5 on the Umnak Plateau, Bering

Iowa State University Capstones, Theses and Graduate Theses and Dissertations Dissertations

2015 Marine Isotope Stage (MIS) 5 on the Umnak Plateau, Bering Sea (IODP Site U1339): diatom taxonomy, grain size and isotopic composition of marine sediments as proxies for primary productivity and sea ice extent Derrick Ray Vaughn Iowa State University

Follow this and additional works at: https://lib.dr.iastate.edu/etd Part of the Geology Commons

Recommended Citation Vaughn, Derrick Ray, "Marine Isotope Stage (MIS) 5 on the Umnak Plateau, Bering Sea (IODP Site U1339): diatom taxonomy, grain size and isotopic composition of marine sediments as proxies for primary productivity and sea ice extent" (2015). Graduate Theses and Dissertations. 14705. https://lib.dr.iastate.edu/etd/14705

This Thesis is brought to you for free and open access by the Iowa State University Capstones, Theses and Dissertations at Iowa State University Digital Repository. It has been accepted for inclusion in Graduate Theses and Dissertations by an authorized administrator of Iowa State University Digital Repository. For more information, please contact [email protected].

Marine Isotope Stage (MIS) 5 on the Umnak Plateau, Bering Sea (IODP Site U1339): Diatom taxonomy, grain size and isotopic composition of marine sediments as proxies for primary productivity and sea ice extent

Derrick Ray Vaughn

A thesis submitted to the graduate faculty

in partial fulfillment of the requirements for the degree of

MASTER OF SCIENCE

Major: Geology

Program of Study Committee: Beth E. Caissie, Major Professor Alan D. Wanamaker, Jr. William J. Gutowski, Jr.

Iowa State University

Ames, Iowa

2015

TABLE OF CONTENTS

Page

LIST OF FIGURES ...... iv

LIST OF TABLES ...... v

NOMENCLATURE ...... vi

ACKNOWLEDGMENTS ...... viii

ABSTRACT………………………………...... ix

CHAPTER 1 INTRODUCTION ...... 1

1.1. Basis for Research...... 1 1.2. Productivity, Nitrate Utilization, and Sea Ice Proxies ...... 2 1.3. Thesis Outline ...... 4

CHAPTER 2 LITERATURE REVIEW ...... 6

2.1. The Bering Sea ...... 6 2.1.1. Bering Sea Circulation ...... 7 2.2.2. Atmospheric Patterns in the Bering Sea ...... 9 2.2. Sea Ice Formation and Transport ...... 11 2.3. Primary Productivity ...... 13 2.3.1. Nitrogen Limitation ...... 16 2.3.2. Iron Limitation ...... 17 2.4.Paleoclimate Proxies ...... 18 2.4.1. Nitrogen Isotopes ...... 18 2.4.2. Carbon Isotopes ...... 20 2.4.3. Grain Size...... 21 2.4.4. Diatom-based Proxies ...... 23 2.5. Marine Isotope Stage 5 ...... 24 2.5.1. Marine Isotope Stage 5 in Beringia ...... 25 2.5.2. Marine Isotope Stage 5 in the Bering Sea ...... 27 2.6. Application ...... 30

CHAPTER 3 MARINE ISOTOPE STAGE (MIS) 5 ON THE UMNAK PLATEAU, BERING SEA (IODP SITE U1339): DIATOMS, GRAIN SIZE AND SEDIMENTARY NITROGEN ISOTOPES AS PROXIES FOR PRIMARY PRODUCTIVITY AND SEA ICE EXTENT ...... 32

3.1. Abstract ...... 32

iii

3.2. Introduction ...... 33 3.3. Study Area ...... 35 3.3.1. Bering Sea ...... 35 3.3.2. Umnak Plateau ...... 36 3.4. Background ...... 36 3.5. Methods...... 38 3.5.1. Sediment Cores ...... 38 3.5.2. Age Model ...... 39 3.5.3. Diatom Analysis...... 39 3.5.4. Isotopic Analysis ...... 41 3.5.5. Grain Size Analysis...... 42 3.6. Results… ...... 42 3.6.1. Diatom Analysis...... 42 3.6.2. Geochemical Analysis ...... 44 3.6.2.1. Nitrogen Isotopes ...... 44 3.6.2.2. Carbon Isotopes ...... 45 3.6.2.3. Total Organic Carbon ...... 45 3.6.2.4. C/N Values ...... 46 3.6.3. Grain Size Analysis...... 46 3.7. Discussion ...... 47 3.7.1. MIS 6 and Deglaciation at the Umnak Plateau ...... 47 3.7.2. MIS 5 at the Umnak Plateau ...... 49 3.7.2.1. MIS 5e and MIS 5d ...... 49 3.7.2.2. MIS 5c and MIS 5b ...... 51 3.7.2.3. MIS 5a ...... 53 3.7.3. MIS 4 at the Umnak Plateau ...... 54 3.8. Conclusions and Future Work ...... 55

APPENDIX A INVESTIGATING VARIATIONS IN DIATOM IDENTIFICATION...... 61

A.1. Abstract ...... 61 A.2. Introduction ...... 61 A.3. Methods ...... 62 A.4. Results ...... 63 A.5. Discussion and Summary ...... 66

APPENDIX B DOES CRUSHING SEDIMENT AFFECT IDENTIFICATION ... 67

B.1. Abstract ...... 67 B.2. Introduction ...... 67 B.3. Methods ...... 67 B.4. Results ...... 69 B.5. Discussion and Summary ...... 75

REFERENCES ...... 76

LIST OF FIGURES

Page

Figure 1 Bering Sea Map...... 31

Figure 2 δ15 N, Chaetoceros RS percentages, Diatom Abundances ...... 58

Figure 3 δ13 C, TOC, C/N, Treated/Untreated Grain Size...... 59

Figure 4 δ15 N Comparisons ...... 60

LIST OF TABLES

Page

Table 1 Tie Points ...... 39

Table 2 Sediment samples used for comparing counts ...... 62

Table 3 Chi-square results for 323-U1339C-4H-3 132-133 cm ...... 63

Table 4 Chi-square results for 323-U1339D-3H-3 36.5-37.5 cm ...... 64

Table 5 Chi-square results for 323-U1339D-4H-3 139-140 cm ...... 64

Table 6 Chi-square results for 323-U1339D-4H-4 69-70 cm ...... 65

Table 7 Chi-square results for 323-U1339C-3H-4 112-113 cm ...... 65

Table 8 Chi-square results for 323-U1339C-3H-6 12-13 cm ...... 66

Table 9 Sediment samples used for crushed/uncrushed comparisons ...... 68

Table 10 Chi-square results for 323-U1339D-3H-4 144-145 cm ...... 69

Table 11 Chi-square results for 323-U1339D-3H-5 54-55 cm ...... 70

Table 12 Chi-square results for 323-U1339D-3H-5 114-115 cm ...... 70

Table 13 Chi-square results for 323-U1339D-4H-2 69-70 cm ...... 71

Table 14 Chi-square results for 323-U1339D-4H-2 129-130 cm ...... 71

Table 15 Chi-square results for 323-U1339D-4H-3 39-130 cm ...... 72

Table 16 Chi-square results for 323-U1339D-4H-3 99-100 cm ...... 72

Table 17 Chi-square results for 323-U1339D-4H-4 8.5-9.5 cm ...... 73

Table 18 Chi-square results for 323-U1339D-4H-4 69-70 cm ...... 73

Table 19 Chi-square results for 323-U1339D-4H-4 108.5-109.5 cm ...... 74

Table 20 Chi-square results for 323-U1339D-4H-5 19-20 cm ...... 74

Table 21 Chi-square results for 323-U1339D-4H-5 59-60 cm ...... 75

NOMENCLATURE

δ13C Stable carbon isotopes

δ15 N Stable nitrogen isotopes

18 δ Ocalcite Stable oxygen isotopes of calcite

ACC Alaskan Coastal Current

AMOC Atlantic Meridional Ocean Circulation

AS Alaskan Stream

C/N Total Organic Carbon to Total Nitrogen

Df Degrees of freedom

EEP Eastern equatorial Pacific

ENSO El Nino Southern Oscillation

GRA Gamma ray attenuation

HCl Hydrochloric Acid

H2O2 Hydrogen Peroxide

IODP International Ocean Discovery Program

IP25 Ice proxy with 25 carbon atoms

IRD Ice-rafted debris

LGM Last Glacial Maximum

MIS Marine Isotope Stage

NaOH Sodium Hydroxide

NGRIP North Greenland Ice Core Project

Psu Practical salinity units

vii

RS Resting spores

SHMP Sodium hexametaphosphate

SST Sea surface temperature

Sv Sverdrup

TOC Total Organic Carbon

WEP Western equatorial Pacific

viii

ACKNOWLEDGMENTS

My time at Iowa State University has been one of the greatest experiences I have ever had. I would first like to thank the Department of Geological and Atmospheric Sciences for giving me the opportunity to pursue my Masters. Prior to this experience, I was working as a lab technician and was trying to find a program to help refine my research interests and provide me with the knowledge and skillset to investigate those interests. I am very grateful to Dr. Beth Caissie who has provided me with so much support and knowledge throughout the course of this research. Without Beth, this research would not be possible. I would also like to offer my appreciation to my committee members Dr. Alan Wanamaker and Dr.

William Gutowski. I have taken several classes from both Dr. Gutowski and Dr. Wanamaker.

Those classes have contributed quite a bit to this research and will continue to do so in my future research. A special thanks goes to the Stable Isotope Lab and Suzanne Ankerstjerne at

Iowa State University for helping me analyze my sediment samples for the geochemical data.

I would also like to thank the other current and former members of the Marine Sediments

Lab consisting of Anna Nesterovich, Natalie Thompson, Bailey Nash, Megan Wolff, Dana

Korneisel, Romina Holder, Sarah Broer, and the polardile. Finally, thanks to my family back at home and my breakfast club family at Crossfit West Ames that helped me keep my sanity whenever my work load started to pile up.

ABSTRACT

The current rapid reduction of sea ice in the arctic has motivated numerous studies to observe how sea ice declines during times of climate warming and its impact on marine ecosystems. Marine Isotope Stage (MIS) 5, the last interglacial prior to the Holocene, is characterized as having higher summer air temperatures and sea level compared to today; however, there is a scarcity of data for how sea ice extent and ecosystems changed during

MIS 5. The Umnak Plateau is not currently covered by sea ice due to the influence of the warm Alaskan Coastal Current entering through the eastern Aleutian Island passes; however, low-resolution studies from the Last Glacial Maximum (LGM) demonstrate that sea ice extended to the Umnak Plateau when sea level dropped and restricted flow of the Alaskan

Coastal Current over the Umnak Plateau. This study uses a multi-proxy approach consisting of grain size, diatom assemblages, and isotopic analyses to determine how environmental conditions changed at the Umnak Plateau (IODP Site U1339) during MIS 5 as well as the end of MIS 6 and the beginning of MIS 4 (146ka – 65ka), both of which are glacial periods.

The research presented in this thesis reveals that the glacials MIS 6 and MIS 4 are both characterized as having decreased primary productivity combined with increased nitrate utilization and increased terrestrial organic matter deposition, suggesting there may have been an extensive sea ice cover at the Umnak Plateau and a limited influence of the Alaskan

Coastal Current. In contrast, MIS 5 is characterized as having higher primary productivity combined with decreased nitrate utilization and decreased sea ice extent. MIS 5e, the warm substage of MIS 5 that has been correlated with the Eemian Interglacial from terrestrial records, shows that decreased productivity at the Umnak Plateau may be related to an

x intensified stratification associated with increased warming of the surface waters that resulted from increased insolation and a prolonged summer season. Comparing the stable nitrogen isotope (δ15 N) record with other sites in the North Pacific reveals noteworthy similarities between δ 15 N patterns during the warm substages of MIS 5 from the Umnak Plateau and from the Gulf of Alaska, the origin of the Alaskan Coastal Current. Thus, the δ 15 N record from the Umnak Plateau may be displaying changes relating to the source of the nitrates over the Umnak Plateau from a more western Bering Sea source during the cold substages of MIS

5 to being sourced from the Alaskan Coastal Current during parts of the warm substages of

MIS 5.

CHAPTER I

INTRODUCTION

1.1. Basis for Research

Sea ice extent in the Arctic and subarctic has been declining in the past few decades, reaching a minimum in November of 2012 since satellite records started in 1979 (National Snow and Ice Data Center, 2012). As sea ice continues to decline, the amount of reflected radiation also declines, leading to increased heating in the Arctic and increased sea ice melt, a process known as Arctic amplification (Serreze and Barry, 2011). If sea ice growth continues to decrease at the current rate, we may see sea ice-free summers by 2030 (Stroeve et al., 2012). This rapid reduction of sea ice in the Arctic, in combination with other environmental changes, has motivated numerous studies to observe how sea ice declines during times of climate warming and its impact on the marine ecosystem. Low-resolution data reveals that in the past 5 Ma, sea ice extent has been low or absent during interglacials whereas glacials have had sea ice extend as far south in the Bering Sea as the Umnak Plateau and Bowers Ridge (Takahashi et al., 2011;

Caissie et al., 2010; Katsuki and Takahashi, 2005). The research presented here explores how environmental conditions changed at the Umnak Plateau during Marine Isotope Stage (MIS) 5, the last interglacial prior to the Holocene, using several proxies including grain size, diatom analyses, and geochemical analyses. Data obtained from the Umnak Plateau is then compared to other sites in the Bering Sea and North Pacific that would have experienced different concentrations of sea ice.

The Umnak Plateau is located in the southeastern region of the Bering Sea, just north of

Bristol Bay (Figure 1). The southeastern Bering Sea currently supports a large ecosystem and is home to a variety of species of fish, crustaceans, molluscs, seabirds and marine mammals (Bhatt

2 et al., 2014). An increase in surface water temperatures and a decrease in sea ice extent could lead to a decline in annual primary productivity and a smaller spring phytoplankton bloom, both of which would greatly impact the southeastern Bering Sea ecosystem (Schumacher et al., 2002).

Reconstructing how primary productivity has changed during previous warm intervals can help us understand how primary productivity may change as the modern climate continues to warm.

We hypothesized that increased sea surface temperatures (SST) during the warm intervals of

MIS 5 would lead to an increase in surface water stratification with surface waters having much higher temperatures than the deeper waters below (Slagstad et al., 2011). This stratification would result in a shallower mixed layer and would limit nitrate exchange between the nitrate- poor surface waters and the nitrate-rich layers below, resulting in a large decrease in productivity

(Slagstad et al., 2011). The proxies that we used to test this hypothesis are explained in the next section.

1.2. Productivity, Nitrate Utilization, and Sea Ice Proxies

Diatom-based proxies have been used in previous studies to infer changes in sea ice extent and other environmental factors in the Bering Sea (i.e. Sancetta, 1983; Caissie et al.,

2010). The Bering Sea is an ideal location for this type of work due to the high concentration of dissolved silica in the North Pacific and seasonal sea ice coverage (Okazaki et al., 2005;

Niebauer et al., 1999). This research uses two such proxies as indicators of productivity: absolute diatom abundances and relative percent abundances of Chaetoceros resting spores (RS). Both are proxies for productivity and Chaetoceros RS are known to be one of the last diatoms to grow, requiring either a large amount of nutrients to sustain multiple diatom blooms throughout the year or an input of nutrients later in the year that allow for secondary blooms (Abrantes, 1988).

Primary productivity in the ocean is supported by both new and recycled nutrients, including nitrate (Stabeno et al., 2002; Fung et al., 2000). The ability of phytoplankton to take up these nitrates is referred to as nitrate utilization. As nitrate becomes depleted or other limiting factors increase (i.e. iron or light), nitrate utilization increases (Brunelle et al., 2007; Galbraith et al., 2008). This can be measured by δ 15 N, calculated using the following equation: δ 15 N =

15 14 15 14 {[( N/ N) sample /( N/ N) standard ] – 1} x 1000 and is reported in parts per thousand (‰; Sigman and Casciotti, 2001). Of of the two stable isotopes of nitrogen, 14 N and 15 N, phytoplankton preferentially take up 14 N due to weaker bonds associated with the lighter isotope (Altabet and

Francois, 1994). Heavy isotopes have extra neutrons, which make the bonds harder to break; therefore relatively more light isotope bonds ( 14 N) will be broken compared to heavier isotope bonds (15 N; Fry, 2006, and references within). As nitrates containing 14 N become depleted, phytoplankton increasingly take up nitrates containing 15 N. Incomplete nitrate utilization in the surface waters leads to lower δ 15 N values of the sinking particulate matter and higher δ 15 N values of the surface water nitrate pool. As nitrate utilization nears completion, δ 15 N of the particulate matter approaches the δ 15 N of the nitrate pool (Galbraith et al., 2008). Glacial periods in the

North Pacific typically have a high nitrate utilization reflected by high δ 15 N, whereas interglacial periods typically have low δ 15 N that is reflective of low nitrate utilization. High nitrate utilization during glacial periods in the Bering Sea most likely resulted from an increase in stratification promoted by sea ice coverage, which inhibited the mixing of nitrate-rich subsurface waters into the surface, combined with a high iron input (Brunelle et al., 2007; Galbraith et al., 2008;

Riethdorf et al., 2013, 2015).

Observing how both the biogenic and terrigenous sediment fractions change with respect to time can reveal changes in sea ice extent as well as biological productivity (Aiello and Ravelo,

2012). Terrigenous matter can be transported into the deep Bering Sea through aeolian and fluvial deposition or by sea ice rafting. Aeolian deposition is restricted to the vicinity of their respective source area and fluvial deposition, the bulk of which in the Bering Sea comes from the

Yukon and Anadyr rivers, mainly occurs on the eastern Bering Sea shelf (Riethdorf et al., 2013).

Sediment from the Yukon River, however, has also been found at the Meiji Drift in the northwest

Pacific (Van-Laningham et al., 2009) and it has been suggested that sea ice during glacial times transported the material from the eastern Bering Sea shelf (Riethdorf et al., 2013). As sea ice forms, several processes work to incorporate sediment into the ice, including tidal and wind mixing that promotes the resuspension of sediments, thermohaline mixing associated with brine rejection during rapid sea ice growth, anchor ice and beach ice formation, suspension freezing, and deposition of sediments onto ice cover via riverine or aeolian sources (Stein, 2008, and references therein; Nurnberg et al., 2011, and references therein). Sea ice coverage is typically identified based on increasing clay-sized terrigenous sediment and decreasing biogenic content

(Aiello and Ravelo, 2012). Additionally, sand- and silt-sized terrigenous particles can be interpreted as seasonal ice-rafted debris (IRD; Sakamoto et al., 2005).

1.3. Thesis Outline

The research presented here explores how environmental conditions changed at the

Umnak Plateau during MIS 5 using several proxies including grain size, diatom analyses, and geochemical analyses. The thesis is divided into three chapters and two appendices. Chapter One is an introduction to the reasoning behind and the proxies used in this research. Chapter Two is a literature review with detailed descriptions of Bering Sea oceanography, productivity, and the proxies used in this study. Chapter Three is a manuscript that will be submitted to the journal

Paleoceanography . It outlines the data collected during this research and includes a discussion of probable environmental conditions at the Umnak Plateau as the Bering Sea transitioned from

MIS 6 (~180-135 ka) to MIS 5 (~130-80 ka) and into MIS 4 (~75-60 ka).

The appendices that follow the three chapters discuss two short experiments performed to evaluate the validity of the data. In Appendix A, I examine whether diatom identifications between researchers varied. In Appendix B, I address if crushing the sediments impacts diatom counts.

CHAPTER TWO

LITERATURE REVIEW

2.1. The Bering Sea

The Bering Sea is a high-latitude sea bounded by Russia to the west, Alaska to the east, and serves as a gateway between the Pacific and Arctic Oceans, to the south and the north, respectively (Figure 1; Stabeno et al., 1999). The eastern side of the Bering Sea is characterized by a broad, shallow continental shelf more than 500 km across, whereas the western side has a much smaller shelf that extends less than 200 m. The deepest basins with a maximum depth of

3,500 m, are also in the west (Stabeno et al., 1999). Several bathymetric highs span the deep basins in the form of ridges, including Shirshov Ridge (750 km long) in the western Bering Sea and Bowers Ridge (700 km long) in the southern Bering Sea, extending north from the Aleutian

Arc. The Umnak Plateau in the eastern Bering Sea is separated from the Bering Sea shelf by the

Bering Canyon (Normak and Carlson, 2003).

The net flow of water in the Bering Sea is from south (Pacific Ocean) to north (Arctic

Ocean). Flow from the Pacific is restricted to the passes in the Aleutian Island Chain, and flow into the Arctic is restricted to the Bering Strait (Stabeno et al., 1999; Niebauer et al., 1999). The shallow Bering Strait (~50 m deep) is responsible for the northward annual transport of approximately 0.8 Sv (Sv = one million cubic meters per second) due to an approximate 0.4 m mean sea level difference between the Bering Sea and the Arctic Ocean (Niebauer et al., 1999;

Coachman, 1993). Tides that enter the Bering Sea from the Pacific Ocean mix the bottom 40 m of water over the southeastern shelf, which provides the shelf with relatively warm saline water and nutrients that allows for high levels of primary production (Stabeno et al., 1999).

2.1.1. Bering Sea Circulation

A large-scale cyclonic surface circulation pattern exists in the Bering Sea bounded by the

East Kamchatka Current in the west, and by the Bering Slope Current and Aleutian North Slope

Current in the east (Figure 1; Stabeno et al., 1999). Bering Sea circulation during glacial periods would have been significantly different because of the closing of the Bering Strait and eastern

Aleutian Island passes in the Aleutian Arc due to continental ice-sheet growth (Takahashi, 1998).

The Aleutian North Slope Current connects the Amukta and Amchitka Passes with the Bering

Slope Current and is therefore strongly influenced by flow through these passes (Stabeno et al.,

1999). Baroclinic transport estimations for the East Kamcahtka Current, the Bering Slope

Current, and the Aleutian North Slope Current are southward with a transport of 7-15 Sv, northwestward with a transport of 3-6 Sv, and northeastward with a transport of 3-5.5 Sv, respectively (Stabeno et al., 1999). Northward transport of Bering Sea water through the Bering

Strait does not impact circulation in the Bering Sea; however, it is important both for the Arctic

Ocean circulation and for circulation on the northern Bering Sea shelf where cold, saline water is produced during ice formation (Stabeno et al., 1999).

The relatively warm Alaskan Coastal Current (ACC) and Alaskan Stream (AS), an extension of the Alaskan Coastal Current, enters the Bering Sea in the south through 14 main passes in the Aleutian Arc and strongly influences the cyclonic gyre circulation in the Bering Sea

(Figure 1; Stabeno et al., 1999; Ladd et al., 2005). Deep mixing associated with the inflow of both the ACC and the AS through these passes allows for the increased nitrate concentrations; however, the ACC itself is nitrate-poor and the AS is nitrate-rich (Ladd et al., 2005). The low nitrate concentrations of the ACC is due to from the accumulation of runoff along the British

Columbia Coast and the consumption of nitrate by phytoplankton in the surface waters (Royer,

1981; Ladd et al., 2005). Both the ACC and the AS flow from the head of the Gulf of Alaska southwestwards along the Alaska Peninsula, extending 1,000 km along the Gulf of Alaska coast

(Stabeno et al., 2002). Part of the ACC flows through the Unimak, Akutan, and Samalga Passes in the eastern Aleutian Islands, whereas part of the AS flows through central and western

Aleutian Island passes (Ladd et al., 2005). The total inflow through the 14 Aleutian Island passes is balanced by the outflow of surface water into the North Pacific through the Kamchatka Strait, located in the southwestern Bering Sea (Stabeno et al., 1999). The Kamchatka Strait is the only site deep enough (> 2,000 m) to allow Deep Pacific Water to flow into the Bering Sea (Schlung et al., 2013). This water mixes with the surrounding waters to form Bering Sea deep water, which lies between 2,000 and 3,000 m in depth and contains the highest concentrations of silica found in the world’s ocean (> 225 mM/L; Reed et al., 1993; Stabeno et al., 1999). In contrast, the eastern Aleutian Islands Passes are much shallower, with the Unimak and Akutan Passes having sill depths less than 100 m (Ladd et al., 2005). Since these eastern passes are shallower than the western passes, decreased sea levels during glacial periods would impact the flow of the ACC through these passes.

Transport through the eastern passes is northward and predominantly baroclinic due to sea level changes that result from winds along the southern side of the Alaska Peninsula

(Schumacher et al., 1982; Ladd et al., 2005). Net baroclinic transport through the Unimak Pass reaches a maximum in the fall and winter when winds are at their strongest (>0.50 Sv) and a minimum in the spring and summer (~0.10 Sv; Stabeno, 1999; Schumacher et al., 1982).

Interaction with the Bering Canyon topography, found along the Aleutian Islands near Unimak

Pass, facilitates shelf-slope exchange and leads to upwelling (Schumacher and Stabeno, 1998).

As this flow continues northward, it intensifies along the east coast of Siberia and forms the

Anadyr Current, which eventually exits the Bering Sea through the Bering Strait (Stabeno,

1999). Mesoscale eddies form within this northward flow, shoaling subsurface waters into the euphotic zone and enhancing biological productivity (Stabeno, 1999; Chavez et al., 2011).

2.1.2. Atmospheric Patterns in the Bering Sea

The Bering Sea’s weather is dominated by surface pressure systems that change seasonally. In the summer, the Bering Sea is bounded by the northern portions of a high-pressure system over the North Pacific and by a low-pressure system over Asia, resulting in weak pressure gradients and winds (Niebauer et al., 1999). In the winter, a low-pressure system dominates the North Pacific in the form of the Aleutian Low (low atmospheric pressure centered on the Aleutian Island Chain; Niebauer et al., 1999). As a result of the pressure gradients formed in the winter, there is an intensification of winds by at least an order of magnitude greater than in the summer (Niebauer et al., 1999; Schumacher et al., 1982). Associated with this increased wind stress is the increased transport of heat, momentum and moisture into the Bering Sea

(Niebauer et al., 1999).

Variability in the Aleutian Low can be linked with Pacific- and global-scale variability, including the El Nino-Southern Oscillation (ENSO) farther south in the Pacific (Niebauer, 1988).

Trades winds in the equatorial Pacific normally blow from the high-pressure area over the eastern equatorial Pacific (EEP) to the normally stable low-pressure area over the western equatorial Pacific (WEP); however, these pressure systems change places at irregular intervals of roughly three to eight years (Garrison, 2005). The trade winds usually drag large volumes of water from the EEP westward, resulting in the accumulation of warm water in the WEP and the upwelling of cooler, deep-water in the EEP to replenish the water lost to the WEP (Garrison,

2005). ENSO refers to the changes in sea level pressure when high pressure builds in the WEP and low pressure dominates the EEP, resulting in a weakening or reversal of the trade winds in the EEP. This change in the trade winds reduces coastal upwelling in the EEP, which results in the reduced transport of cooler deep water to the surface waters and in increased SST. The warmer water causes more evaporation and the low-pressure system over the EEP intensifies, leading to increased precipitation in normally dry areas. In contrast, the WEP experiences a decrease in SST and a decrease in precipitation (Garrison, 2005). A period of SST warming in the eastern equatorial Pacific is known as El Nino and a period of SST cooling as La Nina.

During La Nina events, the EEP cools and the WEP warms rapidly, resulting in stronger currents, increased upwelling in the EEP, and colder-than-normal events in the EEP (Garrison,

2005).

El Nino events in the Northern Hemisphere have modified the Aleutian Low, pulling it slightly southeastward of its normal position in winter months. This has led to both stronger and increased numbers of cyclonic storms that result in a poleward flux of warmer air that warms the

Bering Sea (Niebauer, 1988). Storms enhance sediment reworking and resuspension as well as mixing of the water column (Riethdorf et al., 2013). Conversely, La Nina events involve the enhancement of the normal east to west pressure gradient in the tropical Pacific, which leads to stronger trade winds and is associated with a less intense Aleutian Low that has moved westward of its normal position. This has led to increased anticyclonic storms with an equatorward flux of colder air, resulting in a cooling of the Bering Sea (Niebauer, 1988). A strong Siberian High in the winter also decreases sea surface temperatures (SST) and results in greater sea ice formation and vertical mixing in the mixed layer.

2.2. Sea Ice Formation and Transport

Sea ice is absent from the Bering Sea for about half of the year and either starts to form in the Bering Sea or is transported from the Arctic Ocean and through the Bering Strait in

November (Niebauer et al., 1999). Sea ice usually reaches its maximum extent in March or early

April and annually covers about 37-56% of the Bering Sea, with the heaviest ice concentrations found in the west. Sea ice begins to retreat in April and eventually reaches a minimum extent in

September. Sea ice is currently absent from the Umnak Plateau due to the influx of the warm

ACC; however, substantial sea ice coverage existed during the Last Glacial Maximum (LGM) when sea level was about 100 m lower than today (Caissie et al., 2010). This lower sea level would have restricted the ACC from flowing through the Unimak and Amukta passes, although warm Pacific water could still enter through the deeper passes in the central and western

Aleutian Island passes, including Amchitka Strait (Stabeno et al., 1999).

The formation, advance and retreat of sea ice is influenced by both oceanic and atmospheric factors. Sea ice commonly forms over the continental shelves of the Bering Sea. The production and transport of sea ice depends on which storm track dominates in the winter, with the lowest production of sea ice occurring when the Aleutian Low is intensified (Stabeno et al.,

1999). High frequencies of winter cyclones are correlated with a reduced extent of sea ice due to the northward push of these storms that moves sea ice to the north and brings in warm Pacific waters from the south (Overland and Pease, 1982). Ice formation along coasts occurs in polynyas, which are regions of open water that result from the actions of winds or currents. As sea ice forms, seawater salinity increases (>34 psu) due to brine rejection (Niebauer et al., 1999;

Stabeno et al., 1999). This dense, higher salinity water sinks and results in a low-salinity surface layer and a thick halocline with limited vertical mixing (Warren, 1983). As ice is transported

12 southward, it eventually reaches warmer waters and melts, decreasing surface salinity to values around 30 psu (Stabeno et al., 1999). This progression of sea ice melting leads to the cooling and freshening of the surface waters and advances the ice edge further south (Niebauer et al., 1999).

As sea ice expands southward, regional albedo increases and alters the atmospheric heat budget.

Since the Aleutian Low is sensitive to changes in the latitudinal temperature gradient, this southward ice expansion could force the Aleutian Low westward and into a less intense state, which would allow further sea ice expansion (Addison et al., 2012; Rind, 1998). Whenever sea ice formation is delayed, possibly due to warmer winter temperatures, the cooling and freshening of the surface waters is also delayed. This then leads to a positive feedback loop where this lack of sea ice formation allows relatively warm North Pacific surface waters to move northward and contribute to the decreasing sea ice extent (Serreze et al., 2007).

Terrigenous grains that become incorporated into sea ice and icebergs are known as ice- rafted debris (IRD). IRD found in seasonal sea ice is generally fine grained, composed of quartz and clay minerals and contains occasional dropstones that appear as well-rounded pebbles, which can be 3-5 cm in size (Riethdorf et al., 2013). Anchor ice, in contrast to seasonal sea ice, can incorporate higher percentages of larger sized grains (Aiello and Ravelo, 2003). As ice melts in the summer, the entrained sediment is released, with the coarser sediments being released sooner than the finer sediments (Reimnitz et al., 1998). Lowered sea level during glacials would have exposed greater expanses of continental shelves, which may have led to more IRD incorporation

(Sakamoto et al., 2005). Besides transportation by sea ice, terrigenous matter can also be transported into the Bering Sea through aeolian and fluvial sources (Riethdorf et al., 2013).

Terrigenous materials transported by wind in the North Pacific are restricted to the vicinity of their respective source areas (Riethdorf et al., 2013). The major rivers of the Bering Sea are the

Yukon, Kuskokwim, and Anadyr rivers, with sediment loads from the rivers being 60, 8, and 2 million tons of suspended sediment, respectively (Riethdorf et al., 2013; Brabets et al., 2000;

VanLaningham et al., 2009). The eastern Bering Sea shelf and slope sediments largely consist of sand-sized sediment from the Yukon River whereas clay- and silt-sized particles are restricted to the northeastern shelf. This suggests that the clay- and silt-sized particles are supplied directly from the Yukon River and the sand-sized particles are reworked and transported by storm surges to the outer continental shelf (Nagashima et al., 2012).

2.3. Primary Productivity

The Bering Sea is one of the most productive regions in the world, dominated by diatom blooms in the spring and summer due to high silica concentrations (Okazaki et al., 2005).

Accumulation of diatoms and other planktonic organisms on the sea floor results from these organisms being able to outgrow both their microbial competitors and macrozooplanktonic predators (Chavez et al., 2011). High silica concentrations in the Bering Sea result from many inputs, including: riverine input, hydrothermal and volcanic releases, low temperature submarine weathering, glacial weathering and sediment pore-water dissolution (DeMaster, 1981). Siliceous tests of diatoms and other siliceous organisms (i.e. radiolarians, silicoflagellates, and siliceous sponges) dissolve in the water column once the organisms die and sink, enriching the deeper waters with dissolved silica (DeMaster, 1981). Only 2 to 4% of biogenic silica produced in the oceans is preserved in the sediment with the highest preservation occurring in areas of high siliceous organism productivity (Heath, 1974). The Bering Sea deviates from North Pacific trends with regards to deep water silicate concentrations. In the North Pacific, silicate decreases below 2 km, but in the Bering Sea, silicate increases, reflective of the active silicate regeneration

14 in the deep (Tsunogai et al., 1979). The advection of silicate to the surface waters is through vertical eddy diffusion and upwelling. Regenerated deep sea silicate constitutes about 10% of the silicate taken up by phytoplankton in the surface (Tsunogai et al., 1979). Calcareous organisms are not well preserved in the sediment of the Bering Sea due to the shallow lysocline; therefore, any paleoclimate reconstruction typically associated with these organisms is limited.

Primary productivity in surface waters of the Bering Sea increases from the deep basin to the shelf break and decreases from the shelf break to the coast (Anguilar-Islas et al., 2007). This productivity gradient is due to the concentrations of various nutrients in the water column and the exchange of nutrient-rich deep waters with nutrient-poor surface waters. Nutrients are at a maximum in the late winter when the Bering Sea is covered by sea ice, radiation is low, and there is vertical mixing in the water column (Sigman et al., 2004). Productivity peaks when water column stratification is low, increasing the advection of nutrient-rich deep water into the photic zone. Processes that transport nutrient-rich deep water into the photic zone include wind-driven upwelling and density-driven overturning (Fung et al., 2000). As mentioned earlier, decreases in sea surface temperatures during glacial periods increase the sensitivity of the water column to changes in salinity; therefore, any large inputs of freshwater would increase the stratification of the water column and decrease the vertical transport of nutrients (Sigman et al., 2004; Brunelle et al., 2007). Besides limitations associated with nutrient input, phytoplankton growth can also be limited by light (Grebmeier, 2012) and micronutrients, such as iron (Anguilar-Islas et al., 2007;

Boyd et al., 2004).

The annual production cycle of the southeastern Bering Sea begins with under-ice blooms when insolation is strong enough to reach the base of the ice and induce photosynthesis

(Alexander and Niebauer, 1981). Spring ice retreat introduces meltwater to the water column and

15 results in an intensified stratification, trapping nutrients in the surface layer. The phytoplankton that consume these nutrients then sink to the sea floor while the unused nutrients may be recycled in the water column (Alexander and Niebauer, 1981). If enough nutrients are available to be recycled to the surface waters, or if there is a sufficient amount of nutrients left over from the initial spring bloom, another bloom can occur in the summer due to the weakening of the seasonal thermocline and the action of winds mixing the water column (Stabeno et al., 2002;

Alexander and Niebauer, 1981). Sea ice retreat during deglaciations would have had a similar effect in creating sufficient stratification to trap nutrients in the surface layer, which would result in enhanced productivity (Takahashi, 2005).

Primary productivity in the open ocean is supported by both new and recycled nutrients

(Stabeno et al., 2002; Fung et al., 2000). About half of the nutrients required to support production must be replenished annually (Stabeno et al., 2002). Two sources that are used to resupply nitrogen to the surface waters are the nitrate-rich waters of the deep Bering basin and nitrogen remobilized from the sediments in the form of ammonia and nitrate (Morales et al.,

2014). As particulate organic matter falls below the euphotic zone, it starts to decay and is converted back into inorganic nutrients (Chavez et al., 2011). The Bering Sea deep basinal waters are nitrogen rich (45-46 µM/L at 500 m; Lehmann et al., 2005) due to the influx of organic material from both the surface waters above and the lateral advection of material from the continental shelves. The exchange of water between the iron-rich coastal region and the nitrogen-rich deep basinal waters then leads to high seasonal productivity on the outer shelf

(Addison et al., 2012).

Nutrient recharge begins in the fall and peaks in the winter, when vertical and horizontal transport of these renewed nutrients is possible due to decreased vertical stratification,

16 weakening of hydrographic fronts, and increasing wind strength (Morales et al., 2014). The amount of nutrients mixed up into the surface waters is dependent on the winter mixing depth, which is influenced by both winds and stratification (Chavez et al., 2011). Unlike the outer shelf, the inner Bering Sea shelf is not influenced as significantly by the exchange of nutrient rich waters due to its distance from the shelf break. Instead, nutrients on the inner shelf are remobilized and vertically redistributed directly from the shelf sediments (Morales et al., 2014).

Another source of replenishment includes the transport of water through the Aleutian Island

Passes. Spring flow of water through the Unimak pass contains enough nutrients to support phytoplankton production with nitrate concentrations ranging from 5-14 µM (Stabeno et al.,

2002).

2.3.1. Nitrogen Limitation

Nitrogen is one of the most limited nutrients in the Bering Sea, with its lowest concentrations along the coasts as a result of river runoff and extensive productivity there

(Sukhanova et al., 2008; Whitney et al., 2005). It may have been even more limited during glacial periods due to a highly stratified water column (Brunelle et al., 2007). Nitrogen fixation is the major input of biologically available nitrogen into the ocean and is performed by nitrogen fixers (cyanobacteria and other microorganisms able to convert N 2 into biomass N) in the surface waters. These nitrogen fixers are then remineralized, supplying molecular nitrogen to the dissolved nitrogen pools in the water column (Sigman and Casciotti, 2001).

Nitrogen in the surface waters is primarily lost through denitrification, which is the process by which bacteria reduce nitrate to molecular nitrogen. Denitrification can occur both in the water column and in the sediments when oxygen concentration is low (Sigman and Casciotti,

2001). The decay of organic material into inorganic nitrogen and other nutrients uses oxygen; therefore, when oxygen is low, decay continues with nitrate as the electron acceptor (Chavez et al., 2011). The balance of nitrogen fixation and denitification appears to be in phase over glacial- interglacial timescales with both processes being reduced during glacial conditions (Galbraith et al., 2004). Nitrification, the oxidation of ammonium, in sediment can replenish nitrates into the water column (Sigman and Casciotti, 2001). Ammonium in the sediment is generated by the ammonification of sinking organic material (Lomstein et al., 1989).

2.3.2. Iron Limitation

Nitrogen fixation by phytoplankton in the subarctic Pacific Ocean can be largely limited by iron availability (Anguilar-Islas et al., 2007; Boyd et al., 2004; Chavez et al., 2011). Increases in bioavailable iron could, therefore, lead to large productivity blooms. Modern subarctic regions receive iron mostly through aerial deposition (Martin and Gordon, 1988), volcanic ash (Hamme et al., 2010), and submarine lateral transport from continental margins (Lam et al., 2008).

Despite rivers being a significant source of iron at the coast, the high demand of the coastal phytoplankton for the riverine iron makes the iron inaccessible to phytoplankton in the open ocean (Fung et al., 2000). Mineral aerosols can be transported through the atmosphere by prevailing winds and transported vertically by both convective processes and adiabatic vertical motion accompanying frontal systems (Mahowald et al., 2005). The mineral aerosols that are transported by aerial deposition consist of soil particles picked up by strong winds over erodible surfaces (Mahowald et al., 2005). Of these aerosols, only silt- and clay-sized particles are able to be transported aerially long distances and contribute most of the soluble iron to the open ocean

(Fung et al., 2000; Mahowald et al., 2005). The iron content of the aerosols is dependent on the

18 mineralogy of the source material and of that iron content. Only 0.5% of this type of iron is soluble and bioavailable for primary productivity (Fung et al., 2000). Additionally, ice cover can disrupt the flux of atmospheric iron into the subarctic and Arctic oceans (Tovar-Sanchez et al.,

2010). Material that collects on multiyear ice has been found to contain high amounts of iron, along with other metals and nutrients, which would greatly increase productivity if released

(Tovar-Sanchez et al., 2010). High concentrations of iron have also been associated with IRD incorporated during ice formation (Measures, 1999).

The relationship between iron and productivity can be seen in glacial and interglacial cycles with glacial intervals having higher iron input through atmospheric dust compared to interglacial periods (Martin, 1990). Possible mechanisms for this increased input include stronger surface wind speeds, increased aridity, and lower sea levels exposing erodible continental shelves (Mahowald et al., 2005). Iron concentrations also increase in surface waters during sea ice retreat due to the release of iron during ice melt (Tovar-Sanchez et al., 2010). It has been suggested that iron played only a secondary role in productivity during the recent deglaciation in the North Pacific Ocean, with sequential events of the deepening of the mixed layer through deep convection followed by the stratification of the water column by meltwater input being the main drivers for changing deglaciation production (Lam et al., 2013).

2.4. Paleoclimate Proxies

2.4.1. Nitrogen Isotopes

Nitrogen has two stable isotopes, 14 N and 15 N with 14 N comprising 99.63% of nitrogen found in nature (Sigman and Casciotti, 2001). The ratio of these isotopes ( 15 N/ 14 N) in sediment samples can be compared against a universal standard (atmospheric nitrogen, equation 1; Sigman

19 and Casciotti, 2001). Deviations from this standard are expressed in δ notation and is reported as parts per thousand (‰). δ 15 N can be used to identify changes in surface nitrate utilization, assuming changes in the δ 15 N of the nitrate pool are negligible and the isotope effect for nitrate assimilation is constant (Brunelle et al., 2007).

15 15 14 15 14 δ N = {[( N/ N) sample /( N/ N) standard ] – 1} x 1000 (Equation 1)

δ15 N values of nitrate vary regionally, ranging from 2 to 20 ‰ due to the effects of nitrate assimilation and denitrification (Sigman and Casciotti, 2001). Two sources of nitrogen input into the surface waters, nitrogen fixation and terrestrial runoff, have been shown to produce δ15 N values of -1‰ and 0‰, respectively (Sigman and Casciotti, 2001). Oceanic phytoplankton preferentially assimilate 14 N-bearing nitrate over 15 N-bering nitrate, resulting in lower δ 15 N values (Altabet and Francois, 1994). Elevated sediment δ15 N values can be found in areas where nitrate is completely consumed, such as high latitude regions with high iron concentrations

(Sigman and Casciotti, 2001). δ15 N values of sediments in these regions will primarily record the degree of nitrate consumption by phytoplankton assimilation and how nitrate supply to the surface waters has changed (Sigman and Casciotti, 2001; Brunelle et al., 2007). Nitrate utilization can be increased by either inhibiting nitrate resupply to the euphotic zone or by increasing the ability of the phytoplankton to take up the nitrate in the mixed layer (Galbraith et al., 2008).

Denitrification of marine sediments during warm stages has also been shown to cause very little isotopic enrichment of δ 15 N (~0.5‰), similar to what is observed today. This is attributed to a limited supply of nitrates to denitrifying organisms in denitrification zones where

20 nitrate has been completely consumed (Lehmann et al., 2005). This low isotopic enrichment associated with denitrification changes in δ15 N should therefore not mask the larger isotopic variation expected from the change in nitrate utilization (Brunelle et al., 2007). Nitrification, as opposed to denitrification, generates nitrates of low δ 15 N (Granger et al., 2011).

Diagenesis, in this context, is the combination of chemical, biological and physical processes that alter the quantity and composition of the organic matter in marine sediments

(Henrichs, 1992). Diagenesis begins in the photic zone and continues as the organic matter sinks and at the surface layer of the sediments (Meyers, 1994). At the sediment-water interface, 30% to

99% of the organic matter that has accumulated on the sea floor is decomposed. The oxidation rates of organic matter tend to increase when either temperatures increase or the accumulation rate of the organic matter increases (Henrichs, 1992). Diagenetic alteration of δ 15 N is more pronounced in deep-sea sediments than in continental slopes (Galbraith et al., 2008). One method to gauge the degree of diagenesis on the δ 15 N is to measure the δ 15 N of organic matter within diatom frustules (Galbraith et al., 2008; Brunelle et al., 2007). Similar values between the diatom bound δ 15 N and the bulk δ 15 N would indicate that diagenesis is not the primary factor in the variability of δ 15 N of the bulk sediment (Galbraith et al., 2008).

2.4.2. Carbon Isotopes

Stable carbon isotopes (δ13 C) and the C/N ratio reflect changes in the origin of the organic matter, whether it is terrestrial or marine. Calculating δ 13 C is very similar to calculating

δ15 N with the replacement of 13 C for 15 N and 12 C for 14 N. δ 13 C values also indicate type of photosynthetic pathway used (C 3 vs. C 4 vs. CAM). Marine algae use inorganic carbon from

13 dissolved bicarbonate, which has a δ C value of about 0‰ (Meyers, 1994). Plants using the C3

21 pathway (i.e. plants that grow in cool, wet environments) discriminate against the heavier

13 13 isotope, C, and produce δ C values between -26 to -28 ‰ (Meyers, 1994). The C3 pathway, also known as the Calvin cycle, uses the enzyme Rubisco to catalyze the carboxylation of ribulose-1,5-bisphosphate into 3-phosphoglycerate and 2-phosphoglocolate (Gowik and

Westhoff, 2011). 3-phosphoglycerate is then processed through photorespiration and leads to a net loss of carbon dioxide. When conditions become unfavorable (i.e. higher temperatures and dryness), more carbon dioxide will be lost and would result in a less efficient photosynthetic pathway. Plants using the C 4 photosynthetic pathway have a modified C 3 photosynthetic pathway that involves an adaptation to high light intensities, high temperatures and dryness. This pathway allows for carbon dioxide to be concentrated at the site of Rubisco and results in a more efficient pathway and a better nitrogen-use efficiency compared to the C 3 plants (Gowik and Westhoff,

13 2011). Plants using the C4 (i.e. plants growing in hot, arid climates) pathway producing δ C values with an average of -14‰ (Meyers, 1994). Marine organic matter typically has δ 13 C values between -22 and -20‰ (Meyers, 1994). Coastal regions complicate the isotopic signal in that much of the organic material comes from a variety of sources, including algae, and both C 3 and

C4 plants. C/N can additionally be used to distinguish between algal and land-plant sources of organic matter. Algae C/N values typically fall between 4 and 10, whereas vascular land plants have values greater than 20 (Meyers, 1994).

2.4.3. Grain size

Sediments recovered from the Umnak Plateau (Site U1339; Figure 1) during expedition

323 are composed of three components: biogenic (mainly diatom frustules with varying proportions of calcareous nannofossils, foraminifers, silicoflagellates and radiolarians),

22 volcaniclastic (mainly fine ash), and siliciclastic (clay- to pebble-sized clasts; Expedition 323

Scientists, 2011). Volcaniclastic particles decrease in abundance as distance from volcanic sources, mainly the Aleutian Arc, increases (Aiello and Ravelo, 2012). The most abundant terrigenous grain types include silt-sized quartz and feldspar, clay, mica and rock fragments

(Expedition 323 Scientists, 2011). Seasonal IRD related to sea ice melting in the spring consists of sand- and silt-sized terrigenous particles (Sakamoto et al., 2005). Gravel- to pebble-sized clasts are interpreted as IRD delivered through melting sea ice or icebergs. Clay-sized particles include clay minerals and fragments of siliceous organisms and sand-sized particles are composed of whole diatom valves, siliclastic material, and other biogenics including foraminifera, sponge spicules and fecal pellets (Aiello and Ravelo, 2012). The sediment distribution in the Bering Sea is as follows: mainly terrigenous on the Bering Sea shelf, mixed biogenic and terrigenous on the Bering Slope, and mainly biogenic sediments in the basins and submerged ridges (Aiello and Ravelo, 2012). Diatom valves are most abundant at the southernmost Bering Slope site from the Umnak Plateau (Site U1339; Figure 1), and the least abundant at one of the northernmost Expedition 323 Sites, U1344 (Pervenets Canyon; Aiello and

Ravelo, 2012).

Glacial sediments are characterized as having high abundances of clay-sized siliciclastic grains, reflecting increased terrigenous supply when sea level was low, and poorly preserved diatom valves with silt-sized centric diatoms (Aiello and Ravelo, 2012). Exposed shelf areas during glaciations and lower sea level could have moved the deposition of sediment from the shelf to the basins, creating turbidite sequences (Aiello and Ravelo, 2012). Aleutian and

Kamchatka Basin turbidites show increased siliciclastic material input associated with the melting of sea ice and alpine glaciers (Gardner et al., 1982). Interglacial terrigenous input would

23 be derived mainly from riverine sources; however, the river sediment load would have been limited to the continental shelf (Aiello and Ravelo, 2012). Interglacial sediments are rich in biogenic components with higher abundances and larger particle sizes (i.e., more whole diatom valves) of both centric and pennate diatoms (Expedition 323 Scientists, 2011; Aiello and Ravelo,

2012). Laminated intervals in the sediment cores are correlated with an increase in primary productivity resulting in an increased coarse particle size fraction and the expansion of the oxygen-minimum zone (OMZ). They are often centimeter- and millimeter- scale bands that are differentiated from the rest of the core based on this increased biogenic content, which alters the appearance of the sediments (Aiello and Ravelo, 2012). Observing how both the biogenic and terrigenous sediment fractions change with respect to time can allow for the coupling of sea ice extent with biological production (Aiello and Ravelo, 2012). Increasing clay and decreasing biogenic content would be expected for increased sea ice coverage (Aiello and Ravelo, 2012).

2.4.4. Diatom-based Proxies

Besides the presence of IRD, sea ice conditions can be determined using several diatom- based proxies (Takahashi, 2005). Interglacials have generally recorded lower proportions of sea- ice diatoms and higher proportions of open water diatoms (Katsuki and Takahashi, 2005). Sites closest to the North Pacific consist of high abundances of subarctic diatom species (Takahashi et al., 2011). Neodenticula seminae at the Umnak Plateau has been associated with the inflow of the

AS (Katsuki and Takahashi, 2005). Coscinodiscus marginatus signifies water masses with high nutrient concentrations and low light intensities (Takahashi et al., 2011). Thalassiosira antarctica resting spores (RS), Fragilariopsis oceanica , and Fragilariopsis cylindrus are sea ice indicators and have been found at Umnak Plateau during the LGM (Katsuki and Takahashi,

2005) and deglaciations (Caissie et al., 2010). Chaetoceros is a lightly silicified diatom genus that has been associated with high productivity and upwelling. It blooms when nutrient concentrations are high enough to support multiple blooms throughout the year or when there is an additional pulse of nutrients later in the year that allows a second bloom to occur (Abrantes,

1988; Sancetta, 1982). Hyalochaete Chaetoceros species from refractors resting spores are commonly used as proxies for these highly productive periods (Sancetta, 1982; Caissie et al.,

2010).

2.5. Marine Isotope Stage 5 (180-135 ka)

Marine Isotope Stages are identified based on benthic foraminiferal oxygen isotope

(δ18 O) records. Marine Isotope Stage (MIS) 5 is divided into five substages (Shackleton, 1969).

These substages include three warmer substages (MIS 5a, 5c, and 5e) and two cooler substages

(MIS 5b and 5d). Substage 5e has been associated with the Eemian Interglacial documented in the terrestrial records; however, there is an offset between the peak warming on land compared to the signals in surface waters. This could be due to the suppression of surface water warming when cold fresh meltwater discharged into the ocean (Matthiessen and Knies, 2001). Northern hemisphere summer insolation during the Eemian was at a maximum and the cooling after the

Eemian is related to a decrease in insolation (Max et al., 2014; Cortijo et al., 1999; Berger and

Loutre, 1991). The Eemian is characterized by warmer sea-surface temperatures compared to today and by higher global eustatic sea level, several meters above the current sea level

(Shackleton et al., 2003; Matthiessen and Knies, 2001). Oxygen isotopes from the North

Greenland Ice Core Project (NGRIP) core shows an abrupt climate warming at 115 ka that correlates with MIS 5e and a slow temperature decline that marks the initiation of the last glacial

25 period (Andersen et al., 2004). MIS 5d to 5a in several North Atlantic marine and terrestrial records show six abrupt surface ocean cooling events (Oppo et al., 2006). Foraminiferal δ 18 O data from the North Atlantic indicate that the Atlantic Meridional Ocean Circulation (AMOC) may have weakened during the colder stages of MIS 5 (Max et al., 2014). The massive cooling events in the North Atlantic are attributed to massive glacial ice discharges, supported by a high input of IRD and a drop in δ13 C of the marine sediment.

The last interglacial has been proposed as being comparable to the Holocene, therefore climate variations during this time could provide a good analogue for future climate changes.

Unlike the late Holocene, where warming is influenced by anthropogenic greenhouse gas forcing, warming during MIS 5e was caused by changes in orbital forcing of insolation with

Earth’s orbital eccentricity more than doubled the modern value (Rohling et al., 2008; Berger and Loutre, 1991). Greenhouse gas levels during the last interglacial were comparable to pre- industrial Holocene levels (Petit et al., 1999). Positive feedback processes including increased greenhouse gases and decreased albedo, resulting from a reduced sea ice and snow extent and the expansion of shrub vegetation, would then serve to amplify the warming in the Arctic (Kienast et al., 2011).

2.5.1. Marine Isotope Stage 5 in Beringia

Circulation within the Bering Sea was considerably different during glacial periods than during interglacials due to lower sea levels and intensified continental ice sheets, both of which restricted the flow of North Pacific water through the Aleutian Passes and the flow of Bering Sea water out of the Bering Strait (Okazaki et al., 2005). This lower sea level also led to the exposure of the once submerged continental shelves, uncovering an area known as Beringia (Brigham-

Grette, 2001; Hopkins, 1959). Beringia is considered the shelf and terrestrial regions north of the

Aleutian Islands that expand out to the shelf-slope break in the East Siberia, Chukchi and

Beaufort Seas (Hopkins, 1959). During sea level lowstands, the Yukon and Kuskokwim Rivers may have meandered across the Bering shelf to the Bering Canyon (Carlson and Karl, 1984).

The transition from MIS 6 into MIS 5e led to increased sea levels that left behind large discontinuous sections of deposits representative of MIS 5e along the Alaskan coasts (Dinter et al., 1990). Coastal deposits along the Alaskan coasts determined to be from substage 5e were titled as the Pelukian transgression (Brigham-Grette and Hopkins, 1995; Hopkins, 1967). The preservation of the Pelukian shorelines suggests that sea level reached this position only once during the last interglacial (Brigham-Grette and Hopkins, 1995). The Pelukian shoreline is characterized as containing fossiliferous sand and gravelly beach deposits typically found up to

10 m above sea level (Kaufman et al., 1991). The presence of Northwest Pacific molluscs Natica janthostoma and Protothaca adamsi within these deposits suggests that water temperatures were warmer in the Bering and Chukchi Seas than at present (Hopkins, 1967). Radiolarian and diatom data from the North Pacific support this view of the expansion of warmer waters northward

(Sancetta and Silvestri, 1986; Morley et al., 1987). The winter sea ice limit in the Arctic Ocean has been postulated to have been 800 km north during the last interglacial than it is now and the

Arctic Ocean may have experienced sea ice-free summers (Brigham-Grette and Hopkins, 1995).

The larch-birch tree line also shifted 270 km north of its current position on the Siberian side of the Bering Sea, suggesting the expansion of boreal woodlands to northeast Siberia and to the modern coastline (Kienast et al., 2011). Pollen and spruce cones in Pelukian sediments of northern Alaska and the absence of frost cracks in Pelukian sediments south of the Bering Strait

27 additionally suggest a northward migration of treeline and warmer conditions during the last interglacial (Brigham-Grette and Hopkins, 1995).

The remainder of MIS 5 after the Eemian interglacial consisted of a cooling to more glacial conditions with minor warm intervals. Tide-flat mud from the Ahklun Mountains and the

Alaska Peninsula suggests a major glacial advance to southeast Alaska between 90 and 75 ka

(Kaufman et al., 1996). Flaxman Formation sediments deposited on top of the Pelukian deposits in low-lying areas of the Beaufort Coasts during MIS 5 glaciations have been found to range from 53 ka to 81 ka based on thermoluminescence age estimates and the uranium-series dating of a whale bone found to have the age of 75,000 years (Brigham-Grette and Hopkins, 1995; Carter and Ager, 1989; Dinter et al., 1990). The presence of a whale bone in the Flaxman Member along the north coast of Alaska indicates that sea level during MIS 5a was high enough to submerge the Bering Strait due to the collapse of a high-latitude ice cap over the Canadian islands, allowing seasonal whale migrations from the Pacific Ocean to the Beaufort Sea

(Brigham-Grette and Hopkins, 1995). The Bering Sea during MIS 5a could therefore have been more stratified compared to MIS 5e, possibly resulting from a larger supply of meltwater, and may have had temperatures as warm as MIS 5e (Matthiessen and Knies, 2001). Meltwater discharge into the Bering Sea is supported by the relative abundances of dinoflagellate cysts and the co-occurrence of IRD, lower δ18 O values and higher δ13 C values (Matthiessen and Knies,

2001).

2.5.2. Marine Isotope Stage 5 in the Bering Sea

Similar to terrestrial records, diatom-bound δ 15 N values from Bowers Ridge (Figure 1;

JPC 17) coincide with an abrupt warming at the transition from MIS 6 to MIS 5 followed by a

28 gradual cooling during the remainder of MIS 5 and transition into MIS 4 (Brunelle et al., 2007).

These changes in δ 15 N values were explained as reflecting changes in nitrate utilization between glacials and interglacials (Brunelle et al., 2007). Glacial periods have more complete nitrate consumption through phytoplankton assimilation, increasing the δ 15 N of the surface water nitrate pool and of the biomass. Additionally, glacials have stronger stratification in the upper water column, which impedes the replenishment of nitrates to the surface waters. Decreasing iron input when the water column was still stratified during the deglaciation would have led to iron limitations and decreased nitrate consumption, decreasing δ 15 N (Altabet and Francois, 1994).

During deglaciation, the retreat of alpine glaciers in the Alaskan Range and the Kamchatka

Peninsula would have introduced cold, low-salinity surface waters that would have prevented vertical mixing but still would have provided micronutrients, such as iron, from land (Katsuki and Takahashi, 2005).

Sediment cores collected from Shirshov Ridge (Figure 1; 85 KL) reveal low δ 15 N values during the warm substages of MIS 5 (MIS 5a, 5c, and 5e) that were not as clearly recognized in the Bowers Ridge record and high δ 15 N during both glacials (MIS 6 and 4) and the cold substages (MIS 5b and 5d; Riethdorth et al., 2015). C/N values throughout MIS 5 ranged from

10 to 15, indicating mostly terrestrial input of organic material (Riethdorth et al., 2015).

Deglaciation during the transition from MIS 6 to MIS 5 showed an abrupt 4°C rise in SST and a sharp decline in sea ice concentration (Riethdorth et al., 2013, 2015). High amounts of coarse material during this deglaciation supports the view of a warming ocean and decreasing sea ice concentrations. The coarse material is attributed to the sudden release of IRD in response to the rapid melting of perennial sea ice formed during MIS 6 (Riethdorth et al., 2013). Another potential source for this coarse material is glacial sourced rocks drained by the Yukon River

(Riethdorf et al., 2013; VanLaningham et al., 2009). Cold substages of MIS 5 on Shirshov Ridge were dominated by silt- and clay-sized terrigenous materials, considered to be representative of sea ice rafted debris (Riethdorth et al., 2013). In contrast to the cold substages, the warm substages showed increased marine productivity and decreased terrigenous matter supply

(Riethdorth et al., 2013). Increased productivity at this site during the warm substages is believed to be caused by enhanced vertical mixing of the water column. This occurred due to reduced sea ice extent and rising sea levels that enhanced the flow of the AS into the Bering Sea and induced vertical mixing (Riethdorth et al., 2013). Variations in SST during MIS 5a-5d in the sediment cores at Shirshov Ridge have been correlated to temperature fluctuations in NGRIP cores

(Andersen et al., 2004), which may suggest a common forcing between the northern Atlantic and

Pacific Oceans (Max et al., 2014).

Sediment δ15 N records in the western Subarctic Pacific (Figure 1; ODP 882 and MD

2416) and in the Gulf of Alaska (Figure 1; ODP 887) are nearly identical (Figure 1; Galbraith et al., 2008); however, the MIS 5 record between the two locations are different. This similarity may be driven by the lateral exchange of water within the North Pacific subpolar gyre that maintains a relatively homogenous δ 15 N within <0.5‰ (Galbraith et al., 2008; Ueno and Yasuda,

2003). This subpolar gyre consists of warm, saline water that is transported from the east of

Japan to the northern Gulf Alaska, before finally flowing into the Bering Sea as the AS (Ueno and Yasuda, 2003). Assuming that the homogeneity resulting from the exchange of waters through the gyre is the same during MIS 5 as it is currently, then changes in the δ 15 N could be due to changes in relative nitrate utilization and changes in the δ 15 N of the nitrate pool (Galbraith et al., 2008). The western sites lie beneath surface waters that are more nitrate rich than the Gulf

30 of Alaska and are closer to the East Asian source of aeolian iron (Duce and Tindale, 1991; Fan et al., 2006).

2.6. Application

The research presented in this thesis investigates how environmental conditions changed on the Umnak Plateau during MIS 5 applying the methods and proxies detailed above, including: stable isotopes, grain size and diatom identification. These records are then compared to the other MIS 5 records mentioned above to find any similarities that may indicate whether the changes we observe at the Umnak Plateau are exclusive to the site or if the changes occur regionally. The Umnak Plateau has the potential to provide a record that can reflect how the inflow of the ACC through the eastern Aleutian Island passes changes during MIS 5 (~130-80 ka), given its close proximity to the passes, and how far sea ice may have extended at the end of

MIS 6 (~185-135 ka) and at the beginning of MIS 4 (~75-60 ka).

1 31

Figure 1 . Map of the Bering Sea with surface currents (red arrows; Stabeno et al., 1999) and sites cited in this thesis.

These sites include: Umnak Plateau (dark purple dot, U1339, this study); Gulf of Alaska (yellow dot, ODP 887; Galbraith et al., 2008); Bowers Ridge (dark green, JPC 17; Brunelle et al., 2007); Shirshov Ridge (brown dot, 85KL; Riethdorf et al., 2013, 2015); Western Subarctic (light green dot, MD 2416 and light purple dot ODP 882; Galbraith et al., 2008).

MARINE ISOTOPE STAGE (MIS) 5 ON THE UMNAK PLATEAU, BERING SEA (IDOP SITE U1339): DIATOMS, GRAIN SIZE AND SEDIMENTARY NITROGEN ISOTOPES AS PROXIES FOR PRIMARY PRODUCTIVITY AND SEA ICE EXTENT

A paper to be submitted to Paleoceanography Derrick R. Vaughn 1, Beth E. Caissie 1

3.1. Abstract

The current rapid reduction of sea ice in the Arctic has motivated numerous studies to observe how sea ice declines during times of climate warming and investigate its impact on marine ecosystems. Marine Isotope Stage (MIS) 5, the last interglacial prior to the Holocene, is characterized as having higher summer air temperatures and sea level compared to today; however, there is a paucity of data for how sea ice extent and ecosystems changed during MIS 5.

The Umnak Plateau is not currently covered by sea ice due to the influence of the warm Alaskan

Coastal Current entering through the eastern Aleutian Island passes; however, previous work shows that sea ice has extended to the Umnak Plateau in the past when sea level was low and the passage of the Alaskan Coastal Current across the Umnak Plateau was restricted. This study uses a multi-proxy approach consisting of grain size, diatom assemblages, and isotopic analyses to determine how environmental conditions changed at the Umnak Plateau (IODP Site U1339) during MIS 5 as well as the end of MIS 6 and the beginning of MIS 4, both of which were glacial periods.

The record presented here reveals that glacials MIS 6 and MIS 4 are both characterized as having decreased primary productivity combined with increased nitrate utilization and increased terrestrial organic matter deposition, suggesting there may have been extensive sea ice cover at the Umnak Plateau and limited influence of the Alaskan Coastal Current. In contrast, MIS 5 is generally characterized as having higher primary productivity combined with decreased nitrate utilization and decreased sea ice extent. MIS 5e, the warm substage of MIS 5 that has been

33 compared with the Eemian Interglacial from terrestrial records, has decreased productivity at the

Umnak Plateau, which may be related to an intensified stratification associated with increased warming of the surface waters. This likely resulted from increased insolation and a prolonged summer season. During the warm substages of MIS 5, the δ15 N records from the Umnak Plateau and from the Gulf of Alaska, the origin of the Alaskan Coastal Current, are similar. Thus, the

δ15 N record from the Umnak Plateau may be displaying changes relating to the source of nitrates over the Umnak Plateau with an Alaskan Coastal Current source during parts of the warm substages of MIS 5 and a more western Bering Sea source during the cold substages of MIS 5.

3.2. Introduction

The Bering Sea is one of the most productive regions in the world, dominated by phytoplankton blooms in the spring and summer (Takahashi et al., 2002; Okazaki et al., 2005).

The timing and location of the spring bloom is related to seasonal sea ice extent and the melting of sea ice in the spring (Stabeno et al., 2001). Sea ice melt enhances light availability needed for primary production and traps nitrates in the surface waters by introducing freshwater to the upper layers of the water column and inducing stratification; however, increased stratification also limits the vertical fluxes of nitrates from deeper waters into the surface (Lam et al., 2013; Li et al., 2009). Additionally, phytoplankton productivity may be constrained due to a limited supply of micronutrients, such as iron (Martin, 1990; Anguilar-Islas et al., 2007; Boyd et al., 2004).

34 known as Arctic amplification (Serreze and Barry, 2011). If sea ice continues to decline at the present rate, the maximum extent in sea ice may be more than 500 km north than its current position by the year 2100 (Arzel et al., 2006). Regions that do not have sea ice may then potentially experience decreases in summer productivity owing to increased stratification associated with atmospheric warming and the related decrease in the vertical supply of nitrates in the water column (Slagstad et al., 2011). This rapid reduction of sea ice in the Arctic has motivated numerous studies to observe how sea ice declines during times of climate warming and its impact on the marine ecosystem, including Marine Isotope Stage (MIS) 5, the last interglacial prior to the Holocene.

MIS 5 (~180-135 ka) is divided into five substages based on benthic oxygen isotopic records, consisting of three warm substages (MIS 5e, MIS 5c, and MIS 5a) and two cold substages (MIS 5d and MIS 5d; Shackleton, 1969). MIS 5e is correlated with the Eemian

Interglacial from terrestrial records and has been characterized as having higher summer air temperatures and a higher sea level compared today (Brigham-Grette and Hopkins, 1995;

Shackleton et al., 2003; Matthiessen and Knies, 2001); however, there is a lack of data describing how far sea ice extended into the southeastern Bering Sea during MIS 5 and how surface water conditions changed as a result of the reduction in sea ice. Reconstructing how sea ice and surface water conditions changed during MIS 5 therefore presents an opportunity to reconstruct environmental changes in the Bering Sea during a previous warming that may be used as an analogue for future climate change. Unlike the late Holocene, where warming is influenced by anthropogenic greenhouse gas forcing, warming during MIS 5e was caused by changes in the orbital forcing of insolation with Earth’s orbital eccentricity more than doubled the modern value (Rohling et al., 2008; Berger and Loutre, 1991). The multi-proxy record

35 presented here gives a high-resolution centennial-scale perspective on changes in diatom productivity, nitrate utilization, and in sea ice extent during MIS 5 on the Umnak Plateau.

3.3. Study Area

3.3.1. Bering Sea

The Bering Sea is a high-latitude sea bounded by Russia to the west, Alaska to the east and serves as a gateway between the Pacific and Arctic Oceans (Figure 1; Stabeno et al., 1999).

The eastern side of the Bering Sea is characterized by a broader continental shelf (>500 km across). The western side has a smaller continental shelf (< 200m), but contains the deepest basins with a maximum depth of 3,500 m (Stabeno et al., 1999). The net flow of oceanic water is from the Pacific Ocean, through the Bering Strait, and into the Arctic Ocean. Inflow from the

Pacific Ocean to the Bering Sea is restricted to passes in the Aleutian Island chain (Stabeno et al., 1999; Niebauer et al., 1999). The relatively warm and fresh Alaskan Coastal Current (ACC) enters the Bering Sea in the south through passes in the Aleutian Island Arc and strongly influences the cyclonic gyre circulation in the Bering Sea, which is bounded by the East

Kamchatka Current in the west and by the Bering Slope Current and Aleutian North Slope

Current in the east (Stabeno et al., 1999). Bering Sea circulation during glacial periods would have been considerably different due to the related sea level drop and the closing of the Bering

Strait and the eastern passes in the Aleutian Island Arc (Takahashi, 1998).

Sea ice is absent from the Bering Sea for about half of the year and either starts to form in the Bering Sea or is transported from the Arctic Ocean via winter winds (Niebauer et al., 1999).

Sea ice usually reaches its maximum extent in March or early April and annually covers about

36 half of the Bering Sea, with the heaviest ice concentrations found in the west. Sea ice begins to retreat in April and eventually reaches a minimum extent in September.

3.3.2. Umnak Plateau

The Umnak Plateau is located in the southeastern region of the Bering Sea, just north of

Bristol Bay (Figure 1). Surface and intermediate water flow over the Umnak Plateau is derived from the relatively warm ACC that flows through the eastern Aleutian Island passes (Stabeno et al., 1999; Ladd et al., 2005). This region is not currently covered by sea ice at all due to the influence of the warm ACC; however, sea ice has reached the study area during glacial intervals

(Caissie et al., 2010; Katsuki and Takahashi, 2005). This suggests that the influence of the relatively warm ACC may have been reduced, perhaps due to a drop in sea level, which would have restricted the passage of the ACC through the eastern Aleutian Island passes into the Bering

Sea.

3.4. Background

Observing how both the biogenic and terrigenous sediment fractions change with respect to time can reveal the coupling of sea ice extent with biological production (Aiello and Ravelo,

2012). Phytoplankton productivity is typically low when terrigenous input is high, which can be reflected as sea ice transported grains (Riethdorf et al., 2013). During sea ice formation, several processes incorporate sediment into the ice, including tidal and wind mixing that promotes the resuspension of sediments, thermohaline mixing associated with brine rejection during rapid sea ice growth, anchor ice and beach ice formation, suspension freezing, and deposition of sediments onto ice cover via riverine or aeolian sources (Stein, 2008, and references therein; Nurnberg et

37 al., 2011, and references therein). Sand particles (terrigenous particles greater than 63 μm) can be incorporated into sea ice through these processes and are widely considered to be ice-rafted debris (IRD; Nurnberg et al., 2011; Sakamoto, 2005; Riethdorf et al., 2013). Besides transportation by sea ice, terrigenous matter can also be transported into the Bering Sea through fluvial and aeolian sources (Riethdorf et al., 2013). Terrigenous materials transported by wind in the North Pacific are finer in size compared to terrestrial material transported by sea ice and are typically restricted to the vicinity of their respective source areas (Riethdorf et al., 2013).

Sedimentary nitrogen isotopes (14 N and 15 N) and the ratio of these isotopes ( 15 N/ 14 N) can be compared against an universal standard (atmospheric nitrogen) to identify changes in surface nitrate utilization (Sigman and Casciotti, 2001). Deviations from this standard are expressed in δ notation. δ15 N is calculated using the equation and it reported in ‰ (parts per thousand):

15 15 14 15 14 δ N (= {[( N/ N) sample /( N/ N) standard ] – 1} x 1000)

δ15 N has been used as an indicator of nitrate utilization, which refers to the ability of phytoplankton to utilize the nitrates available to them. Of the two stable isotopes of nitrogen, phytoplankton preferentially take up the 14 N-bearing nitrate due to its lower bond strength

(Altabet and Francois, 1994; Fry, 2006). δ15 N increases as phytoplankton completely consume

14 N-bearing nitrate and increasingly consume 15 N-bearing nitrate (Sigman and Casciotti, 2001).

As nitrates needed for productivity become depleted or when environmental conditions become favorable for high productivity (i.e. increased iron input or increased light availability), nitrate utilization and δ 15 N increases (Brunelle et al., 2007; Galbraith et al., 2008). A δ 15 N of -1‰ and

0.5‰ has been proposed for nitrogen fixation and denitrification, respectively (Sigman and

Casciotti, 2001; Lehmann et al., 2005). Since the isotopic effects of these two processes result in

38 small δ15 N changes compared to changes in nitrate utilization, the δ15 N record may be considered a nitrate utilization record (Brunelle et al., 2007; Riethdorf et al., 2015). Using δ15 N as a nitrate utilization record also assumes that changes in the δ 15 N of the nitrate pool are negligible and the isotope effect for nitrate assimilation is constant (Brunelle et al., 2007). Modern nitrate utilization is incomplete in the North Pacific, a high nutrient-low chlorophyll region, due to low levels of iron (Tsuda et al., 2003). Glacial periods have a reduced supply of nutrients, such as nitrate, to the surface waters due to an intensified stratification, but a continuous supply of atmospheric iron results in a more complete nitrate utilization (Brunelle et al., 2007; Riethdorth et al., 2015).

Primary productivity is interpreted by using the number of diatoms present in a gram of sediment and from relative percent abundances of Chaetoceros resting spores (RS). Chaetoceros

RS are one of the last diatoms to grow in a year and require either a large amount of nutrients to sustain multiple blooms throughout the year or an input of nutrients later in the year that would allow for secondary blooms to occur (Abrantes, 1988).

3.5. Methods

3.5.1. Sediment Cores

Sediment cores used for this study were collected in 2009 from IODP Expedition 323 at the Umnak Plateau (Figure 1; Site U1339) using the advanced piston corer at a water depth of

1867 m. A total of 105 sediment samples were taken from holes U1339C and U1339D in this study with a 20 cm sample interval to achieve an average temporal resolution of approximately

774 years between samples. We measured the grain size of both the full sediment suite as well as the grain size of only the terrigenous grains in all samples as well as basic geochemistry (total

39 organic carbon, total nitrogen, C/N, δ13 C, δ 15 N). Absolute diatom abundance as well as relative percent and absolute abundance of Chaetoceros RS were also measured in 93 samples.

3.5.2. Age Model

18 The age model was constructed by tuning δ Ocalcite of benthic foraminiferal species

Uvigerina peregrina, U. senticosa and Elphidium cf. batialis (Cook et al., In Review) to the global marine benthic foraminiferal isotopic stack (LR04; Lisiecki and Raymo, 2005). Multiple species were used because not one benthic species was present in all samples (Cook et al., In

Review).

Marine Isotope Stages were identified by selecting depths at the mid-point of transitions between glacials and interglacials found in the LR04 global stack. Gamma-ray attenuation

(GRA) density is associated with abrupt decreases in density that could result from increases in opal (Takahashi et al., 2011) and was used as an additional guide for selecting glacial terminations. Dates for each sample were obtained using linear interpolation between each tie point (Table 1; Cook et al., In Review).

Table 1 . Tie points from Cook et al. (In Review) used in this study. Age (ka) Depth Error 14 3.5 0.5 71 20.5 1 130 34.6 1

3.5.3. Diatom Analysis

Diatom slides were made following the methods developed by Scherer (1994), which utilizes a random settling technique with known parameters to calculate absolute diatom abundances. Bulk samples were initially freeze-dried and weighed out in glass sample vials to

40 approximately 0.030 g; however, the more diatom-poor samples required samples greater than

0.030 g. Next, 10% HCl and 30% H2O2 were allowed to react with the samples to remove carbonates and organic matter, respectively. Sodium hexametaphosphate (SHMP) was used to deflocculate the samples after treatment.

After treatment, the samples were allowed to disperse within a 1L sized beaker containing 980 mL of DI water. At the bottom of the beakers was a starched coverslip glued on a modified plastic platform upon which the diatoms settled. One side of the cover slips had starch applied to them from rubbing a gloved finger over an exposed surface on a potato. As the starched side of the cover slip was drying, the other side had rubber cement applied to allow it to stick to the plastic platform. Rubber cement allowed the coverslip to remain in place during settling and does not harden under the water, which makes the cover slip readily removable once the draining is complete (Scherer, 1994). Once the sample had settled, the water was siphoned slowly at rates of less than 4 mL per minute so to avoid the redistribution of particles on the cover slip. Flow of water out of the beakers was restricted by a hose with an attached stopcock.

Once draining was complete, the cover slip was allowed to dry and was mounted onto a labeled slide using several drops of Naphrax in toluene (refractive index = 1.73).

Diatom counts were made following the methods of Schrader and Gersonde (1978). A light microscope was used at 1000x magnification to identify the first 300 randomly encountered diatom valves found within three separate transects across the slide. Diatoms were identified as members of one of three taxa: centric diatoms, pennate diatoms, or Hyalochaete Chaetoceros RS.

The total number of individual diatoms per gram was additionally calculated from the transect data using an equation from Scherer (1994):

T = (NB/A)/M

41 where T is the number of diatoms per unit mass, N is the total number of diatoms counted in the transects, B is the area of the bottom of the beaker (8171.28 mm 2), A is the total area of the transect, and M is the mass of the sample. The transect area is determined by multiplying the field of view for the 100 x objective (0.22 mm) by the transect length from which 300 diatoms were counted.

3.5.4. Isotopic Analysis

Both bulk and acidified samples were measured out to approximately 20 mg and were used to obtain isotopic values for bulk sediment and organic material, respectively. Samples from both methods were freeze dried and then ground into a fine powder using a mortar and pestle prior to analysis. Acidified samples were treated with 1 M HCl following the grinding

13 15 step, which removed inorganic material (i.e. carbonates) from the sample. δ Corg , δ N, %TOC, and C/N (TOC/TN) were measured via a Finnigan MAT Delta Plus XL mass spectrometer in continuous flow mode connected with a Costech elemental analyzer with a zero blank at the

Stable Isotope Laboratory, Department of Geological and Atmospheric Sciences at Iowa State

University. Reference standards (Caffeine [IAE-600], IAEA-N2, Cellulose [IAE-CH-e]) and

Acetaniide, the laboratory standard, were used for isotopic corrections. At least one reference standard was used per every six samples. The average combined uncertaintiy (analytical and regression-based correction) in these data are +/- 0.06‰ (VPDB) for δ 13 C and +/- 0.30‰ (Air) for δ 15 N.

3.5.5. Grain Size Analysis

Freeze-dried bulk sediment samples were deflocculated using a slightly acidic 3M solution of sodium hexametaphosphate (SHMP), agitated, and analyzed using the Malvern

Mastersizer 3000, a laser diffraction particle size analyzer, in the Marine Sediments Laboratory of Iowa State University. This method shows the distribution of grains ranging in size from 0.1 to 2700 µm. The analysis of bulk sediment samples yielded the grain size of organic matter, lithogenic particles, carbonates and biogenic silica. To obtain grain size of just the lithogenic particles, each sample was treated with 30% H 2O2 to remove organic matter, 10% HCl to remove carbonates, and 1 M NaOH to remove biogenic silica. This last addition of NaOH was repeated several times until we observed no diatom valves or sponge spicules present within the sample.

The resulting lithogenic particles were then deflocculated with neutral SHMP, as compared to the slightly acidic SHMP used with the bulk samples, and were analyzed using the Malvern

Mastersizer 3000.

3.6. Results

3.6.1. Diatom Analysis

The total number of diatoms per gram of sediment ranges from 1.0 x 10 6 to 9.9 x 10 7, with a mean of 3.2 x 10 7 (Figure 2). The lowest diatom abundances are consistently found within

MIS 6 and MIS 4, whereas the substages of MIS 5 vary substantially in their diatom abundances.

MIS 6 diatom abundances range from 1.0 x 10 6 to 1.5 x 10 7. MIS 5e starts with relatively diatom-rich sediments, containing the highest diatom abundances found in the record, and then gradually decreases in diatom abundances throughout the rest of MIS 5e, reaching abundances as low as glacial values. Diatom abundances during MIS 5d start low, increase to a high of 7.5 x

10 7 near the middle of the substage, and decrease to low values in the latter half of MIS 5d with a low of 2.4 x 10 7 diatoms per gram of sediment. MIS 5c begins with low diatom abundances and remains relatively low throughout the substage. The largest diatom abundances reached during

MIS 5c was 3.5 x 10 7 and the lowest was 1.8 x 10 7. MIS 5b shows a similar pattern to MIS 5d with regards to diatom abundances with abundances beginning at relatively low values, increasing to a high in the middle of the substage and sharply decreasing to low, glacial values in the latter part of the substage. MIS 5b reaches a high of 9.2 x 10 7 and a low of 8.5 x 10 6 diatoms per gram of sediment. MIS 5a diatom abundances resembles the pattern seen in MIS 5e more than MIS 5c does. MIS 5a experiences a sharp increase in diatom abundances near the beginning of the substage, reaching a high abundance of 9.3 x 10 7, and decreasing in the latter part of MIS

5a, reaching a minimum of 1.3 x 10 7. MIS 4 diatom abundances begin relatively low near 3.0 x

10 7 after the end of MIS 5a and then decrease even more, reaching a low value of 2.6 x 10 6 at the end of the record.

Relative percent abundances of Chaetoceros RS range between 37.7% and 75.7% (Figure

2). The lowest percentages are found during glacials MIS 6 and MIS 4 as well as substage MIS

5e. MIS 6 has percentages ranging from 40% to 59.3%. MIS 5e begins with an increase in relative percent abundances of Chaetoceros RS (74.1%), coeval with an increase in absolute diatom abundances. Diatom abundances and relative percent abundances of Chaetoceros RS then deviate from each other following this increase, as a minimum in Chaetoceros RS percentages (37.7%) in MIS 5e is reached near the middle of the substage when diatom abundances were still relatively high. When Chaetoceros RS rebounds back to high percent abundances (68.3%) near the end of substage MIS 5e, diatom abundances continue to decrease to a minimum. The percentages of Chaetoceros RS are consistently high during substages MIS 5d,

MIS 5c, and MIS 5b, and differ from the variability seen in the absolute diatom abundance data.

MIS 5d starts with low relative percent abundances of Chaetoceros RS (43.9%) following the transition out of MIS 5e and increases substantially to values in the range of 57.7-73.3%. MIS 5c and MIS 5b have high relative abundances (52.2-75.7% and 51.7-69.9%, respectively). As seen in the diatom abundance records, MIS 5a starts with high relative percent abundances of

Chaetoceros RS (54.7-70.8%) of Chaetoceros RS and decreases gradually to low percent abundances (44.8%). Relative percent abundances of Chaetoceros RS during MIS 4 begins with relatively high percentage abundances coming out of MIS 5a (62.6%) but then decreases dramatically to low percent abundances (39-51.8%).

3.6.2. Geochemical Analysis

3.6.2.1. Nitrogen Isotopes

δ15 N values range between -1.96‰ and 5.77‰, with the lowest values found during the warm substages of MIS 5 (MIS 5e, MIS 5c, and MIS 5a) and the highest values in the colder substages of MIS 5 (MIS 5d and MIS 5b) and in MIS 6 and MIS 4 (Figure 2). The large negative excursion of -2.38‰ in MIS 6 was measured in a tephra layer, and so is excluded from our analyses, including the carbon isotopic and grain size data. The mean glacial δ 15 N value from

MIS 4 and MIS 6 is 4.06‰ +/- 0.50‰ and the mean interglacial value from MIS 5 is 3.30‰ +/-

1.89‰. Changes in the δ 15 N values from higher to lower values during the warm substages of

MIS 5 appear to be sudden as opposed to gradual. This is most apparent in MIS 5a, which experiences the lowest δ 15 N values that range from -1.96‰ to -1.43‰. The average δ15 N value for MIS 5a is 2.14‰ +/- 2.90‰. MIS 5c experiences the second lowest period of δ 15 N values, ranging from 0.77‰ to 1.96‰. The average δ15 N value for MIS 5c is 2.26‰ +/- 0.82‰. MIS 5e

45 displays the smallest drop in δ15 N values of the three warm substages with values ranging from

3.34‰ to 4.99‰. The average δ15 N value for MIS 5e is 4.00‰ +/- 0.84‰.Transitioning from the lower δ15 N values to higher glacial values is gradual rather than sudden for the transition between MIS 5e and MIS 5d, whereas the transitions from MIS 5c to MIS 5b and from MIS 5a to MIS 4 are more sudden than gradual. The glacials MIS 6 and MIS 4 have δ 15 N values in the range of 2.94‰ to 5.20‰ and from 3.27‰ to 4.47‰, respectively. The cold substages MIS 5d and MIS 5b have δ 15 N values ranging from 4.30‰ to 5.77‰ (average 4.97‰ +/- 0.48‰) and from 3.34‰ to 5.62‰ (average 4.34‰ +/- 0.68‰), respectively.

3.6.2.2 Carbon Isotopes

δ13 C values are generally consistent throughout this record, ranging from -23.49‰ to -

29.21‰ (Figure 3). Both the highest and lowest δ 13 C value endmembers are found at the end of

MIS 5e and near the middle of MIS 5e, respectively. Large and irregular incursions towards higher values are found throughout MIS 5e and in the transitions between MIS 5e and MIS 5d, between MIS 5d and MIS 5c, between 5c and 5b, and between MIS 5b and MIS 5a. These irregularities fall in the δ 13 C value range of -25.55‰ to -23.49‰. The lowest δ 13 C values can be found consistently throughout MIS 4, -27.45‰ to -26.85‰, during MIS 6, -29.03‰ to -

26.16‰, and in a small interval of substage MIS 5e.

3.6.2.3. Total Organic Carbon

Total organic carbon (TOC) percentages range from 0.37-3.76% (Figure 3). Several large fluctuations in the percentages of TOC occur throughout substages MIS 5e through MIS 5c and during the transitions from MIS 6 to MIS 5 and from MIS 5c to MIS 5b. There is no

46 recognizable trend of TOC within substages MIS 5e to MIS 5b or within MIS 6. The lowest TOC percentages are found in the glacial MIS 6. TOC is persistently low during MIS 5a and MIS 4

(0.85-1.91%).

3.6.2.4. C/N Values

C/N values range between 7.69 and 29.55, vary noticeably throughout the record, and do not reveal any prominent trends or periodicity between different substages and glacials (Figure

3). MIS 5a and MIS 4, however, have persistent lower C/N values compared to MIS 6 and the other substages with values ranging from 10.5 to 20. The lowest C/N value is found in a single sample during the transition from warm substage MIS 5e to cold substage MIS 5d and the highest C/N value is found in a single sample in the transition from warm substage MIS 5c to cold substage MIS 5b.

3.6.3. Grain Size Analysis

Sand percentages range from 9-38% for the full sediment suite and 2- 46% for lithogenic grains (Figure 3). Although the sand fraction does not display any noticeable patterns, sand increases proportionally during deglaciation and at the transition from MIS 5b to MIS 5a. Silt is the predominant grain size in both sample types (Figure 3). Silt also shows no visible trends between the different substages of MIS 5. For the full sediment suite, silt comprises from 61-

88% of the sediments and lithogenic grains are comprised of 49-84% silt. Unlike the previous two grain size fractions, the clay fraction does display visible trends (Figure 3). In general, clay comprises between 0.83-10% of the full sediment suite and 4-18% of lithogenic grains. Clay was highest during the glacials MIS 4 and MIS 6 and showed a pattern of increasing proportion

(fining up) in all MIS 5 warm substages (MIS 5a, MIS 5c, and MIS 5e). MIS 5c and MIS 5e show a sharp drop in clay percentages towards the end of each substage. Clay also showed high concentrations during MIS 5b. The same trend is present in both sample types, but samples with only lithogenic grains tended to be more clay-rich than the full sediment suite.

3.7. Discussion

3.7.1. MIS 6 and Deglaciation at the Umnak Plateau

MIS 6 at the Umnak Plateau was characterized by extensive sea ice, high water column stratification, and low primary productivity. Low insolation combined with weak seasonal contrasts and a shortened summer season during MIS 6 contributed to the expansion of sea ice during this glacial period (Riethdorf et al., 2013, 2015), similar to the Last Glacial Maximum

(LGM), when perennial sea ice may have covered the Umnak Plateau (Caissie et al., 2010;

Sancetta et al., 1983). Additionally, lowered sea level closed the Bering Strait and eastern

Aleutian Island passes, preventing the flow of freshwater out of the Bering Sea (Waelbroeck et al., 2002) and the flow of the warm ACC to the Umnak Plateau (Gorbarenko et al., 2005), essentially isolating the southeastern basin. Blocking the passage of the warm ACC could have allowed sea ice to extend to the Umnak Plateau (c.f. Sancetta, 1983). High proportions of sand and clay particles during MIS 6 (Figure 3) support expanded sea ice but whether the sea ice was seasonal or perennial cannot be determined based on these data. In either case, the effects of the sea ice on the surface waters reduced primary productivity.

Sea ice reduced light availability for photosynthesis to occur below (Elderfield and

Rickaby, 2000) and limited the effects the wind had on mixing the upper water column. In addition, freshwater input from seasonal sea ice melt would have served to increase stratification

(Sigman et al., 2004), limiting nitrate upwelling and replenishment (Francois et al., 1997;

Nurnberg and Tiedemann, 2004; Brunelle et al., 2007; Riethdorf et al., 2013, 2015). If this model is correct, we might expect to see low productivity associated with perennial or nearly perennial sea ice and reduced nitrate availability. In fact, low Chaetoceros RS percentages, low diatom abundances and high δ15 N values during MIS 6 suggest there was low productivity with relatively high nitrate utilization (Figure 2). This suggests that nitrogen-limitation may have been in effect requiring a supply of iron to maintain diatom productivity. Strong winds during MIS 6 would have been able to erode and transport iron containing clay-sized terrigenous material to the site and occasional transport of this iron into the surface waters due to ice melt, allowed for the almost complete consumption of the available nitrates (Brunelle et al., 2007; Galbraith et al.,

2008; Riethdorf et al., 2015).

The deglaciation is a response to an increase in insolation and marks the transition from a period of low productivity and high nitrate utilization to a period of high productivity and low nitrate utilization. Melting of both glacial and sea ice in and surrounding the Bering Sea contributed a considerable amount of terrestrially derived nitrates to the surface waters, which would have triggered increased productivity. This is supported by both high diatom abundances and high Chaetoceros RS percentages along with low nitrate utilization (Figure 2; Nakatsuka et al., 1995). The carbon isotopes (δ13 C) additionally shows a more positive trend during the deglaciation, which indicates a possibly greater marine influence on the organic carbon fraction possibly due to an increase in sea level and a further distance from the shore line (Figure 2;

Meyers, 1994).

3.7.2. MIS 5 at the Umnak Plateau

3.7.2.1. MIS 5e and MIS 5d

High insolation during MIS 5e and a prolonged summer season would have resulted in warmer surface water temperatures, which allowed for continued stratification and limited nitrates upwelled to the surface waters, decreasing primary productivity (Slagstad et al., 2011).

This is supported by a high nitrate utilization and by low Chaetoceros RS percentages and decreasing diatom abundances (Figure 2). Primary productivity may have also declined due to a decrease in aeolian iron deposition, similar to what is seen in the Holocene (Lam et al., 2013); however, a decrease in iron deposition should not lead to high nitrate utilization, as seen here.

Mid- to late-MIS 5e shows an increase in Chaetoceros RS percentages (Figure 2) that might result from an increase in iron or light availability in the summer as compared to the spring, which allowed secondary blooms to flourish. The increase in Chaetoceros RS percentages probably does not result from an increase in nitrate availability since nitrate utilization increases

(Figure 2), opposite what we would expect if more nitrates became available.

Increased insolation and an increased ACC influence, a result of increased sea level that opened of the eastern Aleutian Island passes, during MIS 5e would have both contributed to a decrease in sea ice coverage, supported by low sand and clay percentages (Figure 3); therefore, sea ice transport to the Umnak Plateau was probably absent (Riethdorf et al., 2013). Lithogenic clay percentages increase throughout MIS 5e; however, the reason for this increase is unknown.

Several processes that could lead to an increase in clay percentages include increased terrigenous sediment aerial deposition, stronger surface currents, decreasing sea level, and possibly an increase sea ice cover (Riethdorf et al., 2013). Decreasing insolation, low sand percentages, and increasing Chaetoceros RS percentages during the latter half of MIS 5e (Figure 3) suggest that

50 the increase in clay percentages may be due to increased aeolian iron deposition. The decreasing insolation throughout MIS 5e could also suggest that the increasing lithogenic clay percentages may result from a decrease in sea level, which brings the site closer to land and closer to a terrigenous particle supply.

The opening of the eastern Aleutian Island passes during the deglaciation and the increased influence of the ACC on the Umnak Plateau can be seen in the δ15 N records (Figure 4) from both the Umnak Plateau and from the Gulf of Alaska (Site ODP 887; Figure 1; Galbraith et al., 2008). The ACC flows from the Gulf of Alaska, southwestward along the shelf break, and through several of the eastern Aleutian Island passes (Stabeno et al., 2002). Given the close proximity of the Umnak Plateau to these passes and to the incoming ACC, it would seem reasonable that the Gulf of Alaska and the Umnak Plateau would have similar δ 15 N records during the warm substages of MIS 5. The similarity between the two during MIS 5e can be seen during the beginning of MIS 5e; however, during the latter half of MIS 5e, the Umnak Plateau

δ15 N increases, whereas the Gulf of Alaska δ15 N decreases (Figure 4). This suggests that δ15 N during the beginning of MIS 5e at the Umnak Plateau may primarily record the δ15 N record of the ACC, whereas the latter half of the MIS 5e record reflects either changes in nitrate utilization or in the δ15 N composition of the new source of nitrates. This supports the idea presented earlier that increasing clay percentages during MIS 5e could have indicated decreased sea level, which might have led to the closure of the eastern Aleutian Island passes. Other sites in the North

Pacific (Figure 1; 17 JPC, Brunelle et al., 2007; MD 2416 and ODP 882, Galbraith et al., 2008) show a similar increase in δ15 N during the latter half of MIS 5e, suggesting that the increase in nitrate utilization may be reflecting a regional change (Figure 4).

MIS 5d experienced increasing insolation and a continuing influence of the warm ACC.

Similarities in the δ15 N records between the Umnak Plateau and the Gulf of Alaska during MIS

5e and MIS 5d supports the idea that sea level remained high enough to allow the ACC to influence the Umnak Plateau during MIS 5d (Figure 4). Sea ice was absent during this substage due to the increasing insolation and the warm ACC, as indicated by low lithogenic clay and sand percentages (Figure 3). Since the ACC may have influenced the Umnak Plateau during MIS 5d, the increase in the δ15 N record at the Umnak Plateau (Figure 2) may reflect the changes in the nitrate composition of the ACC as opposed to changes in nitrate utilization at the Umnak Plateau.

Relatively high Chaetoceros RS percentages and diatom abundances suggest that conditions were favorable for high productivity at this time (i.e. high light availability, high iron deposition, and high nitrate replenishment; Lam et al., 2013).

3.7.2.2. MIS 5c and MIS 5b

MIS 5c displays a low δ15 N with high Chaetoceros RS percentages and low diatom abundances (Figure 2). As with MIS 5e, the δ15 N record during MIS 5c at the Umnak Plateau visibly resembles the Gulf of Alaska δ15 N record (Figure 4; Galbraith et al., 2008), which suggests that the decreased δ 15 N could reflect changes in the nitrate composition of the ACC as opposed to changes related to nitrate utilization. As insolation decreases through the latter half of

MIS 5c, lithogenic clay percentages gradually increase (Figure 3). Since clay increases as

Chaetoceros RS percentages increase, the clay could be derived from an increase in aeolian deposition of iron (Riethdorf et al., 2013); however, diatom abundances remain low throughout

MIS 5c. Low diatom abundances may be due to an absent spring bloom in which conditions were not favorable for the bloom (i.e. low light availability or low nitrate/and or iron levels) or

52 may be due to poor preservation of the spring bloom diatoms. More detailed diatom counts are needed to test this hypothesis. The lithogenic clay also appears to increase as insolation decreases, which might indicate an drop in sea level that would have put the Umnak Plateau closer to land.

The beginning of MIS 5b has high primary productivity, supported by high diatom abundances and high Chaetoceros RS percentages, and high δ15 N values (Figure 2). The high

δ15 N could likely reflect the changing nitrate composition of the ACC, shown by the similar δ15 N changes in both the Umnak Plateau record and in the Gulf of Alaska record (Figure 4), and indicates that sea level was high enough to allow for the inflow of the ACC through the eastern

Aleutian Island passes. High productivity might have resulted from an increase in aeolian iron deposition, reflected by the increased lithogenic clay percentages, and from a decrease in surface water temperatures, that allowed for a more uniform distribution of nitrates within the water column.

The middle of MIS 5b exhibits a decrease in primary productivity combined with an increase in δ15 N (Figure 2). Unlike the beginning of MIS 5b, the middle MIS 5b δ15 N record does not follow the Gulf of Alaska record, increasing in value instead of decreasing (Figure 4).

This probably reflects a decrease in sea level that restricted the flow of the ACC through the eastern Aleutian Island passes and onto the Umnak Plateau. Since the ACC influence on the site weakened, the δ15 N record for the latter half of MIS 5b most likely reflects changes in nitrate utilization or in the δ15 N composition of the new nitrate source. When comparing the Umnak

Plateau record from other locations in the North Pacific (85 KL, Riethdorf et al., 2015; ODP 882 and MD 2416, Galbraith et al., 2008), we can see that there is generally an increase in δ15 N during this part of MIS 5b (Figure 4). The increase in nitrate utilization and the decrease in

53 primary productivity may be due a shallower mixed layer that limited nitrate replenishment from the subsurface. This limited nitrate replenishment was combined with a high aeolian iron depositon and adequate light levels that allowed for the more complete consumption of the available nitrates (Brunelle et al., 2007).

3.7.2.3. MIS 5a

Increased insolation and an increase of nitrates in the surface waters at the onset of MIS

5a likely led to an increase in primary productivity, as indicated by the increases in diatom abundances and Chaetoceros RS percentages. Cooler temperatures during this substage may have dampened the effects of a temperature-driven stratification, which allowed for a more even distribution of nitrates in the upper water column. This even distribution of nitrates combined with the increased light availability associated with a prolonged summer season and with a high iron deposition could have consequently led to increased productivity. High nitrate utilization suggests that despite a potential increase in surface water nitrate availability, nitrates were still limiting primary productivity compared to iron and light availability. Comparing δ15 N between the Umnak Plateau and the Gulf of Alaska does not reveal any similarities at the beginning of

MIS 5a (Figure 4); therefore, the δ15 N record during this time period most likely records nitrate utilization. Other sites in the North Pacific (85 KL, Riethdorf et al., 2015; ODP 882 and MD

2416, Galbraith et al., 2008) reveal similar increases in δ15 N during the beginning of MIS 5a; therefore, the cause of this increase in δ15 N may be regional.

The Umnak Plateau and the Gulf of Alaska δ15 N records start to reveal similar patterns near the middle of MIS 5a, almost when insolation reaches a maximum (Figures 2 and 4), suggesting that sea level became high enough to allow the ACC to pass through the eastern

Aleutian Island passes. During MIS 5a, the Gulf of Alaska δ15 N record reaches the lowest δ15 N values found throughout its MIS 5 record. In the middle of MIS 5a the Umnak Plateau record also drops to the lowest values found throughout its record. This similarity could suggest the

δ15 N record during the latter half of MIS 5a reflects the changing nitrate composition of the ACC and not the change in nitrate utilization.

The latter half of MIS 5a also displays a drop in primary productivity soon after insolation reaches a maximum and remains low as insolation decreases, as seen in diatom abundances and Chaetoceros RS percentages (Figure 2). Low primary productivity could be due to a decrease in iron deposition or low light availability. Since insolation is decreasing as primary productivity decreases, decreasing primary productivity may be due to a shortened summer season and decreased light availability. Lithogenic clay gradually increases throughout MIS 5a

(Figure 3) and may indicate increasing deposition of iron-containing clay-sized terrigenous matter through aeolian transport or with decreasing sea level.

3.7.3. MIS 4 at the Umnak Plateau

Low insolation during MIS 4 allowed for the expansion of sea ice to the Umnak Plateau, which limited primary productivity by increasing stratification in the upper water column and decreasing light availability (Sigman et al., 2004; Elderfield and Rickaby, 2000). A potential sea ice cover is supported by an increase in lithogenic clay-sized particles and moderate percentages of sand-sized lithogenic particles (Figure 3). An increase in sea ice, in combination with the pooling of freshwater that resulted from the closing of the Bering Strait when sea level was low, would increase stratification (Sigman et al., 2004). Increased stratification appeared to limit nitrate upwelling and replenishment (Francois et al., 1997; Nurnberg and Tiedemann, 2004;

Brunelle et al., 2007; Riethdorf et al., 2013, 2015). Limited nitrate availability in the surface waters resulted in low primary productivity, as indicated by low diatom abundances and low

Chaetoceros RS percentages (Figure 2). The limited availability of nitrates also combined with high aeolian iron deposition allowed for the near-complete consumption of the available nitrates

(Brunelle et al., 2007; Galbraith et al., 2008; Riethdorf et al., 2015). Comparing the δ15 N record of both the Umnak Plateau and the Gulf of Alaska, we see that they are not similar during MIS 4, which suggests that sea level dropped low enough to close off the eastern Aleutian Island passes

(Figure 4). δ15 N at the Umnak Plateau during MIS 4 may therefore primarily reflect changes in nitrate utilization.

3.8. Conclusions and Future Work

The records presented here generally agree with other Bering Sea records that suggest glacial periods, MIS 6 and MIS 4, have increased terrigenous organic matter deposition, increased nitrate utilization and decreased primary productivity, and that the interglacial, MIS 5, is characterized by decreased sea ice extent, decreased nitrate utilization and increased primary productivity (e.g. Brunelle et al., 2007; Galbraith et al., 2008; Riethdorf et al., 2013, 2015).

Increased terrigenous organic matter deposition during MIS 6 and MIS 4 most likely results from an extensive sea ice coverage that contributes to an increase in high water column stratification and decreased nitrate replenishment (Francois et al., 1997; Nurnberg and Tiedemann, 2004;

Brunelle et al., 2007; Riethdorf et al., 2013, 2015). High iron deposition associated with stronger winds during the glacials allows for the near-complete consumption of the available surface water nitrates (Brunelle et al., 2007; Galbraith et al., 2008; Riethdorf et al., 2015). Increased insolation during the deglaciation resulted in both a retreat of sea ice and an increase in sea level

56 due to the meltwater produced from glacial ice on land, which allowed for the opening of several eastern Aleutian Island passes and the flow of the relatively warm ACC from the North Pacific into the Bering Sea.

Given the general observations between the glacials and MIS 5, MIS 5 was noticeably highly variable itself at the Umnak Plateau. The nitrogen isotope data (δ15 N) between the Umnak

Plateau and the Gulf of Alaska record (ODP 887; Galbraith et al., 2008) are visually similar during the warm substages of MIS 5 (Figure 4), suggesting the δ15 N record from the Umnak

Plateau may reflect changing nitrate sources with a more increased ACC influence during the warm substage and a more westward Bering Sea source during the cold substages. MIS 5e, the warmest substage of MIS 5, had decreased primary productivity and a relatively high nitrate utilization. This might be due to an increase in stratification associated with the increase in surface water temperatures that limited vertical mixing and nitrate replenishment in the water column. High productivity during MIS 5d, MIS 5b, and MIS 5a was possibly a result of high nitrate replenishment combined with high light availability and an adequate supply of iron, which could have been supplied by aeolian deposition. Low productivity during MIS 5c and during the latter halves of MIS 5b and MIS 5a stemmed from low light availability, low iron availability, decreased nitrate replenishment, or from any combination of these factors.

Distinguishing nitrate utilization from a changing δ15 N of the source nitrates will require further testing. This testing could include taking δ15 N measurements from marine sediments at several locations along the pathway of the ACC, at the eastern Aleutian Island passes, and at locations west of the Umnak Plateau. Similar sediment δ15 N values between the Umnak Plateau, the eastern Aleutian Island passes, and from the ACC pathway during most of MIS 5 may further support the idea that the δ15 N record reflects a varying ACC influence, dependent on sea level.

Cold intervals with closed eastern Aleutian Island passes would then be expected to have similar

δ15 N records between the Umnak Plateau and the more westward locations. Data that suggests that the δ15 N records are significantly different may then indicate that the δ15 N record at the

Umnak Plateau reflects changes in nitrate utilization.

Other future work on MIS 5 at the Umnak Plateau can involve a more detailed examination of diatom assemblages (e.g. Sancetta, 1982; Caissie et al., 2010); the use of a biomarker-based sea ice proxy, IP25 (ice-proxy with 25 carbon atoms); and a look at diatom-

15 bound nitrogen isotopes (δ Ndb ; e.g. Belt et al., 2007). Diatom assemblages can be correlated to various environmental conditions (i.e. different sea ice concentrations, influx of the warm ACC, high nutrient concentrations, etc). Interglacials generally record lower concentrations of sea-ice diatoms and higher concentrations of open water diatoms, whereas glacials generally record the opposite (Katsuki and Takahashi, 2005). The presence of sea ice diatoms has also been associated with an organic geochemical lipid, IP25, which has been used as a proxy for seasonal sea ice reconstructions (Belt et al., 2007). IP25 data can therefore further support our interpretations of the grain size data with regards to sea ice transported terrigenous sediment.

δ15 N measurements of organic matter preserved within diatom frustules can be used to gauge the potential degree of diagenesis on the bulk sediment (e.g. Brunelle et al., 2007;

Galbraith et al., 2008). Diagenesis refers to the chemical, biological and physical processes that alter the chemical composition of the organic matter found in marine sediments (Henrichs,

1992). Similar δ15 N values between the diatom-bound record and the bulk sediment record would imply that diagenesis has not altered the chemical composition of the organic matter, whereas significant differences between the two would imply diagenetic alteration (Galbraith et al., 2008).

Insolation (60°N) [W/m2] Chaetoceros RS (%) Berger and Loutre (1991) 30 40 50 60 70 80 440 480 520 560

60000 MIS 4

MIS 5a 80000 58 A

MIS 5b g e

( MIS 5c 100000 y e a r MIS 5d s

B P ) MIS 5e 120000

MIS 6 140000

-4 -2 0 2 4 6 0 20000 40000 60000 80000100000 3 3.5 4 4.5 5 5.5 d15N (‰ Air) Diatoms per unit mass (kg) d18O (‰ VPDB) Lisiecki and Raymo (2005) Cook et al. (2014) - U1339

Figure 2 . δ15 N, Chaetoceros RS percentages, Diatom Abundances. Plotted against insolation (Berger and Loutre, 1991) and δ18 O records of both Lisiecki and Raymo (2005) and Cook et al. (2014). Yellow bars indicate the warm substages: MIS 5e, MIS 5c, and MiS 5a. Also graphed are cold substages MIS 5b and MIS 5d as well as glacials MIS 4 and MIS 6. 59

Figure 3 . δ13 C, TOC, C/N, Treated/Untreated Grain Size. Plotted against insolation (Berger and Loutre, 1991) and δ18 O records of both Lisiecki and Raymo (2005) and Cook et al. (2014). Yellow bars indicate the warm substages: MIS 5e, MIS 5c, and MiS 5a. Also graphed are cold substages MIS 5b and MIS 5d as well as glacials MIS 4 and MIS 6.

15 15 15 Figure 4 . Sediment δ N Comparisons. δ N from the Umnak Plateau plotted with δ N records from Bowers Ridge (bulk and diatom-bound (db) at site 17 JPC; Brunelle et al., 2007), Shirshov Ridge (Site 85 KL; Riethdorf et al., 2015), the Western Subarctic (Sites MD 2416 and ODP 882; Galbraith et al., 2008), and from the Gulf of Alaska (Site ODP 887; Galbraith et al., 2008).

APPENDIX A

INVESTIGATING VARIATIONS IN DIATOM IDENTIFICATION

A.1. Abstract

The analysis presented here tests whether diatom counts were consistent between different researchers or if the counts varied substantially. A total of ten samples were counted from each researcher and were tested for similarity using a chi-square test. Almost all samples were determined to be statistically different between researchers. Future diatom counts should be focused on keeping a consistent counting method to produce viable results.

A.2. Introduction

Diatom identifications can be very useful as proxies for environmental change; however, variations in how diatoms are identified and counted between different researchers and research groups can be a problem. To investigate this possible discrepancy, three separate researchers from the Marine Sediments Lab at Iowa State University each counted the same ten quantitative slides made from Bering Sea sediment samples that depict Marine Isotope Stage (MIS) 5 at the

Umnak Plateau.

A chi-squared test was performed for each sample to test whether or not the differences between the researchers counts are significant. Chi-square tests are used to analyze data that are counts of the number of individuals in different categories and involves calculated expected values (Equation 2; Townend, 2002). The assumptions and requirements for chi-square tests are that random samples and independent measurements and/or observations are taken. Once a chi- square value has been calculated, we can look up the values in tables to get a p-value that will

62 tell us the significance of the measurements. To look up the chi-square values, it is important to know the number of degrees of freedom (df; Equation 3).

A.3. Methods

The sediment samples used are included in table 2. Counting methods followed the methods laid out by Schrader and Gersonde (1979). Diatoms were characterized as being one of the groups: centrics, pennates, and Chaetoceros resting spores (RS; Sancetta, 1982).

Table 2. Sediment samples used for comparing counts

Samples Age (ka) 323-U1339D 3H-3 36.5-37.5cm 74 323-U1339C 3H-4 112-113cm 95 323-U1339D 4H-3 139-140cm 118 323-U1339D 4H-4 69-70cm 121 323-U1339C 4H-3 132-133cm 132 323-U1339C 4H-6 12-13cm 142

The results are presented in contingency tables. The researchers are in the columns and the diatom categories are in the rows. Additionally, the observed counts are located at the top of the table and the expected counts are at the bottom of the table. Each row and column of the observed counts were added to calculate the expected values and are represented in bold. Chi- square values for each column in the expected section were calculated by using equation 4. The final chi-square values is the sum of the three column chi-square values.

Chi-square p-values that fall below 7.779 have a 0.10 level of significance and values that fall below 9.488 have a 0.05 level of significance. Having a 0.05 or 0.10 level of significance means that we have failed to reject the null hypothesis. Any values that lie above the 0.05 level of significance value of 9.488 means we can reject the null hypothesis. The null hypothesis for

63 each sample is that the diatom classifications between the three researchers are similar. Rejecting the null hypothesis means that the counts between researchers are statistically different.

Expected counts = (row total * column total)/grand total ( Equation 2 ) df = (number of rows – 1) x (number of columns -1) ( Equation 3 )

Chi Square: χ² = Σ ((Oij-Eij)² / Eij) (Equation 4)

A.4. Results

The chi-square test (Equation 4) used here has a distribution with 4 degrees of freedom, reflective of three different diatom categories and three different researchers. The results are presented in contingency tables with the final chi-square values on the bottom right. Almost all samples have p-values above the 0.05 level of significance, except for 323-U1339C 4H-3 132-

133cm, which has a p-value of 4.26.

Table 3. Chi-square results for 323-U1339C-4H-3 132-133 cm.

Observed 323-U1339C 4H-3 132-133 cm Bailey Beth Derrick Cen 143 143 132 418 Pen 67 51 47 165 Chae 173 132 121 426 383 326 300 1009 Expected Cen 159 135 124 Pen 63 53 49 Chae 162 138 127 Chi 2.64 0.80 0.82 4.26 Square

Table 4. Chi-square results for 323-U1339D-3H-3 36.5-37.5 cm.

Observed 323-U1339D 3H-3 36.5-37.5cm Bailey Beth Derrick Cen 158 166 88 412 Pen 115 123 78 316 Chae 159 120 135 414 432 409 301 1142 Expected Cen 156 148 109 Pen 120 113 83 Chae 157 148 109 Chi 0.24 8.55 10.38 19.17 Square

Table 5. Chi-square results for 323-U1339D-4H-3 139-140 cm.

Observed 323-U1339D 4H-3 139-140cm Bailey Beth Derrick Cen 116 147 70 333 Pen 77.5 33 39 149.5 Chae 277 149 191 617 470.5 329 300 1099.5 Expected Cen 142 100 91 Pen 64 45 41 Chae 264 185 168 Chi 8.42 32.46 7.92 48.80 Square

Table 6. Chi-square results for 323-U1339D-4H-4 69-70 cm.

Observed 323-U1339D 4H-4 69-70cm Bailey Beth Derrick Cen 115 144 105 364 Pen 70 50 56 176 Chae 150 111 139 400 335 305 300 940 Expected Cen 130 118 116 Pen 63 57 56 Chae 143 130 128 Chi 2.90 9.28 2.08 14.26 Square

Table 7. Chi-square results for 323-U1339C-3H-4 112-113 cm.

Observed 323-U1339C 3H-4 112-113cm Bailey Beth Derrick Cen 98 44 89 231 Pen 49 117 30 196 Chae 176 191 181 548 323 352 300 975 Expected Cen 77 83 71 Pen 65 71 60 Chae 182 198 169 Chi 10.10 49.06 20.66 79.82 Square

Table 8. Chi-square results for 323-U1339C-3H-6 12-13 cm.

Observed 323-U1339C 3H-6 12-13cm Bailey Beth Derrick Cen 143 119 65 327 Pen 23 37 46 106 Chae 178 163 189 530 344 319 300 963 Expected Cen 117 108 102 Pen 38 35 33 Chae 189 176 165 Chi 12.38 2.05 21.90 36.33 Square

A.5. Discussion and Summary

The results of this experiment suggest that each researcher obtained different diatom counts with inconsistent methods of classification on almost all of the samples, with the exception of 323-U1339C 4H-3 132-133cm. This one sample had a 0.10 level of significance, or a 99% confidence. The largest variability was for sample 323-U1339C-3H-4 112-113 cm, which had a chi-square value of almost 80.

Variations in the diatom counts between researchers may be due to differences in how each researcher identified diatoms. A researcher may have incorrectly identified several diatoms and counted them in the wrong group (i.e. counting a pennate as a Chaetoceros RS). One could have also missed small diatoms, which would make one group seem like an overestimate and the other an underestimate. No matter what the issue is, this analysis shows that the diatom counts between different researchers were statistically different and future diatom counts should focus on consistent counting methods.

APPENDIX B

DOES CRUSHING SEDIMENT AFFECT IDENTIFICATION

B.1. Abstract

An experiment was performed to determine whether sediments that were crushed generated different diatom counts compared to uncrushed samples. Samples were chosen to be representative of both cold and warm substages of Marine Isotope Stage (MIS) 5. Using a chi- square test, it was determined that there was no significant difference between the counts from the crushed and uncrushed samples.

B.2. Introduction

During the process of freeze-drying and preparing the sediment samples for making quantitative slides, a mistake was made and a large number of samples were ground and crushed after they were freeze dried. This happened after both grain size and isotopic analyses were performed but before diatoms were counted. To determine whether or not the crushing impacts the diatoms counts, an experiment was conducted using samples that were freeze-dried but not crushed. A chi-square test was used to determine whether or not crushing impacted the diatom counts (Equation 4).

B.3. Methods

Twelve of the uncrushed, freeze-dried samples were chosen to have subsamples taken, which were then crushed. The newly crushed sediment from the twelve samples were then made into quantitative slides following the methods of Scherer (1994). Quantitative slides using an uncrushed subsample had already been made and counted prior to making the crushed quantitative slides. Both sets of slides had the diatoms counted in three different categories:

68 centrics, pennates, and Chaetocers resting spores (RS). Samples were chosen so that there were samples from both warm and cold substages of MIS 5 and can be found in table 9.

P-values that fall below 4.605 have a 0.10 level of significance. Having this level of significance means that we have failed to reject the null hypothesis. Any p-values that lie above the 0.05 level of significance value of 5.991 means we can reject the null hypothesis. The null hypothesis for each sample is that the diatom classifications between the crushed and uncrushed samples are statistically identical. Rejecting the null hypothesis means that the counts between the samples are statistically different.

Table 9. Sediment samples used for crushed/uncrushed comparisons.

Samples Age (ka) U1339D-3H-4 144-145cm 85 U1339D-3H-5 54-55cm 87 U1339D-3H-5 114-115cm 90 U1339D-4H-2 69-70cm 108 U1339D-4H-2 129-130cm 111 U1339D-4H-3 39-40cm 113 U1339D-4H-3 99-100cm 116 U1339D-4H-4 8.5-9.5cm 119 U1339D-4H-4 69-70cm 121 U1339D-4H-4 108.5-109.5cm 123 U1339D-4H-5 19-20cm 125 U1339D-4H-5 59-60cm 127

B.4. Results

The chi-square test (Equation 4) used here has a distribution with 2 degrees of freedom

(Equation 3), reflective of three different diatom categories and two different types of sediment samples. The results are presented in contingency tables with the final chi-square values on the bottom right. Eleven of the twelve samples have p-values below the 0.10 level of significance.

One of the two that does not have a 0.10 level of significance does have a p-value that falls below the 0.05 level of significance (U1339D-4H-2 69-70cm, p-value of 5.54). Only one of the twelve samples falls above the 0.05 level of significance (U1339D-4H-4 69-70cm, p-value of

7.14).

Table 10. Chi-square results for 323-U1339D-3H-4 144-145 cm.

Observed U1339D-3H-4 144-145cm Uncrushed Crushed Cen 75 59 134 Pen 31 26 57 Chae 198 215 413 304 300 604 Expected Cen 67.4 66.6 Pen 28.7 28.3 Chae 207.9 205.1 Chi Square 1.50 1.52 3.02

Table 11. Chi-square results for 323-U1339D-3H-5 54-55 cm.

Observed U1339D-3H-5 54-55cm Uncrushed Crushed Cen 84 80 164 Pen 39 24 63 Chae 190 196 386 313 300 613 Expected Cen 83.7 80.3 Pen 32.2 30.8 Chae 197.1 188.9 Chi Square 1.71 1.78 3.49

Table 12. Chi-square results for 323-U1339D-3H-5 114-115 cm.

Observed U1339D-3H-5 114-115cm Uncrushed Crushed Cen 75 60 135 Pen 32 25 57 Chae 193 215 408 300 300 600 Expected Cen 67.5 67.5 Pen 28.5 28.5 Chae 204 204 Chi Square 1.86 1.86 3.72

Table 13. Chi-square results for 323-U1339D-4H-2 69-70 cm.

Observed U1339D-4H-2 69-70cm Uncrushed Crushed Cen 101 78 179 Pen 26 38 64 Chae 173 184 357 300 300 600 Expected Cen 89.5 89.5 Pen 32 32 Chae 178.5 178.5 Chi Square 2.77 2.77 5.54

Table 14. Chi-square results for 323-U1339D-4H-2 129-130 cm.

Observed U1339D-4H-2 129-130cm Uncrushed Crushed Cen 53 63 116 Pen 27 27 54 Chae 220 210 430 300 300 600 Expected Cen 58 58 Pen 27 27 Chae 215 215 Chi Square 0.55 0.55 1.10

Table 15. Chi-square results for 323-U1339D-4H-3 39-40 cm.

Observed U1339D-4H-3 39-40cm Uncrushed Crushed Cen 62 57 119 Pen 37 47 84 Chae 203 196 399 302 300 602 Expected Cen 59.7 59.3 Pen 42.1 41.9 Chae 200.2 198.8 Chi Square 0.76 0.76 1.52

Table 16. Chi-square results for 323-U1339D-4H-3 99-100 cm.

Observed U1339D-4H-3 99-100cm Uncrushed Crushed Cen 85 88 173 Pen 33 32 65 Chae 184 180 364 302 300 602 Expected Cen 86.8 86.2 Pen 32.6 32.4 Chae 182.6 181.4 Chi Square 0.05 0.05 0.10

Table 17. Chi-square results for 323-U1339D-4H-4 8.5-9.5 cm.

Observed U1339D-4H-4 8.5-9.5cm Uncrushed Crushed Cen 81 69 150 Pen 31 36 67 Chae 191 196 387 303 301 604 Expected Cen 75.2 74.8 Pen 33.6 33.4 Chae 194.1 192.9 Chi Square 0.69 0.70 1.39

Table 18. Chi-square results for 323-U1339D-4H-4 69-70 cm.

Observed U1339D-4H-4 69-70cm Uncrushed Crushed Cen 106 77 183 Pen 56 55 111 Chae 140 168 308 302 300 602 Expected Cen 91.8 91.2 Pen 55.7 55.3 Chae 154.5 153.5 Chi Square 3.56 3.58 7.14

Table 19. Chi-square results for 323-U1339D-4H-4 108.5-109.5 cm.

Observed U1339D-4H-4 108.5-109.5cm Uncrushed Crushed Cen 99 92 191 Pen 54 67 121 Chae 147 143 290 300 302 602 Expected Cen 95.2 95.8 Pen 60.3 60.7 Chae 144.5 145.5 Chi Square 0.85 0.85 1.70

Table 20. Chi-square results for 323-U1339D-4H-5 19-20 cm.

Observed U1339D-4H-5 19-20cm Uncrushed Crushed Cen 108 99 207 Pen 71 59 130 Chae 129 157 286 308 315 623 Expected Cen 102.3 104.7 Pen 64.3 65.7 Chae 141.4 144.6 Chi Square 2.10 2.06 4.16

Table 21. Chi-square results for 323-U1339D-4H-5 59-60 cm.

Observed U1339D-4H-5 59-60cm Uncrushed Crushed Cen 64 74 138 Pen 78 77 155 Chae 165 156 321 307 307 614 Expected Cen 69 69 Pen 77.5 77.5 Chae 160.5 160.5 Chi Square 0.49 0.49 0.98

B.5. Discussion and Summary

Ten of the twelve samples had chi-square values less than 4.605 for a 0.10 level of significance. One of the two that did not have a 0.10 level of significance (U1339D-4H-2 69-70 cm) did have a 0.05 level of significance, falling below the value of 5.991 (5.54). The other sample (U1339D-4H-4 69-70 cm) fell above this value (7.14), and therefore allows us to reject the null hypothesis for that one sample. In total, we have failed to reject the null hypothesis for eleven of the twelve samples. This suggests that there is no significant difference between the uncrushed and crushed samples in MIS 5.

References

Abrantes, F. (1988). Diatom assemblages as upwelling indicators in surface sediments off Portugal. Marine Geology , 85 (1), 15-39

Addison, J.A., B.P. Finney, W.E. Dean, M.H. Davies, A.C. Mix, J.S. Stoner, and J.M. Jaeger (2012). Productivity and sedimentary δ15 N variability for the last 17,000 years along the northern Gulf of Alaska continental slope, Paleoceanography , 27 , PA1206, 1-17.

Aguilar-Islas, A.M., M.P. Hurst, K.N. Buck, B. Sohst, G.J. Smith, M.C. Lohan, K.W. Bruland (2007). Micro- and macronutrients in the southeastern Bering Sea: Insight into iron- replete and iron-depleted regimes. Progress in Oceanography , 73 , 99-126

Aiello, I. W. and A. C. Ravelo (2012). Evolution of marine sedimentation in the Bering Sea since the Pliocene, Geosphere , 8, 1231-1253.

Alexander, V. and H.J. Niebauer (1981). Oceanography of the eastern Bering Sea ice-edge zone in spring. Limnology and Oceanography , 26( 6), 1111-1125.

Altabet, M.A. and R. Francois (1994). Sedimentary nitrogen isotopic ratio as a recorder for surface ocean nitrate utilization, Global Biogeochemical Cycles , 8(1), 103-116.

Andersen, K.K., N. Azuma, J.M. Barnola, M. Bigler, P. Biscayne, N. Caillon, J. Chappellaz, H.B. Clausen, D. Dahl-Jensen, H. Fischer, J. Fluckiger, D. Fritzsche, Y. Fujii, K. Groto- Azuma, K. Gronvold, N.S. Gundestrup, M. Hansson, C. Huber, C.S. Hvidberg, S.J. Johnsen, U. Jonsell, J. Jouzel, S. Kipfstuhl, A. Landais, M. Leuenberger, R. Lorrain, V. Masson-Delmotte, H. Miller, H. Motoyama, H. Narita, T. Popp, S.O. Rasmssen, D. Raynaud, R. Rothlisberger, U. Ruth, D. Samyn, J. Schwander, H. Shoji, M.L. Siggard- Andersen, J.P. Steffensen, T. Stocker, A.E. Sveinbjornsdottir, A. Svensson, M. Takata, J.L. Tison, T.H. Thorsteinsson, O. Watanabe, F. Wilhelms, and J.W.C. White (2004). High-resolution record of Northern Hemisphere climate extending into the last interglacial period, Nature , 431 , 147-151, doi:10.1038/nature02305.

Arzel, O., T. Fichefet, and H. Goosse (2006). Sea ice evolution over the 20 th and 21 st centuris as simulated by current AOGCMs, Ocean Modelling , 12 , 401-415, doi:10.1016/j.ocemod.2005.08.002.

Belt, S.T., G. Masse, S.J. Rowland, M. Poulin, C. Michel, and B. LeBlanc (2007). A novel chemical fossil of palaeo sea ice: IP 25 , Organic Geochemistry , 38 , 16-27.

Berger, A. and M. F. Loutre (1991). Insolation values for the climate of the last 10 million years, Quaternary Science Reviews , 10 , 297-317.

Bertrand, P., T.F. Pedersen, P. Martinez, S. Calvert, and G. Shimmield (2000). Sea level impact on nutrient cycling in coastal upwelling areas during deglaciation: Evidence from nitrogen isotopes. Global Biogeochemical Cycles , 14 (1), 341-355.

Bhatt, U.S., D.A. Walker, J.E. Walsh, E.C. Carmack, K.E. Frey, W.N. Meier, S.E. Moore, F.W. Parmentier, E. Post, V.E. Romanovsky, and W.R. Simpson (2014). Implications for the Arctic sea ice decline for the Earth System, Annual Review of Environment and Resources , 39 , 57-89.

Boyd, P.W., C.S. Law, C.S. Wong, Y. Nojiri, A. Tsuda, M. Levasseur, S. Takeda, R. Rivkin, P.J. Harrison, R. Strzepek, J. Grower, R.M. McKay, E. Abraham, M. Arychuk, J.Barwell- Clarke, W. Crawford, D. Crawford, M. Hales, K. Harada, K. Johnson, H. Kiyosawa, I. Kudo, A. Marchetti, W. Miller, J. Needoba, J. Nishioka, H. Ogawa, J. Page, M. Robert, H. Saito, A. Sastri, N. Sherry, T. Soutar, N. Sutherland, Y. Taira, F. Whitney, S.E. Wong, and T. Yoshimura (2004). The decline and fate of an iron-induced subarctic phytoplankton bloom, Nature , 428 , 549-553.

Brabets, T.P., B. Wang, and R.H. Meade (2000). Environmental and hydrologic overview of the Yukon River Basin, Alaska and Canada, U.S. Geological Survey Water Resources Investigations Report , 99-4204.

Brigham-Grette, J., and D.M. Hopkins (1995). Emergent Marine Record and Paleoclimate of the Last Interglaciation along the Northwest Alaskan Coast. Quaternary Research , 43 , 159- 173.

Brigham-Grette, J. (2001). New perspectives on Beringian Quaternary paleogeography, stratigraphy, and glacial history. Quaternary Science Review 20 , 15-24.

Brunelle, B.G., D.M. Sigman, M.S. Cook, L.D. Keigwin, G.H. Haug, B. Plessen, G. Schettler, and S.L. Jaccard (2007). Evidence from diatom-bound nitrogen isotopes for subarctic Pacific stratification during the last ice age and a link to North Pacific denitrification changes, Paleoceanography , 22 , 1-17.

Carlson, P.R. and H.A. Karl (1984). Discovery of two new large submarine canyons in the Bering Sea, Marine Geology , 56 , 159-179, doi: 10.1016/0025-3227(84)90011-2.

Carter, L.D. and T.A. Ager (1989). Late Pleistocene spruce (Picea) in northern interior basins of Alaska and the Yukon – Evidence from marine deposits in northern Alaska. In “Late Cenozoic History of the Interior Basins of Alaska and the Yukon – Proceedings of a Joint Canadian-American Workshop” (L.D. Carter, T.D. Hamilton, and J.P. Galloway, Eds.), U.S. Geological Survey Circular 1026 , 11-14

Chavez, F.P., M. Messie, and J. T. Pennington (2011). Marine primary production in relation to climate variability and change, Annual Review of Marine Science , 3, 227-260, doi:10.1146/annurev.marine.010908.163917.

Coachman, L.K. (1993). On the flow field in the Chirkov Basin. Continental Shelf Research , 13 , 481-508.

Cook, M.S., A.C. Ravelo, A. Mix, I.M. Nesbitt, and N.V. Miller (In Review). Tracing Bering Sea circulation with benthic foraminiferal stable isotopes during the Pleistocene, Deep Sea Research II .

Cortijo, E., S. Lehman, L. Keigwin, M. Chapman, D. Paillard, and L. Labeyrie (1999). Changes in meridional temperature and salinity gradients in the North Atlantic Ocean (30 ⁰- 72 ⁰N) during the last interglacial period, Paleoceanography , 14 , 23-33, doi:10.1029/1998PA900004.

DeMaster, D.J. (1981). The supply and accumulation of silica in the marine environment, Geochimica et Cosmochimica Acta , 45 , 1715-1732, doi:10/1016/0016-7037(81)90006-5.

Dinter, D.A., L.D. Carter, and J. Brigham-Grette. (1990). Late Cenozoic geologic evolution of the Alaskan North Slope and adjacent continental shelves. In “The Arctic Ocean Region” (A. Grantz, L. Johnson, and J.F. Sweeney, Eds.), 459-490. Geological Society of America, Boulder, CO.

Duce, R.A. and N.W. Tindale (1991). Atmospheric transport of iron and its deposition in the ocean, Limnology and Oceanography , 36(8) , 1715-1726.

Elderfield, H. and R.E.M. Rickaby (2000). Oceanic Cd/P ratio and nutrient utilization in the glacial Southern Ocean, Nature , 405 , 305-310.

Fan, S.M., W.J. Moxim, and H. Levy II (2006). Aeolian input of bioavailable iron to the ocean, Geophysical Research Letters , 33 , L07602, doi: 10.1029/2005GL024852.

Francois, R., M.A. Altabet, E.F. Yu, D.M. Sigman, M.P. Bacon, M. Frank, G. Bohrmann, G. Bareille, and L.D. Labeyrie (1997). Contribution of Southern Ocean surface-water stratification to low atmospheric CO2 concentrations during the last glacial period. Nature , 389 , 929-935.

Fry, B. (2006). Stable Isotope Ecology , New York: Springer.

Fung, I.Y., S.K. Meyn, I. Tegen (2000). Iron supply and demand in the upper ocean, Global Biogeochemical Cycles , 14 (1), 281-295.

Galbraith, E.D., M. Keinast, S.L. Kaccard, T.F. Pedersen, B.G. Brunelle, D.M. Sigman, and T. Kiefer (2008). Consistent relationship between global climate and surface nitrate utilization in the western subarctic Pacific throughout the last 500 ka, Paleoceanography , 23 , PA2212, 1-11.

Gardner, J.V., W.E. Dean, D.H. Klise, and J.G. Baldauf (1982). A climate-related oxidizing event in deep-sea sediment from the Bering Sea, Quaternary Research , 18 , 91-107, doi: 10.1016/0033-5894(82)90023-0.

Gargett, A.E. (1991). Physical processes and the maintenance of nutrient-rich euphotic zones, Limnology and Oceanography , 15 , 361-376.

Garrison, T. (2005). Oceanography: An Invitation to Marine Science , 5th ed., Belmont, CA; Brookes/Cole.

Gorbarenko, S.A., I.A. Basov, M.P. Chekhovskaya, J. Southon, T.A. Khusid, and A.V. Artemova (2005). Orbital and millennium scale environmental changes in the southern Bering Sea during the last glacial-Holocene: geochemical and paleontological evidence, Deep-Sea Research Part II , 52 , 2174-2185, http://dx.doi.org/10.1016/j.dsr2.2005.08.005.

Gowik, U. and P. Westhoff (2011). The path from C 3 to C 4 photosynthesis, Plant Physiology , 155(1) , 56-63.http://dx.doi.org/10.1104/pp.110.165308.

Granger, J., M.G. Prokopenko, D.M. Sigman, C.W. Mordy, Z.M. Morse, L.V. Morales, R.N. Sambrotto, and B. Plessen (2011). Coupled nitrification-denitrification in sediment of the eastern Bering Sea shelf leads to 15N enrichment of fixed N in shelf waters, Journal of Geophysical Research , 116 , 1-18.

Grebmeier, J.M. (2012). Shifting patterns of life in the Pacific Arctic and Sub-Arctic seas. Annual Review Marine Science , 4, 63-78.

Guo, L.D. and R.W. Macdonald (2006). Source and transport of terrigenous organic matter in the upper Yukon River: Evidence from isotope (δ 13 C, ∆14 C, and δ 15 N) composition of dissolved, colloidal and particulate phases, Global Biogeochemistry Cycles , 20 , GB2011, doi: 10.1029/2005GB002593.

Hamme, R.C., P.W. Webley, W.R. Crawford, F.A. Whitney, M.D. DeGradpre, S.R. Emerson, C.C. Eriksen, K.E. Giesbrecht, J.F.R. Gower, M.T. Kavanaugh, M. A. Pena, C.L. Sabine, S.D. Batten, L.A. Coogan, D.S. Grundle, and D. Lockwood (2010). Volcanic ash fuels anomalous plankton bloom in subarctic northeast Pacific, Geophysical Research Letters , 37 , L19604.

Heath, G.R. (1974). Dissolved silica in deep-sea sediment, in Hay, W.W., ed., Studies in Paleo- oceanography, Society of Economic Paleontologists and Mineralogists Special Publication , 20 , 77-93.

Henrichs, S.M. (1992). Early diagenesis of organic matter in marine sediments: progress and perplexity, Marine Chemistry , 39 (1-3), 119-149, doi:10.1016/0304-4203(92)90098-U.

Hopkins, D.M. (1959). Cenozoic History of the Bering Land Bridge. Science , 129 , 1519-1528.

Hopkins, D.M. (1967). Quaternary marine transgressions in Alaska. “ The Bering Land Bridge (D.M. Hopkins, Ed.), 47-90. Stanford Univ. Press, Stanford, C.A.

Hu, A., G.A. Meehl, W. Han, A. Abe-Ouchi, C. Morrill, Y. Okazaki, and M.O. Chikamoto (2012). The Pacific-Atlantic seesaw and the Bering Strait, Geophysical Research Letters, 39, L03702, doi:10.1029/2011GL050567.

Katsuki, K. and K. Takahashi (2005). Diatoms as paleoenvironmental proxies for seasonal productivity, sea-ice and surface circulation in the Bering Sea during the late Quaternary, Deep Sea Research II , 52 , 2110-2130.

Kaufman, D.S., S.L. Forman, P.L. Lea and C.W. Wobus (1996). Age of Pre-late-Wisconsin glacial-estuarine sedimentation, Bristol Bay, Alaska, Quaternary Research , 45 , 59-72.

Kaufman, D.S., R.C. Walters, J. Brigham-Grette, and D.M. Hopkins (1991). Middle Pleistocene age of the Nome River Glaciation, northwestern Alaska. Quaternary Research , 36 , 277- 293.

Keigwin, L.D., G.A. Jones, and P.N. Froelich (1992). A 15,000 year paleoenvironmental record from Meji Seamount far northwestern Pacific, Earth Planetary Science Letters , 111 (2-4), 425-440.

Kienast, F., S. Wetterich, S. Kuzmina, L. Schirrmeister, A. Andreev, P. Tarasov, L. Nazarova, A. Kossler, L. Frolova, and V.V. Kunitsky (2011). Paleontological records indicate the occurrence of open woodlands in a dry inland climate at the present-day Arctic coast in western Beringia during the Last Interglacial, Quaternary Science Reviews , doi:10.1016/j.quascirev.2010.11.024, in press.

Ladd, C., G.L. Hunt, Jr., C.W. Mordy, S.A. Salo, and P.J. Stabeno (2005). Marine environment of the eastern and central Aleutian Islands, Fisheries Oceanography , 14(Suppl. 1) , 22-38,

Lam, P.J. and J.K.B. Bishop (2008). The continental margin is a key source of iron to the HNLC North Pacific Ocean. Geophysical Research Letters , 35 , L07608.

Lam, P.J., L.F. Robinson, J. Blusztajin, C. Li, M.S. Cook, J.F. McManus, L.D. Keigwin (2013). Transient stratification as the cause of the North Pacific productivity spike during deglaciation, Nature Geosciences , 6, 622-626.

Lehmann, M.F., D.M. Sigman, D.C. McCorkle, B.G. Brunelle, S. Hoffmann, M. Keinast, G. Cane, and J. Clement (2005). Origin of the deep Bering Sea nitrate deficit: Constraints from the nitrogen and oxygen isotopic composition of water column nitrate and benthic nitrate fluxes, Global Biogeochemical Cycles , 19 (4), GB4005, doi:10.1029/2005GB002508.

Li, W.K.W., F.A. McLaughlin, C. Lovejoy, and E.C. Carmack (2009). Smallest algae thrive as the Arctic Ocean freshens, Science , 325 , 539, doi:10/1126/science.1179798

Lomstein, B.A., T.H. Blackburn, and K. Henriksen (1989). Aspects of nitrogen and carbon cycling in the northern Bering shelf sediment: 1. The significance of urea turnover in the mineralization of NH4+, Marine Ecology Prog. Ser. , 57 , 237-247.

Mahowald, N.M., A.R. Baker, G. Bergametti, N. Brooks, R.A. Duce, T.D. Jickells, N. Kubilay, J.M. Prospero, and I. Tegen (2005). Atmospheric global dust cycle and iron inputs to the ocean, Global Biogeochemical Cycles, 19 , GB4025, doi:10.1029/2004GB002402

Martin, J.H. (1990). Glacial-Interglacial CO 2 change: The iron hypothesis, Paleoceanography , 5(1), 1-13

Marting, J.H. and R.M. Gordon (1988). Northeast Pacific iron distributions in relation to phytoplankton productivity, Deep Sea Research , 35 , 177-196.

Matthiessen, J. and J. Knies (2001). Dinoflagellate cyst evidence for warm interglacial conditions at the northern Barents Sea margin during marine isotope stage 5. Journal of Quaternary Science , 16 (7), 727-737.

Measures, C.I. (1999). The role of entrained sediments in sea ice in the distribution of aluminum and iron in the surface waters of the Arctic Ocean, Marine Chemistry , 68 , 59-70, doi:10.1016/S0304-4203(99)00065-1.

Meyers, P.A. (1994). Preservation of elemental and isotopic source identification of sedimentary organic matter, Chemical Geology , 114 , 289-302.

Morales, L.V., J. Granger, B.X. Chang, M. G. Prokopenko, B. Plessen, R. Gradinger, D.M. Sigman (2014). Elevated 15N/14N in particulate organic matter, zooplankton, and diatom frustule-bound nitrogen in the ice-covered water column of the Bering Sea eastern shelf, Deep-Sea Research II , 109 , 100-111.

Morley, J., N.G. Pisias, and M. Leinen, (1987). Late Pleistocene time series of atmospheric and oceanic variables recorded in sediments from the subarctic Pacific. Paleoceanography , 2(1), 49-62.

Nagashima, K., Y. Asahara, F. Takeuchi, N. Harada, S. Toyoda, and R. Tada (2012). Contribution of detrital materials from the Yukon River to the continental shelf sediments of the Bering Sea based on the electron spin resonance signal intensity and crystallinity of quartz, Deep-Sea Research II , 61-64 , 145-154.

Nakatsuka, T., K. Watanabe, N. Handa, and E. Matsumoto (1995). Glacial to interglacial surface nutrient variations of Bering deep basins recorded by δ 15 N and δ 15 N of sedimentary organic matter. Paleoceanography , 10 (6), 1047-1061.

National Snow and Ice Data Center (2012). Arctic sea ice settles at record seasonal minimum, Arctic Sea Ice News Analysis , September, 19.

Nedashovskiy, A.P. and V.V. Sapozhnikov (1999). Variability in the components of the carbonate system and dynamics of inorganic carbon in the western Bering Sea in summer, in Loughlin, T.R. and K. Ohtani, eds, Dynamics of the Bering Sea: A summary of physical, chemical, and biological characteristics, and a synopsis of research on the Bering Sea: Fairbanks, University of Alaska Sea Grant Program College Report AK-SG- 99-03 , 311-322.

Niebauer, H.J. (1988). Effects of El Nino-Southern Oscillation and North Pacific weather patterns and interannual variability in the subarctic Bering Sea. Journal of Geophysical Research , 93 , 5051-5068

Niebauer, H.J., N.A. Bond, L.P. Yakunin, and V.V. Plotnikov (1999). An update on the climatology and sea ice of the Bering Sea. In: Loughlin, T.R. and K. Ohtani (Eds.), Dynamics of the Bering Sea . University of Alaska Sea Grant, Fairbanks, Alaska, 22-59.

Normak, W.R. and P. R. Carlson (2003). Giant submarine canyons: Is size any clue to their importance in the rock record?, in Chan, M.A. and A.W. Archer, eds., Extreme depositional environments: Mega end members in geologic time: Geological Society of America Special Paper 370 , 175-190, doi: 10.1130/0-8137-2370-1.175.

Nurnberg, D., D. Dethleff, R. Tiedemann, A. Kaiser, and S.A. Gorbarenko (2011). Okhotsk sea ice coverage and Kamchatka glaciation over the last 350 ka – Evidence from ice-rafted debris and planktonic δ 18 O, Paleogeography, Plaeoclimatology, Palaeoecology , 310 , 191-205, doi:10.1016/j.palaeo.2011.07.011.

Nurnberg, D. and R. Tiedemann (2004). Environmental change in the Sea of Okhotsk during the last 1.1. million years, Paleoceanography , 19(4) , PA4011, doi:10.1029/2004PA001023.

Okazaki, Y., K. Takahashi, H. Asahi, K. Katsuki, J. Hori, H. Yasuda, Y. Sagawa and H. Tokyama (2005). Productivity changes in the Bering Sea during the late Quaternary. Deep-Sea Research II , 52 , 2150-2162

Oppo, D.W., L.D. Keigwin, J.F. McManus, and J.L. Cullen (2001). Persistent suborbital climate variability in Marine Isotope Stage 5 and Termination II, Paleoceanography , 16 , 280- 292, doi:10.1029/2000PA000527.

Oppo, D.W., J.F. McManus, and J.L. Cullen (2006). Evolution and demise of the Last Interglacial warmth in the subpolar North Atlantic, Quaternary Science Reviews , 25 , 3268-3277, doi:10.1016/j.quascirev.2006.07.006.

Overland, J.E. and C.H. Pease (1982). Cyclone climatology of the Bering Sea and its relation to sea ice extent. Monthly Weather Review , 110 , 5-13.

Petit, J.R., J. Jouzel, D. Raynaud, N.I. Barkov, J.M. Barnola, I. Basile, M. Benders, J. Chappellaz, M. Davis, G. Delaygue, M. Delmotte, V.M. Kotlyakov, M. Legrand, V.Y. Lipenkov, C. Lorius, L. Pepin, C. Ritz, E. Saltzman, and M. Stievenard (1999). Climate and atmospheric history of the past 420,000 years from the Vostok ice core, Antarctica., Nature , 399 , 429-436.

Ragueneau, O., P. Treguer, A., Leynaert, R.F. Anderson, M.A. Brzezinski, D.J. DeMaster, R.C. Dugdale, J. Cymond, G. Fisher, R. Fracois, C. Heinze, E. Maier-Reimer, V. Martin- Jezequel, D.M. Nelson, and B. Queguiner (2000). A review of the Si cycle in the modern ocean: recent progress and missing gaps in the application of biogenic opal as a paleoproductivity proxy, Global and Planetary Change , 26 , 317-365.

Rau, G.H., C.W. Sullivan, and L.I. Gordon (1991). δ13 C and δ 15 N variations in Weddell Sea particulate organic matter, Marine Chemistry , 35 , 355-369, doi:10.1016/S0304- 4203(09)90028-7.

Reed, R.K., G.V. Khen, P.J. Stabeno, and A.V. Verkhunov (1993). Water properties and flow over the deep Bering Sea basin, summer 1991. Deep-Sea Research , 40 , 2325- 2334.

Reimnitz, E., M. McCormick, J. Bischof, and D. Darby (1998). Comparing sea-ice sediment load with Beaufort Sea shelf deposits: Is entrainment selective?, Journal of Sedimentary Research, 68 , 777-787, doi:10.2110/jsr.68.777.

Riethdorf, J.R., D. Nurnberg, L. Max, R. Tiedemann, S.A. Gorbarenko, and M.I. Malakhov (2013). Millenial-scale variability of marine productivity and terrigenous matter supply in the western Bering Sea over the past 180 kyr. Climate of the Past , 9, 1345-1373.

Riethdorf, J.R., B. Thibodeau, M. Ikehara, D. Nurnberg, L. Max, R. Tiedemann, and Y. Yokoyama (2015). Surface nitrate utilization in the Bering Sea since 180 ka BP: Insight from sedimentary nitrogen isotopes. Deep-Sea Research II ,

Rind, D. (1988). Latitudinal temperature gradients and climate change, Journal of Geophysical Research , 103 (D6), 5943-5971.

Rohling, E.J., K. Grant, C.H. Hemleben, M. Siddall, B.A.A. Hoogakker, M. Bolshaw, and M. Kucera (2008). High rates of sea-level rise during the last interglacial period, Nature Geoscience , 1, 38-42.

Royer, T.C. (1981). Baroclinic transport in the Gulf of Alaska, II, A fresh water driven coastal current, Journal of Marine Research , 39 , 251-266

Sakamoto, T., M. Ikehara, K. Aoki, K. Iijima, K. Nimurra, T. Nakatsuka, and M. Wakatsuchi (2005). Ice-rafted debris (IRD)-based sea-ice expansion events during the past 100 kyrs in the Okhotsk Sea, Deep-Sea Research II , 52 , 2275-2301.

Sancetta, C. (1982). Distribution of diatom species in surface sediments of the Bering and Okhotsk seas. Micropaleontology , 28 (3), 221-257.

Sancetta, C. (1983). Effect of Pleistocene Glaciation upon oceanographic characteristics of the North Pacific and Bering Sea , Deep-Sea Research , 30 (8a), 851-869.

Sancetta, C. and S.M. Silvestri, (1986). Pliocene-Pleistocene evolution of the North Pacific Ocean-atmosphere system, interpreted from fossil diatoms. Paleoceanography , 1(2), 163- 180.

Scherer, R.P. (1994). A new method for the determination of absolute abundances of diatoms and other silt-sized sedimentary particles. Journal of Paleolimnology , 12 , 171-179.

Schlung, S.A., A.C. Ravelo, I.W. Aiello, D.H. Andreasen, M.S. Cook, K.A. Dyez, T.P. Guilderson, J.P. LaRiviere, Z. Stroynowski (2013). Millenial-scale climate change and intermediate water circulation in the Bering Sea from 90 ka: A high-resolution record from IODP Site U1340. Paleoceacnography , 28 , 54-67, doi: 10.1029/2012PA002365.

Schrader, H.J. and R. Gersonde (1978). Diatoms and silicoflagellates, in Microplaentological Counting Methods and Techniques: An Exercise on an Eight Metres Section of the Lower Pliocene of Capo Rossello, Sicily, Utrecht Micropaleontol. Bull , vol. 17, edited by W.J. Zachariasse et al., 129-176, Utrecht, Schotanus and Jens, Odijk, Netherlands.

Schumacher, J.D., N.A. Bond, R.D. Brodeur, P.A. Livingston, J.M. Napp, and P.J. Stabeno (2002). Climate change in the southeastern Bering Sea and some consequences for biota. In: Hemple, G., and K. Sherman (Eds.), Large Marine Ecosystems of the World: Trends in Exploitation, Protection, and Research , in press.

Schumacher, J.D., C.A. Pearson, and J.E. Overland (1982). On exchange of water between the Gulf of Alaska and the Bering Sea through Unimak Pass. Journal of Geophysical Research , 47 (8), 5785-5795.

Schumacher, J.D. and P.J. Stabeno (1998). The continental shelf of the Bering Sea, in The Sea (1998), edited by A.R. Robinson and K.H. Brink, 789-822, John Wiley and Sands, New York.

Serreze, M., M.M. Holland and J. Stroeve (2007). Perspectives on the Arctic’s shrinking sea-ice cover, Science , 315 , 1533-1536.

Serreze, M.C. and R.G. Barry (2011). Processes and impacts of Arctic amplification: a research synthesis. Global Planetary Change , 77 , 85-96.

Shackleton, N.J. (1969). The last interglacial in the marine and terrestrial records. Proceedings of the Royal Society B174 , 135-154.

Shackleton, N.J., M. F. Sanchez-Goni, D. Pailler, Y. Lancelot (2003). Marine Isotope Substage 5e and the Eemian Interglacial. Global and Planetary Change , 36 , 151-155.

Sigman D.M. and K.L. Casciotti (2001). Nitrogen isotopes in the ocean. In: Steele, J.H., K.K. Turekian, and S.A. Thorpe (eds) Encycopedia of ocean sciences. Academic, London, 2449

Sigman, D.M., S.L. Jaccard, and G.H. Haug (2004). Polar ocean stratification in a cold climate, Nature , 428 , 59-63.

Slagstad, D., I.H. Ellingsen, and P. Wassmann (2011). Evaluating primary and secondary production in an Arctic Ocean void of summer sea ice: An experimental simulation approach, Progress in Oceanography , 90 , 117-131, doi:10.1016/j.pocean.2011.02.009.

Stabeno, P.J., J.D. Schumacher, and K. Ohtani (1999). Physical oceanography of the Bering Sea. In: The Bering Sea: a Summary of Physical, Chemical and Biological Characteristics and a Synopsis of Research . T.R. Loughlin and K. Ohtani (ed). North Pacific Marine Science Organization, PICES, Alaska Sea Grant Press, 1-28.

Stabeno, P.J., R.K. Reed and J.M. Napp (2002). Transport through Unimak Pass, Alaska. Deep- Sea Research II , 49 , 5919-5930

Stein, R (2008). Arctic Ocean Sediments: Processes, proxies, and paleoenvironment, in: Developments in Marine Geology , vol. 2 , edited by: Chamley, H., Elsevier, Amsterdam, 592.

Stroeve, J.C., M. C. Serreze, M.M. Holland, J.E. Kay, J. Malanik and A.P. Barrett (2012). The Arctic’s rapidly shrinking sea ice cover: a research synthesis, Climate Change , 110 , 1005-1027

Sukhanova, I.N., T.E. Whitledge, M.V. Flint, and D.A. Stockwell (2008). Effect of nutrient admendments on the summer phytoplankton dynamics on the Bering Sea Shelf under experimental conditions, Oceanology , 48( 6), 798-812.

Takahashi, K. (1998). The Bering and Okhotsk Seas: modern and past paleoceanographic changes and gateway impact. Journal of Asian Earth science , 16 , 49-58.

Takahashi, K. (2005). The Bering Sea and paleoceanography, Deep-Sea Research II , 52 , 2080- 2091, doi:10.1016/j.dsr2.2005.08.003.

Takahashi, K., A.C. Ravelo, C.A. Zarikian, and the IODP Expedition 323 Scientists (2011). IODP Expedition 323- Pliocene and Pleistocene paleoceanographic changes in the Bering Sea, Scientific Drilling , 11 , 4-13.

Tovar-Sanchez, A., C.M. Duarte, J.C. Alonso, S. Lacorte, R. Tauler, and C. Galban-Malagon (2010). Impacts of metals and nutrients released from melting multiyear Arctic sea ice, Journal of Geophysical Research , 115 , C07003, doi:10.1029/2009JC005685.

Townsend, J. (2002). Practical Statistics for Environmental and Biological Scientists , Chichester: Wiley, Print.

Tsuda, A., S. Takeda, H. Saito, J. Nishioka, Y. Nojiri, L. Kudo, H. Kiyosawa, A. Shiomoto, K. Imai, T. Ono, A. Shimamoto, D. Tsumune, T. Yoshimura, T. Aono, A. Hinuma, M. Kinugasa, K. Suzuki, Y. Sohrin, Y. Noiri, H. Tani, Y. Deguchi, N. Tsurushima, H. Ogawa, K. Fukami, K. Kuma, and T. Saino (2003). A meoscale iron enrichment in the western subarctic Pacific induces a large centric diatom bloom, Science , 300 , 958-961, doi:10.1126/science.1082000

Tsunogai, S., M. Kusakabe, H. Iizumi, I. Koike, and A. Hattori (1979). Hydrographic features of the deep water of the Bering Sea – The Sea of Silica, Deep-Sea Research , 26 , 641-659, doi:0011-7471/79/0601-0641.

Ueno, H. and I. Yasuda (2003). Intermediate water circulation in the North Pacific subarctic and northern subtropical regions, Journal of Geophysical Research , 108 (C11), 3348, doi: 10.1029/2002JC001372.

VanLaningham, S., N.G. Pisias, R.A. Duncan, and P.D. Clift (2009). Glacial-interglacial sediment transport to the Meiji Drift, northwest Pacific Ocean: evidence for timing of Beringian outwash, Earth and Planetary Science Letters , 277 , 64-72.

Warren, B (1983). Why is no deepwater formed in the North Pacific? Journal of Marine Research , 41 , 327-347.

Waelbroeck, C., L. Labeyrie, E. Michel, J.C. Duplessy, J.F. McMannus, K. Lambeck, E. Balbon, and M. Labracherie (2002). Sea-level and deep water temperature changes derived from benthic foraminifera isotopic records, Quaternary Science Reviews , 21 , 295-305, doi:10.1016/S0277-3791(01)00101-9.

Whitney, F.A., D.W. Crawford, and T. Yoshimura (2005). The uptake and export of silicon and nitrogen in HNLC waters of the NE Pacific Ocean, Deep Sea Research Part II , 52 (7-8), 1055-1067.