Recreational Post-Release Mortality and Harvest Slots: Implications for Stock Status and Management of Gulf of Mexico Red Snapper

RECREATIONAL POST-RELEASE MORTALITY AND HARVEST SLOTS: IMPLICATIONS FOR STOCK STATUS AND MANAGEMENT OF GULF OF MEXICO RED SNAPPER

ERIN COLLINGS BOHABOY

A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2019

ACKNOWLEDGMENTS

I would like to acknowledge the National Marine Fisheries Service Cooperative Research

Program for funding my dissertation research and the National Marine Fisheries Service / Sea

Grant Population and Ecosystem Dynamics Graduate Research Fellowship for providing my doctoral funding. I would like to thank my advisor, Dr. William Patterson, who provided me with unwavering guidance, support, good advice, and professional growth opportunities throughout my doctoral journey. I am grateful to my committee members, Dr. Robert Ahrens,

Dr. Shannon Cass-Calay, Dr. Susan Lowerre-Barbieri, and Dr. Miguel Acevedo for providing valuable direction and feedback on my research. I would also like to acknowledge Dr. Ruth

Carmichael and Dr. Alison Robertson who were members of my committee and provided me with valuable guidance at the University of South Alabama before I came to the University of

Florida.

I am grateful to Dr. Tristan Guttridge and Maurits van Zinnicq Bergmann at the Bimini

Biological Field Station in The Bahamas, and Dr. Neil Hammerschlag at the University of Miami

Rosenstiel School of Marine and Atmospheric Science who openly shared information and allowed me to use telemetry data on their tagged bull sharks. I am indebted to cooperating charterboat captains, crews, and customers for collaborating on this research: Captains Troy

Frady, Johnny Greene, Jason Vicars, Gary Jarvis, Sean Kelley, and crew members and customers of F/Vs Distraction, Intimidator, Aquastar, Backdown II, and Total Package. I would like to recognize Dr. Rick Methot and the NOAA Fisheries Stock Synthesis development and user support team for bringing us Stock Synthesis. I would also like to acknowledge Dr. Dan Goethel and Dr. Matthew Smith at NOAA Fisheries who are the authors of the most recent Gulf of

Mexico red snapper stock assessment model which I used as a starting point for many of my analyses.

I would like to thank the many students, technicians, and volunteers who were vital to the field operations and data collection that formed the basis of this dissertation: Dr. Steve Garner,

Joe Tarnecki, Miaya Glabach, Sarah Friedl, Dr. Kristen Dahl, Jordan Bajema, Joe Moss, Jessica

Van Vaerenbergh, Holden Harris, Joe Kuehl, Brian Jones, Matthew Shinego, John Knight,

Beverly Barnett, and Gracie Barnes. I would also like to recognize numerous classmates, co- workers, and friends for supporting me and enriching my life at both the University of Florida and Dauphin Island Sea Lab: Jordan Bajema, Gracie Barnes, Beverly Barnett, Derek

Chamberlain, Pearce Cooper, Dr. Kristen Dahl, Steve and Ashley Dykstra, Dr. Steve Garner,

Miaya Glabach, Holden Harris, Amanda Jefferson, Justin Lewis, Dr. Reid Nelson, Whitney

Scheffel, Joe Tarnecki, and Jessica Van Vaerenbergh, as well as many others.

I am grateful to my family, mainly my mother, Laura Lowder, my father, Edward

Collings, and my brother, Ethan Collings for helping me become a decent human being and for believing that I had the intelligence and perseverance to succeed. I would like to recognize my husband, Erich Bohaboy, who has been my best friend and partner in life for the last 15 years. I am thankful for the many influential people I have had the privilege to know, work with, and be mentored by over the years who supported me with their wisdom and generosity, and encouraged me to become a good scientist. Finally, I'd like to acknowledge the millions of people across the

U.S. who find peace, joy, and purpose in nature. These countless recreational fishers, hunters, outdoorspeople, and everyday citizens are the cornerstone of natural resource conservation, providing us the responsibility to undertake scientific inquiry and employ science-based management so that we may all benefit from healthy functioning ecosystems for generations to come.

TABLE OF CONTENTS

page

ACKNOWLEDGMENTS ...... 3

LIST OF TABLES ...... 7

LIST OF FIGURES ...... 8

LIST OF ABBREVIATIONS ...... 10

ABSTRACT ...... 12

CHAPTER

1 INTRODUCTION ...... 14

2 APPLICATION OF THREE-DIMENSIONAL ACOUSTIC TELEMETRY TO ASSESS THE EFFECTS OF RAPID RECOMPRESSION ON REEF FISH DISCARD MORTALITY ...... 19

Methods ...... 23 Acoustic Telemetry Array ...... 23 Tagging ...... 24 Data Analysis ...... 27 Results...... 30 Discussion ...... 36

3 FINE-SCALE MOVEMENT AND BEHAVIOR OF RED SNAPPER: AN ACOUSTIC TELEMETRY POSITIONING STUDY IN THE NORTHCENTRAL GULF OF MEXICO ...... 52

Methods ...... 55 Acoustic Telemetry Array and Fish Tagging ...... 55 VPS Position Accuracy ...... 56 Data Analysis ...... 58 Results...... 60 VPS Position Accuracy ...... 60 Reef Fidelity and Habitat Overview ...... 61 Proximity of Fish Positions to Reefs ...... 62 Factors Affecting Space Use ...... 63 Individual Variation ...... 65 Discussion ...... 65 Geoposition Accuracy ...... 65 Reef Association ...... 66 Site Fidelity ...... 68 Factors Affecting Space Use ...... 70

Movers vs. Stayers ...... 71 Future Directions ...... 73

4 DISCARDING BEHAVIOR, USAGE, AND ATTITUDES TOWARDS DESCENDER DEVICES ON NORTHERN GULF OF MEXICO REEF FISH CHARTER VESSELS ...... 96

Methods ...... 98 Results...... 100 Discussion ...... 104

5 HARVEST SLOTS AS A MANAGEMENT TOOL TO MAXIMIZE MARINE RECREATIONAL FISHING OPPORTUNITIES FOR GULF OF MEXICO RED SNAPPER ...... 120

Materials and Methods ...... 123 Results...... 129 Discussion ...... 132

6 CONCLUSIONS ...... 140

Discard Mortality ...... 141 Red Snapper Movement and Habitat Use ...... 143 Fisher Behavior and Descender Devices ...... 144 Harvest Slots for GOM Red Snapper ...... 145 Future Directions ...... 146

LIST OF REFERENCES ...... 149

BIOGRAPHICAL SKETCH ...... 163

LIST OF TABLES

Table page

2-1 Summary table of red snapper and gray triggerfish tagged and released during this study ...... 44

2-2 Nonlinear models testing factors affecting discard mortality for red snapper and gray triggerfish ...... 49

2-3 Number of tagged fish in each fate assignment category by time period following release ...... 51

3-1 Results of the drift test to evaluate positional accuracy of the 55-m array ...... 78

3-2 Estimated annual reef fidelity of red snapper ...... 82

3-3 Candidate GAMs explaining hourly 95% 2-D KDE for tagged red snapper ...... 84

3-4 Candidate GAMs explaining hourly 95% 3-D KDE for tagged red snapper ...... 85

3-5 Candidate GAMs explaining distance from the reef (in 2 dimensions) for tagged red snapper ...... 86

3-6 Candidate GAMs explaining depth above bottom for tagged red snapper ...... 87

4-1 Reaction statements presented to charterboat customers ...... 111

4-2 Catch and disposition of fish by red snapper season ...... 113

4-3 Catch rates by target species or group ...... 115

4-4 Species and groups with the highest prevalence of barotrauma signs ...... 115

4-5 Survey responses by group ...... 119

LIST OF FIGURES

Figure page

2-1 Maps of the study area ...... 43

2-2 Digital images of external acoustic transmitter attachments ...... 44

2-3 Calculated swim speeds of fish within the array ...... 45

2-4 Depth and positions of a red snapper consumed by a predator, a red snapper that lost the acoustic transmitter tag, and a gray triggerfish that emigrated from the array ...... 46

2-5 Estimated 48-hr mortality of red snapper and gray triggerfish ...... 47

2-6 Estimated red snapper discard mortality rates by depth and release method from previous studies ...... 48

2-7 Cox proportional hazards model-estimated survival for surface- and descender- released fish ...... 50

3-1 Map of the northern Gulf of Mexico indicating acoustic array locations ...... 75

3-2 Map of the 60-receiver shallow acoustic array deployed at 28−35 m depth from February 2016 to March 2017...... 76

3-3 Map of the 46-receiver deep acoustic array deployed at 48−55 m depth from August 2017 to July 2018 ...... 77

3-4 Map and figure describing the drift test performed in the 55-m acoustic array to confirm the positional accuracy of the array ...... 78

3-5 Summary of fish reef residency and fate for the 30-m array ...... 79

3-6 Images of common artificial reef types in the 30-m array ...... 80

3-7 Summary of fish reef residency and fate for the 55-m array ...... 81

3-8 An example large single steel and tire pyramid reef (reef m) in the 55-m array ...... 82

3-9 Distance of tagged fish from reefs 2 and Z in the 30-m array and distance of tagged fish from reefs a and c in the 55-m array ...... 83

3-10 Best GAMs predicting the effects of time of year on red snapper space use and movement ...... 88

3-11 Best GAMs predicting the effects of time of day on red snapper space use and movement ...... 89

3-12 Best GAMs predicting the effects of time of day category on red snapper space use and movement ...... 90

3-13 Best GAMs predicting the effects fish length on red snapper space use and movement ...... 91

3-14 Best GAMs predicting the effects of environmental variables on red snapper space use and movement...... 92

3-15 Distance between tagged fish and reefs ...... 93

3-16 Best GAM predicting space utilization by fish tagged in the 30-m array ...... 94

3-17 Best GAM predicting space utilization by fish tagged in the 55-m array ...... 95

4-1 Map showing observed charterboat fishing sites in the northern GOM ...... 111

4-2 Total observed effort by target and open/closed red snapper season...... 112

4-3 Observed catch of red snapper by length, disposition, season, and fishing target ...... 114

4-4 Condition of surface-released fish by species ...... 116

4-5 Digital images from GoPro3 camera footage of descender device deployments ...... 116

4-6 Surveyed charterboat customers' responses to questions from the survey ...... 117

5-1 Example length-based selectivity and retention functions...... 136

5-2 Effects of harvest slots and reduction of future discard mortality rate on model outputs ...... 137

5-3 Relative change in model outputs for five prospective management scenarios and two discard mortality rate reduction levels ...... 138

5-4 Relative change in abundance of fish by age in the beginning of year population and recreational catch per unit effort ...... 139

LIST OF ABBREVIATIONS

2DRMS Twice distance root mean squared

ACL Annual catch limit

AIC Akaike Information Criterion

AICc Small-sample Akaike Information Criterion

CI Confidence interval

CPUE Catch per unit effort

FES Federal Effort Survey

FL Fork length

FWC Florida Fish and Wildlife Conservation Commission

GAM Generalized additive model

GMFMC Gulf of Mexico Fisheries Management Council

GOM Gulf of Mexico

HPE Horizontal position error iTAG Integrated Tracking of Aquatic Animals in the Gulf of Mexico

KDE Kernel density estimate

MPA Marine protected area

MSE Management strategy evaluation

NMFS National Marine Fisheries Service

NOAA National Oceanic and Atmospheric Administration

OFL Overfishing limit

OTN Ocean Tracking Network

ROV Remote operated vehicle

SD Standard deviation

SE Standard error

SEDAR Southeast Data, Assessment, and Review

SS Stock Synthesis

TL Total length

VPS Vemco Positioning System

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

RECREATIONAL POST-RELEASE MORTALITY AND HARVEST SLOTS: IMPLICATIONS FOR STOCK STATUS AND MANAGEMENT OF GULF OF MEXICO RED SNAPPER

Erin Collings Bohaboy

December 2019

Chair: William F. Patterson III Co-chair: Robert N. M. Ahrens Major: Fisheries and Aquatic Sciences

Recreational fishing is a culturally and economically significant component of many U.S.

Gulf of Mexico (GOM) reef fish fisheries, including the red snapper (Lutjanus campechanus) fishery. Recreational fishers discard the majority of red snapper they catch, with a portion of those fish suffering post-release (discard) mortality. I evaluated the efficacy and likely impact of employing descender (rapid recompression or weighted return-to-depth) devices to reduce discard mortality in the GOM red snapper recreational fishery with a multifaceted approach that included 1) large-scale (~15 km2) three-dimensional acoustic telemetry experiments conducted to estimate the discard mortality of released red snapper; 2) observation of for-hire recreational fishing trips to examine fisher discarding behaviors, descender device use, and attitudes regarding catch-and-release fishing and descender devices; and 3) population dynamics model simulations to examine whether harvest slot regulations combined with descender device use could be an effective management approach in the GOM recreational red snapper fishery.

Descender devices reduced discard mortality of red snapper by 40%. For-hire fishing boat customers generally had positive experiences and attitudes towards descender devices, but most respondents had not previously used a descender device. Assuming descender devices become

widely used by recreational fishers, resulting in a 25–40% reduction in the discard mortality rate of red snapper, harvest slot regulations could satisfy several management objectives while detracting only slightly from others. In particular, harvest slots with a large maximum length limit would lead to longer recreational fishing seasons, reduced dead discarded biomass, increased recreational catch rates, and increased abundance of older fish in the population. My findings suggest discard mortality reduction outreach and education efforts targeting recreational fishers, for-hire operators, and crew, combined with incentives to recreational fishers to reduce red snapper discard mortality rates, would enable fisheries managers to use harvest slot regulations to increase recreational fishing opportunities and improve stock status of GOM red snapper.

CHAPTER 1 INTRODUCTION

Recreational fishing accounts for a significant portion of the annual harvest and bycatch of many fish stocks in U.S. waters of the Gulf of Mexico (GOM) (National Oceanic and

Atmospheric Administration 1998; Coleman et al. 2004b). These recreational fisheries are culturally important and contribute significantly to the national and local economies through fisher expenditures on fishing equipment, boats, and trips (Lovell et al. 2013). Numerous GOM reef fishes have been exploited since at least the 1950s and 1960s and some stocks are depleted or overfished (Hood et al. 2007; SEDAR 2013, 2014a, 2015a). Recreational fishers discard the majority of the catch of several reef fish species including red snapper (Lutjanus campechanus), gray triggerfish (Balistes capriscus), gag (Mycteroperca microlepis), greater amberjack (Seriola dumerili), and sharks (Sauls 2012; Sauls and Cermak 2013; SEDAR 2013, 2014a, 2014b, 2015a;

Garner and Patterson 2015). Unfortunately, traditional fisheries management measures, such as closed seasons, bag limits, and minimum sizes often exacerbate the issue of discarding in recreational fisheries (National Oceanic and Atmospheric Administration 1998; Campbell et al.

2014).

Discarded fish may suffer immediate or delayed mortality, thus diminishing benefits of release and contributing to wasted harvest and the national bycatch problem (Bartholomew and

Bohnsack 2005; Rummer and Bennett 2005; Strelcheck and Hood 2007). The percentage of live discarded fish assumed to die following release is a key parameter in stock assessment models, particularly for the eastern subarea of the GOM red snapper stock where the recreational fishery is the predominant source of mortality of adult fish (SEDAR 2013). The results of numerous studies indicate barotrauma is a significant contributor to reef fish post-release mortality

(Bartholomew and Bohnsack 2005; Campbell et al. 2010b). Barotrauma occurs when fish

experience a rapid drop in pressure as they are brought to the surface by anglers and the gas inside the swim bladder expands. Signs of barotrauma include everted stomach, protruding eyes

(exophthalmia), prolapsed intestines and gonadal tissue, bulging scales, and an inability to submerge once released. Fish suffering from barotrauma may experience immediate, observable mortality, such as being preyed upon at the surface by seabirds or marine mammals, or the harmful effects of capture and release may take days or weeks to result in mortality (Rummer and Bennett 2005; Campbell et al. 2010a).

Recent research suggests recompressed fish, such as those released with descender devices, may have improved post-release survival (Hochhalter and Reed 2011; Butcher et al.

2012; Benaka et al. 2014). Descender devices include several different designs of weighted hooks, clamps, or cages which are used to return fish to depth. Forcing fish that would otherwise be buoyant at the surface to re-submerge may reduce predation by delivering the fish as close as possible to the relative safety of reefs where they were captured. Studies on the efficacy of descender devices to increase post-release survival of recreationally caught GOM red snapper are limited, but their results suggest potential benefits of using descender devices over venting or surface release depend on factors such as temperature, the presence of predators, water depth, and hook injury (Diamond et al. 2011; Drumhiller et al. 2014; Curtis et al. 2015; Williams et al.

2015).

A key to estimating acute and chronic effects of catch and release on fishes and evaluating the efficacy of descender devices to reduce post-release mortality is the ability to estimate survival following release. Continuing advances in underwater acoustic telemetry have provided researchers with a reliable means to track tagged fish in situ for months to years, providing a novel means to estimate long-term discard mortality and understand the effects of

handling on survival (Pollock and Pine 2007; Bacheler et al. 2009b; Espinoza et al. 2011;

Decelles and Zemeckis 2014). Acoustic telemetry systems consist of numerous receivers which detect and log acoustic transmissions emitted from transmitters. Transmitter tags are either attached externally or implanted inside fish and regularly transmit a brief pattern of ultrasonic pulses which identify unique tags. Transmitter tags with advanced sensors are also capable of transmitting environmental information, such as pressure (depth), temperature, and acceleration.

A particularly powerful telemetry approach relies on a large number of acoustic receivers spaced closely enough to provide a high probability that each acoustic tag transmission will be simultaneously detected by multiple receivers. With time-difference-of-arrival positioning, the speed of sound and time delay between detections of a common acoustic transmission on at least

3 receivers is used to triangulate the location of the transmission (Smedbol et al. 2014). The resulting dataset yields geoposition estimates for tagged fish every few minutes, offering rare insight into the spatial and temporal fine-scale behavior, movement, and survival of tagged fish.

The geoposition estimates are of sufficiently high resolution (over both time and space) to allow inference about post-release fates of tagged fish, including predation, which is rarely observed and may be an underestimated component of discard mortality (Raby et al. 2014).

Making informed management decisions regarding descender devices requires empirical data not only on the efficacy of these devices to reduce discard mortality, but also on the practical application of the prospective regulations within the recreational fishery and the willingness of fishers to follow them. Through cooperative research, for-hire fishing vessels, which harvest approximately 42% of the GOM recreational red snapper quota annually (SEDAR

2018), provide a valuable platform for at-sea fisheries observers to examine the fishing and discarding behavior of recreational fishers and make inferences about descender devices and how

catch-and-release regulations may apply within the fishery. Traditionally, estimating discarded catch from the GOM recreational reef fish fishery relied on self-reporting by fishers (SEDAR

2005, 2013), but results of several studies have demonstrated observers can collect detailed data on recreational fishing practices (depth, gear, handling) and discarded catch (species, length, condition) (Garner and Patterson 2015; Sauls et al. 2017; SEDAR 2018). Through interviews and surveys, observers can investigate the receptivity of customers, captains, and deckhands towards catch-and-release fishing and using descender devices, which may give a prediction of compliance with a best practices recommendation or management regulation.

Decreased red snapper discard mortality could potentially enable the Gulf of Mexico

Fishery Management Council (GMFMC) to consider harvest slot regulations. A harvest slot regulation specifies that fishers may only harvest intermediate-sized fish which are greater than a minimum and less than a maximum length, requiring all fish outside of the slot size to be discarded. Harvest slot regulations have been successful in increasing the numbers of valuable older and larger female fish, decrease harvest rates, and increase opportunities for anglers to catch very large fish in freshwater and nearshore fisheries where discard mortality rates are generally low (Dotson et al. 2013; Long et al. 2015). However, catch-and-release regulations such as harvest slots can lead to increased numbers of dead discards if the mortality of released fish is too high (Farmer et al. 2014). Simulations of the effect of reduced discard mortality under the assumption of widespread descender device usage and harvest slot regulations in the recreational GOM red snapper fishery would provide valuable guidance to the GMFMC.

The overall goal of my dissertation research was to test factors affecting discard mortality of red snapper and determine whether employing descender devices within the recreational fishery would likely improve post-release survival of released fish. The ultimate objective was to

provide management advice on actions that could improve sustainability of the GOM red snapper fishery and increase recreational fishing opportunities. In Chapter 2, I investigated whether using descender devices to release recreationally caught red snapper improves survival, and I quantified how season (air/water temperature), capture depth, handling time, fish size, and fish condition modulate these effects. The objective of Chapter 3 was to provide insight into the movement and three-dimensional space use of red snapper, yielding valuable information on the effects of factors such as time of day, water temperature, sea state, fish size, and reef characteristics on patterns of red snapper habitat use. These high-accuracy spatial habitat behavior data can be integrated in future analyses on the design and placement of artificial reefs or spatial closed areas. In Chapter 4, my objective was to quantify how fishing behavior

(including variables such as target species, season, depth fished, and vessel fished from) influenced catch and discarding rates in the GOM for-hire recreational reef fish fishery, and I also increased observations of how fishers use descender devices to release fish. The observer study described in Chapter 4 was conducted to evaluate the practicality of descender device regulations or recommendations within the GOM for-hire recreational reef fish fishery. In

Chapter 5, I integrated updated estimates of red snapper discard mortality using descender devices (Chapter 2) to estimate how harvest slot regulations would affect recreational fishing opportunities (season length, catch rates, or size of fish in catch).

CHAPTER 2 APPLICATION OF THREE-DIMENSIONAL ACOUSTIC TELEMETRY TO ASSESS THE EFFECTS OF RAPID RECOMPRESSION ON REEF FISH DISCARD MORTALITY

Marine recreational fishing is an important economic activity and source of non- monetary benefits for people around the world (Arlinghaus et al. 2007; Lovell et al. 2013). In addition to fish they harvest, recreational fishers often release (discard) a large portion of their catch. In the U.S., Canada, and Europe, marine recreational discarding rates often exceed 50% of the total catch and approach 100% for some species (Ferter et al. 2013; National Marine

Fisheries Service 2017; Fisheries and Oceans Canada 2019). Catch-and-release fishing has historically been practiced in recreational fisheries as a conservation and management strategy

(Radonski 2002). However, discarded fish that die as a result of being captured and released are still removed from the population, reducing the intended conservation benefits of release

(Bartholomew and Bohnsack 2005; Rummer and Bennett 2005; Strelcheck and Hood 2007). The proportion of live discarded fish that die following release (discard mortality rate) may approach

100% in some recreational fisheries and is influenced by multiple factors including environmental conditions (e.g., depth of capture, water temperature), fish condition (e.g., species, size, presence of hook-related injuries), and handling (e.g., time out of water, time required for an angler to land a fish once hooked) (Muoneke and Childress 1994; Bartholomew and Bohnsack

2005; Brownscombe et al. 2017). In fisheries where recreational catches, discarding rates, or discard mortality are high, dead discards can represent a significant portion of stock removals. In these cases, reducing uncertainty of discard mortality estimates, improving understanding of the factors that affect discard mortality, and developing methods to minimize discard mortality may be instrumental in successful management.

Estimating discard mortality and quantifying the factors affecting it are particularly challenging for marine fish species, largely due to the difficulty of monitoring the fate of

released fish. Traditional approaches to studying discard mortality, such as mark-recapture, enclosure, and inferences from physiological studies, often require numerous assumptions

(including emigration/immigration rates of fish, angler reporting rates, and natural and fishing mortality), expose fish to the deleterious effects of prolonged captivity, or fail to account for the effects of predation. Ultrasonic underwater acoustic telemetry can provide estimates of the three- dimensional positions of tagged fish, and is a reliable means to track the movement, behavior, and survival of tagged fish for months to years in their natural environment. In particular, predation may be a greatly underestimated contributor to discard mortality because predation events are rarely observed using approaches traditionally applied in post-release mortality studies

(Raby et al. 2014). Acoustic transmitter tags often are surgically implanted in the abdominal cavity of fish, requiring sedation, extended handling, and may rupture over-inflated swim bladders of fish suffering from barotrauma, or otherwise add extraneous variables to the process of estimating release mortality. Alternatively, external attachment of acoustic transmitter tags

(e.g., Curtis et al. 2015; Capizzano et al. 2016; Dance et al. 2016; Runde and Buckel 2018) reduces handling trauma and preserves barotrauma symptoms, thus improving estimates of barotrauma-related mortality.

Red snapper (Lutjanus campechanus) are highly sought-after by recreational fishers in the U.S. Gulf of Mexico (GOM) who annually discard more than 70% of red snapper they catch

(SEDAR 2018). Discard mortality rate estimates for recreationally caught red snapper vary from near zero to greater than 80% depending on factors such as release method, handling, season, capture depth, and fish condition (Campbell et al. 2014; Drumhiller et al. 2014; Curtis et al.

2015). Gray triggerfish (Balistes capriscus) are another popular recreational reef fish species in the northern GOM which are targeted by recreational fishers with hook and line over artificial

reefs where they are also caught incidentally to red snapper. GOM gray triggerfish have experienced long-term stock declines concurrent with reductions in recreational fishing seasons and increased discards; 70% of recreationally caught gray triggerfish in the eastern GOM are discarded annually (SEDAR 2015a). Discard mortality of recreationally caught-and-released gray triggerfish is rarely studied and most survival estimates of caught-and-released gray triggerfish are fairly high based on recapture rates of tagged fish and observations of fish behavior at release (85–100%, Patterson et al. 2002; Rudershausen et al. 2013). Discard mortality was assumed to be 5% in the most recent GOM gray triggerfish stock assessment

(SEDAR 2015a). However, Runde et al. (2019) used underwater tagging to control for the effects of barotrauma and estimated discard mortality of gray triggerfish in the southeastern U.S. recreational fishery was much higher (65–66%) and could account for extensive stock removals when considered together with the magnitude of annual discards.

Fish experience a rapid drop in pressure as they are brought to the surface by anglers, leading to barotrauma which can be a significant contributor to post-release mortality of discarded fish (Bartholomew and Bohnsack 2005; Campbell et al. 2010a). Venting (the practice of releasing gas from a fish's swim bladder with a large gauge hypodermic needle) has been proposed as a means to improve post-release survival of marine fish suffering from barotrauma, but results are equivocal regarding the benefits of venting and some investigators suggest improperly venting fish can damage organs and increase discard mortality (Wilde 2009; Eberts and Somers 2017). Rapid recompression of released fish using weighted return-to-depth tools, also known as descender devices, is an alternative means to alleviate barotrauma symptoms and potentially improve post-release survival. Descender devices include several different designs of weighted hooks, clamps, or cages. Forcing fish that would otherwise be buoyant at the surface to

re-submerge may reduce predation by delivering reef fishes as closely as possible to the relative safety of the reef structure from which they were captured while avoiding the internal trauma of venting. Evidence supporting the efficacy of descender devices in reducing discard mortality is largely based on studies which show fish recompressed in the laboratory or in cages (absent predation) have increased survival (Parker et al. 2006; Jarvis and Lowe 2008; Overton et al.

2008; Pribyl et al. 2012). In very few studies has the effect of cage-less descender devices on discard mortality of marine fish been examined (Sumpton et al. 2010; Hochhalter and Reed

2011; Curtis et al. 2015). In contrast to the U.S. recreational Pacific rockfish fishery, where the use of descender devices has been widely advocated (Chen 2012; California Sea Grant et al.

2014), descender devices have yet to gain widespread use among GOM recreational reef fish fishers, and no management regulations exist to require or encourage their use.

The goals of this study were to investigate whether descender devices reduce discard mortality in recreationally caught-and-released red snapper and gray triggerfish, and to evaluate the effects of other variables that might affect survival, including season (air/water temperature), capture depth, handling time, and fish condition. Large-area (>15 km2) three-dimensional geopositioning acoustic telemetry arrays were used to monitor fine-scale movement and behavior of tagged fish for up to 1 year, providing insight into the importance of predation on post-release survival, which has traditionally been overlooked in studies of discard mortality (Raby et al.

2014). During this study, techniques were developed for quickly applying external acoustic transmitter tags to red snapper and gray triggerfish without the need for sedation and surgery, thus more closely approximating the handling of recreationally caught-and-released fish.

Methods

Acoustic Telemetry Array

An array of 60 Vemco (Bedford, Nova Scotia, Canada) VR2 acoustic receivers was deployed 28 km south of Pensacola Beach, Florida from February 2016 to March 2017 at 28 to

35 m depth (hereafter "30-m array”). In September 2016, the array was shifted approximately 0.5 km to the south and expanded in the southeast corner to include additional artificial reefs (Fig. 2-

1). The array was reduced to 46 receivers and moved to a deeper location (48–55 m) approximately 80 km south of Orange Beach, Alabama from August 2017 to July 2018

(hereafter "55-m" array). Habitat within the study areas of each array deployment consisted of open sand bottom interspersed with numerous artificial reef structures (cement pyramids, reef balls, and chicken coops) and likely also included some natural low-relief limestone hard-bottom habitat in the 55-m array. Acoustic receivers were placed in a grid that provided geopositioning capability of tagged fish in an area >15 km2 at each array. The maximum distance between receivers was 600 m, allowing for >50% probability that acoustic tag transmissions within the array could be detected by at least 3 receivers simultaneously based on previous range testing

(Patterson 2013). Receivers within each array were a mix of Vemco model VR2Tx and VR2W receivers. Model VR2Tx receivers were deployed with the internal synchronization (sync) transmitters set to very high output (160 dB), while each model VR2W receiver was deployed with a Vemco V16-5x sync transmitter (set to output 162 dB) suspended 2 m above the receiver on a line attached to a foam buoy. Each receiver was attached to the top of a 2-m tall PVC support pipe that was set in a 36-kg cement base. Grab lines were attached between the cement base and the support pipe that could be used as hoist points during deployment or retrieval. Most receiver bases had a line attached to a floating buoy approximately 2 m above the receiver to increase visibility or suspend a V16-5x sync transmitter (for model VR2W receivers). Each

receiver-base unit was lowered to the sea floor on a hook and line from the side of the boat and then located using the boat's sonar depth sounder to measure the GPS coordinates and ensure the acoustic receiver was properly deployed in an upright position.

Acoustic receivers were retrieved midway through each array deployment in September

2016 and March 2018, respectively, to clean any fouling organisms from the acoustic hydrophone and offload the logged acoustic transmission data from each receiver. During retrieval, the vessel’s captain would locate the receiver-base unit using the boat’s sonar depth sounder. A heavy retrieval line connected to a large hook was attached to a VideoRay

(Pottstown, Pennsylvania, U.S.) Pro4 mini remotely operated vehicle (ROV). The pilot maneuvered the ROV to attach the retrieval hook to one of the base’s grab lines. The hook was mounted on the ROV so that it would easily detach when the ROV was flown away from the receiver base, thus leaving the retrieval line and hook attached to the base. After ensuring the

ROV was free from the retrieval line, the receiver base was raised to the surface and brought onboard the boat with the assistance of a stainless steel davit and an electric winch. Using the

ROV enabled the retrieval of 20−30 receiver-base units in a single day and retrieving the cement base in addition to the receiver avoided leaving marine debris at study sites.

Tagging

Fish were tagged with Vemco V13P-1x acoustic transmitter tags which transmitted a 153 dB unique acoustic ID code and pressure value at random intervals between 1 and 3 minutes

(expected battery life = 468 days). Tags were attached to fish externally to minimize handling time and avoid rupturing the swim bladder which often occurs when fish suffering from barotrauma are subjected to tag implantation in the abdominal cavity. There were 2 different tag attachment methods used in this study. For the majority of tagged fish, the tag was secured to a

2-mm diameter threaded stainless steel bar with a 6.35-mm nylon-lined locking stainless steel

hex nut. The stainless steel bar was sharpened on one end and inserted through the fish’s dorsal pterygiophores and secured with a second 6.35-mm stainless steel hex nut behind a 3.2-mm thick polyethylene disk. Silicon disks (2.4−3.2 mm thick) were placed under the tag and under the polyethylene disk to minimize abrasion (Fig. 2-2A–B).

This stainless steel bar acoustic tag attachment device was tested on wild-caught red snapper (n = 3) that were held in captivity at Dauphin Island Sea Lab (Dauphin Island, Alabama,

U.S.), with the goal of achieving greater than 2-week tag retention. These tagging trials were performed with "dummy" V13P tags, which had the same dimensions, weight, and buoyancy as regular V13P tags but did not transmit acoustic signals. One fish shed its tag at 39 days which was the result of becoming tangled in the net cover of the aquaculture tank, causing the entire tag attachment to tear dorsally through the fish's back. Another fish lost its tag at 54 days when the stainless steel hex nut unscrewed from the threaded bar. This potential issue was controlled for in the field by using only new hex nuts where the nylon liner was not compressed from previous use or by slightly bending the ends of the stainless steel rod after affixing tags to fish such that nuts could not spin off the end of the threaded bar. The tag retention trial was terminated after

220 days and the remaining fish which had retained its tag for the duration was euthanized.

An alternate tag attachment method was used to attach acoustic tags to red snapper during late summer 2017 in the deep array. Tags were attached with this second method to a medium- sized (20 mm length x 10 mm width) Domeier dart head (Domeier et al. 2005) with approximately 3 cm of polymer-coated braided stainless steel fishing line. Marine heat-shrink tubing was applied over the tag cap to reduce movement and friction against the side of the fish

(Fig. 2-2C). Domeier dart heads are constructed of soft polymer and polyethylene terephthalate surgical fibers and are designed to heal into the muscle tissue of tagged fish. These dart heads

have been used to quickly attach acoustic and satellite tags to tunas (Domeier et al. 2005), sharks

(Rogers et al. 2013), and groupers (Angela Collins, pers. comm.), minimizing handling time and producing high tag retention estimates. However, after examining the acoustic detection data from fish tagged with Domeier tag attachment device, it was apparent that red snapper tag retention, while adequate to estimate release mortality, did not always allow us to track fish for longer (>1 month) time periods. Therefore, I returned to the initial tagging method for the final tagging event in spring 2018.

Red snapper and other reef fishes were captured from artificial reefs within the study area using hook and line baited with cut squid or herring. Red snapper were the primary species tagged in this study (n = 141); however, several gray triggerfish (n = 26) were also tagged. Fish were tagged during 4 events: 1) spring 2016 and 2) late summer 2016 in the 30-m array, and 3) late summer 2017 and 4) spring 2018 in the 55-m array. Each fish was held in a damp V-shaped silicone-covered measuring board, measured to the nearest mm fork or total length (FL or TL) and tagged externally with a V13P acoustic transmitter tag using either of the methods described above. Acoustically tagged fish also were tagged with a Floy (Seattle, Washington, U.S.) dart tag that advertised a $50 reward and toll-free phone number to report tagged fish. The presence/absence of traumatic hooking injury, any signs of barotrauma (exophthalmia, pronounced bloating, prolapsed intestine or gonads, protruding scales, or everted stomach), and total time out of water for dehooking and tagging were recorded for each tagged fish. The fight time (time between when a fish was hooked and reached the surface) was recorded for most fish based on verbal indication of the hooking event by the angler. During the final tagging event in spring 2018, a Reefnet (Mississauga, Ontario, Canada) Sensus Ultra depth logger was attached to the terminal fishing tackle used to capture fish (Murie and Parkyn 2013). Depth profiles were

examined to identify when a fish was hooked and subsequently reached the surface to determine hooking depth and fight time. Bottom temperature at each artificial reef where fish were captured was taken from the average daily logged temperature by the closest VR2Tx receiver. Air temperature on each day when fish were tagged was acquired from the National Data Buoy

Center at Station 42012 (approximately 48 km west of the 30-m array; National Oceanic and

Atmospheric Administration and National Weather Service 2017) and Station 42040

(approximately 77 km west/southwest of the 55-m array; National Oceanic and Atmospheric

Administration and National Weather Service 2019).

Each tagged fish was released either at the surface or with a descender device over the reef where it was captured. A SeaQualizer (Davie, Florida, U.S.) descender device was primarily used to release fish at depth. However, a second type of descender device was also used to return

7 red snapper to depth in spring 2016, but its use was discontinued when some fish prematurely detached from the device at the surface. A downward looking GoPro (San Mateo, California,

U.S.) Hero3 camera was mounted above each descender device to record fish descent and release to evaluate the performance of the descender device, behavior of released fish, and possible predator interactions. Fish released at the surface were observed and assessed for release condition following (Patterson et al. 2002): condition-1 = fish immediately oriented to the bottom and swam down rapidly; condition-2 = fish oriented to the bottom and swam down slowly or erratically; condition-3 = fish remained on the surface; and condition-4 = fish was apparently dead at the surface, including from predation.

Data Analysis

Detection data were offloaded from acoustic receivers in September 2016, March 2017,

April 2018, and July 2018 and sent to Vemco for Vemco Positioning System (VPS) geolocation estimation. Position estimates with horizontal position error (HPE; Smith 2013) in the upper 5th

percentile for each of the 4 datasets were excluded from further analyses. Successive geoposition estimates separated by less than 10 minutes (600 seconds) were used to calculate swim speeds of tagged fish. The average swim speed (v) in meters per second of a tagged fish as it moved from position 1 with coordinates (Lat1, Lon1) at time t1 to position 2 with coordinates (Lat2, Lon2) at time t2, where coordinates are in radians, was calculated as

퐿푎푡 − 퐿푎푡 퐿표푛 − 퐿표푛 √ 2 2 1 ( ) ( ) 2 2 1 2푟 ∗ 푎푟푐푠𝑖푛 푠𝑖푛 ( 2 ) + 푐표푠 퐿푎푡1 푐표푠 퐿푎푡2 푠𝑖푛 ( 2 ) 푣 = (2-1) (푡2 − 푡1) where r = 6.371 × 106 m is the assumed mean radius of the Earth.

Tagged fish were assigned a fate based on estimated swim speeds, geographical movements, and depth below the surface. The days to each fate were calculated and fish fates were binned for some analyses over each of 3 time periods: immediate (within 48 hours of release), short-term (48 hours to 14 days after release), and long-term (greater than 14 days after release). The possible assigned fates were predation, emigration, tag loss, surface mortality, harvest, survival, and unknown. Predation and tag loss were indicated by an abrupt change in tag movement or depth. Tags from fish that were preyed upon moved faster than 0.5 m/s through the array and did not center around reef locations (see below for rationale). Tags that were stationary on the bottom were assumed to have detached from fish and were classified as tag loss. Fish that were classified as surface mortalities were either observed dead after release or were detected only at the surface as the fish drifted from the array. Harvested fish disappeared from within the center of the array and were reported by fishers, while emigrating fish moved toward and then disappeared from the edge of the array. Tagged fish that were still alive at the end of each time period were classified as alive and present within the array. In instances when position and depth

data of tagged fish were insufficient to differentiate among potential fates, the fate was classified as unknown.

Mortality (predation or surface mortality) was attributed to capture and release (i.e., discard mortality) if it occurred during the immediate (0−48 hrs.) time period following release.

Point estimates and SEs of discard mortality were estimated assuming a binomial error distribution (as by Pollock and Pine 2007). The point estimate of mortality is 푀̂ = 푑/푛 where known mortalities (d = number of fishes assigned to either predation or surface mortality fates) divided by the total number of "at-risk" tagged fish (n = number of tagged fish with known fates excluding individuals assigned the fate of harvest or emigration). The standard error of estimated mortality is 푆퐸푀̂ = √푀̂(1 − 푀̂)/푛.

The influence of release method (surface versus at depth with a descender device), season, reef depth, fish length, presence/absence of trauma to the mouth or gills from hooking or handling, presence/absence of barotrauma symptoms, handling time, fight time, difference between water and air temperature (ΔT, where ∆푇 = 푇푏표푡푡표푚 − 푇푎𝑖푟), and inadvertent venting

(caused by insertion of the anchor tag or a fish's everted esophagus or stomach being punctured by its teeth) on an individual's probability of mortality in the immediate time period was explored with a linear modeling approach. The predicted probability of mortality for each individual (푀̂𝑖 for i = 1 … n) was modeled as a linear function of potential explanatory variables

푒훽0+훽1푋1+훽2푋2+⋯+훽푘푋푘 (X1, X2, …, Xk) and model parameters (β1, β2, …, βk), where 푀̂ = , with 𝑖 1+푒훽0+훽1푋1+훽2푋2+⋯+훽푘푋푘

푀 푛 ̂ 푖 ̂ (1−푀푖) the total binomial loglikelihood being ∑𝑖=1 푀𝑖 (1 − 푀𝑖) . Categorical variables such as release method, season, capture depth, presence/absence of hooking trauma, and presence/absence of barotrauma were coded as {0,1} and Mi was the assigned fate of each tagged fish from the telemetry data (where mortality = 1, survival to next time period = 0). Continuous

variables (length, time out of water, fight time, and ΔT) were scaled to have mean = 0 and variance = 1. For red snapper where only FL was measured, FL was converted to TL following

푇퐿 = 1.0812 ∗ 퐹퐿 − 0.950 (in mm; SEDAR 2018). Candidate models were evaluated for parsimony with the small-sample Akaike Information Criterion (AICc; Burnham and Anderson

2002).

The risk of mortality over time for fish released with a descender device relative to fish released at the surface was evaluated using Cox proportional hazards models (Cox and Oakes

1984). Cox proportional hazards models describe the risk of mortality as a function of elapsed time (i.e., are well suited to staggered-entry designs such as this study where individuals are tagged and released during multiple events). Proportional hazards models are also well suited to describe the relative risk of mortality between 2 different treatments (e.g., fish released at the surface or at depth with a descender device) and can accommodate datasets where individuals leave the study and must be censored from the model. Fish that were assigned fates of emigration, harvest, tag loss, or were still alive when the acoustic array was retrieved were censored at the date of the event. Fish reported harvested or recaptured either outside the array or after having lost the acoustic tag were censored from the model on the date of harvest or recapture, not on the earlier date of tag loss or emigration. I included the maximum number of covariables (thus minimizing risk of falsely rejecting variables) found to be informative in the linear models of discard mortality (were within 2 AICc from the lowest AICc models; Burnham and Anderson 2002). Cox proportional hazards models were fit, and model performance evaluated with the "survival" package in R (Therneau and Lumley 2015; R Core Team 2016).

Results

In total, 141 red snapper ranging from 30.5 to 89.0 cm TL (mean ± SD: 52.1 ± 14.1 cm) and 26 gray triggerfish ranging from 32.1 to 50.0 cm FL (mean ± SD: 41.5 ± 5.5 cm) were

tagged with acoustic transmitter tags and released into the acoustic array (Table 2-1). Fish were captured from artificial reefs, tagged, and released into the acoustic array during 4 events: 1) spring 2016 (26 April−3 May); 2) late summer 2016 (14 September); 3) late summer 2017 (2

September); and 4) spring 2018 (11−24 April). Fight time range was 9–220 s (mean ± SD: 96 ±

55 s) and tagged fish were out of the water from 39 to 330 s (mean ± SD: 99 ± 38 s). Fight times were mistakenly not recorded for the majority (n = 30) of red snapper tagged in late summer

2017. Thirteen (13) red snapper showed signs of traumatic hooking (i.e., the fish hook was removed from the fish’s throat, gills, or gut, or bleeding from the mouth or gills was observed).

Approximately half of red snapper (n = 74) and gray triggerfish (n = 10) were released at depth using a descender device. There were no observed instances of a predator removing a tagged fish from the descender device, although 3 red snapper of 71 descended fish successfully recorded on video were observed being consumed by a shark (n = 2) or dolphin (n = 1) shortly after release from the descender. Sharks or dolphins were present (i.e., visually identifiable either during fish descent or while the descender device was being retrieved) in 25 observed descender releases

(35%).

Offloaded detection data from acoustic receivers (n = 60 receivers in the 30-m arrays and n = 46 receivers in the 55-m arrays) included 10.3 million detections of 165 tagged fish that yielded 1.33 million position estimates for 154 tagged fish. Fish were tracked within the acoustic array up to 330 days following release. Fate and time to fate (days post-release) were assigned to each tagged fish by comparing the individual's movement, swim speed, and depth to known or inferred behavior patterns indicating a unique fate. Movement, swim speed, and depth data available for large sharks present within the arrays was provided by serendipitous detection of bull sharks (Carcharhinus leucas) tagged elsewhere, as well as examination of the position data

from tagged red snapper observed being consumed by large sharks after tag and release. A tagged bull shark (2.55 m female tagged February 2016 in the Bahamas by Tristan Guttridge at

Bimini Biological Field Station Foundation) passed through the 30-m acoustic array on 3 occasions (11 September, 13 September, and 8 October 2016) providing 59 position estimates.

Another tagged bull shark (2.50 m female, tagged off Key Biscayne, Florida by Neil

Hammerschlag at University of Miami) moved through the 55-m array on 24 November 2017, yielding 31 position estimates. The calculated mean swim speed of these 2 tagged bull sharks within the array was 0.95 m/s (range 0.10 to 1.54 m/s; Fig. 2-3A). There are no depth data for these sharks as neither fish’s tag contained a pressure sensor. In addition, movement and depth of feeding sharks were provided from tags attached to red snapper (45.2 and 46.0 cm TL) that were observed being preyed upon by large sharks (species uncertain, estimated length 1.5−2.5 m TL) on 2 September 2017. Both acoustic tags continued transmitting for several days following consumption by sharks and provided 100 position estimates from several minutes to 50 hours following release. The estimated mean speed of these 2 consumed tags, hence sharks, within the array was 0.57 m/s (range 0.00 − 1.44 m/s; Fig. 2-3B). Two tagged red snapper (65.7 and 54.9 cm TL) were also observed being consumed by bottlenose dolphins (Tursiops turncatus) following release on 2 September 2017, but there were no position estimates available for those transmitter tags following consumption. The 65.7 cm TL red snapper was observed by the tagging crew being taken by a dolphin at the surface and the transmitter tag was detected on the bottom immediately following consumption, suggesting the dolphin did not consume the tag together with the fish. The 54.9 cm TL red snapper was visible on the GoPro video footage being taken into the mouth of a dolphin at depth following release from the descender device. That acoustic tag transmitted from depths above the bottom 4 times within approximately 2 minutes

following release, after which no acoustic transmissions were detected. Although there were no instances where tagged fish were observed being preyed upon by dolphins that yielded sufficient detection data to draw inferences on dolphin movement or behavior, depth data for a 34.0 cm TL red snapper released at the surface on 26 April 2016 show the tag moving frequently (several times per hour) from the bottom to within several meters of the surface for approximately 4 days following release, suggesting the tag may have been moving with a dolphin as it traveled frequently to the surface to breathe.

Normal behavior indicative of living red snapper and gray triggerfish was inferred from fish that were harvested or recaptured after spending time at liberty within the array. Between

November 2017 and February 2019, 10 tagged red snapper were recaptured or harvested after being at large between 30 and 844 days, providing confirmation that these fish were alive within the array. Although 8 of 10 red snapper had lost their acoustic tags prior to being recaptured, review of the VPS position data revealed that prior to tag loss, harvest, or emigration from the array, median calculated swim speed was 0.02 m/s (range 0.00−0.75 m/s, based on 95 thousand positions; Fig. 2-3C). The frequency distribution of swim speeds was roughly lognormal with a mode between 0.0 and 0.1 m/s. Two gray triggerfish were harvested by spearfishers on 15 May

2016 after 16 and 20 days at large, respectively. Prior to being harvested, median calculated swim speed was 0.02 m/s (range 0.00−0.40 m/s, based on 8.1 thousand positions; Fig. 2-3D).

Similar to the ten recaptured red snapper, the frequency distribution of swim speeds was roughly lognormal with a mode between 0.0−0.1 m/s.

A sudden shift in depth or position over time indicated predation, tag loss, or emigration from the array. For example, a 33.5 cm TL red snapper was tagged and released in the shallow array on 14 September 2016. After approximately 8 hours, the fish moved 600 m away from the

reef where it was released to an adjacent reef where it remained at depths greater than 25 m until

19 September 2016 03:00 UTC. A sudden shift in depth (ranging from the surface to the bottom) and movement (exiting and entering the array and moving several kilometers in a single direction) indicated the occurrence of a predation event (Fig. 2-4A). Tag loss was apparent by a shift from variable to constant depth and position (Fig. 2-4B). Emigration events were often assumed to occur when tagged fish left the acoustic array without first displaying the abrupt shift in speed, depth, or movement patterns indicative of a predation event (Fig. 2-4C).

Point estimates of red snapper discard mortality (0−48 hours following release) were lowest for descender-released fish in the 30-m array in late summer 2016 (22.7%, SE = 8.9%) and highest for surface-released fish in the 55-m array in late summer 2017 (80%, SE = 12.6%)

(Fig. 2-5). Overall, red snapper discard mortality for all tagging events combined was 56.6% (SE

= 6.8%) for surface released fish and 36.1% (SE = 6.1%) for descender released fish. Predation accounted for 77% of all red snapper mortalities and 83% of discard mortalities. My estimates of discard mortality are within the upper range of estimates by depth from previous acoustic telemetry and discard mortality studies of red snapper (Fig. 2-6; Campbell et al. 2014; Piraino and Szedlmayer 2014; Curtis et al. 2015; Williams et al. 2015; Williams-Grove and Szedlmayer

2016a).

Release method (surface or descender), presence/absence of traumatic hooking injury, fish length, time out of water, and the change in temperature from the bottom water to the air were all informative variables in linear regression models of red snapper discard mortality (Table

2-2). The Cox proportional hazards model including these variables indicated using a descender device to release red snapper significantly reduced discard mortality by a ratio of 0.40 (95% CI =

0.23 − 0.71; Fig. 2-7A) and red snapper suffering from traumatic hooking were 5 times more

likely to experience discard mortality (95% CI = 2.2 – 11.2). Fish length, time out of water, and change in temperature from bottom water to the air did not significantly affect probability of mortality in the Cox proportional hazards model at alpha = 0.05 (P = 0.054, 0.283, and 0.630 respectively). However, the full parameter set chosen a priori based on the nonlinear model selection was retained. When presence/absence of traumatic hooking injury, fish length, time out of water, and change in temperature from water to air were controlled, estimated discard mortality of surface-released red snapper was 60.7% (95% CI = 44.0% – 72.4%) and for descender-released red snapper was 31.2% (95% CI = 18.8% – 41.8%).

Gray triggerfish (n = 26) were tagged primarily in spring 2016 in the 30-m array (n = 22).

Point estimates of discard mortality were higher (60.0%; SE = 15.5%) for descender-released fish than for surface-released fish (26.7%; SE = 11.4%). Predation accounted for 85% of total gray triggerfish mortalities and 100% of discard mortality. The dataset did not include enough tagged gray triggerfish to evaluate the effects of depth, season, barotrauma symptoms, traumatic hooking, or accidental venting on gray triggerfish discard mortality. Of the remaining variables, I selected the linear regression model for gray triggerfish discard mortality that included fight time, time out of the water, and fish length, while additional variables such as change in temperature from water to air, or whether or not fish were released with the descender device did not inform the model sufficiently to warrant inclusion (Table 2-2). The Cox proportional hazards model did not indicate that fight time, time out of water, fish length, or release method had a significant effect on gray triggerfish survival at the alpha = 0.05 level (P = 0.0516, 0.828, 0.098, and 0.339, respectively). When all 3 covariates were included, 95% confidence intervals based on the Cox proportional hazards modeling of predicted gray triggerfish survival overlapped for surface (95% CI = 55 – 100%) and descender (CI = 22 – 100%) released fish (Fig. 2-7B).

Few mortalities of red snapper (7 predation and 6 reported harvests of fish after tag loss or emigration from the array) or gray triggerfish (1 predation, 2 harvest) were observed beyond the first 48 hours following release (Table 2-3). Emigration of tagged fish from an array within

14 days following release was rare, as only 4 red snapper were inferred to have emigrated from an array within 48 hours following release, but there were an additional 4 red snapper and 2 gray triggerfish that emigrated from an array within 14 days. Ultimately, 14 red snapper and 7 gray triggerfish (25 and 54% of at-risk fish for each species) emigrated from an array 14−269 days following release. The mean time to emigration was 44 days (± SD: 46 days) for red snapper and

118 days (± SD: 99 days) for gray triggerfish. Acoustic tag losses were highest for the Domeier dart attachment used in late summer 2017 (13.8% tag loss within 48 hours, 20.7% within 14 days). Otherwise, the highest 14-day cumulative tag loss using the threaded bar attachment was

5.1% for red snapper tagged in spring 2018. The threaded bar attachment for gray triggerfish also had low tag loss rates. Of 26 gray triggerfish that were tagged in 2016, only 1 was estimated to have lost its acoustic tag (330 days after release).

Discussion

Predation by highly mobile predators within several hours of release was the dominant source of discard mortality for acoustically tagged red snapper (83%) and gray triggerfish

(100%). Traditional mark-recapture, laboratory, and enclosure approaches utilized to estimate discard mortality do not explicitly account for predation, although authors of several studies have drawn inferences of fish behavior or included observations indicating predation may be an important contributor to discard mortality, at least in the case of red snapper (Campbell et al.

2010a; Drumhiller et al. 2014). Authors of most discard mortality studies using acoustic telemetry methods identified mortality events by lack of movement (freshwater/estuarine:

Hightower et al. 2001; Bacheler et al. 2009b, marine: Curtis et al. 2015; Jackson et al. 2018;

Runde and Buckel 2018), implying that following release fish succumbed directly to handling injury, starvation, cold-kill, or disease. Data from the current study suggest, at least in the marine environment where large predators may be present or abundant, fish that are compromised following capture and release may be more likely to be consumed by predators before they die and settle to the bottom. The large spatial coverage and positioning accuracy of the acoustic telemetry arrays, combined with observations of large tagged sharks moving within the arrays or of sharks consuming tagged fish, provided the unique ability to explicitly identify predation events. Researchers using acoustic telemetry to estimate mortality of tagged fish have inferred the occurrence of predation events from movement, depth, or acceleration data (Heupel and

Simpfendorfer 2002; Friedl et al. 2013; Ellis et al. 2017; Runde and Buckel 2018); however, data in this study indicate predation accounted for the majority of discard mortality observed in acoustically tagged red snapper and gray triggerfish. This result may be location-specific, with an abundance of large coastal sharks or dolphins occurring in our region, or it may indicate that the large spatial coverage of the study and the inclusion of depth sensors on tags enabled the identification of predation events that otherwise would have gone undetected.

These large-area acoustic arrays enabled the differentiation between tagged fish which emigrated from an array under their own volition and transmitters which moved out of an array with a predator that had consumed a tagged fish. I observed that following predation by a large shark, transmitters moved away from the tagging site immediately, and often moved out of the detection area of the array within several hours. Surviving fish, in contrast, rarely moved away from the reef within hours of being tagged. Results of several acoustic telemetry studies have indicated a large number of tagged red snapper and gray triggerfish apparently leave the detection area surrounding tagging reefs within 2–6 days post-tagging (Szedlmayer and

Schroepfer 2005; Piraino and Szedlmayer 2014; Herbig and Szedlmayer 2016). These individuals, which can be >30% of all tagged animals, typically have been classified as tagging- induced emigrations and censored from further analyses when the objective of a study was to examine fish behavior or survival separate from discard or tagging mortality. However, when the primary objective is to estimate discard mortality with acoustic telemetry data, it is vital to accurately distinguish between emigration from the tagging site and predation, as censoring these individuals from the post-release survival data may greatly underestimate discard mortality.

My overall red snapper discard mortality estimates are within the range of discard mortality estimates reported by other authors but also generally higher than most previous estimates at comparable depths. This may be because my approach to estimating discard mortality, with large-scale three-dimensional tracking of reef fish over weeks to years, was better able to detect predation events than in previous studies. If I exclude predation from discard mortality (by censoring all tagged fish that were identified as predation mortalities), estimated red snapper overall discard mortality in the 30-m array (regardless of release method or season) would drop from 36.8% (including predation) to 5.3% (excluding predation), corresponding to a reduction from approximately the 90th percentile to the 10th percentile of estimates from previous studies at depths from 25 to 35 m (Fig. 2-6). For red snapper released in the 55-m array, excluding predation as a source of mortality reduced estimated discard mortality from 54.4%

(70th percentile of previous studies at depths from 50 to 60 m) to 21.2% (lower estimate than previous studies at comparable depths; Fig. 2-6). Many coastal and offshore shark populations in the southeast U.S. and GOM have begun to recover in recent years, with positive trends expected to persist (Peterson et al. 2017; SEDAR 2017). Increased shark abundance, including bull sharks

(Froeschke et al. 2012), could lead to higher predation rates on recreationally released fish,

explaining the high estimates in this study and suggesting that predation may be an increasingly important driver of discard mortality of reef fish in coming years as shark populations continue to recover.

Descender devices approximately halved estimated red snapper discard mortality in this study. There have been very few comparable studies of descender devices where released fish were at risk of predation. In several of these investigations, the authors concluded that releasing fish with descender devices substantially decreased discard mortality (e.g., by 67% for red snapper, Curtis et al. 2015; by 98% for yelloweye rockfish Sebastes ruberrimus, Hochhalter and

Reed 2011). However, there was no significant benefit of releasing fish with descender devices for 6 additional species investigated in mark-recapture and acoustic telemetry studies (Sumpton et al. 2010; Eberts et al. 2018). Laboratory and enclosure studies that exclude predation are much more numerous and are more likely to conclude little or no effect of descender devices on discard mortality (Roach et al. 2011; Butcher et al. 2012; Ng et al. 2015), with Drumhiller et al.

(2014) being a noticeable exception (red snapper survival was estimated to be 17% for surface- released and 83% for experimentally recompressed fish). I found the effect of descender devices on discard mortality of gray triggerfish was not statistically significant, largely due to small sample size. Increased handling (fight time and time out of water) and decreased fish size appeared to have a positive effect on discard mortality; however, none of the variables investigated were deemed to have a significant effect on gray triggerfish discard mortality. All gray triggerfish tagged in this study were captured from shallow (30 m) depth and none were observed to be suffering from barotrauma, so it is possible that stress from additional handling and delayed return to the water obscured any benefits of being returned to depth with the descender device. My overall discard mortality estimate for gray triggerfish (40%) was lower

than recent estimates by Runde et al. (2019) who included fish captured from greater depths in their analyses, which I believe warrants concern that these fish may be much more susceptible to barotrauma-induced discard mortality than previously assumed.

Aside from release method, the presence of traumatic hooking injury greatly affected discard mortality of red snapper, which is widely supported by results of previous investigations that suggest throat-, gut-, or gill- hooked fish have low survival probability (Muoneke and

Childress 1994; Murphy et al. 1995). Reef depth, presence/absence of barotrauma symptoms, fish size, and season did not have an apparent effect on discard mortality of red snapper or gray triggerfish. However, many barotrauma symptoms may be cryptic and not detectable without internal examination of fish tissues and organ systems (Rummer and Bennett 2005). Fishing depth profiles from spring 2018 tagging also indicated fish were rarely captured near the bottom and instead were hooked and retrieved from a range of mid-water depths. The absence of a depth effect may be due to this observation error because I had to rely on reef depth in the discard mortality models since I did not have capture depth data for most of the tagged fish. Similarly, the categorical variable season may not have adequately described the physiological stressors that fish were exposed to during each of the 4 tagging events. It is possible that seasonal variation in predation pressure obscured any apparent effect of temperature-induced physiological stress that tagged fish experienced. For example, fish tagged and released in the early spring may have experienced less temperature-induced physiological stress but were exposed to more actively feeding sharks. Instead of season, I chose to use the change in temperature between bottom water and air temperature to reflect the amount of temperature stress that fish experienced, which was significant in the red snapper discard mortality linear regression model (a larger increase in temperature from the bottom water to the air increased

mortality). The overall lowest red snapper discard mortality occurred in late summer 2016 when the bottom water temperature was greater than the air temperature (i.e., fish experienced a 2–3°C temperature drop when they were brought to the surface and removed from the water for tagging). The highest overall discard mortality occurred in late summer 2017 when air temperature was comparable to late summer 2016 but the bottom water temperature was 7 °C colder than the air temperature on the day of tagging.

I believe my estimates of discard mortality reduction due to descender devices may be conservative since seasoned fishers, in particular crew on for-hire fishing vessels, could streamline the rigging of the descender device (reducing the amount of weight and excluding the video cameras to reduce drag and increase retrieval speed) while also reducing handling time relative to fishes tagged in this study. I did not examine the effect of venting fish on discard mortality and chose instead to evaluate the efficacy of descender devices compared to unvented surface-released fish. Results are equivocal regarding the benefits of venting and some investigators suggest improperly venting fish can damage organs and increase discard mortality

(Wilde 2009; Eberts and Somers 2017). In contrast, neither I nor previous researchers to my knowledge have presented evidence that descender devices cause harm to fish. Recreational fishers will undoubtedly play the primary role in efforts to reduce discard mortality, and although venting may continue to be the preferred method of discard mortality reduction among fishers in many instances (Crandall et al. 2018), further evidence supporting the efficacy of descender devices could facilitate acceptance among GOM reef fish fishers.

Continuing advances in geopositioning acoustic technology, including reduced costs, enable researchers to deploy more receivers covering larger areas. Greater spatial coverage within studies, combined with the proliferation of cooperative networks that foster equipment

sharing and information exchange among researchers, will be instrumental to improving studies of discard mortality and other ecological processes of marine fishes (Lowerre-Barbieri et al.

2019). Future investigations in the marine environment should be designed to measure the effects of predation on the survival of discarded fish or else risk ignoring this potentially significant driver of mortality. The quantification of dead discards and efforts to reduce discard mortality will be increasingly vital considerations in recreational fisheries around the world where recovering population abundances, harvest prohibitions, or non-consumptive attitudes of fishers results in large numbers of discarded fish.

A B

Figure 2-1. Maps of the study area: A) Northern Gulf of Mexico indicating acoustic array locations (labeled b and c). B) The 60-receiver shallow acoustic array was deployed at 28−35 m depth from February 2016 to March 2017. Receiver locations and the approximate extent of the array from February 2016 to September 2016 are shown as solid triangles (▲) and a solid line (─), while receiver locations and array extent from September 2016 to March 2017 are shown as open triangles (Δ) and a broken line (--- )). C) The deep array was deployed at 48−55 m depth from August 2017 to July 2018. Receivers and array extent are shown by solid triangles (▲)) and a solid line (─)). Artificial reef locations are denoted by squares (■) in both panels (B) and (C).

A B C

Figure 2-2. Digital images of A) a 43-cm TL red snapper held in captivity with an external acoustic tag attached with the stainless steel bar method, B) an acoustic tag with the stainless steel bar external attachment device, and C) an acoustic tag with a Domeier dart attachment device. Photos courtesy of author.

Table 2-1. Summary table of red snapper (RS) and gray triggerfish (GT) tagged and released during 4 events over the course of this study. Length for red snapper is total length and gray triggerfish is fork length. Conditions Fish tagged Tagging Depth Air Botto Species Release Length Fight Time out % with n Event (m) temp m method (cm, time (s, of water baro- (°C) temp mean ± mean ± (s, mean trauma (°C) SD) SD) ± SD) Spring 2016 28−31 22.7− 20.0− RS S 45 ± 17 75 ± 61 78 ± 28 30 10 26 Apr − 24.3 21.4 RS D 55 ± 17 63 ± 29 118 ± 39 0 10 3 May GT S 41 ± 5 67 ± 42 100 ± 35 8 13 GT D 41 ± 6 68 ± 44 126 ± 27 0 9 Summer 2016 28−31 26.6 28.8− RS S 39 ± 5 57 ± 32 80 ± 17 11 18 14 Sep 29.9 RS D 45 ± 12 81 ± 60 95 ± 32 23 22 GT S 45 51 90 0 2 Summer 2017 51−57 28.5 20.4− RS S 57 ± 78 ND 74 ± 20 53 15 2 Sep 21.7 RS D 59 ± 14 ND 117 ± 74 61 18 Spring 2018 51−57 19.8− 21.0− RS S 56 ± 12 146 ± 39 94 ± 26 48 23 11−24 20.3 21.8 RS D 58 ± 13 134 ± 36 114 ± 25 83 24 April GT S 50 177 85 100 1 GT D 45 168 131 100 1

A B

C D

Figure 2-3. Calculated swim speeds of A) tagged bull sharks (n = 2), B) red snapper that were observed being preyed upon by large sharks in September 2017 (n = 2), C) red snapper prior to recapture (n = 10), and D) gray triggerfish prior to harvest (n = 2).

Figure 2-4. Depth (left) and positions (right) of A) 33.5-cm TL red snapper released 14 September 2016 and consumed by a predator 19 September 2016, B) 37.6-cm TL red snapper released 26 April 2016 and lost the acoustic transmitter tag 13 June 2016, and C) 43.8-cm FL gray triggerfish released 14 September 2016 and emigrated from the array 30 December 2016. Normal behavior of each tagged fish is noted with black symbols (●) whereas red squares (■) indicate the shifted depths and movements associated with each fate (predation, tag loss, and emigration, respectively). Depth is in m and dates are UTC.

Figure 2-5. Estimated 48-hr percent mortality (± 95% CIs) for red snapper (RS) and gray triggerfish (GT) released at the surface versus at depth with descender devices. Sample size (number of fish with known fates at the end of 48 hours post-release) is above each point estimate.

Figure 2-6. Estimated red snapper discard mortality (%) by depth and release method. Data from prior studies (open symbols) were compiled in Campbell et al. (2014), with more recent estimates from Piraino and Szedlmayer (2014), Curtis et al. (2015), Williams et al. (2015), and Williams-Grove and Szedlmayer (2016).

Table 2-2. Nonlinear models testing factors affecting discard mortality for red snapper and gray triggerfish. ΔAICc values are shown relative to the model with the lowest AICc. Model n k -LL ΔAICc Red snapper 푀̂ ≈ 훽0 + ∆푡푒푚푝 + 푑푒푠푐푒푛푑푒푟 + 푡푟푎푢푚푎 + 푙푒푛𝑔푡ℎ 140 5 65.9 0 푀̂ ≈ 훽0 + ∆푡푒푚푝 + 푑푒푠푐푒푛푑푒푟 + 푡푟푎푢푚푎 + 푙푒푛𝑔푡ℎ 139 6 65.7 1.7 + 푡𝑖푚푒표푢푡 푀̂ ≈ 훽0 + ∆푡푒푚푝 + 푑푒푠푐푒푛푑푒푟 + 푡푟푎푢푚푎 140 4 68.3 2.6 푀̂ ≈ 훽0 + ∆푡푒푚푝 + 푑푒푠푐푒푛푑푒푟 + 푡푟푎푢푚푎 + 푙푒푛𝑔푡ℎ 139 7 65.5 3.6 + 푡𝑖푚푒표푢푡 + 푠푒푎푠표푛 푀̂ ≈ 훽0 + ∆푡푒푚푝 + 푑푒푠푐푒푛푑푒푟 140 3 71.4 6.7 푀̂ ≈ 훽0 + ∆푡푒푚푝 141 2 74.5 10.8 푀̂ ≈ 훽0 141 1 79.2 18.1 Gray triggerfish1 푀̂ ≈ 훽0 + 푡𝑖푚푒푓𝑖푔ℎ푡 + 푡𝑖푚푒표푢푡 + 푙푒푛𝑔푡ℎ 21 4 8.5 0 푀̂ ≈ 훽0 + 푡𝑖푚푒푓𝑖푔ℎ푡 + 푡𝑖푚푒표푢푡 21 3 10.1 0.2 푀̂ ≈ 훽0 + 푡𝑖푚푒푓𝑖푔ℎ푡 + 푡𝑖푚푒표푢푡 + 푙푒푛𝑔푡ℎ + ∆푡푒푚푝 21 5 8.3 3.1 푀̂ ≈ 훽0 + 푡𝑖푚푒푓𝑖푔ℎ푡 22 2 13.6 4.4 푀̂ ≈ 훽0 + 푡𝑖푚푒푓𝑖푔ℎ푡 + 푡𝑖푚푒표푢푡 + 푙푒푛𝑔푡ℎ + ∆푡푒푚푝 21 6 7.9 6.4 + 푑푒푠푐푒푛푑푒푟 푀̂ ≈ 훽0 26 1 16.8 8.3 1For gray triggerfish not all levels of some variables contained samples, thus effects of depth, barotrauma symptoms, accidental venting, traumatic hooking, and season on discard mortality could not be explored.

A B

Figure 2-7. Cox proportional hazards model-estimated survival (± 95% CIs) for surface-released (red ○) and descender-released (black ▲) fish. A) Red snapper models included presence/absence of traumatic hooking injury, fish length, time out of water, and the change in temperature from the bottom water to the air as covariables, and B) gray triggerfish models included fight time, time out of the water, and fish length as covariables.

Table 2-3. Number of tagged fish in each fate assignment category by time period following release. "Alive and present" includes tagged fish that were identified as alive with acoustic tag attached and within a study array at the end of each time period (the end of the 14+ days time period was the retrieval of the acoustic array). Time post-release Fate 0–48 hrs 2–14 days 14+ days Red snapper: surface released Predation mortality 21 2 4 Surface mortality 9 0 0 Harvest mortality 0 0 0 Emigration 3 1 3 Tag lost 3 0 11 Unknown 7 0 0 Alive and present 23 20 2 Red snapper: descender released Predation mortality 22 1 0 Surface mortality 0 0 0 Harvest mortality 0 0 0 Emigration 1 3 12 Tag lost 3 3 13 Unknown 9 0 0 Alive and present 39 32 6 Gray triggerfish: surface released Predation mortality 4 0 1 Surface mortality 0 0 0 Harvest mortality 0 0 1 Emigration 0 2 6 Tag lost 0 0 0 Unknown 1 0 0 Alive and present 11 9 1 Gray triggerfish: descender Predation mortality 6 0 0 Surface mortality 0 0 0 Harvest mortality 0 0 1 Emigration 0 0 1 Tag lost 0 0 1 Unknown 0 0 1 Alive and present 4 4 0

CHAPTER 3 FINE-SCALE MOVEMENT AND BEHAVIOR OF RED SNAPPER: AN ACOUSTIC TELEMETRY POSITIONING STUDY IN THE NORTHCENTRAL GULF OF MEXICO

Gulf of Mexico (GOM) reef fish, including adult red snapper (Lutjanus campechanus) commonly occur at artificial reefs (e.g., petroleum platforms, scuttled ships, and manufactured reef modules) where they are targeted by fishermen. Fisheries managers have used artificial reefs to improve fishing opportunities and also to provide refuge to adult fish from fishermen by deploying artificial reefs in regulatory no-fishing zones (i.e., marine protected areas, MPAs) or unpublished locations (Coleman et al. 2004a; Addis et al. 2016). Areas of artificial reefs may also form de facto no-trawl zones, reducing fishing mortality of juvenile reef fish and other ecologically important species (Cowan et al. 2011). The role artificial reefs will play in the future management of GOM red snapper depends on how closely adult red snapper associate with reefs, how they move between artificial reefs and nearby natural habitat, and the factors that affect these movements. To best address these questions, researchers will likely need spatially and temporally fine-scale data, such as hourly observations of fish locations accurate to within several meters, that have previously been unavailable due to the difficulty of tracking fish over large areas of the open ocean.

Advanced geopositioning acoustic telemetry, such as Vemco Positioning System (VPS) arrays, rely on time-difference-of-arrival data processing of acoustic detection data [see Smedbol et al. (2014)] to provide near-continuous fine-scale position (within several meters; Espinoza et al. 2011; Biesinger et al. 2013; Piraino and Szedlmayer 2014; Roy et al. 2014; Guzzo et al. 2018) and movement data on tagged animals. An array consisting of 10s of acoustic receivers can provide consistent spatial sampling coverage of animal positions over large areas (10s of km2), providing the novel ability to use position and movement data of aquatic organisms to test hypotheses regarding spatial ecology and habitat use. Movement and habitat preference studies

of marine fish using large geopositioning acoustic telemetry arrays are sparse and mostly focused in estuaries (Espinoza et al. 2011; Furey et al. 2013; Dean et al. 2014; Özgül et al. 2015). Some acoustic telemetry researchers have used time-averaged activity centers of tagged animals based on raw detection data (Heupel and Simpfendorfer 2002; Simpfendorfer et al. 2002; Lowe et al.

2003; Szedlmayer and Schroepfer 2005; Bellquist et al. 2008; Heupel et al. 2010; Farmer and

Ault 2011; Topping and Szedlmayer 2011; Froehlich et al. 2019). However, when evaluated for accuracy, that approach results in reduced spatial and temporal data resolution even in shallow estuaries (positions accurate to no less than 200 m, typically at intervals no less than 15 minutes;

Simpfendorfer et al. 2002; Friedl et al. 2013). Geopositioning acoustic telemetry arrays, on the other hand, provide temporal and spatial fine-scale data necessary for investigating the spatial ecology of fishes at higher resolution.

Knowledge of fine-scale temporal/spatial habitat use of GOM red snapper would contribute to a basic understanding of the factors affecting habitat preferences and could guide future management or conservation actions. For example, understanding trends in fish proximity to reef structure or residence on a given reef may provide guidance on artificial reef design and placement. Similarly, marine protected areas and spatial fishing closures have been used with varying degrees of success in the GOM (Coleman et al. 2004a). Future efforts to establish similarly protected or no-harvest areas could benefit from improved understanding of the factors affecting spatial distributions and movements of these fish. In addition, from a reactive standpoint, failure to appreciate the variability in individual behavior and movement within marine fish populations has been implicated in poor management outcomes for several fish stocks (Petitgas et al. 2010).

Deployment of large-scale acoustic telemetry arrays also provides unique opportunities to investigate temperament-dependent behaviors in marine fish. Ecologists have suggested that animals within a population may consistently display sets of behavioral traits (a temperament) across time and situations (Huntingford 1976; Sih et al. 2004; Reale et al. 2010). For example, individuals displaying active ("bold", "fast", or "risk-prone") temperaments may be more aggressive towards prey and conspecifics, less cautious of predators, farther ranging, and more likely to disperse to new habitats (Wilson et al. 1994; Sih et al. 2004; Harrison et al. 2015).

Temperament, as a characteristic of within-species variation, provides a link to evolution via differential selection (Gosling 2001). Given their behavior traits, more active individuals may, in turn, be faster growing and more likely to be harvested (Biro and Post 2008; Reale et al. 2010).

Some researchers have found evidence of different movement behaviors, and perhaps temperaments, in red snapper. Results of mark-recapture studies of GOM red snapper suggest that some individuals move great distances (100s of km: Beaumariage 1969; Patterson et al.

2001; Strelcheck et al. 2007) and others do not. Diamond et al. (2007) suggested the dichotomy of "movers" vs. "stayers" to describe these behavioral differences among red snapper in the western GOM, noting that movers were more likely to be found in natural hard-bottom habitat while stayers were more often smaller fish found over artificial reefs. A goal of this chapter was to investigate whether red snapper display a range in movement behaviors that may be indicative of "mover" and "stayer" temperament types (as evidenced by fine-scale movement and habitat use).

The overall goal of this chapter of my dissertation was to estimate the range in three- dimensional activity patterns observed among acoustically tagged red snapper, from which fish temperament might be estimated. Specific objectives were to: 1) characterize the three-

dimensional movement patterns and activity centers of red snapper at spatial scales from m to km for up to twelve months; 2) quantify variables which may affect fish movement and behavior; and 3) investigate whether individuals display heterogeneity in individual movement patterns and habitat preferences over extended time periods which might be indicative of bold

("mover") vs. shy ("stayer") temperaments.

Methods

Acoustic Telemetry Array and Fish Tagging

The specific dates, locations, and configurations of the acoustic telemetry arrays used in this study are described in detail in Chapter 2. Briefly, a ~15 km2 array of acoustic receivers was deployed in the northern GOM from February 2016 to March 2017 at 28–35 m depths (hereafter

"30-m" array) and then from August 2017 to July 2018 at 48–55 m depths (hereafter "55-m array"; Figs. 3-1 – 3-3). Habitat within the study areas of each array consisted of open sand bottom interspersed with artificial reef structures (cement or tire pyramids, reef balls, and chicken coops, all typically 2–5 m tall) and likely also included some natural low-relief limestone hard-bottom habitat. The locations of artificial reefs within the study area was determined based on knowledge of previous Florida Fisheries and Wildlife Conservation

Commission (FWC) funded artificial reef constructions, local knowledge from fishermen, or partial coverage sidescan sonar surveys between 2014 and 2019 of the 30-m array which were provided by FWC (Anthony Knapp, pers. comm.). In several instances, the location of previously unknown artificial reefs was inferred from the activity centers of tagged fish.

Oceanographic conditions were monitored throughout the study period. Hourly average atmospheric pressure (mbar), significant wave height (m), wind speed (m/s), and peak wind speed (a.k.a. gust speed, m/s) were acquired from the National Data Buoy Center at Station

42012 (approximately 48 km west of the 30-m array; National Oceanic and Atmospheric

Administration and National Weather Service 2017) and Station 42040 (approximately 77 km west/southwest of the 55-m array; National Oceanic and Atmospheric Administration and

National Weather Service 2019). Average bottom temperatures (°C) were compiled from the data logs of the VR2Tx receiver located closest to the center the acoustic array.

Red snapper were externally tagged with Vemco V13P depth-sensing acoustic transmitters and released into the acoustic array during 4 tagging events: 1) spring 2016 (26

April−3 May, n = 20); 2) late summer 2016 (14 September, n = 40); 3) late summer 2017 (2

September, n = 33); and 4) spring 2018 (11−24 April, n = 47). Each tagged fish was measured to the nearest mm fork or total length (FL or TL) and then returned to the water as quickly as possible. Fish were released either at the surface or at depth using a descender device. Three to 6 months after each tagging event, detection data were offloaded from acoustic receivers and sent to Vemco for Vemco Positioning System (VPS) geolocation estimation, yielding 4 datasets with

900 thousand position (with depth) estimates for 126 tagged red snapper that were alive and present within the array for up to 324 days. Vemco provided position estimates as both latitude/longitude coordinates and dataset-specific northing (Y) and easting (X) in meters.

VPS Position Accuracy

Vemco provides a relative estimate of error sensitivity (HPE, unitless) for all VPS position estimates of tagged animals and transmitters as well as an estimate of horizontal position error (HPEm, m) for VR2Tx receivers which have measured GPS coordinates. Smith (2013) provides a statistical method to establish a relationship between HPEm and HPE for VR2Tx receivers which can be applied to HPE estimates of tagged animals in order to estimate absolute

* * position error (HPEm , m) at the 95% confidence level. Calculation of HPEm as a function of

HPE was performed for each of the 4 VPS datasets in this study separately corresponding to each time the array was retrieved and detection data offloaded (September 2016, March 2017, March

2018, and July 2018). Twice distance root mean squared (2DRMS) was calculated as twice the square root of the combined variance in X and Y for each 1-m increment of HPE averaged over all VRT2x receivers. The slope and intercept of the linear regression of HPE (the mid-point of each bin, i.e., each half meter) and 2DRMS was then applied to HPE estimates for transmitter

* tagged animals to produce an estimate of position error in m (HPEm ). Position estimates of

* transmitter tagged fish with HPEm values greater than 5 m in the 30-m array and 6 m in the 55- m array were eliminated from each dataset.

A "drift test" was performed to validate positional accuracy within the 55- m acoustic array by comparing VPS calculated coordinates to those recorded directly from the GPS positioning system on the boat. Note that even the best commercially available GPS antennas generally have intrinsic error of 2–3 m. A VEMCO V13T-1x-069k high power transmitter

(identical to the transmitter used on tagged animals in this study, but with a fixed 10-second delay between transmissions and a temperature rather than pressure sensor) was suspended approximately 3 m below the boat from the port side of the stern. The boat was allowed to drift without propulsion or steering for at least 10 minutes while the time, boat heading, and GPS coordinates from the ship's navigation system were recorded. The drift test was repeated at 5 different sites within the acoustic array: the center and each corner (NW, NE, SW, and SE; Fig.

3-4A). The GPS antenna is located on top of the wheelhouse in the center of the boat used for the drift test (F/V Intimidator), which caused an offset between the recorded GPS location and the actual position of the transmitter. VPS transmitter positions were therefore corrected to the GPS antenna location based on the geometry of the boat (9.6 m and +18.4° relative to the bow) and the heading of the bow (relative to north) during each drift test (Fig. 3-4B).

Data Analysis

Only fish that were alive and present within the acoustic array for longer than 48 hours following release were included in behavioral analysis in order avoid confounding effects of capture/tagging and post-release mortality on fish behavior. Position estimates occurring at or after the assigned time of fate (e.g., predation, emigration, tag loss; see Chapter 2) were excluded from analysis. Variables potentially influencing fish movement and behavior where calculated for each position estimate. Time of day relative to the time of sunrise, sunset, and nautical twilight for each day at the array location was determined using the R package 'suncalc'

(Thieurmel and Elmarhraoui 2019). Time of day was defined as: day is between sunrise (top edge of the sun appears on the horizon) and sunset (sun disappears below the horizon); evening twilight occurs between sunset and the end of nautical dusk / twilight (the sun is 12° below the horizon); night occurs between the end of nautical dusk and the beginning of nautical dawn / twilight; and morning twilight occurs between the beginning of nautical dawn and sunrise. Moon phase was quantified as the percent of the moon visible each day as calculated with the R package 'lunar' (Lazaridis 2015). Positions estimates were assigned to the hour of the day (0–23) and day of the year (1–365 or 366). General fish locations were classified by reef during periods when tagged fish returned to a given reef regularly (i.e., spending intervals off the reef for less than 1 day). To eliminate bias which may be caused by error in the boat's GPS navigation system, the relative easting and northing (X and Y) of each reef within the study area was estimated by dataset as the overall mean of all positions of fish that were classified as being near that particular reef. The "nearest reef" was then assigned to each estimated fish position to more accurately describe the movements of fish between neighboring reefs. The study area is known to contain a variety of types and sizes of artificial reefs typically ranging from 2–5 m tall, however, detailed characteristics of individual reefs such as design (pyramids, reef balls, and

chicken coops) and size were not consistently available for all reefs, so these variables could not be incorporated into the model. Instead, reef depth and the distance to the closest adjacent reef

(distance to the nearest neighbor) were calculated from reef location.

I used kernel density estimates (KDEs) to characterize the space utilization of tagged red snapper in two- and three- dimensions (2- and 3-D). KDEs describe the probability of an animal occupying a position in space at a given confidence level. KDEs rely on several assumptions, including an underlying probability distribution ("kernel") of animal locations around one or more centers of activity. Most KDEs described in the literature (Don and Rennolls 1983; Worton

1987, 1989) were developed for mark-recapture datasets that were typically smaller and carried the assumption of independence among observations (i.e., position data were not auto- correlated). In this study, the time delays between successive VPS position estimates were often only a few minutes, thus the location of a tagged fish in any time (t) was not independent of the location of the fish at the previous time (t-1). Movement-based KDE methods include modeled paths connecting timeseries of positions to more realistically estimate the true movements of animals and are more appropriate for dealing with auto-correlated timeseries of animal positions.

In this study, I calculated KDEs from Brownian Bridge modeled movements between

VPS positions (Horne et al. 2007; Benhamou 2011) based on a bivariate-normal probability density kernel for 2 dimensions and extended to a trivariate-normal kernel for 3 dimensions by

Tracey et al. (2014b), as implemented in the R package 'MKDE' (Tracey et al. 2014a). Grid cells were 1 m2 or m3 with maximum delay between positions retained in the movement model of

2 2 1500 seconds and assumed X-Y and depth movement variances (휎푥푦 and 휎푧 in Tracey et al.

2014b) equal to 0.14 m2/s (the mid-point of observed red snapper swim speeds, see Chapter 2). I calculated hourly 95% KDEs (i.e., the area or volume over which an animal has a 95%

probability of occurring, often referred to as home range) for each fish, providing KDE estimates for approximately 52,000 fish × hours. Independent from these movement-modeled KDEs, for each estimated VPS fish position (n = 400,000), the distance (m) to the nearest reef (in 2 dimensions) and the depth above the bottom (m) were calculated.

I used generalized additive models (GAMs) implemented in the R package 'mgcv' (Wood

2019) to examine the influence of independent variables on 4 response metrics of fish movement and habitat use: hourly 2- and 3-D 95% KDEs, distance from the nearest reef, and distance above bottom. A natural log transformation was applied to KDEs and the distance from the closest reef for analyses. Depth above bottom data were truncated to exclude any negative values (likely caused by depth sensor error or incorrect bottom depth estimates) and transformed using a natural log (x+1) transformation. Independent variables were evaluated for both linear and smoothed (thin plate regression splines) effects. Smoothed time of day (hour) and day of year were further penalized to be circular such that the beginning and end predicted values were equal. Time of day, array dataset, fish ID, and reef were evaluated as fixed categorical variables for some analyses. Each candidate GAM was evaluated for parsimony based on the Akaike

Information Criterion (AIC).

Results

VPS Position Accuracy

Drift tests were performed July 30, 2018 between approximately 14:00 and 15:30 UTC

(9:00 to 10:30 AM local time). The drift test at the northwest corner of the array was shifted to the east to avoid a vessel that was operating sidescan sonar in the area on that day. The mean positional error of the VPS array ranged from less than 1.0 m at the center of the array to 7.9 m at the NW site (Table 3-1). The number of tag transmissions resulting in VPS position estimates

(positioning efficiency) was lowest in the center of the array and highest in the southern corners.

Overall, mean positional accuracy was within 4.2 m with 75% of positions accurate to within 6 m.

Reef Fidelity and Habitat Overview

Thirty-five (35) red snapper (mean TL = 41 cm, range 33–63 cm TL) were alive and present within the 30-m acoustic array for at least 48 hours (up to 195 days) following release

(Fig. 3-5). Fate estimates (methods and results presented in Chapter 2) indicated 14 fish emigrated from the array 4–165 days following release, 13 fish lost the acoustic tag 11–123 days following release, 6 fish were consumed by predators 5–147 days following release, and 2 fish were still alive and present in the 30-m array when receivers were retrieved in February 2017.

Tagged fish returned frequently to reef 2 where the majority of fish were tagged in April–May

2016 and September 2016, and reef Z where many fish were tagged in September 2016. Reef 2 is a pair of cement pyramids (each approximately 2 m tall and 3 m wide at the base) and reef Z includes at least 2 connected chicken coops (each approximately 4 m long, 3 m wide, and 3 m high; Fig. 3-6). Sidescan data from 2019 suggest there may be a second structure 30 m southeast of the chicken coops where fish were caught and released at reef Z. There were at least 5 additional artificial reefs (α, β, δ, γ, and ε) spaced 120–250 m apart in the southeast corner of the

30-m array that were not within positioning range until the array was shifted to the south/southeast in September 2016. These 5 reefs were likely single or paired chicken coops similar to reef Z. Although some tagged fish favored a particular reef within the α–ε cluster, most fish moved frequently among them. Positions of tagged fish were generally clustered around artificial reefs, with individuals making frequent forays 70–100 m away from the reef and less frequently moving more than 500 m away before returning.

Twenty-five (25) red snapper (mean TL = 55 cm, range 41–76 cm TL) were alive and present within the 55-m acoustic array for at least 48 hours (up to 324 days) following release

(Fig. 3-7). Tag loss was a pervasive problem, especially for fish tagged in September 2017, with

56% of fish (n = 14) estimated to have lost their acoustic tags 10–105 days following release.

Fate estimates indicated 6 fish emigrated from the array 4–103 days following release, 1 fish was consumed by a predator 8 days after release, and 4 fish were still alive and present in the 55-m array when receivers were retrieved in July 2018. Tagged fish returned frequently to reefs a and c (single steel and tire pyramids approximately 4 m tall and 6 m wide) where the majority of fish were tagged in September 2017 and April 2018 (Fig. 3-8). In general, fish changed reefs less often in the 55-m array compared to the 30-m array, gravitating towards a single reef for periods exceeding several weeks.

Estimates of reef fidelity were highly variable depending on the tagging reef and timescale of the observations used to estimate an annual rate (Table 3-2). In the 30-m array, reef fidelity was highest for reef 2 when only the first 30 days were used to calculate an annual rate

(47.8%); however, after 60 days only 8 fish remained on the reef and estimated annual reef fidelity = 2.2%. Fish were much less likely to stay on reef Z (annual reef fidelity < 1% regardless of whether 30 or 60 day time periods were considered). Due to high discard mortality and tag loss for fish tagged in the 55-m array (Chapter 2), sample sizes available for reef fidelity estimates in the deeper array are small and estimates ranged from < 1% for fish tagged on reefs a and m to 100% for fish tagged on reef c.

Proximity of Fish Positions to Reefs

Tagged fish on reef Z in the 30-m array in September 2016 (n = 12) stayed closer to the reef (median distance = 12.0 m) than fish on reef 2 (median distance = 14.4 m) during the same time period (Fig. 3-9A). However, fish frequently made brief forays farther from each reef: the greatest 5% of positions were farther than 44.2 and 41.5 m away from reefs Z and 2, respectively. Tagged fish (n = 5) stayed closer to reef 2 during May: median distance was 9.2 m

with 5% of positions greater than 21.3 m (Fig. 3-9B). The median distance between tagged fish and the reef was similar for reefs a (11.7 m, n = 3 tagged fish) and c (10.9 m, n = 5 tagged fish) in the 55-m array during September 2017, with fish around reef c making more trips farther from the reef (the greatest 5% of positions were 25.7 and 29.7 m from reefs a and c, respectively; Fig.

3-9C). Similarly, in May the median distance of tagged fish to reef a (16.4 m) was greater than for fish around reef c (13.3 m) while fish around reef c made more trips farther from the reef (the greatest 5% of positions were 35.3 and 41.8 m from reefs a and c, respectively; Fig. 3-9D).

Factors Affecting Space Use

The best GAM of hourly 95% 2-D KDE included smooth terms for hour, distance to the nearest reef neighbor, day of year, fish length, water temperature, wind gust speed, atmospheric pressure, and wave height, and linear terms for reef depth and array (dataset) as a categorical variable. This model explained 14.2% of variance in the data (Table 3-3). The best GAM of hourly 95% 3-D KDE explained much more of the observed variance (25.0%) and included smoothed terms for moon visibility in addition to the same variables as the best hourly 95% 2-D

KDE GAM (Table 3-4). The most influential variable in the GAM describing distance from the reef was time of day, which explained 16.6% of the total variance, with the best model including

11 additional variables and explaining 21.8% of total variance (Table 3-5). Fish length was the most influential variable affecting distance above the bottom; whereas the full GAM included fish length, reef depth, hour of the day, temperature, day of the year, distance to the nearest reef neighbor, moon visibility, wave height, atmospheric pressure, wind gust speed, average wind speed, and array/dataset (Table 3-6).

Space utilization (hourly 2- and 3-D 95% KDE), distance from the reef, and depth above bottom were highest during the winter, peaking in February and March while activity was greatly depressed during the summer months (Fig. 3-10). Tagged fish showed spikes in space utilization

and two-dimensional distance from the reef during the crepuscular periods (morning twilight and evening twilight) while activity was greatly reduced during the mid-day hours (Figs. 3-11 and 3-

12). Tagged fish were closest to the bottom during the nighttime hours and moved higher into the water column during the day. Although fish length was significant in both GAMs for 2- and 3-D space utilization, there was no clear trend in space utilization with length, except that the smallest fish (<35 cm TL) and largest fish (>75 cm TL) were apparently less active (Figs. 3-

13A–B). Both distance to reef and distance above bottom increased with length (Fig. 3-13C–D).

Three dimensional space use was lowest at 18-20 °C and increased with temperature, also showing a positive trend with wave height (Fig. 3-14A–B). Space use was inversely related to the fullness of the moon (although the effect was small) and wind gust speed (Fig. 3-14C–D) while the effects of distance to reef nearest neighbor and atmospheric pressure did not show any apparent overall trend (Fig3-14E–F).

A comparison of fish space utilization between reefs was only possible for a subset of reefs within each array, since fish were not observed using many of the available reefs.

Controlling for variables affecting space use, expected distance between fish and reefs was estimated by adding a reef variable to the best GAM model. Among the 2 most commonly used reefs in the 30-m array, fish generally stayed a significantly greater distance from reefs Z, α, and

β (chicken coops) and a closer distance to reef 2 (a multiple pyramid reef; Fig. 3-15A). Among the small number of fish that were observed around reef O (a single pyramid), fish were generally 2-3 times as far from the reef than they were when around reef 2. Note that the existence and location of reef 5 was inferred from fish movements and there is no information regarding the type of reef, whether artificial or natural. In the 55-m array, despite that reefs a, c, and m were each of similar construction (Fig. 3-8), fish were generally much greater distances

from reefs a and m than from reef c (Fig. 3-16B). Distance of fish from the reef was not significantly different for reefs m, n (a chicken coop), and o (of unknown construction).

Individual Variation

Some tagged red snapper frequently engaged in reef switching behavior where they moved frequently between reefs, often spending several days at one reef before moving to another (e.g., RS42, RS50, RS59, RS69, RS74, RS75, RS78, RS80, and RS82 in the 30-m array;

Fig. 3-5, and RS136 in the 55-m array; Fig. 3-7). Most tagged fish, especially in the 55-m array, were observed using only 1 or 2 reefs. Directed movements between reefs occurred quickly and were not captured in the calculated hourly 95% KDEs of space utilization. When Fish ID was added as a fixed effect in the best-fit GAM for hourly 95% 2-D KDEs, the amount of variance explained by the model increased modestly from 14.17 to 17.29 %. Controlling for all variables in the model (time of day, time of year, atmospheric pressure, temperature, wave height, and wind gust speed) and comparing individuals within reefs, some fish appeared to occupy larger hourly 95% 2-D KDEs, especially in the 30-m array (e.g., RS52, RS64, RS75, RS77, RS80,

RS84; Fig. 3-16). In the 55-m array, two-dimensional space use across individuals displayed more of a continuum rather than a dichotomy, ranging from <450 m2 for RS128 and RS129 to

>550 m2 for RS99, RS100, and RS132 (Fig. 3-17).

Discussion

Geoposition Accuracy

Data from this study indicate acoustic telemetry can provide high accuracy geoposition estimates of marine fish over large (>15 km2) areas of open ocean habitat. Independent positioning tests, such as the drift test in this study, are important to confirm relative position error estimates (HPE) provided for tagged animals with VPS positions estimates and the

* associated HPEm estimates are realistic. I did not evaluate geoposition accuracy in the 30-m

array, or over a range of environmental conditions, but I found all areas of the 55-m array with the exception of the northwest corner provided geoposition accuracy to within the 6-m threshold specified a priori. Independent geoposition accuracy tests are rarely performed and described in animal movement studies (exceptions include Espinoza et al. 2011; Piraino and Szedlmayer

* 2014; Roy et al. 2014). Instead, researchers often rely on HPE and derived HPEm estimates of tagged animals, borrow accuracy estimates from other studies or arrays, or select an HPE cutoff that preserves a chosen percentile of position data (Reubens et al. 2013; Özgül et al. 2015;

Williams et al. 2015). Availability of high accuracy (within several meters) position data covering a large area is advantageous for studies of fish survival (Chapter 2) but is a necessity when making inferences on specific habitat use of fish.

Reef Association

Red snapper were most active and farthest from the reef during both crepuscular periods.

During the daytime, especially in the late morning to early afternoon, fish were much less active, staying closer to the reef, but moving higher in the water column. It is likely that red snapper were foraging over large areas of habitat and feeding on demersal organisms from dusk to dawn, which is supported by findings of other studies that red snapper and other reef fish often feed some distance away from, not directly at, artificial reefs (Tarnecki and Patterson 2015; Dance et al. 2018). Red snapper moved higher in the water column and had much smaller space utilization during the day, suggesting that fish were not actively searching for food but may have been opportunistically taking pelagic prey from the water column during daytime. This particular behavior, whereby red snapper concentrate into a smaller area above artificial reefs, likely explains why they are easily caught over artificial reefs during the day by fishers (as suggested by high catch rates on daytime charter fishing trips, see Chapter 4). There was a positive relationship between fish length and height above the bottom as well as between fish length and

distance to the reef, suggesting that smaller red snapper are more dependent on reefs as a means to avoid predation, possibly because their small size leaves them more vulnerable to predators

(Patterson et al. 2001).

Red snapper used much smaller 2- and 3-D space during summer (April to August) than winter. Similarly, there was almost a 10-fold difference in KDEs between the 30-m and 55-m arrays. Larger 95% hourly KDEs suggest fish are actively covering more space, likely in response to lower density of food organisms during winter months or at deeper water depths as they search for prey. The effect of season on distance from reefs and distance above the bottom is much more nuanced and may indicate seasonal prey switching or shifts in fish behavior to other activities such as spawning. In previous studies, red snapper exhibited a variety of patterns in use, movement, and depth as a function of time of day and season. For example, in studies of red snapper around single reefs, Williams-Grove and Szedlmayer (2017) observed red snapper were more active during the daytime and summer and Piraino and Szedlmayer (2014) concluded red snapper were most active during nighttime and summer. In a recent acoustic telemetry study of red snapper over a 0.8 km2 area of artificial reefs in the western GOM, there were only slight differences in activity over time of day (fish were relatively active at all hours with a peak during the early morning; Froehlich et al. 2019). It is possible that red snapper alter movement patterns and prey preference to maximize exploitation of available food resources, which may be sufficiently variable in space and time (possibly over multiple years) that no single study has yet captured the full range of these behaviors.

GAMs for 95% hourly KDEs revealed a large range of variation in space use by reef, depth, season, and time of day. These patterns in space use may indicate that fish movement is the product of trade-offs between competition for food resources and shelter from predators. For

example, when comparing reefs within the same depth and season, I found when fish had smaller median distances to the reef, the upper quartile and 5% of positions were farther away from the reef than when fish had larger median distances to the reef. It is possible that this space-use pattern where fish were spending most time relatively close to the reef but making longer distance forays could be associated with high predator presence at the reefs. If overall predator presence within the area is high (as might occur with higher population densities of predators), red snapper might capitalize on the times when predators are least present at a particular reef, moving farther away from other competitors at the reef to access higher food densities and maximize feeding rates, thus minimizing the amount of time that they spend exposed to risk of predation (as occurs with "bout feeding", Ahrens et al. 2012). In contrast, when fish occupy intermediate-sized areas around reefs and do not make long-distance excursions, it is possible that overall predation presence within the area is lower, allowing fish to spend more time foraging at closer distances to the reef where competition with other reef fishes for food is higher

(which is indicative of a "continuous exchange" foraging arena, Ahrens et al. 2012). As densities of predators and reef fishes change over time, perhaps as large pelagic sharks, red snapper, and other reef fish populations recover, the space-use patterns of fishes at artificial reefs may also shift, complicating decisions by managers regarding the placement of artificial reefs.

Site Fidelity

Estimates of site fidelity of red snapper (the tendency of fish to remain at or leave the reef, often the reef where they were originally captured and released) are variable over time and hence sensitive to the timescale of observations. Estimates from multiple-year studies would be expectedly less sensitive to intra-annual variation. Several long-term mark-recapture studies have estimated site fidelity ranging from 12.8%–26.5% per year (Patterson and Cowan 2003; Addis et al. 2016) to 46% (Ingram and Patterson 2001; Diamond et al. 2007). Some acoustic telemetry

studies of red snapper movement resulted in similar site fidelity estimates (e.g., 67% over 117 days; Szedlmayer and Schroepfer 2005, equivalent to 29% per year) while other acoustic telemetry studies have yielded estimates as high as 72%–82% per year (Topping and Szedlmayer

2011; Williams-Grove and Szedlmayer 2016b). I found large differences in estimated fidelity across reefs, possibly explaining disagreements in fidelity estimates between studies for dissimilar reef types. Although sample size was small, it appears that fish were less likely to stay on tagging reefs if there were other reefs close by, possibly because the distance that fish must traverse when they are most susceptible to predation) to reach a neighboring reef was low and fish would be more likely to redistribute in response to limiting resources at the tagging reef.

Several researchers have suggested red snapper and other reef fish may make long distance movements in response to tropical storms or hurricanes (Patterson et al. 2001; Secor et al. 2019). During this study, there were 4 named storms that passed within 300 km of the study site in addition to Hurricane Irma which was a large storm with the center passing within 500 km of the study site. Of the 6 red snapper that were tagged in spring 2016 and survived for at least

48 hours, 4 fish emigrated from the acoustic array 3 days or less before Tropical Storm Colin passed 240 km southeast of the array on 6 June 2016. It is also possible that storms may increase loss rates of external tags: of the 8 fish that were tagged in September 2017, 4 fish lost the acoustic transmitter tag either during or shortly after Hurricanes Irma and Nate passed through the area on 11 September and 7 October, respectively. It is possible that tagged fish were exposed to greater contact with reef structure, other fish, or simply experienced greater strain on the external tag attachment during storms due to water movement that increased tag loss. Strong storms may also have effects on fish behavior and space use aside from influencing emigration rates. The data from this study are being included in a meta-analysis of acoustic transmitter

tagged red snapper that were at large during tropical storms and hurricanes in the GOM during

2016–2018, including Hurricane Harvey, a category 4 hurricane that affected the western GOM in August 2017 and Hurricane Michael, a category 5 hurricane that moved through the eastern

GOM in October 2018. By pooling data across multiple studies, we hope to increase sample size sufficiently to address specific hypotheses regarding the effects of strong storms on red snapper large-scale movements, space use, and behavior.

Factors Affecting Space Use

Tagged red snapper used only a small subset of the artificial reefs within each array.

Although the target was to tag 80 red snapper in each array, the number of fish with available long-term movement data was greatly reduced by discard mortality (Chapter 2), difficulty in catching red snapper in the 30-m array, fish moving to reefs that were on the edge of the array where VPS accuracy was not within the 5- or 6- m cut-off, fish emigrating from the array, and tag loss. Since the majority of tagged fish were captured from just a few artificial reefs, the sample of individuals included in this movement study likely does not fully represent all red snapper in the study area. Among the roughly 50% of reefs that fish used that also had known configurations, there was large variation in the expected distance from tagged fish to the reef for reefs of similar type and size. The spacing between reefs (e.g., distance to the nearest neighboring reef) had a significant effect on space use of fish and there was apparently an optimum distance of 600–700 m. Given the limited number of reefs that red snapper were observed using in this study (<10), the apparent correlation between red snapper space use and reef spacing warrants further investigation.

Reef size, represented as volume of the reef, likely affects how fish use the surrounding space. I observed fish generally used larger spaces and stayed further from smaller reefs. For example, fish occupied large hourly 95% 3-D KDEs within the 55-m array at reefs a, c, and m,

which are 21 m3 pyramids, while fish stayed closer to the relatively large (36–72 m3) chicken coops in the 30-m array. This apparent correlation between red snapper space use and reef volume may have been amplified by confounding reef depths, as prey availability was likely lower in the 55-m array than in the 30-m array, requiring fish to range over a larger area to find demersal organisms while greater depths afforded closer access to a larger volume of water close to the reef where red snapper may have been feeding on pelagic fish and zooplankton

(McCawley and Cowan 2007; Tarnecki and Patterson 2015; Schwartzkopf et al. 2017; Dance et al. 2018). The fish and invertebrate communities that artificial reefs support may be influenced by reef size and type (Dance et al. 2011; Ajemian et al. 2015), which in turn may influence red snapper movement, space use, and foraging activity around reefs. There were no species composition or fish density monitoring surveys conducted during this study, however, some reefs in the 30-m array were cleared of lionfish (Pterois volitans) in January 2016. Patterson (2013) observed red snapper using larger home ranges at artificial reefs that had high lionfish densities. I did not observe a difference in space use or depth of fish around artificial reefs that were cleared of lionfish. Small sample size (6 reefs) and confounding reef characteristics (type of reef and distance to nearest neighboring reef) may be responsible for the apparent lack of relationship between lionfish and red snapper space use.

Movers vs. Stayers

I observed that some fish visited more reefs than others, and commonly made farther excursions from a given reef, while other fish seem to use only one reef for extended times.

Interestingly, fish that visited more reefs, and could therefore be classified as movers, did not have larger KDEs of space utilization, suggesting the move from one reef to another took place rarely and quickly enough to be overlooked in the large sample size of hourly KDE estimates.

Among fish that stayed on only one reef, some routinely used larger spaces and some had

relatively small hourly 95% KDEs, which may reflect the ability of an individual fish to find sufficient food and avoid predation while minimizing activity and energy expenditure.

Unfortunately, the relatively short timeframe of my observations (several months) is much less than the lifetime of red snapper, which can live over 50 years, thus making it difficult to identify variation in behavior patterns between fish amid variation that likely occurs through the lifetime of the individuals.

When the effects of time of day, time of year, and environmental conditions (atmospheric pressure, temperature, wave height, and wind gust speed) were controlled for and individuals were compared within reefs, some individual red snapper displayed more variable and generally larger hourly 95% 2-D KDEs (e.g., hourly 95% KDE >350 m2 in the 30-m array and >550 m2 in the 55-m array). These more active fish routinely used larger spaces and made trips farther from the reef, increasing the amount of time at risk of predation by large predators. In general, especially in the 55-m array where greater depths afford fish easier access to pelagic habitat close to the reefs, space use between fish followed a continuum more than a dichotomy. Based on a fine-scale acoustic telemetry study in a North Carolina estuary, Fodrie et al. (2015) observed that red drum (Scieanops ocellatus) displayed habitat preferences over short timescales

(days or weeks) but not over longer timescales (months), suggesting that specialization was not a sign of individuality but rather shifting behaviors over time. In this study, fish were monitored for up to 12 months, such that short-term shifts in space use around artificial reefs and movement between reefs blended to reveal that specialized behaviors in red snapper are not consistent among individuals.

I observed many tagged fish leave the array, hence observations of these active fish were truncated whereas fish that stayed within the array for long periods had many more observations.

Because fish that emigrated from the array may have been inherently more active individuals

(movers) it is likely that the remaining observations are skewed towards less movement-likely fish, possibly biasing conclusions. This sampling bias where more sedentary individuals are more likely to be observed is a central limitation of acoustic telemetry. The large size (>15 km2) of the acoustic array in this study reduces detection bias relative to other studies relying on small

(<1 km2) acoustic arrays centered on single reefs. However, given that red snapper are known to move 10s to 100s of km over several months (Beaumariage 1969; Patterson et al. 2001), very large acoustic telemetry arrays or consortiums of arrays will be necessary to sample the full breadth of movements in red snapper populations. An additional source of sampling bias in my data may be caused by the small number of reefs (2 in each array) where red snapper were caught and tagged in the study area. There was a noticeable difference in activity levels depending on which reef fish were on and many fish stayed on the reef where they were tagged.

If fish that aggregate together on a given reef have common individual characteristics, then the effective sample size available to investigate variation in these types of individual characteristics may have been decreased.

Future Directions

The potential application of fine-scale positioning acoustic telemetry to address ecological questions is expanding as technology has advanced, costs have decreased, and analytical capabilities, such as VPS and computationally intensive 3-D movement-based models, are becoming more accessible. Ideally, future studies should be designed to apply acoustic telemetry methods towards broader ecological principles in the marine environment, such as predator-prey interactions in foraging arenas, niche overlap or modification among competing species, or the ramifications of invasive species and climate change. A thorough understanding of habitat within a large study area is necessary to answer these broader questions, providing a

valuable opportunity for multi-disciplinary collaboration among researchers. For example, in the northern GOM, the distribution of artificial reefs can change over time as reefs are buried and uncovered by sediment or new artificial reefs are added. Full-coverage surveys of bottom habitat such as sidescan performed regularly within the study period would provide a thorough understanding of reef structure, while reliable continuous oceanographic data collection within the study area would indicate more accurately the conditions that fish are exposed to as they interact with their environment. The increase of collaborative networks of acoustic telemetry researchers holds the promise that researchers could pool acoustic telemetry resources and tagging expertise to deploy larger positioning arrays and track multiple species and their interactions simultaneously. Future acoustic telemetry studies may also increase our understanding of how much is poorly understood regarding marine fish spatial ecology and behavior. As our ability to collect and analyze vast amounts of fine-scale spatial and temporal movement data improves, perhaps we will find that the magnitude of variation among individual fish and dynamic processes such as a changing environment may require ever-increasing sample sizes and complexity in our approach.

Figure 3-1. A map of the northern Gulf of Mexico indicating the locations of the 30-m (▲) and 55-m (■) acoustic arrays. Depth contours are in meters.

Figure 3-2. A map of the 60-receiver shallow acoustic array deployed at 28−35 m depth from February 2016 to March 2017. Receiver locations and the approximate extent of the array from February 2016 to September 2016 are shown as solid triangles and a dark border while receiver locations and array extent from September 2016 to March 2017 are shown as white symbols and a white border.

Figure 3-3. A map of the 46-receiver deep acoustic array deployed at 48−55 m depth from August 2017 to July 2018. Receivers and array extent are shown solid triangles and a dark border.

A B

Figure 3-4. Map and figure describing the drift test performed in the 55-m acoustic array to confirm the positional accuracy of the array. A) Map showing the track of each drift test (white circles) within the array. The broken line indicates that approximately array extent and triangles denote acoustic receiver positions. B) A diagram of the F/V Intimidator illustrating how transmitter locations were corrected to correspond to the GPS antenna according to the geometry of the boat and bow heading through the test.

Table 3-1. Results of the drift test to evaluate positional accuracy of the 55-m array. Positional efficiency is the number of test tag transmissions that yielded VPS position estimates. Start End n Positioning Accuracy (m) HPE Site (UTC) (UTC) Positions efficiency mean range mean range NW 15:21 15:33 7 10% 7.9 4.1–10.4 3.9 3.8–4.0 NE 15:04 15:14 13 22% 5.1 3.1–7.6 4.6 3.3–6.9 Center 13:57 14:11 9 11% 1.0 0.0–3.7 3.9 3.0–5.4 SW 14:39 14:49 40 67% 4.3 0.5–8.2 3.2 2.4–6.2 SE 14:20 14:30 31 52% 3.7 0.0–6.4 3.3 2.6–4.4 Overall 100 30% 4.1 0.0–10.5 3.5 2.4–6.9

* RS65 re-entered the array on 26 January 2017 displaying a movement pattern consistent with having been consumed by a large predator.

Figure 3-5. Summary of fish reef residency and fate for the 30-m array. Fish fates include emigration (em, the fish left the array), tag loss (ls, tag detached from the fish and was lying on the bottom), predation (p), and at-large (al, the fish was still present in the array when the acoustic receivers were retrieved).

Figure 3-6. Images of common artificial reef types in the 30-m array. A) paired cement pyramid Reef 2 in April 2016 and B) Paired chicken coop Reef Z in October 2016. Photos courtesy of Steven B. Garner.

Figure 3-7. Summary of fish reef residency and fate for the 55-m array. Fish fates include emigration (em, the fish left the array), tag loss (ls, tag detached from the fish and was lying on the bottom), predation (p), and at-large (al, the fish was still present in the array when the acoustic receivers were retrieved).

Figure 3-8. An example large single steel and tire pyramid reef (reef m) in the 55-m array photographed in July 2018. Reefs a and c are of similar construction. Photo courtesy of Steven B. Garner.

Table 3-2. Estimated annual reef fidelity of red snapper. 30 days 60 days Remain Moved Tag lost / Annual reef Remain Moved Tag lost / Annual reef Reef (n) (n) predation fidelity (%) (n) (n) predation fidelity (%) 30-m array 2 16 1 5 47.8 8 7 7 2.2 Z 4 5 4 < 1 1 8 4 < 1 55-m array a 5 3 3 < 1 2 6 3 < 1 c 5 0 5 100 5 0 5 100 m 0 3 0 < 1 0 3 0 < 1

A B

C D

Figure 3-9. Distance of tagged fish from reefs 2 and Z in the 30-m array during A) September and B) May, and distance of tagged fish from reefs a and c in the 55-m array during C) September, and D) May.

Table 3-3. Candidate GAMs explaining hourly 95% 2-D KDE for tagged red snapper. Categorical variables (array and fish ID) are treated as fixed effects and smoothed terms are noted as s(). ΔAIC values are shown relative to the model with the lowest AIC.

Model % Variance Residual ΔAICc explained df

β0 + s(hour) + s(d_reef_neighbor) + s(day) + array + 14.17 51919 0 reef_depth + s(fish_length) + s(temperature) + s(gust_speed) + s(atm_pressure) + s(wave_height)

β0 + s(hour) + s(d_reef_neighbor) + s(day) + array + 14.14 51922 11 reef_depth + s(fish_length) + s(temperature) + s(gust_speed) + s(atm_pressure)

β0 + s(hour) + s(d_reef_neighbor) + s(day) + array + 14.09 51929 28 reef_depth + s(fish_length) + s(temperature) + s(gust_speed)

β0 + s(hour) + s(d_reef_neighbor) + s(day) + array + 13.94 51931 113 reef_depth + s(fish_length) + s(temperature)

β0 + s(hour) + s(d_reef_neighbor) + s(day) + array + 13.68 51939 254 reef_depth + s(fish_length)

β0 + s(hour) + s(d_reef_neighbor) + s(day) + array + 13.27 51947 483 reef_depth

β0 + s(hour) + s(d_reef_neighbor) + s(day) + array 12.11 51948 1173

β0 + s(hour) + s(d_reef_neighbor) + s(day) 10.48 51952 2122

β0 + s(hour) + s(d_reef_neighbor) 8.06 51959 3494

β0 + s(hour) 5.10 51968 5121

β0 + fish_ID 4.27 51901 5710

β0 0.00 51976 7827

Table 3-4. Candidate GAMs explaining hourly 95% 3-D KDE for tagged red snapper. Categorical variables (array and fish ID) are treated as fixed effects and smoothed terms are noted as s(). ΔAIC values are shown relative to the model with the lowest AIC. Model % Variance Residual ΔAICc explained df β0 + array + s(hour) + s(day) + s(d_reef_neighbor) + 24.96 51914 0 reef_depth + s(fish_length) + s(temperature) + s(gust_speed) + s(atm_pressure) + s(wave_height) + s(moon_visibility) β0 + array + s(hour) + s(day) + s(d_reef_neighbor) + 24.95 51916 4 reef_depth + s(fish_length) + s(temperature) + s(gust_speed) + s(atm_pressure) + s(wave_height) β0 + array + s(hour) + s(day) + s(d_reef_neighbor) + 24.92 51920 14 reef_depth + s(fish_length) + s(temperature) + s(gust_speed) + s(atm_pressure) β0 + array + s(hour) + s(day) + s(d_reef_neighbor) + 24.86 51927 41 reef_depth + s(fish_length) + s(temperature) + s(gust_speed) β0 + array + s(hour) + s(day) + s(d_reef_neighbor) + 24.76 51929 106 reef_depth + s(fish_length) + s(temperature) β0 + array + s(hour) + s(day) + s(d_reef_neighbor) + 24.57 51937 224 reef_depth + s(fish_length) β0 + array + s(hour) + s(day) + s(d_reef_neighbor) + 24.00 51945 599 reef_depth β0 + array + s(hour) + s(day) + s(d_reef_neighbor) 23.16 51946 1171 β0 + array + s(hour) + s(day) 21.65 51955 2159 β0 + array + s(hour) 19.42 51963 3605 β0 + fish_ID 18.06 51901 4600 β0 + array 14.99 51971 6368 β0 0.00 51974 14805

Table 3-5. Candidate GAMs explaining distance from the reef (in 2 dimensions) for tagged red snapper. Categorical variables (array and fish ID) are treated as fixed effects and smoothed terms are noted as s(). ΔAIC values are shown relative to the model with the lowest AIC. Model % Variance Residual ΔAICc explained df β0 + s(day) + array + reef_depth + s(fish_length) + 21.79 395124 0 d_reef_neighbor + s(hour) + s(temperature) + s(wave_height) + s(moon_visibility) + s(atm_pressure) + s(wind_speed) + s(gust_speed) β0 + s(day) + array + reef_depth + s(fish_length) + 21.77 395131 109 d_reef_neighbor + s(hour) + s(temperature) + s(wave_height) + s(moon_visibility) + s(atm_pressure) + s(windspeed) β0 + s(day) + array + reef_depth + s(fish_length) + 21.71 395139 396 d_reef_neighbor + s(hour) + s(temperature) + s(wave_height) + s(moon_visibility) + s(atm_pressure) β0 + s(day) + array + reef_depth + s(fish_length) + 21.62 395148 854 d_reef_neighbor + s(hour) + s(temperature) + s(wave_height) + s(moon_visibility) β0 + s(day) + array + reef_depth + s(fish_length) + 21.41 395156 1899 d_reef_neighbor + s(hour) + s(temperature) + s(wave_height) β0 + s(day) + array + reef_depth + s(fish_length) + 21.14 395165 3222 d_reef_neighbor + s(hour) + s(temperature) β0 + s(day) + array + reef_depth + s(fish_length) + 20.74 395173 5217 d_reef_neighbor + s(hour) β0 + s(day) + array + reef_depth + s(fish_length) + 18.90 395181 14251 d_reef_neighbor β0 + s(day) + array + reef_depth + s(fish_length) 16.85 395182 24103 β0 + s(day) 16.55 395194 25509 β0 + s(day) + array + reef_depth 14.09 395191 36992 β0 + fish_ID 13.01 395104 42087 β0 + s(day) + array 11.15 395192 50298 β0 0.00 395203 96994

Table 3-6. Candidate GAMs explaining depth above bottom for tagged red snapper. Categorical variables (array and fish ID) are treated as fixed effects and smoothed terms are noted as s(). ΔAIC values are shown relative to the model with the lowest AIC. Model % Variance Residual ΔAICc explained df β0 + s(fish_length) + reef_depth + s(hour) + 32.64 342730 0 s(temperature) + s(day) + d_reef_neighbor + s(moon_visibility) + s(wave_height) + s(atm_pressure) + s(gust_speed) + s(wind_speed) + array β0 + s(fish_length) + reef_depth + s(hour) + 32.50 342733 708 s(temperature) + s(day) + d_reef_neighbor + s(moon_visibility) + s(wave_height) + s(atm_pressure) + s(gust_speed) + s(wind_speed) β0 + s(fish_length) + reef_depth + s(hour) + 32.38 342740 1267 s(temperature) + s(day) + d_reef_neighbor + s(moon_visibility) + s(wave_height) + s(atm_pressure) + s(gust_speed) β0 + s(fish_length) + reef_depth + s(hour) + 32.28 342748 1790 s(temperature) + s(day) + d_reef_neighbor + s(moon_visibility) + s(wave_height) + s(atm_pressure) β0 + s(fish_length) + reef_depth + s(hour) + 31.99 342757 3212 s(temperature) + s(day) + d_reef_neighbor + s(moon_visibility) + s(wave_height) β0 + s(fish_length) + reef_depth + s(hour) + 31.70 342766 4651 s(temperature) + s(day) + d_reef_neighbor + s(moon_visibility) β0 + s(fish_length) + reef_depth + s(hour) + 31.16 342774 7324 s(temperature) + s(day) + d_reef_neighbor β0 + s(fish_length) + reef_depth + s(hour) + 30.16 342775 12297 s(temperature) + s(day) β0 + s(fish_length) + reef_depth + s(hour) + 29.13 342783 17298 s(temperature) β0 + fish_ID 28.01 342713 22816 β0 + s(fish_length) + reef_depth + s(hour) 26.98 342792 27515 β0 + s(fish_length) + reef_depth 23.54 342800 43275 β0 + s(fish_length) 19.90 342801 59231 β0 0.00 342810 135270

A B

C D

Figure 3-10. Best GAMs predicting the effects of time of year on A) hourly 95% 2-D KDE, B) hourly 95% 3-D KDE, C) distance from reef, and D) depth above bottom. In each figure, the solid line shows expected values and the dashed lines represent the upper and lower 95% confidence intervals.

A B

C D

Figure 3-11. Best GAMs predicting the effects of time of day on A) hourly 95% 2-D KDE, B) hourly 95% 3-D KDE, C) distance from reef, and D) depth above bottom. In each figure, the solid line shows expected values and the dashed lines represent the upper and lower 95% confidence intervals.

A B

Figure 3-12. Best GAMs predicting the effects of time of day category on A) hourly 95% 3-D KDE, and B) depth above bottom. For each response variable, the best GAM was modified to include time of day as the categorical variable instead of the continuous cyclic variable hour. In each figure, the solid line shows expected values and the dashed lines represent the upper and lower 95% confidence intervals.

A B

C D

Figure 3-13. Best GAMs predicting the effects fish length (TL, cm) on A) hourly 95% 2-D KDE, B) hourly 95% 3-D KDE, C) distance from reef, and D) depth above bottom. In each figure, the solid line shows expected values and the dashed lines represent the upper and lower 95% confidence intervals.

A B

C D

E F

Figure 3-14. Best GAMs predicting the effects of environmental variables on hourly 95% 3-D KDE: A) bottom temperature, B) wave height, C) percent moon visible, D) wind gust speed, E) distance to nearest neighbor reef, and F) atmospheric pressure. In each figure, the solid line shows expected values and the dashed lines represent the upper and lower 95% confidence intervals

A B

Figure 3-15. Distance between tagged fish and reefs in A) the 30-m array, and B) the 55-m array. In each figure, the solid line shows expected values and the dashed lines represent the upper and lower 95% confidence intervals.

A B

Figure 3-16. Best GAM predicting hourly 95% 2-D KDE by fish tagged on A) reef 2, and B) reef Z. Time of day, time of year, and environmental conditions (atmospheric pressure, temperature, wave height, and wind gust speed) have been controlled in the model. In each figure, the solid line marks the median and the box includes the 95% confidence interval.

Figure 3-17. Best GAM predicting hourly 95% 2-D KDE by fish tagged on reefs a and c. Time of day, time of year, and environmental conditions (atmospheric pressure, temperature, wave height, and wind gust speed) have been controlled in the model. In each figure, the solid line marks the median and the box includes the 95% confidence interval. The solid line marks the median and the box includes the 95% confidence interval.

CHAPTER 4 DISCARDING BEHAVIOR, USAGE, AND ATTITUDES TOWARDS DESCENDER DEVICES ON NORTHERN GULF OF MEXICO REEF FISH CHARTER VESSELS

Recreational fishers in the eastern Gulf of Mexico (GOM) discard 70–80% of the red snapper (Lutjanus campechanus) they catch, which accounts for over 75% of all discarded red snapper, including shrimp trawl bycatch (2012–2015; SEDAR 2018). Recreational discard rates for other reef fish species, such as gray triggerfish (Balistes capriscus), gag (Mycteroperca microlepis), and greater amberjack (Seriola dumerili), are also high, ranging from 63% to 93% of total recreational catch by species (SEDAR 2014a, 2014b, 2015a). Discard data that are self- reported by fishers and surveyed by dockside agents typically only include basic information such as the species, number, and status at release (live versus dead). Results of empirical studies suggest multiple factors influence survival probability of discarded reef fish, including gear configuration, depth of capture, hook location, fish length, barotrauma impairment, time out of water, release method, release condition, and season (Bartholomew and Bohnsack 2005;

Campbell et al. 2010a; Benaka et al. 2014). Because these parameters are lacking for self- reported discards, stock assessments for many GOM reef fish apply a point estimate of discard mortality to all sectors of the recreational fishery (SEDAR 2014a, 2014b, 2015a, 2015b, 2016,

2018), causing a considerable degree of uncertainty in estimates of total removals (landed catch plus dead discards). Given the complexity of factors which affect discard mortality and the variability in discard mortality estimates, NOAA has repeatedly called for further study on discard rates and mortality within GOM reef fish fisheries.

The Gulf of Mexico Fishery Management Council (GMFMC) is currently pursuing policies and outreach programs to encourage GOM fishers to use descender devices to reduce mortality of discarded reef fish that are suffering from barotrauma (Gulf of Mexico Fishery

Management Council 2017, 2018a, 2018b, 2018c, 2019a). However, results of studies conducted

to examine the efficacy of descender devices to reduce discard mortality are equivocal (Wilde

2009; Campbell et al. 2014; Eberts and Somers 2017), although recent research suggests descender devices can reduce discard mortality of red snapper (Chapter 2; Bohaboy et al. 2019).

Regardless, the ability of descender devices to influence GOM reef fish stocks will rely on if and how they are employed by fishers. The efficacy of descender devices in the recreational fishery, including the for-hire sector, is poorly understood because there is a scarcity of data on the prevalence of their usage in the fishery. The effectiveness of descender devices in improving post-release survival of discarded fish may also be heavily dependent on where and when anglers fish (depth, season, and presence of predators), the behavior of fishers (targeted species, gear configuration, and handling technique such as time out of water), and characteristics of fish that are discarded (species, size, hooking injury, or barotrauma impairment). The frequency of using descender devices to release fish within the charterboat sector, as well as the factors influencing fishers' choice to use these tools, is currently not understood. Finally, the success of a recommendation or regulation to use descender devices hinges on the willingness of fishers and deckhands to adopt to this new technology.

Cooperative research using at-sea observers on recreational fishing boats has been undertaken in several U.S. fisheries (Stephens et al. 2006; Bochenek et al. 2012; Garner and

Patterson 2015; Sauls et al. 2017; SEDAR 2018), and can be invaluable for understanding the fishing practices, discarding behavior, and attitudes of fishers, while building stakeholder support for fisheries management (Mackinson et al. 2011). For example, the Florida Fish and Wildlife

Conservation Commission (FWC) began an at-sea observer program on for-hire recreational vessels in 2005. In recent years, observers have collected data on discarded reef fish on approximately 80 trips annually from northwest Florida. However, there have been only a

limited number of trips from the western Florida Panhandle and Alabama, where fishers typically target red snapper over numerous small artificial reefs (Dominique Lazarre, FWC, personal communication). The FWC observer program also does not evaluate fisher behavior and attitudes regarding descender devices, suggesting the need for a targeted study on discarding and descender device use in the northern GOM charterboat fleet.

The overall goal of this portion of my dissertation was to understand how descender devices could be used on northern GOM reef fish charterboats to reduce discard mortality of red snapper and other reef fishes. Specific objectives of this study were to: 1) collect detailed data on fishing behavior, catch, and discards of reef fish by fishers participating in the northern GOM for-hire recreational reef fish fishery, particularly in relation to discard mortality; 2) evaluate the prevalence of and factors influencing descender device use; 3) characterize the burden associated with descender device use (i.e., the amount of time and effort required to release discarded fish with the devices); 4) increase observations on predator interaction with descender devices

(predation of fish by sharks or dolphins); and 5) collect information on fishers' opinions towards non-consumptive (catch-and-release) reef fish fishing and using descender devices to release discarded fish.

Methods

An observer accompanied fishing trips on charter fishing vessels (ranging in capacity from 6 to 16 passengers) to collect information on fishing effort and catch. For each site where customers fished for reef fish, the observer recorded the number of fishing rods or rigs used simultaneously, terminal tackle used, general fish targeted (e.g., red snapper, gray triggerfish, grouper, vermilion snapper Rhomboplites aurorubens, etc.), time spent fishing, sea and weather conditions, water depth, and GPS coordinates. All fish were identified to species, measured, and observed for any symptoms of barotrauma (exophthalmia, pronounced bloating, prolapsed

intestine or gonads, everted stomach). Hooking location (mouth, throat, gills) and any additional trauma, such as rough handling or being dropped on the deck, were also noted. The decision to retain or discard fish was made by the captain, deckhand, or customers. Fish that were released on the surface were observed and assessed for release condition following Patterson et al. (2002): condition-1 = fish immediately oriented to the bottom and swam down rapidly; condition-2 = fish oriented to the bottom and swam down slowly or erratically; condition-3 = fish remained on the surface; and condition-4 = fish was apparently dead at the surface, including from predation.

Some fish were released using a descender device (SeaQualizer) under the direction of the captain or deckhand. A downward looking GoPro Hero3 camera was mounted above the descender device to record fish during release to evaluate the performance of the descender device, behavior of released fish, and predator interaction. The total time out of water and the time required to put fish on the descender device, when appropriate, were recorded for all discarded fish.

At the end of each fishing trip, either during the return to port or at the dock, charterboat customers were offered a 12-question survey to assess their beliefs and attitudes towards catch- and-release fishing, post-release survival of fish, using descender devices, and conservation in general (Table 4-1). Respondents were asked to react to each of the 12 statements on a Likert scale that included: 1 ("strongly disagree"), 2 ("disagree"), 3 ("no opinion"), 4 ("agree"), and 5

("strongly agree"). Median response scores for each question were compared across prospective respondent groups including whether or not a descender device was used on that trip, whether the trip occurred in red snapper season, and which charter vessel the trip occurred on. The observer avoided influencing customers' responses to the survey. For example, if a customer asked the observer what a descender device was, the observer would answer only after the survey was

completed. The observer also did not interfere with communication between the customers and the deckhands and captains, who routinely share opinions and information regarding fishing, regulations, and conservation to clients as part of the overall charterboat experience. The Mann-

Whitney U test and Kruskal-Wallis one-way analysis of variance were used to assess whether the responses were significantly different among respondent group assignments.

Results

Observers accompanied 56 trips on 4 for-hire recreational fishing vessels between June

2016 and September 2018. Thirty (30) trips took place during open red snapper seasons and 26 trips took place during closed red snapper seasons. The majority of the 296 fishing sites were

<30 m deep and <20 km from shore (Fig. 4-1). The mean site depth was 30.6 m for open season and 35.4 m for closed season fishing trips. The mean distance to shore was 26.2 km for open season and 31.0 km for closed season fishing trips. The 5 primary types of fishing that were observed are: 1) targeted red snapper fishing mid-water with single-hook rigs; 2) bottom fishing for vermilion snapper, red porgy (Pagrus pagrus), or other small reef fishes with 2- or 3-hook bottom rigs; 3) grouper fishing using large single-hook rigs with a long leader and live bait; 4) targeted gray triggerfish fishing mid-water using similar gear as for red snapper; and 5) fishing for sharks or jacks with single-hook 3-way swivel rigs. During open seasons, red snapper were targeted most often (>86% of observed effort; Fig. 4-2), and fishers typically only targeted other species after the red snapper bag limit for all customers had been retained. During closed seasons, vermilion snapper and red porgy were most commonly targeted (>43% of observed effort; Fig. 4-2). The crew of one of the observed vessels conducted targeted red snapper catch- and-release fishing during the closed season. As a result, 31% of the observed closed season effort in this study was targeted red snapper fishing.

100

The observed catch was dominated by red snapper (50%), vermilion snapper (16%), gray triggerfish (12%), and red porgy (9%); the remaining catch included 37 other fish species (Table

4-2). The percentage of catch that was discarded varied among species, from zero for less commonly encountered species with long seasons and liberal minimum size regulations, such as gray snapper (Lutjanus griseus), gag, lane snapper (L. synagris), and tilefishes (Caulolatilus cyanops, C. microps, and Lopholatilus chaemaeleonticeps), to 100% for prohibited closed season species (e.g., gray triggerfish and greater amberjack) and undesirable species (e.g., sharksucker,

Echeneis naucrates).

During red snapper open season trips, over half of the total red snapper catch was discarded (n = 410), the majority of which were legal harvest size (≥406 mm total length, TL)

(Fig. 4-3). The decision to retain or discard a legal-sized red snapper was made most often by boat captains, who rarely consulted the customers on individual fish. The motivations associated with open season red snapper discarding could be categorized by common patterns: 1) "live high grading" when captains expected to catch and preferred to harvest larger fish; 2) customers did not want to harvest the maximum number of red snapper; or 3) the bag limit of 2 red snapper per person had already been harvested. During red snapper closed seasons, the size distribution of discards varied depending on targeted species. When fishers were targeting red snapper during closed red snapper season for catch-and-release, the length distribution of discards was similar to the overall distribution of catch during the open season. When fishers were not targeting red snapper and were primarily targeting vermilion snapper, red porgy, and other small reef fish, discarded red snapper were generally smaller (<400 mm TL) but were less commonly encountered (n = 158). Catch rates (fish per angler × hour) by species were generally much higher when the species was the primary target (Table 4-3).

101

Almost 16% of red snapper showed signs of barotrauma (Table 4-4), representing the largest number of afflicted fish because red snapper were abundant in the total catch. The most common signs of barotrauma in red snapper were bloating and everted stomach. Of the total surface-released red snapper with condition observations (n = 955), 4.3% were severely impaired at the time of release (condition-3 or -4; Fig. 4-4). Tilefishes had the highest occurrence (27.3%) of observed barotrauma symptoms but were caught relatively rarely since they were only targeted on longer trips farther offshore. Tilefishes were never discarded. Groupers (e.g.,

Epinephelus adscensionis, E. morio, Mycteroperca microlepis, and M. phenax) were also likely to show signs of barotrauma (23.5% of total grouper catch). Of the 33 greater amberjack that were observed during release at the surface, 2 (6.1%) were in release condition 2 as they struggled at the surface for some time (typically less than 1 minute) before swimming down deep enough to stay submerged. An additional 5 greater amberjack (15.2%) were unable to submerge

(condition-3), 4 of which were retrieved and then released with a descender. Gray triggerfish displayed barotrauma symptoms 8.6% of the time (primarily prolapsed intestines or prolapsed gonads, but rarely inverted stomach) and were rarely released in poor condition.

All greater amberjack and over 50% of red snapper whose orientation and swimming were severely impaired upon release showed signs of barotrauma, suggesting these fish may benefit most from release with descender devices. Amberjack was targeted on some trips despite the season for greater amberjack being closed during nearly all observed trips in this study.

Targeted catch-and-release fishing for red snapper also appears to be gaining popularity among some captains who believe their customers may prefer the opportunity to catch and release trophy-sized red snapper over the opportunity to harvest smaller species. In these instances,

102

releasing fish with descender devices may be more effective and better received by captains, deckhands, and customers.

Charterboat captains elected to release 30 red snapper, 8 greater amberjack, and 1 gray triggerfish with a SeaQualizer descender device over 25 observer trips (Fig. 4-5A–B), suggesting that descender device use is relatively rare (approximately 3 fish were descended for every 100 fish that were discarded), at least among observed vessels. The mean time required to attach the fish to the descender device was 29 s (range 7–60 s). Captains frequently retrieved and then descended fish that were unable to swim down on their own (condition-3). Most captains and deckhands preferred to set the SeaQualizer to release fish at 15 m (50 ft) because A) it minimized the time required to descend the fish and retrieve the descender; B) they believed that

15 m is sufficient to increase survival; and C) the fish was removed from view of the customers who may have been distressed by releasing an apparently dead fish. Some deckhands preferred to vent fish over descending them because venting requires less time and achieves similar objectives (a perceived increase in post-release survival and a better experience for customers).

Approximately 80% of fish that were released with the descender device on observer trips were successfully recorded by a downward-facing GoPro3 camera. During 2 releases, sharks were observed investigating the descender device or camera after the fish had been released (Fig. 4-5C–D). However, sharks or dolphins were never observed preying on released fish, despite being present (i.e., observed at the surface or evidenced by fish that were preyed upon as customers reeled them in) at 30.8% of reef fish fishing sites.

Only 21 of 204 (10.3%) charterboat customers who completed the 12-question survey after their fishing trip reported they had previously used descender devices. Most participants responded positively ("agree" or "strongly agree") when asked questions regarding the ease of

103

using a descender device (questions 1 and 10), whether employing the descender device enhanced their fishing experience (questions 2 and 3), whether descender devices were beneficial to released fish (question 5), and whether they supported regulations for descender devices

(question 8; Fig. 4-6). Attitudes between participants regarding descender devices were significantly different between those who had seen a descender device used on their trip (n = 99) and those who did not see one used (n = 105; Table 4-5). Respondents who had not seen a descender device used on their fishing trip were most likely to reply "no opinion" to statements referring to descender devices, indicating that many of them were either apathetic towards descender devices and their potential benefits, or did not know what a descender device was.

There was a significant difference between fishing boats in survey responses to 8 of the 12 questions.

Discussion

Discarding fish that are undersized, out of season, or undesirable is widespread in the

GOM for-hire recreational reef fish fishery. For some types of fishing (combinations of technique and gear configuration), charterboat captains are effective at catching intended species while avoiding non-target species. When fishing for groupers and amberjacks, catch of non- target species such as red snapper, gray triggerfish, and vermilion snapper is relatively rare. In contrast, high catch rates of non-targeted species when fishing for gray triggerfish suggest this type of fishing has relatively high bycatch. For example, red snapper were caught at a rate of 3.5 fish per angler × hour when gray triggerfish were being targeted, which is nearly as high as catch rates when fishers were targeting red snapper (4.17 fish per angler × hour) and over twice the rate at which gray triggerfish were caught when fishers were targeting gray triggerfish. Red snapper and gray triggerfish are often caught together as both species readily take baited hooks and inhabit shallow (<40 m) artificial reefs in densities where they are vulnerable to charterboat

104

fishers, so it is most challenging for charterboat captains to catch one while avoiding the other.

Typically, the harvest seasons for red snapper and gray triggerfish do not co-occur, which was true for all but 3 of the trips observed in this study. It is possible that aligning the harvest seasons for these 2 species could reduce discards if fishers redirected effort to other types of fishing that have lower catch rates of red snapper and gray triggerfish, such as targeting jacks or groupers.

Aligning harvest seasons for different species is in conflict with current management approaches in the GOM where managers typically seek to have harvest seasons for popular reef fish species out of phase to increase the overall fishing season length. Catches of non-target species when fishing for vermilion snapper and red porgy were also high: fishers caught nearly half as many non-targeted fish (mostly gray triggerfish) as vermilion snapper and red porgy. Three of the 4 fishing vessels observed in this study primarily targeted vermilion snapper and red porgy over natural hard-bottom and artificial reefs during the red snapper closed seasons because these were the only reef fish species legal to harvest during most of the closed-season trips observed in this study.

Previous studies reporting recreational reef fish catch per unit effort as a function of target species are limited, and often researchers must assume fishing boats target species that are open for harvest. Garner and Patterson (2015) reported the relative percent contribution of red snapper to total catch and discards was greater during red snapper open season, suggesting fishing boats could, at least to some degree, effectively target red snapper when they chose to.

However, in this study, I observed targeted catch-and-release fishing for red snapper and greater amberjack when the harvest seasons for each of these species were closed. Though catch-and- release fishing for greater amberjack was done on 3 boats, only 1 routinely targeted red snapper during closed season. Interestingly, in the case of red snapper, catch rates (number of fish per

105

angler × hour) were 38% higher in the closed season (when the intention was to catch and release fish) than in the open season (when fish were able to be harvested), suggesting that some level of local depletion of red snapper may occur during the open season, or that elements of fish or fisher behavior cause increased catchability of red snapper during the closed season. If targeted catch-and-release fishing for reef fishes gains popularity, the assumption that fishers target species only during open seasons may become less valid, further complicating attempts to quantify and minimize discards.

I observed 94.7% of gray triggerfish catch was discarded (over all types of fishing for open and closed red snapper season combined), which is higher than discard rates reported in several previous studies (maximum 81.7%; SEDAR 2015a, range 67.4 to 86.9 % on charter fishing boats in northwest Florida; Sauls et al. 2014, and 90.7%; Garner and Patterson 2015). In the current study, higher discard rates were largely regulatory since almost all of the observed trips took place when gray triggerfish harvest was prohibited. Observed open season red snapper discard rates (51.1%) were similar to estimates from other studies of for-hire fishing boats in the eastern GOM (50.7%; Garner and Patterson 2015) but overall higher than rates reported when private recreational fishing boats were included (17.6 to 39.9% from 2011 to 2016; SEDAR

2018), thus suggesting private recreational fishers may be less adept at catching red snapper, discard fewer fish, and harvest a greater portion of the fish they catch. As a result, applying regulations or management guidelines (such as the use of descender devices to reduce discard mortality) may be more effective for for-hire fishing boats than for private fishing boats.

The majority of discarded fish appeared to be in good condition at the surface (i.e., did not display barotrauma symptoms and swam down vigorously), thus descender devices were rarely used to return fish to depth. Descender devices were never used on one of the fishing

106

vessels in this study, as crew members preferred to vent fish that were suffering from obvious signs of barotrauma. Deckhands indicated the primary reasons why fish were sometimes released on the surface in poor condition included the belief that the fish would die regardless of whether or not it was descended (as for gut-hooked fish), a descender device was not readily available, or it was not convenient to use one to release a given fish. Although the time required to attach a fish to a descender device was generally less than 30 seconds, retrieving the descender device from depth following release can require several minutes of extra effort, especially if the fish was large enough to require a heavier weight. However, using electric reels or delegating operation of the descender to a passenger could alleviate demands on crewmembers' time. I found no evidence that descender devices increased susceptibility to predation as no predators were observed interacting with any fish that were released with descender devices, allaying any potential criticism that fish attached to descender devices would be more likely to be consumed by predators. A criticism of descender devices in that past has been that they were not commercially available to the recreational fishing community (73 FR 5117). The descender device used in this study (the SeaQualizer), as well as several other designs, are now widely available in the U.S. and have also been distributed as part of several outreach programs to encourage descender device use in the GOM and South Atlantic U.S.

Benefits of descender devices would be maximized on charter fishing boats if they were used: 1) to release red snapper during targeted red snapper fishing, especially if done for catch and release, since catch rates and barotrauma incidence were high, and 2) to release greater amberjack during targeted greater amberjack catch-and-release fishing when discarding is expected and a large portion of fish are in poor condition upon release. Although small red snapper that are caught incidentally during targeted vermilion snapper and red porgy fishing are

107

often retrieved from the bottom (as opposed to shallower mid-water depths where fishers target larger red snapper) and have a relatively high incidence of barotrauma symptoms and being released in poor condition, the fast-paced nature of vermilion snapper and red porgy fishing makes it less likely that deckhands (especially on boats with a larger number of fishers) would have the time to spare for releasing fish with a descender device.

Releasing fish with descender devices would likely be practical and advantageous in the private sector of the GOM recreational reef fish fishery, which accounts for a significant portion of fishing effort and likely discards in the recreational sector, although differences between the private and charterboat recreational fleets are poorly understood (SEDAR 2018). Recreational boaters and less experienced fishermen in the private reef fish fishery may experience higher bycatch rates compared to successful charterboat captains who have access to and knowledge of a greater number and diversity of fishing sites. Even smaller charterboats will have many different types of rods, reels, and terminal tackle on board, as well as the expertise to switch gear types and techniques mid-trip. For example, during red snapper open season trips, once the bag limit had been reached on observed trips, charterboats often switched fishing practices to target other reef fish species, (e.g., slow trolling for mackerel or fishing for greater amberjack using live bait), when red snapper were rarely caught. Many charterboats also have superior bottom sonar that allows captains to target more dense reef fish aggregations in the water column. In contrast, private fishermen are more likely to visit a limited number of heavily-fished nearshore public artificial reefs regardless of whether they intend to retain fish, perhaps affording a greater opportunity to release fish using descender devices. There might also be more time available to use a descender device to release fish on smaller private boats with fewer passengers and lower catch rates.

108

The west coast rockfish fishery includes a large recreational component conducted largely from for-hire vessels in deep water and is characterized by high discards and discard mortality. Extensive outreach by NOAA and state fisheries management agencies, and most importantly, large-scale voluntary support from angler groups and sportfishing organizations, has increased awareness of barotrauma and descender device use while discouraging the practice of venting fish (Benaka et al. 2014; California Sea Grant et al. 2014; Sportfishing Association of

California 2019). A recent study of California rockfish fishers suggests 42% use descender devices (Bellquist et al. 2019), which is certainly higher than estimated descender device use rates in the northern GOM (10.3% in the current study and 12–13% of Florida recreational reef fish fishers, Crandall et al. 2018). Using west coast rockfish as an example, fisheries managers and angler organizations in the GOM could address the apparent lack of familiarity with descender devices and increase post-release survival of GOM reef fish suffering from barotrauma. Such efforts are already underway to introduce descender devices to fishers and help them feel comfortable using them (Florida Sea Grant and University of Florida 2017; Gulf of

Mexico Fishery Management Council 2019a). My observations suggest captains and crewmembers, not customers, typically direct fishing operations on for-hire fishing trips and decide when to release fish, thus efforts to increase descender device use in the for-hire sector should be targeted towards this influential group.

Finally, climbing descender device use rates among west coast rockfish fishers may also be partially due to outreach efforts that associate angler behavior (e.g., using descender devices to return rockfish) to favorable outcomes (e.g., higher catch limits for some species). In 2013, the

Pacific Fishery Management Council Groundfish Management Team approved using decreased discard mortality rates in rockfish stock assessment models as a result of descender device use

109

for recreational catch accounting of cowcod, canary, and yelloweye rockfish (Pacific Fishery

Management Council 2013). Recreational discard mortality rates of red snapper in the GOM have been adjusted down to account for fisher behavior in the past (SEDAR 2018) and have had an effect on stock-wide total allowable catch, however, as discussed in Chapter 5, dead discards are not attributed to specific fishery sectors, likely reducing incentives for recreational fishers to take action to reduce discard mortality.

110

Table 4-1. Reaction statements presented to charterboat customers who were requested to respond on a scale of ranging from 1 (strongly disagree) to 5 (strongly agree). 1) It was easy to employ the descender device to return fish to depth. 2) Using the descender device enhanced my fishing experience today. 3) Knowing a future fishing trip would utilize a descender device would make me less likely to participate. 4) Practicing conservation is as important to me as having a successful fishing trip. 5) Employing the descender device likely enhanced survival of released fish. 6) The most important feature of a successful trip is number of fish caught. 7) I am an experienced angler. 8) I would support a regulation requiring the use of descender devices. 9) I have used descender devices on other fishing trips before today. 10) The additional time required to return fish to depth placed an undue burden on my fishing party. 11) Using descender devices should be voluntary not required. 12) I would be more likely to use a descender device if the additional conservation meant a longer red snapper season.

Figure 4-1. Map showing observed charterboat fishing sites (n = 287) in the northern GOM between June 2016 and September 2017. Depth contours are in meters.

111

Figure 4-2. Total observed effort (angler × hours) by target and open/closed red snapper season.

112

Table 4-2. Catch and disposition of fish by red snapper open vs. closed season. Open season Closed season Common name Scientific name Catch (n) Discard (%) Catch (n) Discard (%) red snapper Lutjanus campechanus 803 51.1 585 100 gray triggerfish Balistes capriscus 100 100 256 92.6 vermilion Rhomboplites 65 18.5 368 20.4 snapper aurorubens little tunny Euthynnus alletteratus 17 5.9 13 0 cobia Rachycentron canadum 16 100 5 80 sharksucker Echeneis naucrates 15 100 16 100 red porgy Pagrus pagrus 13 38.5 238 0 greater Seriola dumerili 11 100 25 100 amberjack king mackerel Scomberomorus 7 14.3 10 20 cavalla tomtate Haemulon 6 0 107 14.0 aurolineatum almaco jack Seriola rivoliana 5 0 5 20 scamp Mycteroperca phenax 5 40 8 0 Spanish Scomberomorus 4 25 6 33.3 mackerel maculatus gray snapper Lutjanus griseus 3 0 4 0 gag Mycteroperca 2 0 0 0 microlepis spotted moray Gymnothorax moringa 2 100 0 0 banded Seriola zonata 1 0 1 0 rudderfish blue runner Caranx crysos 1 0 5 0 bank sea bass Centropristis ocyurus 1 0 5 40 lane snapper Lutjanus synagris 1 0 3 0 rock hind Epinephelus 1 0 0 0 adscensionis short bigeye Pristigenys alta 1 0 0 0 sand perch Diplectrum formosum 1 0 9 55.6 wahoo Acanthocybium 1 0 0 0 solandri blackline Caulolatilus cyanops 0 0 6 0 tilefish tattler Serranus pheobe 0 0 6 0 tilefish Lopholatilus 0 0 4 0 chaemaeleonticeps creolefish Paranthias furcifer 0 0 3 0 dolphin Coryphaena hippurus 0 0 3 0

113

Table 4-2. Continued Season open Season closed Discarded Discarded Common name Scientific name Catch (n) (%) Catch (n) (%) great barracuda Sphyraena barracuda 0 0 2 0 rock sea bass Centropristis 0 0 2 50 philadelphica sharks Charcharhinus spp. 0 0 3 100 spotted soapfish Rypticus subbifrenatus 0 0 2 100 blueline tilefish Caulolatilus microps 0 0 1 0 crevalle jack Caranx hippos 0 0 1 100 jolthead porgy Calamus bajonado 0 0 1 0 knobbed porgy Calamus nodosus 0 0 1 0 red grouper Epinephelus morio 0 0 1 0 skilletfish Gobiesox strumosus 0 0 1 100 Spanish flag Gonioplectrus hispanus 0 0 1 0 whitebone Calamus leucosteus 0 0 1 0 porgy

Figure 4-3. Observed catch of red snapper by length (mm total length, TL), disposition, season, and fishing target.

114

Table 4-3. Catch rates in number per angler × hour by target species or group. Target species or group Vermilion Red snapper / Gray Overall Catch species snapper red porgy Groupers triggerfish Jacks non-target Red snapper 4.17 0.73 1.48 3.49 0 1.19 Vermilion snapper 0.2 3.27 0 0.71 0 0.22 Red porgy 0.01 2.19 0 0.29 0 0.03 Groupers 0.01 0.08 0.27 0 0 0.03 Gray triggerfish 0.48 1.57 0.16 1.54 0 0.74 Greater amberjack 0.05 0.08 0.11 0 1.08 0.06

Table 4-4. Species and groups with the highest prevalence of barotrauma signs. Proportion w/ signs Species or group Observed catch (n) of barotrauma Tilefishes 11 0.273 Groupers 17 0.235 Sea bassesa 28 0.179 Greater amberjack 36 0.167 Red snapper 1,382 0.159 Gray triggerfish 351 0.086 Other species 955 0.009 a Sea basses include smaller serranids, such as sand perch, bank sea bass, and tattler, that are not targeted, thus are caught incidentally.

115

Figure 4-4. Condition of surface-released fish by species. Fish that were released at the surface but were not observed are excluded (n = 39 red snapper, n = 3 greater amberjack, n = 9 gray triggerfish, n = 6 vermilion snapper, and n = 10 other species).

A B

C D

Figure 4-5. Digital images from GoPro3 camera footage of descender device deployments. A) a 762 mm fork length (FL) greater amberjack being descended after the fish failed to submerge following a surface release B) a 435 mm TL red snapper swims away from the descender device after release, C) as the descender device is retrieved, a large shark investigates the descender device, and D) the same large shark returns and approaches the camera. Photos courtesy of author.

116

Figure 4-6. Surveyed charterboat customers' responses to questions from the survey. SD = "strongly disagree", D = "disagree", N = "no opinion", A = "agree", SA = "strongly agree".

117

Figure 4-6. Continued

118

Table 4-5. Survey responses by group. Questions are listed in Table 4-1. Median Median Median response response response Descender used Mann- Red snapper Mann- Kruskal- during trip Whitney season Whitney Boat Wallis Question Yes No p-value Open Closed p-value A B C D p-value 1 4 3 <0.01 4 4 0.89 3 4 4 3 <0.01 2 4 3 <0.01 3 3 0.27 3 3 4 3 0.48 3 1 3 <0.01 2 2 0.99 2.5 2 1 2 <0.01 4 5 4 0.41 5 5 0.01 5 4 5 5 0.03 5 5 3 <0.01 4 4 0.01 4 4 5 4 <0.01 6 2 2 0.94 2 2 0.87 2 4 2 2 <0.01 7 3 3 0.32 3 3 0.08 2 4 4 3 <0.01 8 4 3 0.04 4 4 0.02 3.5 3 4 4 0.67 9 1 3 <0.01 2 2 0.77 2 3 1 2 <0.01 10 1 2 <0.01 2 2 0.35 1.5 2 1 2 0.07 11 3 3 0.66 3 3 0.01 3 3 3 3 <0.01 12 4 4 <0.01 4 4 0.10 3.5 4 5 4 <0.01

119

CHAPTER 5 HARVEST SLOTS AS A MANAGEMENT TOOL TO MAXIMIZE MARINE RECREATIONAL FISHING OPPORTUNITIES FOR GULF OF MEXICO RED SNAPPER

Recreational fishing accounts for significant harvest and removals from many marine fish stocks (Coleman et al. 2004b; Cooke and Cowx 2004; Radford et al. 2018). Recreational fishery managers often use a combination of minimum fish length limits, closed seasons, gear restrictions, and bag (possession) limits in order to constrain harvests and limit the impact of recreational fisheries on stocks. Maximum fish length limits (or harvest slots when implemented together with minimum length limits) discourage fishers from harvesting larger fish, thus preserving older spawners while allowing recreational fishers to harvest smaller, younger individuals. In many fish species, larger, older females expend disproportionately more energy towards reproduction than younger fish, producing more larger or higher quality eggs (Hixon et al. 2014; Barneche et al. 2018). In species that spawn multiple times per year, older, larger female fish may spawn more times, more frequently, and over a longer spawning season, hence are more likely to have at least a few successful spawns in unfavorable conditions and many successful spawns in favorable conditions that produce the occasional large year classes on which populations of many marine fish rely. In this way, older, larger fish would be valuable for increasing the resiliency of the population to exploitation, habitat degradation, and environmental perturbation (Berkeley et al. 2004; Hixon et al. 2014; Lowerre-Barbieri et al.

2015). In North American recreational fisheries, harvest slots have most commonly been employed in freshwater bass (Micropterus spp.) fisheries to decrease overall harvest rates and increase opportunities for anglers to catch large fish (Dotson et al. 2013; Long et al. 2015).

Marine examples of harvest slot regulations are mostly limited to shallow water inshore fisheries where discard mortality is low, including several stocks of drum (e.g., red drum, Sciaenops

120

ocellatus, and spotted seatrout, Cynoscion nebulosus) in the U.S. Gulf of Mexico (GOM), and striped sea bass (Morone saxatilis) in the U.S. mid-Atlantic states.

Marine fish stocks can support a diverse range of stakeholders who may have competing interests, posing a challenge to managers who must prioritize management objectives while maintaining equity across user groups. Although maximum sustainable yield and economic efficiency of harvest fleets may be an effective management objective for exclusively extractive fisheries (e.g., commercial fisheries), many marine fisheries include a growing recreational component where fishers may be willing to forego some harvest in favor of non-consumptive benefits such as higher catch rates, extended or more flexible fishing seasons, or the opportunity to catch very large fish (Cooke and Cowx 2006). Even among recreational users, the priorities of individuals may be variable and complex, ranging from fishers who have a strong personal normative towards releasing fish to those who always harvest fish a food source and are functioning as subsistence fishers (Anderson et al. 2007; Stensland and Aas 2014). Trends towards ecosystem-based fisheries management and awareness of a changing climate are further expanding management objectives to include a wider range of ecosystem services (Lynch et al.

2018).

Northern red snapper (Lutjanus campechanus) support a diverse and valuable fishery in the U.S. GOM. Annual harvest of red snapper is allocated roughly equally between commercial fishers (which have been managed under an individual fishing quota program since 2007) and recreational fishers. The recreational fishery has been managed primarily with a minimum fish length regulation, closed seasons, and bag limits, and is largely open-access (with the exception of for-hire recreational fishing vessels operating in the offshore waters under Federal jurisdiction). Maximum length regulations have never been used in the management of the GOM

121

red snapper recreational fishery, mostly due to concerns that high discard mortality would eliminate any benefits of discarding larger fish. In addition, quantitative analyses of the potential effects of harvest slot regulations on future stock population dynamics and fish catches have not been performed within the peer-reviewed red snapper stock assessment model, leaving large uncertainty in how the stock would be affected by such regulations. Emerging technologies and best practices such as rapid recompression of fish by releasing them with descender devices may effectively reduce discard mortality (Chapter 2; Curtis et al. 2015; Bohaboy et al. 2019) and recent Gulf of Mexico Fishery Management Council (GMFMC) outreach and restoration initiatives have begun to encourage recreational fishers to minimize discard mortality (Gulf of

Mexico Fishery Management Council 2019a).

The objectives of this study are: 1) using GOM red snapper as an example, demonstrate the trade-offs between outcomes of prospective management regulations in a large-scale fishery comprised of diverse stakeholders; and 2) predict the results of recreational harvest slot regulations, informed by recent research regarding red snapper discard mortality and discard mortality reduction, on the GOM red snapper stock and catches by the fishery. I performed the analyses using the integrated age-structured Stock Synthesis (SS) population dynamics model from the most recent stock assessment of GOM red snapper. This relatively complex population model has undergone extensive public review and has been accepted by U.S. fisheries managers as the best science available for the management of GOM red snapper. The transparency of the process through which the model was developed, combined with the high public availability of the model and supporting research, will facilitate the adaptation of these methods and findings to other fish stocks.

122

Materials and Methods

I simulated the population dynamics of GOM red snapper for years 2017–2076 under a range of recreational length (harvest slot) regulations and assumptions regarding historic recreational discard mortality and potential reductions in future discard mortality. I modified the population model from the most recent Southeast Data, Assessment, and Review (SEDAR) of

GOM red snapper as the basis for the simulation analyses. SEDAR is a cooperative process managed by the three Southeastern U.S. Fishery Management Councils (Caribbean, Gulf of

Mexico, and South Atlantic) and includes close coordination with NOAA Fisheries and the regional Fisheries Commissions (Atlantic and Gulf States). The SEDAR process was developed to ensure that the highest level of scientific quality and peer-review of stock assessments (i.e., input catch and survey data, biological parameters, and population dynamics model construction and performance diagnostics) are used for the management of U.S. fishery resources. SEDAR completed a comprehensive stock assessment of GOM red snapper in June 2013 (SEDAR 2013), which established all data inputs, biological information, and the population dynamics model that was accepted by the GMFMC. The population dynamics model was constructed in the integrated modeling environment Stock Synthesis (SS), a free and open-source software used for many stock assessments in the U.S. and has been extensively peer-reviewed (Methot and Wetzel 2013; software, documentation, and user support are available at https://vlab.ncep.noaa.gov/group/stock-synthesis/home). SS is a statistical catch-at-age modeling environment that incorporates multiple data and error sources for catch and survey data (e.g., indices of abundance, catch, length, and age), biological information (e.g., growth, reproduction, and stock-recruitment), fishery characteristics (length- or age- based selectivity, seasonality, or spatial structure), mark-recapture data, and environmental variability. The GOM red snapper

123

population dynamics model was updated and peer-reviewed during the most recent stock assessment released in 2017 (SEDAR 2018; henceforth "SEDAR 52").

SEDAR 52 was constructed in SS version 3.24 and is described in detail in SEDAR

(2013) and (2018). A brief description of the SEDAR 52 model with special emphasis on specifications unique to GOM red snapper is included here. The age-structured population model begins in 1872 with the assumption of a previously fished stock and includes ages 0 through the plus group of 20+ years. Growth, maturity, age-specific fecundity, and age-specific natural mortality are fixed time-invariable inputs in the model. The number of age-0 fish produced by the spawning stock in year y (Ry) follows a Beverton-Holt function with steepness (h) = 0.99 and is allowed to vary from the base Beverton-Holt relationship via lognormally distributed annual

2 recruitment deviations (푅̃푦), a deviation magnitude parameter (휎푅 = 0.3), and annual bias adjustment factors (By) to account for bias due to the estimation of recruitment variability in fishery assessment models (Methot and Taylor 2011; Methot and Wetzel 2013; equation 5-1).

Unfished recruitment (R0 which gives rise to the corresponding unfished spawning stock biomass, S0) is an estimated parameter in the model and is time-varying to allow for an apparent

offset shift in productivity that occurred in 1984 (R0,1872–1983 = e · R0,1984–2016, where offset is estimated). Stock-wide annual recruits (Ry) are distributed between the east (Ry,east) and west

(Ry,west) GOM subareas (which are divided roughly by the Mississippi River at -89° E longitude) as a function of the model-estimated parameter poverall,east and a vector of lognormally distributed

2 annual deviations 푝̃푦,푒푎푠푡 (where σ푝 was fixed in SEDAR 52; equations 5-2–5-4). Following recruitment, population dynamics are independent between the two subareas.

4ℎ푅0푆푦−1 2 푅̃푦−0.5퐵푦σ푅 2 푅푦 = 푒 푅̃푦 ~푁(0; σ푅) (5-1) 푆0(1−ℎ)+푆푦−1(5ℎ−1)

124

푝̃푦,푒푎푠푡 2 푝푦,푒푎푠푡 = 푝표푣푒푟푎푙푙,푒푎푠푡푒 푝̃푦,푒푎푠푡 ~푁(0; σ푝) (5-2)

푒푝푦,푒푎푠푡 푅 = 푅 pwest = 0 (5-3) 푦,푒푎푠푡 푦 푒푝푦,푒푎푠푡+푒푝푤푒푠푡

푒푝푤푒푠푡 푅 = 푅 pwest = 0 (5-4) 푦,푤푒푠푡 푦 푒푝푦,푒푎푠푡+푒푝푤푒푠푡

SEDAR 52 incorporates data from 6 fishery-independent survey indices (summer bottom trawl, fall bottom trawl, video, ichthyoplankton, bottom longline, and remote operated vehicle/ROV surveys), 2 fishery-dependent survey indices (private/charterboat and headboat recreational fleets), 2 directed recreational fishing fleets (open season private/charterboat and headboat), 1 bycatch recreational fishing fleet (closed season combined for private and for-hire),

2 directed commercial fishing fleets (longline, handline), and 2 bycatch commercial fishing fleets

(closed season longline/handline and shrimp trawl bycatch). Fishery independent surveys and fishing fleets are both referred to as fleets in SS, while most have separate data streams for the east and west subareas (i.e., there are 29 fleets in the SEDAR 52 population model). Catch age composition is available for the latter years of most fleets.

The relative vulnerability of fish to catch (selectivity, Selex), harvest (retention, Ret), and discard (Disc) are functions of age or length, are specific to each fleet, are time-varying to allow for changes in fishing regulations, and can follow several potential forms. Selex is influenced mostly by gear type, as well as where and how fishers target fish. For example, if fish are targeted throughout their range using gears that catch all fish above a given length, selectivity at length (SelexL) could be described using a simple logistic function, which increases to 1 at the length where fish are susceptible to the fishing gear and remains at 1 for all larger fish (equation

5-5; Fig. 5-1A). Ret is largely a function of fishers' choice to either harvest or release fish once they are caught and can be heavily influenced by length regulations. For example, if there is a minimum length regulation below which fishers are prohibited from keeping a fish, as well as a

125

maximum length regulation above which fishers are prohibited from keeping a fish (i.e., a harvest slot), then retention could be described using a double-logistic function, which increases above the minimum length regulation and decreases above the maximum length regulation

(equation 5-6; Fig. 5-1B).

푃3 푆푒푙푒푥퐿 = (5-5) 1 + 푒−푃2(퐿−푃1)

푃 1 푅푒푡 = 3 × (1 − ) 퐿 푃5−퐿 (5-6) 푃1 −퐿 1 + 푒 푃6 1 + 푒 푃2

In equations 5-5 and 5-6, P1 is length at the inflection of the ascending portion of SelexL or RetL (the length at 50% of asymptotic selectivity or retention), P2 influences the steepness of the ascending portion of SelexL or RetL (larger values produce a more gradual slope), and P3 is the retention or selectivity value at the asymptote (set equal to 1 for full retention or selectivity).

Parameters P5 and P6 describe the length at inflection and relative steepness (respectively) of the descending portion of SelexL or RetL. For clarity, P4 and P7 are omitted from equations 5-5 and 5-

6 because they do not apply to GOM red snapper where selectivity or retention of male and female fish are identical (see Methot et al. 2017 for the complete list of selectivity and retention functions available within SS). Discards at length (DiscL) are simply the inverse of RetL:

퐷𝑖푠푐퐿 = 1 − 푅푒푡퐿. (5-7)

The process model in SS generates the population numbers at age (푁푎푔푒), population numbers at length (푁퐿), and an age-length probability function to relate numbers at age to length

(푃퐿|푎푔푒). More detailed descriptions of all components within the SS process model can be found

푓푙푒푒푡 in Methot and Wetzel (2013). The number of fish caught (퐶푎푡푐ℎ퐿 ), harvested

푓푙푒푒푡 푓푙푒푒푡 (퐻푎푟푣푒푠푡퐿 ), and discarded (퐷𝑖푠푐푎푟푑푠퐿 ) at length by fleet are estimated in SS as:

126

푛 푓푙푒푒푡 푎푔푒 퐶푎푡푐ℎ퐿 = ∑ (푆푒푙푒푥푓푙푒푒푡 × 푃퐿|푎푔푒 × 푁퐿 × 퐹푓푙푒푒푡) (5-8) 푓푙푒푒푡=1

푛 푓푙푒푒푡 푓푙푒푒푡 푓푙푒푒푡 퐻푎푟푣푒푠푡퐿 = ∑ (퐶푎푡푐ℎ퐿 × 푅푒푡퐿 ) (5-9) 푓푙푒푒푡=1

푛 푓푙푒푒푡 푓푙푒푒푡 푓푙푒푒푡 퐷𝑖푠푐푎푟푑푒푑퐿 = ∑ (퐶푎푡푐ℎ퐿 × 퐷𝑖푠푐퐿 ) (5-10) 푓푙푒푒푡=1

푓푙푒푒푡 The number of discarded fish that die (dead discards) is the product of 퐷𝑖푠푐푎푟푑푠퐿 and the fleet specific discard mortality rate.

I obtained the SEDAR 52 SS version 3.24 input files, as well as the updated version 3.30 input files (D. Goethel, NOAA Fisheries, Miami, Florida, personal communication, 2018). I verified the SEDAR 52 SS version 3.30 population model was producing similar parameter estimates, modeled outputs (i.e., annual stock biomass, recruitment, and catches for each fleet), and displayed the same model performance as for the SS version 3.24 population model. I used

SS version 3.30 in the analyses because it includes the ability to specify double-logistic retention

(equation 5-6) whereas in version 3.24 retention was constrained to be simple logistic. In

2 addition, the change from SS version 3.24 to 3.30 included the added ability to estimate σ푝

(equation 5-2), which improved the GOM red snapper model fit and stability. I made several modifications to the SEDAR 52 SS version 3.30 population model: 1) converted RetL to double- logistic for the recreational fishing fleets but fixed P5 at a large value (>> theoretical maximum length of red snapper) to mimic simple logistic retention as in the SEDAR 52 SS version 3.24 model; 2) added a new parameter set for RetL for the recreational fleets to allow for independence of parameter values in the terminal model year (2016) from previous model years;

127

and 3) re-estimated all parameters for the base years (1872–2016) of the assessment model for 4 recreational discard mortality rates. I explored a range of recreational discard mortality rates

(25%, 50%, and 75%) within the base years of the model in addition to the recreational discard mortality rates assumed in SEDAR 52 (22% in the east subarea and 21% in the west subarea from 1981–2007 and 11.8% for both subareas from 2008–2016) because recent research suggests discard mortality rates of GOM red snapper and other reef fish may be historically underestimated and higher than assumed in SEDAR 52 (Chapter 2; Bohaboy et al. 2019; Runde et al. 2019).

I projected the GOM red snapper population for a range of prospective harvest slots by modifying the P1 and P5 parameters of RetL (equation 5-6) for 2016–2076: minimum length = 16 or 18 in. (40.64 and 45.72 cm) total length (TL) and maximum length = 22, 24, 26, 28, 30, 32, 34 in. (55.88–86.36 cm) TL, or no maximum length. For each harvest slot, I modified the value of the discard mortality rate for each recreational fleet assuming relative reductions beginning in

2016 equal to 0% (no change), 25%, 50%, 75%, and 100% (0% discard mortality or 100% survival of released fish). The following assumptions regarding the forecast period (2017–2076) were made: 1) annual stock-wide recruitment was equal to the average of the recent time period

(R0,1984–2016); 2) recruitment was distributed between east and west subareas as the average of

1984−2016; 3) overall annual fishing mortality (F) was set to achieve 26% spawning potential ratio (SPR26%) in 2032 (consistent with the rebuilding plan established for GOM red snapper by the GMFMC); and 4) F was partitioned among all fleets according to the 2011–2015 average, with scaling applied to F for the targeted fleets in order to maintain the current GOM red snapper regulatory harvest weight allocations of 49% recreational and 51% commercial.

128

Model outputs are reported for year 2032 and include changes in the length of the recreational open fishing season, dead discarded biomass, catch (numbers and biomass), harvest

(numbers and biomass), average weight per fish in the catch or harvest, proportion of total dead discard biomass attributed to the recreational fishery, recreational catch per unit effort (number of fish, CPUE), and the age distribution of fish within the beginning of year population and recreational catch. Results for each prospective management scenario (Yi) are given as percent change relative to the forecasted 2032 value assuming current management regulations (16 in. minimum length with no maximum length) and recreational discard mortality rate are unchanged

(a.k.a. business as usual, YBAU: equation 5-11). Recreational open fishing season length (effort, E) was assumed proportional to total recreational open fishing season F (i.e. F = qE where catchability, q, is assumed constant).

푌 − 푌 𝑖 퐵퐴푈 × 100% (5-11) 푌퐵퐴푈

Results

Assuming recreational discard mortality rate was higher in the base years (1981–2016) of the population model reduced goodness of fit and was associated with a modeled increase in stock productivity (which was most apparent from the upward trend in the fitted value of the

GOM-wide unfished recruitment parameter, Ln(R0)). Modeled spawning stock biomass in 2016 decreased whereas the absolute spawning stock biomass at the rebuild target of SPR26% in 2032 increased with greater assumed discard mortality rate in the base years of the model. There was no change in general trends of the effects of future reductions in discard mortality rate or harvest slots across the range of assumed discard mortality rates in the base years. However, effects of relative changes in discard mortality rate were amplified with higher assumed base mortality

129

rate. Results presented here are for assumed recreational discard mortality rate = 25% in the base years of the population model.

Recreational open fishing season length increased with more narrow harvest slots, such that even without reducing discard mortality rate, adopting a very narrow harvest slot regulation

(16–22 in.) would increase the season length by 145% (Fig. 5-2A). Season length was also extended when discard mortality rate was reduced, though the effect was less pronounced than for harvest slots. Harvest slots increased the total weight of dead discards (Fig. 5-2B) while reducing recreational (Fig. 5-2C) and commercial (Fig. 5-2D) harvested biomass. However, by combining a wide harvest slot (e.g., 16–30 in.) with reduced discard mortality rate (50%) it was possible to reduce biomass of dead discards (-38%) while increasing recreational and commercial harvested biomass (2.5% for each) and recreational season length (22%). Harvest biomass allocations between the recreational and commercial sectors were held constant in the simulation modeling (hence changes in harvested biomass resulting from harvest slots and reduced discard mortality appear identical Figs. 5-2C and 5-2D). As a result, the total share of dead biomass taken by the recreational fishing fleets displays similar trends as for dead discard biomass: harvest slots without reductions in discard mortality increase the recreational share of dead biomass while reductions in discard mortality reduce the recreational share of dead biomass

(Fig. 5-2E). The recreational average weight per harvested fish decreased markedly for decreasing maximum harvest lengths, however the effect of reductions in discard mortality rate on average weight per harvested fish were negligible (Fig. 5-2F). CPUE (in numbers) of fish greater than 30 in. showed only modest changes in response to harvest slots and reduced discard mortality rate. The greatest gain in CPUE of fish larger than 30 in. occurred with no maximum size limit for reductions in discard mortality rate up to 50%; however, if discard mortality rate

130

were reduced by 75%, a wide harvest slot of 16 – 34 in. would result a greater gain in CPUE of fish larger than 30 in. (Fig. 5-2G). Harvest slot regulations and reduced discard mortality rates increased total recreational catch (both numbers and biomass) but overall increases in catch were outpaced by increases in effort, hence CPUE decreased for many management scenarios. For example, even with reductions in discard mortality rate of 50%, overall recreational CPUE decreased for all harvest slots with a lower length limit = 16 in. and an upper length limit ≤ 28 in.

(Fig. 5-2H). The general trends in model outputs were the same while the magnitude of change was generally magnified when minimum harvest length was 18 in. (Figs. 5-2I–P).

It is not possible to maximize all model outputs with any single combination of recreational harvest slot regulations, even under the unrealistic assumption of 100% reduction in recreational discard mortality rate. Regulations leading to a large increase in season length would also result in reduced harvest and total CPUE while greatly increasing the weight of dead discards (Fig. 5-3). Regulation scenarios with very large harvest slots (e.g., 16 in. minimum and

30 in. maximum length) would allow for a small increase in season length and CPUE of fish larger than 30 in., considerable reduction in dead discard biomass, and cause only minor changes in the other management outputs (harvest, total CPUE, and average size of harvested fish).

Model projections show harvest regulations decrease the population abundance of fish age 15 and younger while increasing the abundance of fish age 16 and older (Figs. 5-4A–D). In some instances, narrow harvest slots increased the abundance of very large, old fish in the plus group

(age-20+) greater than 8.0% relative to the BAU management scenario. However, increases of older fish in the population were not reflected strongly in total recreational CPUE at age (Figs. 5-

4E–H). Even though catch at age increased for all management scenarios in Fig. 5-4, increases in effort outpaced the changes such that CPUE decreased for fish ages 2–18.

131

Discussion

My analyses demonstrate it is not possible to simultaneously maximize all management objectives for GOM red snapper through recreational harvest slot regulations, but some harvest slots can satisfy several objectives while detracting only slightly from others. Relatively subtle management changes such as implementing very wide harvest slots combined with modest reductions in discard mortality rate would be provide some benefits while posing only a small change from current fisher behavior (from 2014–2018 only 4.3% of red snapper harvested by private and charterboat recreational fishers were over 32 in. TL; NOAA Office of Science and

Technology Recreational Fisheries Statistics Query [online database], accessed 6 November

2019). The danger of considering a narrow set of management objectives when evaluating the effects of management actions is quite clear: maximizing season length with a narrow slot regulation (especially absent significant reductions in discard mortality rate) would result in undesirable outcomes, some quite severe, in most management objectives including increases in dead discards, fishery waste, and lost harvest. In general, reducing discard mortality without implementing harvest slots would be beneficial to all the management objectives examined. The efficacy of harvest slots in meeting management objectives pivots on the ability of recreational fishers to reduce discard mortality, suggesting recent efforts to encourage recreational fishers to minimize discard mortality (Gulf of Mexico Fishery Management Council 2019a) may contribute to the success of harvest slot regulations should they be implemented in the future.

Increased abundance of older ages in a fished population is the basic predicted result of harvest slots (Berkeley et al. 2004). I projected a large increase in red snapper ages 16 and older in most harvest slot management scenarios, however, the benefits of these added older females are likely under-accounted for in my analysis for several reasons. First, the spawner-recruit relationship has essentially been decoupled in the SEDAR 52 population model (through h =

132

0.99 and 푅̃푦, equation 5-1) such that a larger spawning stock does not result in a greater number of recruits (in forecast model years, recruitment is constant). These added recruits would be expected to contribute to high catch rates and harvests (e.g., age 3–4 fish that are captured by recreational fishers and are larger than the 16 in. minimum length regulation, assuming fishers chose to harvest smaller fish). However, the benefits of a larger spawning stock do manifest through regulations, as stock rebuilding, overfishing limits, and hence harvest quotas are based on the stock SPR. Second, GOM red snapper populations are characterized by occasional very strong year classes which, as with many marine fish species that are periodic life history strategists, may be vital to population persistence in variable environments (Winemiller and Rose

1992; Cowan et al. 2011). It is possible the size of the GOM red snapper spawning stock is less important than the ability of the spawning stock to take advantage of spatially or temporally sporadic favorable conditions to maintain long-term reproductive success, referred to as

"reproductive resilience" by Lowerre-Barbieri et al. (2015). Reproductive resilience is heightened by greater numbers of older fish in the population, which spawn more times, more frequently, over a longer spawning season, and over a larger area. Managing fishery resources for resilience to climate change, habitat alteration, and environmental variation is prescribed in

U.S. fisheries management legislature and numerous programmatic directives (50 CFR Chapter

VI). These analyses suggest harvest slots may be an effective tool to help meet these objectives.

Although the benefit of extended harvest seasons would be felt strongly among recreational fishers, harvest slots and increased dead discards would have deleterious effects on the commercial sector of the GOM red snapper fishery. Currently, GOM red snapper are managed by the GMFMC with a GOM-wide harvest quota (annual catch limit; ACL) derived from the overfishing limit (OFL) as reduced by scientific uncertainty. The harvest allocation is

133

currently 49% to the recreational sector and 51% to the commercial sector. Stock removals resulting from discards are not assigned to either the commercial or recreational sector. If harvest slots are implemented in the recreational fishery, the number of discarded fish would increase, and any discard mortality would result in wasted biomass, which would reduce the biomass available GOM-wide for harvest, with 51% of such a reduction being borne by the commercial sector. Similarly, the motivation for recreational fishers to reduce discard mortality rates or implement wide harvest slots would be reduced as the benefits of those actions to the recreational sector would be diluted given they are shared with the commercial sector.

In this analysis, I applied harvest slots and reduced discard mortality rates to all GOM recreational fleets equally. However, for 2018 and 2019, NOAA Fisheries issued exempted fishing permits allowing management authority over the private angling component to each of the 5 Gulf States for red snapper landed from State and Federal waters in that state. The

GMFMC is still responsible for setting the GOM-wide red snapper ACL, which is then allocated between the commercial (51%) and recreational (49%) sectors. Following Amendment 40 to the

GOM reef fish management plan, since 2015 the recreational harvest quota has been allocated between the Federal for-hire sector (42.3%) and the private recreational sector (57.7%). With the recent State authority over the management of the private recreational sector, the private recreational quota is further divided among Gulf states: Florida (44.8%), Alabama (26.3%),

Mississippi (3.6%), Louisiana (19.1%), and Texas (6.2%; Gulf of Mexico Fishery Management

Council 2019b). The result being that GOM red snapper now has 7 distinct management units, which do not correspond to fishing fleets in the stock assessment, greatly complicating analyses of state-level management actions. With 6 management units in the recreational fishery, combined with the continued practice where wasted harvest from discards of any single unit are

134

shared among all others, the externality apparent between the commercial and recreational fleets is now amplified within the recreational sector. For instance, if Florida adopted harvest slots and decreased discard mortality, positive consequences (such as increased season length) and negative consequences (waste) would be shared among the Federal for-hire fishery and the other states as the GOM-wide OFL would be adjusted before the ACL was allocated among management units.

SS provides a good model application of harvest slots in a marine fish stock. Although fisheries population modelers must balance the dangers of over-simplifying models with over- parameterizing models and making them too complex, SS provides the flexibility to accommodate a range of data availability, species biology, and fishery characteristics of marine fish stocks. Many U.S. stock assessments rely on SS models and the growing user community has embraced the continued improvements to the program, including the migration to the newest version 3.30. To my knowledge, this investigation represents one of the first demonstrations of the newly added double-logistic retention function in SS version 3.30 that enables investigations of harvest slots. I expect the overall conclusions of this chapter will apply directly to other fish stocks: the main benefits of harvest slot regulations (reduction in dead discard biomass, increased catch rates, and increased abundance of older fish) will only occur if discard mortality rate is simultaneously reduced. Stock-specific analyses, including identifying the level of discard mortality reduction required to achieve these benefits, can be repeated using these same methods for other fish stocks that have working SS models.

135

A B

Figure 5-1. Example length-based selectivity and retention functions. A) Simple logistic selectivity at length assuming fish are 50% selected at 25 cm (P1=25), and B) simple logistic retention at length depicting current recreational regulations of a 16 in. (40.64 cm) minimum length and no maximum length (solid line) and double-logistic retention at length reflecting an 18–26 in. (45.72–66.04 cm) harvest slot (broken line).

136

Figure 5-2. Effects of harvest slots and reduction of future discard mortality rate on model outputs relative to current regulations and discard mortality rate. For 16 in. (40.6 cm) minimum length: A) Recreational open fishing season length, B) total weight of dead discards, C) recreational harvest weight, D) commercial harvest weight, E) recreational share of dead biomass, F) recreational average weight per harvested fish, G) recreational CPUE (in numbers) of fish greater than 30 in., and H) total recreational CPUE (in numbers). Panels I–P are for the same model outputs as for A– H with 18 in. (45.7 cm) minimum length.

137

Figure 5-3. Relative change in model outputs for five prospective management scenarios and two discard mortality rate reduction levels: A) 25% discard mortality rate reduction and B) 50% discard mortality rate reduction for minimum length = 16 in. (40.6 cm) and either no maximum length or maximum lengths between 24 and 32 in. (61.0 and 81.3 cm); C) 25% discard mortality rate reduction and D) 50% discard mortality rate reduction with minimum length = 18 in. (45.7 cm) and either no maximum length or maximum lengths between 24 and 32 in. (61.0 and 81.3 cm). BAU = business as usual (current regulations and discard mortality rate are unchanged). Note the scales for each variable axis within a panel are different but are identical for each variable between panels.

138

Figure 5-4. Relative change in abundance of fish by age in the beginning of year population (A– D) and recreational catch per unit effort (CPUE, E–H) for several prospective management scenarios. Change is relative to model predicted population and CPUE under the current regulations and discard mortality rate (business as usual, BAU). 16– none = 16 in. (40.6 cm) minimum length and no maximum length, 16–30 = 16 in. (40.6 cm) minimum length and 30 in. (76.2 cm) maximum length, 18–26 = 18 in. (45.7 cm) minimum length and 26 in. (66.0 cm) maximum length, and 16–24 = 16 in. (40.6 cm) minimum length and 24 in. (61.0 cm) maximum length. Each management scenario assumes 25% reduction in future discard mortality rate.

139

CHAPTER 6 CONCLUSIONS

The goals of this dissertation research were to address vital data needs put forth by Gulf of Mexico (GOM) fisheries managers. These goals include: further understanding red snapper

(Lutjanus campechanus) discard mortality and the factors that affect it; determining whether descender devices are effective tools to reduce discard mortality in red snapper; investigating the practicality of using descender devices and the willingness of fishers to use descender devices aboard charter fishing vessels; and quantifying how reductions in discard mortality, combined with modified harvest size regulations, would affect the GOM red snapper stock and fishery. I used highly accurate three-dimensional positioning acoustic telemetry to monitor the movements and fates of fish in the natural environment following capture, tagging, and release in the northern GOM. The spatial extent of the acoustic telemetry array allowed me to estimate discard mortality of red snapper more completely than previous studies, largely by fully accounting for mortality due to predation, which was the primary driver of discard mortality of tagged fish. The fine-scale positioning acoustic telemetry approach provided the unique opportunity to investigate additional questions regarding red snapper movement, behavior, and habitat selection, as well as discard mortality of gray triggerfish (Balistes capriscus), which has been largely overlooked in previous studies of the recreational GOM reef fish fishery. Findings will provide guidance to

GOM fisheries managers who aim to optimize benefits gained from the GOM reef fish fisheries, including maximizing recreational fishing opportunities. In addition, the successful application of a large positioning acoustic telemetry array, the findings regarding the importance of predation in discard mortality, and the novel insight regarding the fine-scale movements of red snapper will provide guidance and motivation for future investigations of reef fish ecology in the northern GOM.

140

Discard Mortality

Estimates of discard mortality of red snapper (56.6% over all tagging events) and gray triggerfish (40% over all tagging events) were within the range of, but higher than most, estimates reported in most previous studies, and certainly higher than the values assumed in the most recent stock assessments for each species (11.8% for red snapper, 5% for gray triggerfish;

SEDAR 2015a, 2018). The primary reason for the discrepancy, at least for red snapper, is likely the ability to discern predation of tagged fish; whereas, approaches taken in previous studies either precluded predators as a source of mortality (laboratory and enclosure studies) or censored fish that may have been consumed by predators that quickly left detection ranges of deployed receivers (limited coverage of individual receivers or small, <1 km2 arrays). For example, if I excluded predation as a source of mortality and re-estimated discard mortality from the data (by censoring all tagged fish identified as predation mortalities), estimated red snapper overall discard mortality in the 30-m array (regardless of release method or season) would drop from

36.8% (including predation) to 5.3% (excluding predation), corresponding to a reduction from approximately the 90th percentile to the 10th percentile of estimates from previous studies at depths from 25 to 35 m. There may have been a "tag effect" that interacted with predation, whereby red snapper bearing acoustic telemetry tags were more susceptible to predation mortality than red snapper that had simply been captured and released, but release mortality estimates in the current study were still high relative to other acoustic telemetry-based estimates reported in the literature. Furthermore, precautions were taken to minimize effects of acoustic tags on fish, mainly by using an external tag attachment to minimize handling time, observing tagged fish in captivity to ensure there were no deleterious effects of acoustic tags on fish behavior, and using the smallest acoustic tag practicable. However, it is possible the internal

141

power supply of the acoustic transmitter tags could stimulate the predatory response of large sharks with electrosensory capabilities.

Discard mortality of gray triggerfish has potentially been underestimated among the small number of previous studies. A recent study by Runde et al. (2019) of discard mortality of gray triggerfish in the southeastern U.S. recreational fishery is a notable exception. Using underwater tagging to control for the effects of barotrauma, the investigators estimated that discard mortality of gray triggerfish was 65–66% and suggested that previous studies relying on surface condition as a proxy for survival excluded delayed mortality and greatly underestimated discard mortality. My results are similar because long-term monitoring with acoustic telemetry accounted for sub-surface mortalities: even though all gray triggerfish that were released at the surface were able to re-submerge and none showed signs of barotrauma, overall discard mortality was 40%. Underestimated discard mortality combined with high discarding rates of gray triggerfish in the GOM may be causing removals for this heavily exploited and valuable stock that are unaccounted for in the stock assessment. My results, in agreement with Runde et al.

(2019), warrant further investigation of discard mortality of gray triggerfish and a reconsideration of assumed discard mortality rates used in the stock assessments.

Stock assessments for GOM reef fish often include sensitivity analyses to investigate uncertainty in parameters such as natural mortality or discard mortality on estimates of stock biomass or overfishing limits. The most recent stock assessment of GOM red snapper included a sensitivity analysis of discard mortality (for the recent time period beginning in 2008) equal to

15.8% while the base model included 10.0% discard mortality. When I investigated higher discard mortality rates (up to 75%) in the base years of the assessment in Chapter 5, I found current (terminal year, 2016) stock biomass was lower whereas absolute spawning stock biomass

142

at the rebuild target of SPR26% in 2032 was greater as compared to the base SEDAR 52 discard mortality rates. This apparent shift in productivity would be expected to have regulatory consequences for GOM red snapper, as a period of reduced fishery yields may be necessary to for the stock to increase to SPR26% in 2032. Although I did not investigate the implications of historically underestimated discard mortality rate in the base years of the GOM gray triggerfish assessment model, I expect the conclusions would be similar as for red snapper. The most recent stock assessment included a base assumption of 5% discard mortality with a sensitivity analysis of 10% discard mortality. My findings, together with those of Runde et al. (2019) suggest discard mortality values ranging to at least 40% should be investigated for GOM gray triggerfish, as the change in our perceptions of stock status and trajectory towards the rebuilding target would likely reduce allowable catch limits from this valuable stock.

Red Snapper Movement and Habitat Use

The image of red snapper as sedentary residents of northern GOM artificial reefs has been scrutinized over past decades, largely in the context of debates regarding the importance of artificial reefs on the population dynamics of this iconic species. Although conclusions regarding the population-level benefits of artificial reefs are outside the focus of this dissertation, findings suggest red snapper are highly mobile, and even over periods less than 1 year (which is short relative to the lifetime of an adult red snapper), these fish switch frequently between reefs and some individuals routinely use open habitat away from reefs. To my knowledge, this application of three-dimensional positioning acoustic telemetry was the first attempt to cover the movements and behavior of red snapper with high accuracy at artificial reefs and in open-bottom expanses spanning between reefs. The results of previous investigations suggest red snapper have seasonally and demographically variable diets, often consisting of sediment-dwelling and pelagic organisms found away from artificial reefs (McCawley and Cowan 2007; Tarnecki and Patterson

143

2015; Schwartzkopf et al. 2017; Dance et al. 2018). My observations of red snapper movements over open sand-bottom habitat between artificial reefs support the possibility that the events and behaviors having the most influence on red snapper ecology may be taking place in the vast spaces away from artificial reefs that have formerly been excluded from artificial reef-focused studies. Investigations combining diet or stable isotope analyses and acoustic telemetry (e.g., if acoustic transmitter tagged fish were fin-clipped before release and sacrificed for stomach content analyses upon recapture) could provide novel insight into how fine-scale habitat use and movements reflect the feeding ecology of red snapper on both short-term and life-long timescales.

Fisher Behavior and Descender Devices

Descender devices reduced discard mortality of red snapper by almost 50% relative to surface-released fish. Despite high catch and discard rates of red snapper, it appears that descender devices were rarely used on the vessels observed during this research. Only 21 fishers surveyed (10.3% of respondents) indicated they had ever used a descender before, which is comparable to the findings of Crandall et al. (2018) that 12–13% of Florida recreational fishers surveyed had used a descender device within the previous 12 months to release reef fish suffering from barotrauma. Annually, between 1.9 and 3.8 million red snapper are discarded by recreational fishers in the GOM, which, using the relatively low discard mortality rate of 10% assumed in the most recent stock assessment, still accounts for 190,000–380,000 wasted red snapper (SEDAR 2018). A widespread shift in behavior of GOM recreational fishers, including the acceptance and use of descender devices, could result in a significant reduction in wasted fish, potentially increasing the number of landed red snapper by up to 19% in some years. I observed that, at least in the charterboat fishing fleet, captains and crews were typically responsible for deciding to harvest red snapper and whether to release fish using a descender

144

device, suggesting success of any future regulations or recommendations regarding descender device use in the for-hire sector will depend less on fishers themselves and more on captains and crew.

Discrete choice surveys administered to a large number of recreational fishers, from both the private and for-hire sectors, would enable a greater understanding of fisher support for harvest slot regulations and descender device use (e.g., Murphy 2018; Bellquist et al. 2019). The

Likert-scale survey in this study established that most charterboat customers were open to catch- and-release fishing regulations and using descender devices, especially if doing so would result in a longer fishing season. Further investigation of fishers' willingness to adopt non-consumptive fishing regulations such as harvest slots should be undertaken with a larger number of GOM reef fish fishers.

Harvest Slots for GOM Red Snapper

Implementing very wide harvest slots combined with modest reductions in discard mortality rate would allow for a small increase in season length and catch per unit effort (CPUE) of fish larger than 30 in., considerable reduction in dead discard biomass, and only minor changes in the other management outputs (harvest, total CPUE, and average size of harvested fish). The associated increased abundance of fish age 16 and older in the population would likely contribute to the reproductive resiliency of the GOM red snapper stock, meeting additional management objectives such as anticipating climate variability and ecosystem factors. The benefits of these added older females are likely under-accounted for in my analysis because the resulting increased recruitment, which may be very large in some years, is not reflected in the spawner-recruit relationship or fecundity schedule used in the base population dynamics model.

This dissertation also highlights the influence of the current harvest allocation scheme on the likely results of prospective management actions. Any savings in reduced dead discards in the

145

recreational fishery (e.g., through using descender devices and reducing discard mortality) would be shared between the recreational and commercial sectors. In particular, increased dead discards would reduce harvests in the commercial sector of the GOM red snapper fishery under the current management paradigm where harvest, not removals or dead biomass is allocated between the two sectors. Similarly, the motivation for recreational fishers to reduce discard mortality rates or implement wide harvest slots would be reduced as the benefits of those actions to the recreational sector would be diluted given they are shared with the commercial sector.

Future analyses of the effects of modified discard mortality and harvest slot regulations should account for states-based management, where regulations (e.g., minimum length limits, harvest slots, and bag limits) will be applied independently among the 5 GOM states and the

Federal for-hire sector. As with many assessments of valuable fish stocks in the U.S., the GOM red snapper stock assessment is continually increasing in complexity to accommodate improvements in landings and discard estimation, further understanding of fish life history (e.g., natural mortality, fecundity, or individual growth), and advances in population modeling capabilities (e.g., long-term, seasonal, or spatial variability of assessment parameters). In particular, greatly increased estimates of historic recreational catches stemming from the incorporation of state surveys and the Federal Effort Survey (FES) will be incorporated in upcoming assessments of many GOM fish stocks.

Future Directions

Future investigations of reef fish discard mortality should be placed in the context of broader ecological questions. For example, given that predation accounted for most of the discard mortality observed in this study, inquiries may focus on whether interspecies dynamics and population trends of large sharks are possibly driving reef fish discard mortality. The importance of density-dependent factors to post-release survival of reef fishes may also add

146

complexity to efforts to quantify discard mortality. For example, fish occupying artificial reefs at high density might spend more time farther from the reef to obtain food or might travel long distances seeking reefs with a lower density of reef fish. In this case, fish suffering from the deleterious effects of capture and release may be more susceptible to mortality from predation.

Three-dimensional acoustic telemetry is an ideal tool to approach these challenging questions. Improved communication, collaboration, and data sharing through networks of acoustic telemetry researchers (e.g., iTag, FACT, and OTN) also should bring together researchers with species-specific expertise and interests, enabling investigations of simultaneous movements and interactions among multiple species. For example, the bull sharks observed moving through my experimental array (Chapter 2) were tagged by researchers who belong to the iTag network who were investigating their own questions regarding the biology of sharks off the east coast of Florida. For this dissertation research, I benefitted from fortuitously detecting these tagged sharks, but planned concurrent tagging of multiple species in a common area would yield more observations. In a similar vein, drawing on the skills and resources of physical and geological oceanographers would provide an expanded understanding of the spatially and temporally dynamic environment where fishes (and fishers) interact.

The multi-faceted approach of my dissertation research, including the focused studies on discarding and descender device use by fishers (Chapter 4) and implications of potential management actions within the stock assessment (Chapter 5), highlight the spreading understanding in fisheries research that fishers are an integral part of the ecosystem and must be included in ecological investigations. Even with regulations mandating the use of descender devices, the outcome in the GOM would likely be similar to some sections of the west coast rockfish fishery where legal enforcement of such regulations is negligible and success hinges on

147

the voluntary use of descender devices (Fisheries and Oceans Canada 2018). Engaging social scientists and human dimensions experts within the fisheries research community will improve survey study design and methods to understand fisher reactions and support for prospective regulations. Quantifying the likely effects of updated understanding of fishery processes (e.g., new estimates of discard mortality rates) and potential management measures may be best performed with stock assessment simulation studies or within the context of a full management strategy evaluation (MSE). Such analyses are not always accessible to investigators who specialize in individual-based studies of fish biology or ecology, providing another opportunity for collaboration, in this case between ecologist performing basic fisheries research and stock assessment analysts.

148

LIST OF REFERENCES

73 FR 5117. 2008. Final rule: reef fish fishery and shrimp fishery of the Gulf of Mexico; amendment 27/14.

Addis, D. T., W. F. Patterson III., and M. A. Dance. 2016. The potential for unreported artificial reefs to serve as refuges from fishing mortality for reef fishes. North American Journal of Fisheries Management 36(1):131–139.

Ahrens, R. N. M., C. J. Walters, and V. Christensen. 2012. Foraging arena theory. Fish and Fisheries 13:41–59.

Ajemian, M. J., J. J. Wetz, B. Shipley-Lozano, J. Dale Shively, and G. W. Stunz. 2015. An analysis of artificial reef fish community structure along the northwestern gulf of Mexico shelf: potential impacts of “rigs-to-reefs” programs. PLoS ONE 10(5):1–23.

Anderson, D. K., R. B. Ditton, and K. M. Hunt. 2007. Measuring angler attitudes toward catch- related aspects of fishing. Human Dimensions of Wildlife 12:181–191.

Arlinghaus, R., S. J. Cooke, J. Lyman, D. Policansky, A. Schwab, C. Suski, S. G. Sutton, and E. B. Thorstad. 2007. Understanding the complexity of catch-and-release in recreational fishing: an integrative synthesis of global knowledge from historical, ethical, social, and biological perspectives. Reviews in Fisheries Science 15:75–167.

Bacheler, N. M., J. A. Buckel, J. E. Hightower, L. M. Paramore, and K. H. Pollock. 2009a. A combined telemetry – tag return approach to estimate fishing and natural mortality rates of an estuarine fish. Canadian Journal of Fisheries and Aquatic Sciences 66:1230–1244.

Bacheler, N. M., L. M. Paramore, S. M. Burdick, J. A. Buckel, and J. E. Hightower. 2009b. Seasonal variation in age-specific movement patterns of red drum Sciaenops ocellatus inferred from conventional tagging and telemetry. Fishery Bulletin 107(4):405–419.

Barneche, D. R., R. D. Robertson, C. R. White, and D. J. Marshall. 2018. Fish reproductive- energy output increases disproportionately with body size. Science 360:642–645.

Bartholomew, A., and J. A. Bohnsack. 2005. A review of catch-and-release angling mortality with implications for no-take reserves. Reviews in Fish Biology and Fisheries 15:129–154.

Beaumariage, D. S. 1969. Returns from the 1965 Schlitz Tagging Program including a cumulative analysis of previous results. Technical Series No. 59. St. Petersburg, Florida.

Bellquist, L., S. Beyer, M. Arrington, J. Maeding, A. Siddall, P. Fischer, J. Hyde, and N. C. Wegner. 2019. Effectiveness of descending devices to mitigate the effects of barotrauma among rockfishes (Sebastes spp.) in California recreational fisheries. Fisheries Research 215(March):44–52.

149

Bellquist, L. F., C. G. Lowe, and J. E. Caselle. 2008. Fine-scale movement patterns, site fidelity, and habitat selection of ocean whitefish (Caulolatilus princeps). Fisheries Research 91:325– 335.

Benaka, L. R., L. Sharpe, L. Anderson, K. Brennan, J. E. Budrick, C. Lunsford, E. Meredith, M. S. Mohr, and C. Villafana. 2014. Fisheries release mortality: identifying, prioritizing, and resolving data gaps. NOAA Technical Memorandum NMFS-F/SPO-142.

Benhamou, S. 2011. Dynamic approach to space and habitat use based on biased random bridges. PLoS ONE 6(1).

Berkeley, S. A., M. A. Hixon, R. J. Larson, and M. S. Love. 2004. Fisheries sustainability via protection of age structure and spatial distribution of fish populations. Fisheries 29(8):23– 32.

Biesinger, Z., B. M. Bolker, D. Marcinek, T. M. Grothues, J. A. Dobarro, and W. J. Lindberg. 2013. Testing an autonomous acoustic telemetry positioning system for fine-scale space use in marine animals. Journal of Experimental Marine Biology and Ecology 448:46–56.

Biro, P. A., and J. R. Post. 2008. Rapid depletion of genotypes with fast growth and bold personality traits from harvested fish populations. Proceedings of the National Academy of Sciences 105(8):2919–2922.

Bochenek, E. A., E. N. Powell, and J. DePersenaire. 2012. Recall bias in recreational summer flounder party boat trips and angler preferences to new approaches to bag and size limits. Fisheries Science 78:1–14.

Bohaboy, E. C., T. L. Guttridge, N. Hammerschlag, M. P. M. Van Zinnicq Bergmann, and W. F. Patterson III. 2019. Application of three-dimensional acoustic telemetry to assess the effects of rapid recompression on reef fish discard mortality. ICES Journal of Marine Science. doi: 10.1093/icesjms/fsz202.

Brownscombe, J. W., A. J. Danylchuk, J. M. Chapman, L. F. G. Gutowsky, and S. J. Cooke. 2017. Best practices for catch-and-release recreational fisheries – angling tools and tactics. Fisheries Research 186:693–705.

Burnham, K. P., and D. R. Anderson. 2002. Model selection and multimodel inference, 2nd edition. Springer-Verlag, New York.

Butcher, P. A., M. K. Broadhurst, K. C. Hall, B. R. Cullis, and S. R. Raidal. 2012. Assessing barotrauma and release methods among angled snapper Pagrus auratus. Fisheries Research 127–128:49–55.

California Sea Grant, Oregon Sea Grant, and University of Southern California Sea Grant. 2014. Bring that rockfish down. Pamphlet VS - 01/31/2014M.Patyten. California Department of Fish and Game.

150

Campbell, M. D., W. B. Driggers III., B. Sauls, and J. F. Walter. 2014. Release mortality in the red snapper (Lutjanus campechanus) fishery: a meta-analysis of 3 decades of research. Fishery Bulletin 112:283–297.

Campbell, M. D., R. Patino, J. Tolan, R. Strauss, and S. L. Diamond. 2010a. Sublethal effects of catch-and-release fishing: measuring capture stress, fish impairment, and predation risk using a condition index. ICES Journal of Marine Science 67:513–521.

Campbell, M. D., J. Tolan, R. Strauss, and S. L. Diamond. 2010b. Relating angling-dependent fish impairment to immediate release mortality of red snapper (Lutjanus campechanus). Fisheries Research 106:64–70. Elsevier B.V.

Capizzano, C. W., J. W. Mandelman, W. S. Hoffman, M. J. Dean, J. Kneebone, E. Jones, M. J. Stettner, D. R. Zemeckis, H. P. Benoit, N. J. Buchan, J. A. Langan, and J. A. Sulikowski. 2016. Estimating and mitigating the discard mortality of Atlantic cod (Gadus morhua) in the Gulf of Maine recreational rod-and-reel fishery. ICES Journal of Marine Science 73(9):2342–2355.

Chen, I. 2012. Putting rockfish back where they belong. Science 338:600–601.

Coleman, F. C., P. B. Baker, and C. C. Koenig. 2004a. A review of Gulf of Mexico marine protected areas: successes, failures, and lessons learned. Fisheries 29(2):10–21.

Coleman, F. C., W. F. Figueira, J. S. Ueland, and L. B. Crowder. 2004b. The impact of United States recreational fisheries on marine fish populations. Science 305:1958–1960.

Cooke, S. J., and I. G. Cowx. 2004. The role of recreational fishing in giobal fish crises. BioScience 54(9):857–860.

Cooke, S. J., and I. G. Cowx. 2006. Contrasting recreational and commercial fishing: Searching for common issues to promote unified conservation of fisheries resources and aquatic environments. Biological Conservation 128(1):93–108.

Cowan, J. H. Jr., C. B. Grimes, W. F. Patterson III, C. J. Walters, A. C. Jones, W. J. Lindberg, D. J. Sheehy, W. E. Pine III, J. E. Powers, M. D. Campbell, K. C. Lindeman, S. L. Diamond, R. Hilborn, H. T. Gibson, and K. A. Rose. 2011. Red snapper management in the Gulf of Mexico: science- or faith-based? Reviews in Fish Biology and Fisheries 21:187–204.

Cox, D. R., and D. Oakes. 1984. Analysis of survival data. Chapman and Hall, London, UK.

Crandall, C. A., T. M. Garlock, and K. Lorenzen. 2018. Understanding resource-conserving behaviors among fishers: barotrauma mitigation and the power of subjective norms in Florida’s reef fisheries. North American Journal of Fisheries Management 38(2):271–280.

Curtis, J. M., M. W. Johnson, S. L. Diamond, and G. W. Stunz. 2015. Quantifying delayed mortality from barotrauma impairment in discarded red snapper using acoustic telemetry. Marine and Coastal Fisheries 7:434–449.

151

Dance, K. M., J. R. Rooker, J. B. Shipley, M. A. Dance, and R. J. D. Wells. 2018. Feeding ecology of fishes associated with artificial reefs in the northwest Gulf of Mexico. PLoS ONE 13(10):1–26.

Dance, M. A., D. L. Moulton, N. B. Furey, and J. R. Rooker. 2016. Does transmitter placement or species affect detection efficiency of tagged animals in biotelemetry research? Fisheries Research 183:80–85.

Dance, M. A., W. F. Patterson III, and D. T. Addis. 2011. Fish community and trophic structure at artificial reef sites in the northeastern Gulf of Mexico. Bulletin of Marine Science 87(3):301–324.

Dean, M. J., W. S. Hoffman, D. R. Zemeckis, and M. P. Armstrong. 2014. Fine-scale diel and gender-based patterns in behaviour of Atlantic cod (Gadus morhua) on a spawning ground in the western Gulf of Maine. ICES Journal of Marine Science 71(6):1474–1489.

Decelles, G., and D. Zemeckis. 2014. Acoustic and radio telemetry. Pages 397–428 in S. X. Cadrin, L. A. Kerr, and S. Mariani, editors. Stock identification methods, 2nd edition.

Diamond, S., T. Hedrick-Hopper, G. Stunz, M. Johnson, and J. Curtis. 2011. Reducing discard mortality of red snapper in the recreational fisheries using descender hooks and rapid recompression. Final report grant number NA07NMFS4540078. Lubbock, Texas.

Diamond, S. L., M. D. Campbell, D. Olson, Y. Wang, J. Zeplin, and S. Qualia. 2007. Movers and stayers: individual variability in site fidelity and movements of red snapper off Texas. Pages 163–187 in W. F. Patterson III., J. H. Cowan Jr., G. R. Fitzhugh, and D. L. Nieland, editors. Red snapper ecology and fisheries in the US Gulf of Mexico. American Fisheries Society Symposium 60, Bethesda, Maryland.

Domeier, M. L., D. Kiefer, N. Nasby-Lucas, A. Wagschal, and F. O’Brien. 2005. Tracking Thunnus thynnus orientalis with the aid of an automated algorithm. Fishery Bulletin 103:292–306.

Don, B. A. C., and K. Rennolls. 1983. A home range model incorporating biological attraction points. Journal of Animal Ecology 52(1):69–81.

Dotson, J. R., M. S. Allen, J. A. Kerns, and W. F. Pouder. 2013. Utility of restrictive harvest regulations for trophy largemouth bass management. North American Journal of Fisheries Management 33(3):499–507.

Drumhiller, K. L., M. W. Johnson, S. L. Diamond, M. Megan, R. Robillard, and G. W. Stunz. 2014. Venting or rapid recompression increase survival and improve recovery of red snapper with barotrauma. Marine and Coastal Fisheries 6(1):190–199.

Eberts, R. L., and C. M. Somers. 2017. Venting and descending provide equivocal benefits for catch-and-release survival: study design influences effectiveness more than barotrauma relief method. North American Journal of Fisheries Management 37(3):612–623.

152

Eberts, R. L., M. A. Zak, R. G. Manzon, and C. M. Somers. 2018. Walleye responses to barotrauma relief treatments for catch-and-release angling: short-term changes to condition and behavior. Journal of Fish and Wildlife Management 9(2):415–430.

Ellis, T. A., J. A. Buckel, and J. E. Hightower. 2017. Winter severity influences spotted seatrout mortality in a southeast US estuarine system. Marine Ecology Progress Series 564:145–161.

Espinoza, M., T. J. Farrugia, D. M. Webber, F. Smith, and C. G. Lowe. 2011. Testing a new acoustic telemetry technique to quantify long-term, fine-scale movements of aquatic animals. Fisheries Research 108:364–371.

Farmer, N. A., and J. S. Ault. 2011. Grouper and snapper movements and habitat use in Dry Tortugas, Florida. Marine Ecology Progress Series 433:169–184.

Farmer, N. A., J. C. Tetzlaff, and A. J. Strelcheck. 2014. 2014 Gulf of Mexico red snapper recreational slot limit analysis. SERO-LAPP-2014-05.

Ferter, K., M. S. Weltersbach, H. V. Strehlow, J. H. Vølstad, J. Alos, R. Arlinghaus, M. Armstrong, M. Dorow, M. de Graaf, T. van der Hammen, K. Hyder, H. Levrel, A. Paulrud, K. Radtke, D. Rocklin, C. R. Sparrevohn, and P. Veiga. 2013. Unexpectedly high catch- and-release rates in European marine recreational fisheries: implications for science and management. ICES Journal of Marine Science 70(7):1319–1329.

Fisheries and Oceans Canada. 2018. A review of the use of recompression devices as a tool for reducing the effects of barotrauma on rockfishes in British Columbia. Canadian Science Advisory Secretariat Science Response 2018/043. Nanaimo, British Columbia.

Fisheries and Oceans Canada. 2019. Survey of recreational fishing in Canada, 2015. Ottawa, Ontario.

Florida Sea Grant, and University of Florida. 2017. Catch & Release Fishing [online]. Available from http://catchandrelease.org/descending-and-venting.html [accessed 13 August 2019].

Fodrie, F. J., L. A. Yeager, J. H. Grabowski, C. A. Layman, G. D. Sherwood, and M. D. Kenworthy. 2015. Measuring individuality in habitat use across complex landscapes: approaches, constraints, and implications for assessing resource specialization. Oecologia 178:75–87.

Friedl, S. E., J. A. Buckel, J. E. Hightower, F. S. Scharf, and K. H. Pollock. 2013. Telemetry- based mortality estimates of juvenile spot in two North Carolina estuarine creeks. Transactions of the American Fisheries Society 142(2):399–415.

Froehlich, C. Y. M., A. Garcia, and R. J. Kline. 2019. Daily movement patterns of red snapper (Lutjanus campechanus) on a large artificial reef. Fisheries Research 209:49–57.

Froeschke, J. T., B. F. Froeschke, and C. M. Stinson. 2012. Long-term trends of bull shark (Carcharhinus leucas) in estuarine waters of Texas, USA. Canadian Journal of Fisheries and Aquatic Sciences 70(1):13–21.

153

Furey, N. B., M. A. Dance, and J. R. Rooker. 2013. Fine-scale movements and habitat use of juvenile southern flounder Paralichthys lethostigma in an estuarine seascape. Journal of Fish Biology 82:1469–1483.

Garner, S. B., and W. F. Patterson III. 2015. Direct observation of fishing effort, catch, and discard rates of charter boats targeting reef fishes in the northern Gulf of Mexico. Fishery Bulletin 113(2):157–166.

Gosling, S. D. 2001. From mice to men: what can we learn about personality from animal research? Psychological Bulletin 127(1):45–86.

Gulf of Mexico Fishery Management Council. 2017. Possessing venting tools and descending devices when fishing for reef fish. Options paper for a framework action to the fishery management plan for the reef fish resources of the Gulf of Mexico. Award No. NA15NMF4410011. Tampa, Florida.

Gulf of Mexico Fishery Management Council. 2018a. January 2018 council update [online]. Available from http://gulfcouncil.org/press/2018/january-2018-council-update/ [accessed 26 February 2018].

Gulf of Mexico Fishery Management Council. 2018b. Proposed outreach plan policy on the use of venting tools and descending devices. Tab O, No. 4, June 2018 council meeting.

Gulf of Mexico Fishery Management Council. 2018c. Agency efforts to outreach barotrauma and the use of venting and/or descending devices. Tab O, No. 4(a) June 2018 Council Meeting.

Gulf of Mexico Fishery Management Council. 2019a. Fishing for our future [online]. Available from https://gulfcouncil.org/fishing-for-our-future/ [accessed 4 November 2019].

Gulf of Mexico Fishery Management Council. 2019b. Gulf Council votes to allow state management of recreational red snapper for private anglers. For immediate release April 4, 2019. Tampa, Florida.

Guzzo, M. M., T. E. Van Leeuwen, J. Hollins, B. Koeck, M. Newton, D. M. Webber, F. I. Smith, D. M. Bailey, and S. S. Killen. 2018. Field testing a novel high residence positioning system for monitoring the fine-scale movements of aquatic organisms. Methods in Ecology and Evolution 9:1478–1488.

Harrison, P. M., L. F. G. Gutowsky, E. G. Martins, D. A. Patterson, S. J. Cooke, and M. Power. 2015. Personality-dependent spatial ecology occurs independently from dispersal in wild burbot (Lota lota). Behavioral Ecology 26(2):483–492.

Herbig, J. L., and S. T. Szedlmayer. 2016. Movement patterns of gray triggerfish, Balistes capriscus, around artificial reefs in the northern Gulf of Mexico. Fisheries Management and Ecology 23:418–427.

154

Heupel, M. R., and C. A. Simpfendorfer. 2002. Estimation of mortality of juvenile blacktip sharks, Carcharhinus limbatus, within a nursery area using telemetry data. Canadian Journal of Fisheries and Aquatic Sciences 59(4):624–632.

Heupel, M. R., B. G. Yeiser, A. B. Collins, L. Ortega, and C. A. Simpfendorfer. 2010. Long-term presence and movement patterns of juvenile bull sharks, Carcharhinus leucas, in an estuarine river system. Marine and Freshwater Research 61(1):1–10.

Hightower, J. E., J. R. Jackson, and K. H. Pollock. 2001. Use of telemetry methods to estimate natural and fishing mortality of Striped Bass in Lake Gaston, North Carolina. Transactions of the American Fisheries Society 130(4):557–567.

Hixon, M. A., D. W. Johnson, and S. M. Sogard. 2014. BOFFFFs: on the importance of conserving old-growth age structure in fishery populations. ICES Journal of Marine Science 71(8):2171–2185.

Hochhalter, S. J., and D. J. Reed. 2011. The Effectiveness of deepwater release at improving the survival of discarded yelloweye rockfish. North American Journal of Fisheries Management 31(5):852–860.

Hood, P. B., A. J. Strelcheck, and P. Steele. 2007. A history of red snapper management in the Gulf of Mexico. Pages 267–284 in W. F. Patterson III., J. H. Cowan Jr., G. R. Fitzhugh, and D. L. Nieland, editors. Red snapper ecology and fisheries in the US Gulf of Mexico. American Fisheries Society Symposium 60, Bethesda, Maryland.

Horne, J. S., E. O. Garton, S. M. Krone, and J. S. Lewis. 2007. Analyzing animal movements using Brownian bridges. Ecology 88(9):2354–2363.

Huntingford, F. A. 1976. The relationship between anti-predator behavior and aggression among conspecifics in the three-spined stickleback, Gasterosteus aculeatus. Animal Behavior 24:245–260.

Ingram, G. W. Jr., and W. F. Patterson III. 2001. Movement patterns of Red Snapper (Lutjanus campechanus), Greater Amberjack (Seriola dumerili), and Gray Triggerfish (Balistes capriscus) in the Gulf of Mexico and the utility of marine reserves as management tools. Proceedings of the Gulf and Caribbean Fisheries Institute 52:686–699.

Jackson, L. S., J. M. Drymon, T. R. Nelson, and S. P. Powers. 2018. Biotelemetry based estimates of greater amberjack (Seriola dumerili) post-release mortality in the northern Gulf of Mexico. Fisheries Research 208(July 2017):239–246.

Jarvis, E. T., and C. G. Lowe. 2008. The effects of barotrauma on the catch-and-release survival of southern California nearshore and shelf rockfish (Scorpaenidae, Sebastes spp.). Canadian Journal of Fisheries and Aquatic Sciences 65(7):1286–1296.

Lazaridis, E. 2015. R Package “lunar”: lunar phase & distance, seasons and other environmental factors.

155

Long, J. M., M. S. Allen, W. F. Porak, and C. D. Suski. 2015. A Historical perspective of black bass management in the United States. American Fisheries Society Symposium 82:99–122.

Lovell, S. J., S. Steinback, and J. Hilger. 2013. The economic contribution of marine angler expenditures in the United States, 2011. NOAA Technical Memorandum NMFS-F/SPO- 134.

Lowe, C. G., D. T. Topping, D. P. Cartamil, and Y. P. Papastamatiou. 2003. Movement patterns, home range, and habitat utilization of adult kelp bass Paralabrax clathratus in a temperate no-take marine reserve. Marine Ecology Progress Series 256:205–216.

Lowerre-Barbieri, S., L. Crabtree, T. Switzer, S. W. Burnsed, and C. Guenther. 2015. Assessing reproductive resilience: an example with South Atlantic red snapper Lutjanus campechanus. Marine Ecology Progress Series 526:125–141.

Lowerre-Barbieri, S. K., R. Kays, J. T. Thorson, and M. Wikelski. 2019. The ocean’s movescape: fisheries management in the bio-logging decade (2018-2028). ICES Journal of Marine Science 76(2):477–488.

Lynch, P. D., R. D. Methot, and J. S. Link. 2018. Implementing a next generation stock assessment enterprise. NOAA Technical Memorandum NMFS-F/SPO-183.

Mackinson, S., D. C. Wilson, P. Galiay, and B. Deas. 2011. Engaging stakeholders in fisheries and marine research. Marine Policy 35:18–24.

McCawley, J. R., and J. H. Cowan Jr. 2007. Seasonal and size specific diet and prey demand of Red Snapper on Alabama artificial reefs. Pages 77–104 in W. F. Patterson III, J. H. Cowan Jr., G. R. Fitzhugh, and D. L. Nieland, editors. Red Snapper ecology and fisheries in the U.S. Gulf of Mexico. American Fisheries Society Symposium 60, Bethesda, Maryland.

Methot, R. D., T. A’mar, C. Wetzel, and I. Taylor. 2017. Stock Synthesis user manual version 3.30. NOAA Fisheries, Seattle, Washington. 181 pp.

Methot, R. D., and I. G. Taylor. 2011. Adjusting for bias due to variability of estimated recruitments in fishery assessment models. Canadian Journal of Fisheries and Aquatic Sciences 68:1744–1760.

Methot, R. D., and C. R. Wetzel. 2013. Stock synthesis : A biological and statistical framework for fish stock assessment and fishery management. Fisheries Research 142:86–99.

Muoneke, M. I., and W. M. Childress. 1994. Hooking mortality: a review for recreational fisheries. Reviews in Fisheries Science 2(2):123–156.

Murie, D. J., and D. C. Parkyn. 2013. Release mortality of Gulf of Mexico greater amberjack from commercial and recreational hand-line fisheries: integration of fishing practices, environmental parameters, and fish physiological attributes. SEDAR33-DW29. North Charleston, South Carolina.

156

Murphy, M. D., R. F. Heagey, V. H. Neugebauer, M. D. Gordon, and J. L. Hintz. 1995. Mortality of spotted seatrout released from gill-net or hook-and-line gear in Florida. North American Journal of Fisheries Management 15:748–753.

Murphy, R. D. J. 2018. Social and ecological dimensions of the Striped Bass (Morone saxatilis) fisheries in southern New England. Northeastern University.

National Marine Fisheries Service. 2017. Fisheries of the United States, 2016. Silver Spring, Maryland.

National Oceanic and Atmospheric Administration. 1998. Managing the Nation’s bycatch: priorities, programs and actions for the National Marine Fisheries Service. Washington D.C.

National Oceanic and Atmospheric Administration, and National Weather Service. 2017. National Data Buoy Center: Station 42012- Orange Beach [online]. Available from http://www.ndbc.noaa.gov/station_page.php?station=42012 [accessed 07 November 2017].

National Oceanic and Atmospheric Administration, and National Weather Service. 2019. National Data Buoy Center: Station 42040- Luke Offshore Test Platform [online]. Available from https://www.ndbc.noaa.gov/station_page.php?station=42040 [accessed 07 January 2019].

Ng, E. L., J. P. Fredericks, and M. C. Quist. 2015. Effects of gill-net trauma, barotrauma, and deep release on postrelease mortality of lake trout. Journal of Fish and Wildlife Management 6(2):265–277.

Overton, A. S., J. Zabawski, and K. L. Riley. 2008. Release mortality of undersized fish from the snapper–grouper complex off the North Carolina coast. North American Journal of Fisheries Management 28:733–739.

Özgül, A., A. Lök, A. Ulaş, F. O. Düzbastilar, T. T. Tanrikul, and C. Pelister. 2015. Preliminary study on the use of the Vemco Positioning System to determine fish movements in artificial reef areas: A case study on Sciaena umbra Linnaeus, 1758. Journal of Applied Ichthyology 31:41–47.

Pacific Fishery Management Council. 2013. Groundfish Management Team report on proposed discard mortality for cowcod, canary rockfish, and yelloweye rockfish released using descending devices in the recreational fishery. Agenda Item D.5.b GMT Report.

Parker, S. J., H. I. McElderry, P. S. Rankin, and R. W. Hannah. 2006. Buoyancy regulation and barotrauma in two species of nearshore rockfish. Transactions of the American Fisheries Society 135(5):1213–1223.

Patterson, W. F. III, and J. H. Cowan Jr. 2003. Site fidelity and dispersion of tagged red snapper, Lutjanus campechanus, in the northern Gulf of Mexico. Fisheries, Reefs, and Offshore Development: American Fisheries Society 36:181–193.

157

Patterson, W. F. III, G. W. Ingram Jr., R. L. Shipp, and J. H. Cowan Jr. 2002. Indirect estimation of red snapper (Lutjanus campechanus) and gray triggerfish (balistes capriscus) release mortality. Proceedings of the 53rd Gulf and Caribbean Fisheries Institute2:526–536.

Patterson, W. F. III, J. C. Watterson, R. L. Shipp, and J. H. Cowan Jr. 2001. Movement of tagged red snapper in the Northern Gulf of Mexico. Transactions of the American Fisheries Society 130:533–545.

Patterson, W. F. III. 2013. Acoustic tracking of reef fishes associated with EE-LAARS Refugia Reefs: Project Final Report Grant No. FWC-10340.

Peterson, C. D., C. N. Belcher, D. M. Bethea, W. B. Driggers, B. S. Frazier, and R. J. Latour. 2017. Preliminary recovery of coastal sharks in the south-east United States. Fish and Fisheries 18(5):845–859.

Petitgas, P., D. H. Secor, I. McQuinn, G. Huse, and N. Lo. 2010. Stock collapses and their recovery: mechanisms that establish and maintain life-cycle closure in space and time. Journal of Marine Science 67:1841–1848.

Piraino, M. N., and S. T. Szedlmayer. 2014. Fine-scale movements and home ranges of red snapper around artificial reefs in the northern Gulf of Mexico. Transactions of the American Fisheries Society 143:988–998.

Pollock, K. H., and W. E. Pine III. 2007. The design and analysis of field studies to estimate catch-and-release mortality. Fisheries Management and Ecology 14:123–130.

Pribyl, A. L., C. B. Schreck, M. L. Kent, K. M. Kelley, and S. J. Parker. 2012. Recovery potential of black rockfish, Sebastes melanops Girard, recompressed following barotrauma. Journal of Fish Diseases 35(4):275–286.

R Core Team. 2016. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.

Raby, G. D., J. R. Packer, A. J. Danylchuk, and S. J. Cooke. 2014. The understudied and underappreciated role of predation in the mortality of fish released from fishing gears. Fish and Fisheries 15:489–505.

Radford, Z., K. Hyder, L. Zarauz, E. Mugerza, K. Ferter, R. Prellezo, H. V. Strehlow, B. Townhill, W.-C. Lewin, and M. S. Weltersbach. 2018. The impact of marine recreational fishing on key fish stocks in European waters. PLoS ONE:1–16.

Radonski, G. C. 2002. History and application of catch-and-release fishing: the good, the bad, and the ugly. Pages 3–10 in J. A. Lucy and A. L. Studholme, editors. Catch and release in marine recreational fisheries. American Fisheries Society Symposium 30, Bethesda, Maryland.

158

Reale, D., D. Garant, M. M. Humphries, P. Bergeron, V. Careau, and P. Montiglio. 2010. Personality and the emergence of the pace-of-life syndrome concept at the population level. Philosophical Transactions of the Royal Society B: Biological Sciences 365(1560):4051– 4063.

Reubens, J. T., F. Pasotti, S. Degraer, and M. Vincx. 2013. Residency, site fidelity and habitat use of atlantic cod (Gadus morhua) at an offshore wind farm using acoustic telemetry. Marine Environmental Research 90:128–135.

Roach, J. P., K. C. Hall, and M. K. Broadhurst. 2011. Effects of barotrauma and mitigation methods on released Australian bass Macquaria novemaculeata. Journal of Fish Biology 79(5):1130–1145.

Rogers, P. J., C. Huveneers, S. D. Goldsworthy, J. G. Mitchell, and L. Seuront. 2013. Broad- scale movements and pelagic habitat of the dusky shark Carcharhinus obscurus off Southern Australia determined using pop-up satellite archival tags. Fisheries Oceanography 22(2):102–112.

Roy, R., J. Beguin, C. Argillier, L. Tissot, F. Smith, S. Smedbol, and E. De-Oliveira. 2014. Testing the VEMCO Positioning System: spatial distribution of the probability of location and the positioning error in a reservoir. Animal Biotelemetry 2(1).

Rudershausen, P. J., J. A. Buckel, and T. Burgess. 2013. Estimating discard mortality of black sea bass (Centropristis striata) and other reef fish in North Carolina using a tag-return approach. Combined Final Report North Carolina Sea Grant FRG 07-FEG-01 and 09-FEG- 04.

Rummer, J. L., and W. A. Bennett. 2005. Physiological effects of swim bladder overexpansion and catastrophic decompression on red snapper. Transactions of the American Fisheries Society 134:1457–1470.

Runde, B. J., and J. A. Buckel. 2018. Descender devices are promising tools for increasing survival in deepwater groupers. Marine and Coastal Fisheries: Dynamics, Management, and Ecosystem Science 10:100–117.

Runde, B. J., P. J. Rudershausen, B. Sauls, C. S. Mikles, and J. A. Buckel. 2019. Low discard survival of gray triggerfish in the southeastern US hook-and-line fishery. Fisheries Research 219(May):105313.

Sauls, B. 2012. Release mortality estimates for recreational hook-and-line caught red snapper derived from a large-scale tag-recapture study in the eastern Gulf of Mexico. SEDAR31- DW23. SEDAR, North Charleston, South Carolina. 20 pp.

Sauls, B., O. Ayala, and R. Cody. 2014. A directed study of the recreational red snapper fisheries in the Gulf of Mexico along the West Florida Shelf – Final project report. SEDAR42-RD- 01. SEDAR, North Charleston, South Carolina. 100 pp.

159

Sauls, B., and B. Cermak. 2013. Characterization of greater amberjack discards in recreational for-hire fisheries. SEDAR33-DW04. SEDAR, North Charleston, South Carolina. 24 pp.

Sauls, B. J., O. Ayala, R. Germeroth, J. Solomon, and R. Brody. 2017. Red snapper discard mortality in Florida’s recreational fisheries. SEDAR52-WP-09. SEDAR, North Charleston, South Carolina. 15 pp.

Schwartzkopf, B. D., T. A. Langland, and J. H. Cowan Jr. 2017. Habitat selection important for red snapper feeding ecology in the northwestern Gulf of Mexico. Marine and Coastal Fisheries 9(1):373–387.

Secor, D. H., F. Zhang, M. H. P. O’Brien, and M. Li. 2019. Ocean destratification and fish evacuation caused by a Mid-Atlantic tropical storm. ICES Journal of Marine Science 76(2):573–584.

SEDAR. 2005. Stock assessment report of SEDAR 7 Gulf of Mexico red snapper. SEDAR, North Charleston, South Carolina. 480 pp.

SEDAR. 2013. SEDAR 31 – Gulf of Mexico red snapper stock assessment report. SEDAR, North Charleston, South Carolina. 1103 pp.

SEDAR. 2014a. SEDAR 33 – Gulf of Mexico gag stock assessment report. SEDAR, North Charleston, South Carolina. 609 pp.

SEDAR. 2014b. Sedar 33 – Gulf of Mexico greater amberjack stock assessment report. SEDAR, North Charleston, South Carolina. 490 pp.

SEDAR. 2015a. SEDAR 43 – Stock assessment report: Gulf of Mexico gray triggerfish. SEDAR, North Charleston, South Carolina. 174 pp.

SEDAR. 2015b. SEDAR 42 – Gulf of Mexico red grouper stock assessment report. SEDAR, North Charleston, South Carolina. 612 pp.

SEDAR. 2016. SEDAR 45 – stock assessment report: Gulf of Mexico vermilion snapper. SEDAR, North Charleston, South Carolina. 188 pp.

SEDAR. 2017. SEDAR 54 Stock Assessment Report: HMS Sandbar Shark. North Charleston, South Carolina. 193 pp.

SEDAR. 2018. SEDAR 52 – Stock Assessment Report Gulf of Mexico Red Snapper. SEDAR, North Charleston, South Carolina. 435 pp.

Sih, A., A. Bell, and J. C. Johnson. 2004. Behavioral syndromes: An ecological and evolutionary overview. Trends in Ecology and Evolution 19(7):372–378.

Simpfendorfer, C. A., M. R. Heupel, and R. E. Hueter. 2002. Estimation of short-term centers of activity from an array of omnidirectional hydrophones and its use in studying animal movements. Canadian Journal of Fisheries and Aquatic Sciences 59:23–32.

160

Smedbol, S., F. Smith, D. Webber, R. Vallée, and T. King. 2014. Using underwater coded acoustic telemetry for fine scale positioning of aquatic animals. VEMCO, Bedford, Nova Scotia, Canada.

Smith, F. 2013. Understanding HPE in the VEMCO Positioning System (VPS). DOC-005457- 01.

Sportfishing Association of California. 2019. Sportfishing Association of California [online]. Available from https://www.californiasportfishing.org/ [accessed 13 August 2019].

Stensland, S., and O. Aas. 2014. The role of social norms and informal sanctions in catch-and- release angling. Fisheries Management and Ecology 21:288–298.

Stephens, J., D. Wilson-Vandenberg, D. Wendt, J. Carroll, R. Nakamura, E. Nakada, S. J. Reinecke, and J. Wilson. 2006. Rockfish resouces of the south central California coast: analysis of the resource from partyboat data, 1980 – 2005. The California Cooperative Oceanic Fisheries Investigations Reports 47:140–155.

Strelcheck, A. J., J. H. Cowan Jr., and W. F. Patterson III. 2007. Site fidelity, movement, and growth of red snapper Lutjanus campechanus: implications for artificial reef management. Pages 147–162 in W. F. Patterson III., J. H. Cowan Jr., D. A. Nieland, and G. R. Fitzhugh, editors. Population ecology and fisheries of U.S. Gulf of Mexico red snapper. American Fisheries Society Symposium 60. Bethesda, Maryland.

Strelcheck, A. J., and P. B. Hood. 2007. Rebuilding red snapper: recent management activities and future management challenges. Pages 385–396 in W. F. Patterson III., J. H. Cowan, Jr., G. R. Fitzhugh, and D. L. Nieland, editors. Population ecology and fisheries of U.S. Gulf of Mexico red snapper. American Fisheries Society Symposium 60. Bethesda, Maryland.

Sumpton, W. D., I. W. Brown, D. G. Mayer, M. F. Mclennan, A. Mapleston, A. R. Butcher, D. J. Welch, J. M. Kirkwood, B. Sawynok, and G. A. Begg. 2010. Assessing the effects of line capture and barotrauma relief procedures on post-release survival of key tropical reef fish species in Australia using recreational tagging clubs. Fisheries Management and Ecology 17(1):77–88.

Szedlmayer, S. T., and R. L. Schroepfer. 2005. Long-term residence of red snapper on artificial reefs in the northeastern Gulf of Mexico. Transactions of the American Fisheries Society 134:315–325.

Tarnecki, J. H., and W. F. Patterson III. 2015. Changes in red snapper diet and trophic ecology following the Deepwater Horizon Oil Spill. Marine and Coastal Fisheries 7:135–147.

Therneau, T. M., and T. Lumley. 2015. R Package “survival”: survival analysis.

Thieurmel, B., and A. Elmarhraoui. 2019. R Package “suncalc”: compute sun position, sunlight phases, moon position and lunar phase.

161

Topping, D. T., and S. T. Szedlmayer. 2011. Site fidelity, residence time and movements of red snapper Lutjanus campechanus estimated with long-term acoustic monitoring. Marine Ecology Progress Series 437:183–200.

Tracey, J. A., J. Sheppard, J. Zhu, R. Sinkovts, A. Chourasia, G. Lockwood, and R. N. Fisher. 2014a. R Package “mkde”: 2D and 3D movement-based kernel density estimates (MKDEs).

Tracey, J. A., J. Sheppard, J. Zhu, F. Wei, R. R. Swaisgood, and R. N. Fisher. 2014b. Movement- based estimation and visualization of space use in 3D for wildlife ecology and conservation. PLoS ONE 9(7).

Wilde, G. R. 2009. Does venting promote survival of released fish? Fisheries 34(1):20–28.

Williams-Grove, L. J., and S. T. Szedlmayer. 2016a. Mortality estimates for red snapper based on ultrasonic telemetry in the Northern Gulf of Mexico. North American Journal of Fisheries Management 36:1036–1044.

Williams-Grove, L. J., and S. T. Szedlmayer. 2016b. Acoustic positioning and movement patterns of red snapper, Lutjanus campechanus, around artificial reefs in the northern Gulf of Mexico. Marine Ecology Progress Series 553:233–251.

Williams-Grove, L. J., and S. T. Szedlmayer. 2017. Depth preferences and three-dimensional movements of red snapper, Lutjanus campechanus, on an artificial reef in the northern Gulf of Mexico. Fisheries Research 190:61–70.

Williams, L. J., J. L. Herbig, and S. T. Szedlmayer. 2015. A cage release method to improve fish tagging studies. Fisheries Research 172:125–129.

Wilson, D. S., A. B. Clark, K. Coleman, and T. Dearstyne. 1994. Shyness and boldness in humans and other animals. Trends in Ecology and Evolution 9(11):442–446.

Winemiller, K. O., and K. A. Rose. 1992. Patterns of life-history diversification in North American fishes: implications for population regulation. Canadian Journal of Fisheries and Aquatic Sciences 49(10):2196–2218.

Wood, S. N. 2019. Package “mgcv”: mixed GAM computation vehicle with automatic smoothness estimation.

Worton, B. J. 1987. A review of models of home range for animal movement. Ecological Modelling 38:277–298.

Worton, B. J. 1989. Kernel methods for estimating the utilization distribution in home-range studies. Ecology 70(1):164–168.

162

BIOGRAPHICAL SKETCH

Erin (Collings) Bohaboy grew up in Andover, Massachusetts. From an early age, she spent most of her time outside in the woods or at local ponds and lakes, catching or feeding fish, frogs, and turtles. Fascinated by aquatic animals and freshwater ecosystems, she was an avid aquarium keeper. Erin harbored her love of fish and furthered her appreciation and respect for nature by fishing frequently with her father at rivers, on lakes, at the beach, and on the ice. Erin graduated from Andover Public High School in 2000 and was offered generous scholarships at several universities and colleges in New England.

Erin graduated summa cum laude with a B.S. in zoology (concentrating in ichthyology and aquaculture) from the University of New Hampshire in May 2004. Beginning her junior year, she worked as an undergraduate research assistant with Dr. Thomas Kocher at the Hubbard

Center for Genome Studies on several projects including pigmentation development in juvenile

African cichlids, hatchery selection of Atlantic cod in the aquaculture industry, and genetic sex determination in zebrafish. Her time was divided between research and caring for thousands of

African cichlids, most from Lake Malawi, that were bred and maintained for genetics research.

During her final undergraduate year, she took a class taught by Dr. Andrew Rosenberg in fisheries science and stock assessment where she discovered that she had a penchant for quantitative data analyses and fish population modeling. Encouraged by Dr. Rosenberg to pursue a career with the National Marine Fisheries Service, after graduating, Erin volunteered as a scientist on the NOAA research vessel Albatross IV doing sea scallop and groundfish surveys out of Woods Hole, Massachusetts. Between research surveys, she worked for the natural resources consulting firm MRAG Americas compiling bycatch estimates of U.S. Fisheries. In

2004 Erin met her future husband, Erich, who was a NOAA Corps officer stationed on the

Albatross IV.

163

Over the following decade, Erin and Erich traversed the United States 4 times as Erich was routinely transferred between ship and land assignments around the country. Erin continued her career at a variety of positions ranging from working on contract with NOAA on Endangered

Species Act Research Permitting in California and later Ocean Acidification Research in Seattle; doing field surveys of riverine fishes with the Wyoming Department of Game and Fish; and at the Seattle Aquarium where she interned as a tropical fish aquarist. Erin worked on the NOAA

Deepwater Horizon Oil Spill Natural Resources Damage Assessment at the consulting company

RPS ASA in Rhode Island from 2011–2012. Erin earned her M.S. in oceanography with Dr.

Jeremy Collie at the University of Rhode Island Graduate School of Oceanography in 2010. Her

M.S. thesis was on multi-species population dynamics (surplus production modeling) of Georges

Bank fishes. From 2010–2012 she was part of CAMEO (Comparative Analysis of Marine

Ecosystems), a collaboration between government, private industry, and academic researchers across the U.S., Canada, and Norway which concluded with the publication of a special volume in Marine Ecology Progress Series.

Erin began her Ph.D. with Dr. William Patterson at the University of South Alabama /

Dauphin Island Sea Lab in fall 2015. For 3 years, she was the recipient of the National Marine

Fisheries Service / Sea Grant Population and Ecosystem Dynamics Graduate Research

Fellowship, which provided doctoral funding and the tremendous opportunity to travel to and meet scientists at the NOAA Fisheries laboratories in Beaufort, North Carolina, Seattle,

Washington, and Honolulu, Hawaii. Erin transferred to the University of Florida in fall 2017.

Upon completion of her Ph.D., Erin plans to work as a fisheries biologist performing research and stock assessments for NOAA Fisheries.

164