales passed Northstar at times when the effect was quite subtle and limited to the southern edge of the migration corridor. The LoA authorized the “taking by harassment” of up to 717 bowheads annually, or as many as 1533 as long as this larger number applied to no more than 2 of 5 years. The occurrence of a small “Northstar effect” on a low number of bowheads in each of 2001 and 2002 would be well within the provisions of the LoA. There was no specific information on numbers of gray or beluga whales (if any) that may have been close enough to Northstar to be disturbed by Northstar production and drilling operations in 2002. Given that gray whale sightings have been infrequent this far to the east in previous years, it is most likely that no gray whales were affected by Northstar in 2002. For belugas, the estimated numbers that might approach within an assumed 1–2 km (0.6–1.2 mi) disturbance radius are 10–20 in an average year. The LoA authorized the “taking by harassment” of up to 5 gray whales and 45 belugas. Availability for Subsistence

Disturbance from production activities at Northstar during the winter of 2001–02 did not displace seals to the extent that their availability to hunters from Nuiqsut would be reduced. However, some 11-28 Monitoring at Northstar, 1999–2002 concern has been expressed that seals near Northstar may be avoided by hunters given the perception that these seals would be at greater risk of exposure to contaminants. The bowhead hunt at Cross Island has had a quota of four whales per year in recent years, and the numbers of whales landed were 4 in 2000, 3 in 2001, and 4 in 2002. The hunt was more protracted and apparently more difficult in 2001 than in some other recent years. The difficulties in the 2001 hunt were partly because of weather and ice problems, but in addition the whales were described as being skittish and difficult to hunt in 2001. The acoustic monitoring was not designed to address the question of skittish whales. However, this monitoring indicates that it is unlikely that displacement effects extended far enough to the east and northeast to affect whales being hunted near Cross Island. This study was not designed to assess how far west of Northstar the apparent displacement effect extended, nor to assess whether this effect persists far enough west to affect accessibility of bowheads to hunters from Barrow, 250–300 km (160–190 mi) to the west. Given the natural variability in the offshore distances and headings of bowheads, a small offshore displacement seems unlikely to persist that far, but there is no specific evidence on that point. Even if displacement or other disturbance effects did persist far to the west, they would involve only the low percentage of the whales that passed close to Northstar at high-noise times.

ACKNOWLEDGEMENTS

We thank Drs. Bill Streever and Ray Jakubczak of BP, plus Wilson Cullor and Allison Erickson of OASIS Environmental (on behalf of BP), for their support during this project. We also thank the authors of all other chapters of this report, who provided the data used in estimating numbers of seals and whales potentially affected. In particular, we thank Drs. Susanna Blackwell and Charles Greene (Greeneridge Sciences) for providing the necessary acoustic data and associated analyses. Dr. Bill Streever of BP and Robert Suydam of the North Slope Borough Dept of Wildlife Management commented on an earlier draft of this chapter. Ted Elliott of LGL assisted with data management and analysis.

LITERATURE CITED Angliss, R.P. and K.L. Lodge. 2002. Alaska marine mammal stock assessments, 2002. NOAA Tech. Memo. NMFS-AFSC-133. Nat. Mar. Fish. Serv., Nat. Mar. Mamm. Lab., Seattle, WA. 224 p. Blackwell, S.B. and C.R. Greene Jr. 2001. Sound measurements, 2000 break-up and open-water seasons. p. 7-1 to 7-55 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of industrial sounds, seals, and whale calls during construction of BP's Northstar Oil Development, Alaskan Beaufort Sea, 2000. [Draft, April 2001.] LGL Rep. TA2429-2. Rep. from LGL Ltd., King City, Ont., and Greenridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 316 p. BPXA. 1999. Petition for regulations pursuant to section 101 (a) (5) of the Marine Mammal Protection Act covering taking of marine mammals incidental to offshore oil and gas development and production in the Beaufort Sea, Sept. 1999 ed. Submitted by BP Explor. (Alaska) Inc., Anchorage, AK, to Nat. Mar. Fish. Serv. [Silver Spring, MD]. Prepared by LGL Alaska Res. Assoc., Anchorage, AK. 121 p. BPXA. 2001. Monitoring and reporting plan. Section 13 In: Request for a Letter of Authorization ... covering taking of marine mammals incidental to construction and operation of offshore oil and gas facilities in the U.S. Beaufort Sea, Alaska (50 C.F.R. Part 216, Subpart R). Request from BP Exploration (Alaska) Inc., Anchorage, AK, to Nat. Mar. Fish. Serv. [Silver Spring, MD]. Prepared by LGL Alaska Res. Assoc. Inc., Anchorage, AK. Chapter 11: Seals & Whales Potentially Affected 11-29

Born, E.W., F.F. Riget, R. Dietz and D. Andriashek. 1999. Escape responses of hauled out ringed seals (Phoca hispida) to aircraft disturbance. Polar Biol. 21(3):171-178. Frost, K.J., L.F. Lowry, J.R. Gilbert and J.J. Burns. 1988. Ringed seal monitoring: relationships of distribution and abundance to habitat attributes and industrial activities. U.S. Dep. Commer., NOAA, OCSEAP Final Rep. 61 (1989):345-445. NTIS PB89-234645. Frost, K.J., L.F. Lowry, G. Pendleton and H.R. Nute. 2002. Monitoring distribution and abundance of ringed seals in northern Alaska. OCS Study MMS 2002-043. Final rep. from the Alaska Dept. of Fish and Game, Juneau, AK, for U.S. MMS, Anchorage, AK. 66 p. + Appendices. George, J.C., J. Zeh, R. Suydam and C. Clark. 2002. Population size of the Bering-Chukchi-Beaufort Seas stock of bowhead whales, Balaena mysticetus, based on the 2001 census off Point Barrow, Alaska. SC/54/BRG5. Intern. Whal. Comm., Cambridge, U.K. Greene, C.R. Jr., M.W. McLennan, T.L. McDonald and W.J. Richardson. 2001. Acoustic monitoring of bowhead whale migration, autumn 2000. p. 9-1 to 9-57 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of industrial sounds, seals, and whale calls during construction of BP's Northstar Oil Development, Alaskan Beaufort Sea, 2000. [Draft, April 2001.] LGL Rep. TA2429-2. Rep. from LGL Ltd., King City, Ont., and Greenridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 316 p.

Greene, C.R., Jr., M.W. McLennan, T.L. McDonald and W.J. Richardson. 2002. Acoustic monitoring of bowhead whale migration, autumn 2001. p. 8-1 to 8-79 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of industrial sounds, seals, and whale calls during construction of BP's Northstar oil development, Alaskan Beaufort Sea, 2001, Draft, Oct. 2002 ed. LGL Rep. TA2573-2. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. Kelly, B.P., and L.T. Quakenbush. 1990. Spatiotemporal use of lairs by ringed seals (Phoca hispida). Can. J. Zool. 68(12):2503-2512. Kelly, B.P., L.T. Quakenbush and B.D. Taras. 2002. Correction factor for ringed seal surveys in northern Alaska. p. 3-15 In: Annu. Rep. 8, Coastal Marine Institute, Univ. Alaska Fairbanks. OCS Study MMS 2002-001.

LGL and Greeneridge. 1996. Northstar marine mammal monitoring program, 1995: baseline surveys and retrospective analyses of marine mammal and ambient noise data from the central Alaskan Beaufort Sea. LGL Rep. TA2101-2. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK. 104 p. Link, M.R., T.L. Olson and M.T. Williams. 1999. Ringed seal distribution and abundance near potential oil development sites in the central Alaskan Beaufort Sea, spring 1998. LGL Rep. P-430. Rep. from LGL Alaska Res. Assoc. Inc., Anchorage, AK, for BP Explor. (Alaska) Inc., Anchorage, AK. 58 p. Long, F., Jr. 1996. History of subsistence whaling by Nuiqsut. p. 73-76 In: Proc. 1995 Arctic Synthesis Meeting, Anchorage, AK, Oct. 1995. OCS Study MMS 95-0065. U.S. Minerals Manage. Serv., Anchorage, AK. 206 p. + Appendices. Miller, G.W., R.E. Elliott and W.J. Richardson. 1996. Marine mammal distribution, numbers and movements. p. 3-72 In: LGL and Greeneridge, Northstar marine mammal monitoring program, 1995: Baseline surveys and retrospective analyses of marine mammal and ambient noise data from the central Alaskan Beaufort Sea. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK. 104 p. Miller, G.W., R.E. Elliot and W.J. Richardson. 1998. Ringed seal distribution and abundance near potential oil development sites in the central Alaskan Beaufort Sea, spring 1997. LGL Rep. TA2160-3. Rep. from LGL Ltd., King City, Ont., for BP Explor. (Alaska) Inc., Anchorage, AK. 36 p. 11-30 Monitoring at Northstar, 1999–2002

Miller, G.W., R.E. Elliott, W.R. Koski, V.D. Moulton and W.J. Richardson. 1999. Whales. p. 5-1 to 5-109 In: W.J. Richardson (ed.), Marine mammal and acoustical monitoring of Western Geophysical's open-water seismic program in the Alaskan Beaufort Sea, 1998. LGL Rep. TA2230-3. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for Western Geophysical, Houston, TX, and U.S. Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 390 p. Moulton, V.D., T.L. McDonald, W.J. Richardson, R.E. Elliott and M.T. Williams. 2000. Fixed-wing aerial surveys of seals near BP’s Northstar and Liberty sites in 1999 (and 1997-99 combined). p. 3-1 to 3-71 In: W.J. Rich- ardson and M.T. Williams (eds.), Monitoring of ringed seals during construction of ice roads for BP’s Northstar oil development, Alaskan Beaufort Sea, 1999. LGL Rep. TA2349-3. Final Rep. from LGL Ltd., King City, Ont., and LGL Alaska Res. Assoc. Inc., Anchorage, AK, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 153 p. Moulton, V.D., R.E. Elliott, T.L. McDonald and W.J. Richardson. 2001. Fixed-wing aerial surveys of seals near BP’s Northstar and Liberty sites in 2000 (and 1997-2000) combined). p. 5-1 to 5-66 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of industrial sounds, seals, and whale calls during construction of BP's Northstar Oil Development, Alaskan Beaufort Sea, 2000. [Draft, April 2001.] LGL Rep. TA2428-3. Rep. from LGL Ltd., King City, Ont., and Greenridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 316 p. Moulton, V.D., R.E. Elliott, W.J. Richardson and T.L. McDonald. 2002. Fixed-wing aerial surveys of seals near BP’s Northstar and Liberty sites in 2001 (and 1997-2001 combined). p. 5-1 to 5-60 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of industrial sounds, seals, and whale calls during construction of BP’s Northstar Oil Development, Alaskan Beaufort Sea, 2001. [Oct. 2002 ed.] LGL Rep. TA2570-3. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 309 p. NMFS. 2000a. Taking marine mammals incidental to construction and operation of offshore oil and gas facilities in the Beaufort Sea/Final rule. Fed. Regist. 65(102, 25 May):34014-34032. NMFS. 2000b. Small takes of marine mammals incidental to specified activities; oil and gas exploration drilling activities in the Beaufort Sea. Fed. Regist. 65(197, 11 Oct.):60407-60411. NMFS. 2001. Taking and importing marine mammals; taking marine mammals incidental to construction and operation of offshore oil and gas facilities in the Beaufort Sea. Fed. Regist. 66(246, 21 Dec.):65923-65935. Patenaude, N.J., W.J. Richardson, M.A. Smultea, W.R. Koski, G.W. Miller, B. Würsig and C.R. Greene Jr. 2002. Aircraft sound and disturbance to bowhead and beluga whales during spring migration in the Alaskan Beaufort Sea. Mar. Mamm. Sci. 18(2):309-335. Richardson, W.J., M.A. Fraker, B. Würsig and R.S. Wells. 1985. Behaviour of bowhead whales Balaena mysticetus summering in the Beaufort Sea: reactions to industrial activities. Biol. Conserv. 32(3):195-230. Richardson, W.J., C.R. Greene Jr., C.I. Malme and D.H. Thomson. 1995. Marine mammals and noise. Academic Press, San Diego, CA. 576 p. Suydam, R.S., L.F. Lowry, K.J. Frost, G.M. O'Corry-Crowe and D. Pikok Jr. 2001. Satellite tracking of eastern Chukchi Sea beluga whales into the Arctic Ocean. Arctic 54(3):237-243. Suydam, R., L. Lowry, K. Frost, G. O'Corry-Crowe and G. VanBlaricom. 2003. Satellite tracking of eastern Chukchi Sea beluga whales in the Beaufort Sea and Arctic Ocean, 1998-2002. Abstr. 15th Bienn. Conf. Biol. Mar. Mamm., Greensboro, NC, Dec. 2003:159. Treacy, S.D. 2000. Aerial surveys of endangered whales in the Beaufort Sea, fall 1998-1999. OCS Study MMS 2000-066. U.S. Minerals Manage. Serv., Anchorage, AK. 135 p. Treacy, S.D. 2002a. Aerial surveys of endangered whales in the Beaufort Sea, fall 2000. OCS Study MMS 2002- 014. U.S. Minerals Manage. Serv., Anchorage, AK. 111 p. Chapter 11: Seals & Whales Potentially Affected 11-31

Treacy, S.D. 2002b. Aerial surveys of endangered whales in the Beaufort Sea, fall 2001. OCS Study MMS 2002-061. U.S. Minerals Manage. Serv., Anchorage, AK. 117 p. Urick, R.J. 1972. Noise signature of an aircraft in level flight over a hydrophone in the sea. J. Acoust. Soc. Am. 52(3, Pt. 2):993-999. Williams, M.T., J.A. Coltrane and C.J. Perham. 2001a. On-ice location of ringed seal structures near Northstar, December 1999 and May 2000. p. 4-1 to 4-22 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of industrial sounds, seals, and whale calls during construction of BP's Northstar Oil Development, Alaskan Beaufort Sea, 2000. [Draft, April 2001.] LGL Rep. P485-2. Rep. from LGL Ltd., King City, Ont., and Greenridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 316 p. Williams, M.T., W.J. Richardson, J.W. Lawson and V.D. Moulton. 2001b. Estimated numbers of seals and whales potentially affected by Northstar activities, Nov. 1999 – Oct. 2000. p. 10-1 to 10-23 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of industrial sounds, seals, and whale calls during construction of BP's Northstar Oil Development, Alaskan Beaufort Sea, 2000. [Draft, April 2001.] LGL Rep. TA2432-8. Rep. from LGL Ltd., King City, Ont., and Greenridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 316 p. Williams, M.T., V.D. Moulton and W.J. Richardson. 2002a. Estimated numbers of seals and whales potentially affected by Northstar activities, Nov. 2000 – Oct. 2001. p. 9-1 to 9-27 In: W.J. Richardson and M.T. Wil- liams (eds.), Monitoring of industrial sounds, seals, and whale calls during construction of BP’s Northstar Oil Development, Alaskan Beaufort Sea, 2001. [Oct. 2002 ed.] LGL Rep. TA2563-7. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchorage, AK, and Silver Spring, MD. 309 p. Williams, M.T., T.G. Smith and C.J. Perham. 2002b. Ringed seal structures in sea ice near Northstar, winter and spring of 2000-2001. p. 4-1 to 4-33 In: W.J. Richardson and M.T. Williams (eds.), Monitoring of industrial sounds, seals, and whale calls during construction of BP’s Northstar Oil Development, Alaskan Beaufort Sea, 2001. [Oct. 2002 ed.] LGL Rep. P557-2. Rep. from LGL Ltd., King City, Ont., and Greeneridge Sciences Inc., Santa Barbara, CA, for BP Explor. (Alaska) Inc., Anchorage, AK, and Nat. Mar. Fish. Serv., Anchor- age, AK, and Silver Spring, MD. 309 p.