University of Florida Thesis Or Dissertation Formatting Template

ABUNDANCE AND DISTRIBUTION OF ESTUARINE BOTTLENOSE DOLPHINS (TURSIOPS TRUNCATUS) IN THE NORTHWESTERN GULF OF MEXICO

ERROL ISAAC RONJE

A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE

UNIVERSITY OF FLORIDA

2019

To my kids, Jewel and Roxy, who woke me up

ACKNOWLEDGMENTS

I owe many people gratitude for their support during this research. First, I thank my committee members for their guidance and direction during the analysis and writing of this thesis. These are, from the University of Florida at Gainesville (UF): Juliane

Struve, Vincent Lecours, and Randall Wells; and from the National Marine Fisheries

Service (NMFS): Lance Garrison and Keith Mullin. Each one of them offered a unique perspective and skillset that helped me to grow as a person and scientist, and I will always remember their efforts. I especially thank Juliane for her patience and flexibility when I changed directions (more than once) during my studies. I thank Bill Lindberg and Rob Ahrens of UF for challenging me philosophically and rigorously testing my mettle for critical thinking. Thanks also to Amy Abernethy, UF graduate coordinator, for her quick and clear responses to my frequent administrative questions as I navigated my way through the graduate program.

I could not have completed this research without the following people: I thank

Keith Mullin and Lance Garrison for putting their confidence in me to lead this field work;

Jeff Adams (NMFS) for his Finbase wizardry and irreplaceable assistance with my queries; and Heidi Whitehead of the Texas Marine Mammal Stranding Network

(TMMSN) for field and lab support, logistics, and provision of resources that were directed towards the NMFS-TMMSN collaboration, from which this thesis was born.

I thank those colleagues that endured my many and varied questions about the finer details of ArcGIS, photo-ID methods, or abundance estimation: among these are

Jason Allen, David Hanisko, Annie Gorgone, Jenny Litz, Shauna McBride-Kebert, Adam

Pollack, and Todd Speakman. I also thank Caitlin Barclay, Amy Brossard, Clay

Carabajal, Katie Clark, Scott Hall, Jaymie Reneker, Cory Teague, Mel Thompson,

Whitnie Walton, and Nicole Willis for their assistance during long days in the field or with photo processing and verification. I thank Bernd Würsig and Randall Davis of Texas

A&M Galveston for their support with lab space and logistics.

Many thanks to Kevin Barry for his field work ethic, and for grinding through

Finbase verification in the lab. Thanks to Paxton Secker for her assistance during the long and often uneventful Sabine Lake surveys, and for completing a Finbase verification marathon in early 2019. Thanks to Christina Toms for her versatility and positive demeanor in the field, functioning as my sounding board, and for setting me straight on more than one occasion. I thank the incomparable Sarah Piwetz for her refreshing point of view, outstanding helpfulness, initiative, scientific integrity, and for putting up with me in general.

I thank my wife Tanya, for her understanding of my absent-mindedness while I attempted to balance work and graduate school, and for taking up my slack when I failed to remember some event or responsibility. I credit her support as the linchpin that held me together through a time of chronic sleep-deprivation. Finally, I thank my mom for giving me a place to crash during field work and for her unwavering support over the years.

This work was conducted under MMPA Permit No. 14450 issued to the SEFSC by the NMFS Office of Protected Resources, Louisiana Wildlife and Fisheries Scientific

Collecting Permit SCP#46 and approved by the NMFS Atlantic Institutional Animal Care and Use Committee.

TABLE OF CONTENTS

page

ACKNOWLEDGMENTS ...... 4

LIST OF FIGURES ...... 9

LIST OF ABBREVIATIONS ...... 10

ABSTRACT ...... 12

CHAPTER

1 INTRODUCTION ...... 14

Estimating Abundance ...... 15 Modeling Spatial Distribution ...... 16

2 BACKGROUND AND METHODS ...... 21

Survey Area Background ...... 21 West Bay ...... 22 Galveston Bay ...... 23 Sabine Lake ...... 24 Visual Surveys ...... 25 Survey Procedure ...... 25 Survey Design ...... 26 West Bay...... 28 Galveston Bay ...... 28 Sabine Lake ...... 29 Photo Analysis ...... 30

3 ABUNDANCE ESTIMATION ...... 33

Estimating Abundance and Identifying Potential Transients ...... 33 Model Selection ...... 34 Correction for Unmarked Fins ...... 35 Results ...... 35 Survey Effort and Dolphin Group Encounters ...... 35 Photo Analysis ...... 36 Abundance Estimates ...... 37 West Bay...... 38 Galveston Bay ...... 38 Sabine Lake ...... 39 Discussion ...... 39 Limitations ...... 45

4 HABITAT ANALYSIS ...... 55

Modeling Dolphin Distribution and Abiotic Factors ...... 55 Water Quality ...... 55 GIS Data Preparation ...... 57 Habitat Modeling ...... 59 Results ...... 63 West Bay and Galveston Bay ...... 63 Sabine Lake ...... 65 Discussion ...... 67 Limitations ...... 70

5 CONCLUSIONS ...... 100

APPENDIX

A TABLE OF INTER-BAY BOTTLENOSE DOLPHIN MATCHES ...... 105

B ABUNDANCE MODEL RANK ...... 107

C CAPTURE PROBABILITIES ...... 110

LIST OF REFERENCES ...... 114

BIOGRAPHICAL SKETCH ...... 126

LIST OF TABLES Table page

3-1 Summary of 2014–2018 bottlenose dolphin photo-ID survey effort ...... 47

3-2 Bottlenose dolphin abundance estimates ...... 48

4-1 West Bay–Galveston Bay water quality data ...... 71

4-2 Sabine Lake water quality data...... 72

4-3 Example of raw water quality data ...... 73

4-4 Summary results of visual survey and environmental data ...... 74

4-5 Generalized additive models attempted for West Bay–Galveston Bay...... 74

4-6 Generalized additive models attempted for Sabine Lake...... 75

4-7 Summary of mean observed and predicted dolphin density ...... 75

A-1 Sighting histories for inter-bay bottlenose dolphin matches ...... 105

B-1 Ranked scores of abundance models in program MARK ...... 107

C-1 Estimated capture probability parameters for selected models ...... 110

LIST OF FIGURES

Figure page

1-1 Northwestern Gulf of Mexico estuaries surveyed in this study ...... 20

2-1 Photo-ID mark-recapture survey routes ...... 32

3-1 West Bay dolphin groups and survey track-line...... 49

3-2 Galveston Bay dolphin groups and survey track-line...... 50

3-3 Sabine Lake dolphin groups and survey track-line...... 51

3-4 Discovery curves of estimated individuals ...... 52

3-5 Sighting locations for bottlenose dolphin inter-bay matches...... 53

3-6 Best estimated abundance corrected for unmarked animals ...... 54

4-1 Habitat analysis data for West Bay and Galveston Bay ...... 76

4-2 Habitat analysis data for Sabine Lake ...... 83

4-3 Fishnet grid for West Bay and Galveston Bay ...... 90

4-4 Fishnet grid for Sabine Lake ...... 91

4-5 Smooth plot of West Bay–Galveston Bay GAM ...... 92

4-6 West Bay–Galveston Bay observed and predicted bottlenose dolphin density .. 93

4-7 Smooth plots of the Sabine Lake GAM ...... 96

4-8 Sabine Lake observed and predicted bottlenose dolphin density ...... 97

LIST OF ABBREVIATIONS adj. R2 Adjusted r-squared

AIC Akaike’s Information Criterion

AICc Akaike’s Information Criterion, corrected for small sample size

CI Confidence interval

D Distinctiveness

DEM Digital Elevation Model edf Estimated degrees of freedom

ESRI Environmental Systems Research Institute

GAM Generalized additive model

GIS Geographic information system

GIWW Gulf Intracoastal Waterway

GPS Global positioning system

HSC Houston Ship Channel

MMPA Marine Mammal Protection Act

MSE Mean squared error

NAD North American Datum

NMFS National Marine Fisheries Service

NOAA National Oceanic and Atmospheric Administration

NorTex North Texas catalog of bottlenose dolphin dorsal fins nwGoMx Northwestern Gulf of Mexico photo-ID Photographic-identification

PQ Photo quality

TSP Tentatively Selected Plan

TWDB Texas Water Development Board

U.S. United States

UTM Universal Transverse Mercator

VIF Variance Inflation Factor

Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science

ABUNDANCE AND DISTRIBUTION OF ESTUARINE BOTTLENOSE DOLPHINS (TURSIOPS TRUNCATUS) IN THE NORTHWESTERN GULF OF MEXICO

Errol Isaac Ronje

December 2019

Chair: Juliane Struve Major: Fisheries and Aquatic Sciences

Recent assessments of estuarine common bottlenose dolphin (Tursiops truncatus, Montagu, 1821) abundance and habitat use are lacking throughout portions of the southeastern United States, northwestern Gulf of Mexico, including areas of

Texas and western Louisiana. In total, I led the efforts of 91 small-boat photographic identification surveys covering ~1,800 km2 and comprising ~11,000 km of track-line in winter and summer seasons from 2014–2018 in West Bay (n = 25), the Galveston Bay system (n = 49), Sabine Lake (n = 17), and adjacent coastal waters. Abundance estimates were calculated based on individual dolphin encounter histories constrained by spatiotemporal parameters to approximately represent 1) the total population encountered in each embayment, 2) the population limited to the interior of each embayment, and 3) the population filtered for potential transient dolphins. The best estimates for dolphins were (winter / summer, respectively) 37.9 / 36.5 for West Bay,

841.9 / 1131.5 for Galveston Bay, and 121.6 / 162.2 for Sabine Lake. Photo-identified dorsal fin matches among survey areas suggest the estuarine and coastal waters of northern Texas and western Louisiana may be inhabited by a metapopulation of

bottlenose dolphins, where in this study, a range of 4–15% of marked individuals in each study area were identified as transients. A generalized additive model was used to predict the density of individual bottlenose dolphins relative to habitat characteristics and provide maps of the observed and model-predicted bottlenose dolphin density.

Habitat covariates in the species distribution model included water quality data

(temperature, pH, dissolved oxygen, salinity) and bathymetry (water depth and sea floor slope) and the proximity of dolphin group locations to the coast. Both the observed and predicted density estimates clearly point to deep engineered channels as primary habitat selections for dolphins in each embayment, and together, the results of the abundance and habitat analyses suggest the Gulf of Mexico passes into Galveston Bay and Sabine Lake may be population mixing areas where bottlenose dolphins in the region are particularly abundant.

CHAPTER 1 INTRODUCTION

Common bottlenose dolphins (Tursiops truncatus, Montagu, 1821) are globally distributed between the 60th parallels in temperate to tropical oceanic and coastal waters, and estuarine populations have been documented throughout their range

(Jefferson et al., 2008; Wells & Scott, 2018). In the Gulf of Mexico, common bottlenose dolphins (hereafter referred to as “bottlenose dolphins” or “dolphins”) are widely distributed in bays, sounds, estuaries, and coastal waters, including the waters of Texas

(Bräger et al., 1994; Gunter, 1942; Lynn & Würsig, 2002; Maze & Würsig, 1999; Shane,

1977; Würsig & Lynn, 1996). In the United States (U.S.), bottlenose dolphins are included in the Marine Mammal Protection Act (MMPA) of 1972, which mandates that current information on abundance and geographic distribution be collected for each stock unit to assess risks to the populations and inform conservation measures (Wade

& Angliss, 1997). Bottlenose dolphin stock units are defined for 31 northern Gulf of

Mexico embayments based on the nature of their geographic separation, and the assumption that each embayment likely represents a population of dolphins with long- term site-fidelity and social bonds that would be difficult to replace should the stock be significantly depleted or exterminated (Mullin et al., 2007; Wells et al., 1987). Periodic assessments of abundance and distribution are needed for long-term monitoring programs to detect population trends, updating the geographic range of stock areas, and providing baseline data for population-level impacts associated with habitat modifications, sea-level rise, catastrophic events (e.g., major hurricanes), unusual mortality events, and other natural and anthropogenic pressures. However, long-term population trends for the northwestern Gulf of Mexico (nwGoMx) bottlenose dolphin

stocks have not been assessed due to large temporal or spatial data gaps (Phillips &

Rosel, 2014). Out of the seven Texas Gulf Coast estuarine bottlenose dolphin stocks delineated by the National Marine Fisheries Service (NMFS), three are in the upper

Texas coast of the nwGoMx: West Bay, Galveston Bay, and Sabine Lake (Figure 1-1).

Up-to-date abundance estimates are part of a suite of population information required by the NMFS to assess the viability of each dolphin stock (Hayes et al., 2017).

Additionally, a characterization of the abiotic factors influencing dolphin spatial distribution and density are critical to understanding the potential threats to this protected species.

Estimating Abundance

I conducted photographic identification (photo-ID) mark-recapture surveys for bottlenose dolphins between December 2014 and June 2017 in estuarine and coastal waters along the Texas nwGoMx coast to estimate dolphin abundance and investigate connectivity among the three stocks. Photo-ID mark-recapture methods take advantage of unique, naturally occurring marks on the dorsal fins of wild dolphins to identify individual animals across a time series of sampling occasions (Würsig & Würsig, 1977).

The mark-recapture methodology has been used for a wide range of population studies and can serve as a foundation for insights into population size, behavior, migration, survival, fecundity, and the delineation of stock boundaries (Wells, 2018). When applied to adjacent habitat areas, photo-ID studies provide opportunities to characterize the relationships among spatially defined bottlenose dolphin populations comprising a metapopulation (Chabanne et al., 2017). A metapopulation is composed of a network of populations or subpopulations interacting at some level, characterized by similar population dynamics and intraspecific tendencies of distribution or dispersal between

spatially heterogeneous habitats (Kritzer & Sale, 2004). The work presented in this thesis provides the first geographically comprehensive abundance estimates for West

Bay, Galveston Bay, and Sabine Lake populations of bottlenose dolphins in the nwGoMx. I provide three different abundance estimates for each area and season: A) the total number of animals encountered in the entire survey area for each of the embayments (including adjacent coastal waters), B) the abundance of only those individuals encountered in the interior waters of each embayment, and C) estimates that account for transient animals that were observed in more than one stock area and are limited to a smaller subset of “residents” (i.e., animals seen in interior waters of only one embayment or its adjacent coastal waters in both summer and winter seasons). This identification of transient individuals suggests the nwGoMx estuaries and coastal waters of this study may be host to a metapopulation of bottlenose dolphins with some individuals that range between discrete geographic areas. The practical importance of these different approaches to abundance estimation is not only in the provision of baseline population information for each dolphin stock but also in providing insight into the connectivity of spatially and demographically separated populations.

Modeling Spatial Distribution

The data collected during photo-ID and other types of visual surveys may be paired with habitat information to examine the environmental and geographic factors influencing marine mammal habitat choices (Harvey et al., 2017; Keller et al., 2012;

Redfern et al., 2006). The concept of a species’ ecological niche is the theoretical underpinning for the species-environment descriptions resulting from species distribution models [SDMs, (Elith & Leathwick, 2009)]. The species ecological niche was first described as an n-dimensional environmental space (Hutchinson, 1957), in

which the behavior of animals is not driven by a single external factor, but by a synthesis of environmental factors operating simultaneously (Friedrich, 1969). A fundamental understanding of how to approach species conservation requires knowledge of intrinsic and external factors operating on the species of interest and includes population growth rates, foraging ecology, and environmental conditions necessary to support a viable population. Habitat models or SDMs correlate biological, environmental, and physiographic information with observed species location or abundance data to predict spatiotemporal population patterns, species vulnerabilities

(e.g., invasive species and disease), and to test hypotheses on factors influencing migrations and habitat use (Robinson et al., 2011). The relevance of any SDM is in its power to translate species-environment connectivity into geographic space, so SDMs should also include a biogeographic context (Elith & Leathwick, 2009; Miller, 2010).

Species distribution models may explain observations within the spatial and temporal framework of the study or extrapolate species habitat use into new areas and future climates, using data collected from previously studied environments. Ecologically relevant predictors for a given species are important parts of a model used for extrapolation and should be rooted in what is known of the biology of the subject species. Abiotic factors as proxies for biological information have proven useful to predict marine species biodiversity (McArthur et al., 2010; Torres et al., 2008). Habitat variables influencing biota include topographic complexity, habitat heterogeneity, vegetative density, and a wide variety of chemical, bathymetric, and geographic features (Bosch et al., 2018; Bouchet et al., 2015).

Population studies of marine mammals are conducted on different spatial and temporal scales and with differing objectives and availability of data for external factors.

Some researchers have suggested prey distribution is the most likely determinant of cetacean distribution, yet acknowledge that it is challenging to obtain prey distribution data that is paired with the spatiotemporal resolution of cetacean observations (Lambert et al., 2014; Redfern et al., 2006; Rogan et al., 2017). Species distribution models focused on estuarine common bottlenose dolphins found the presence of abiotic factors such as nitrate, salinity, and distance to shore as potentially influential predictors of dolphin presence [Mississippi Sound, (Pitchford et al., 2014)], and bathymetry (benthic slope and depth, e.g., deep channels with steep embankments) was noted as a significant factor influencing dolphin habitat use or foraging tactics off the Florida Gulf of

Mexico coast, the Moray Firth, and the Irish Shannon estuary (Allen et al., 2001; Hastie et al., 2003; Ingram & Rogan, 2002; Wells, 2019). Miller and Baltz (2010) found water temperature to be an overall more informative factor of bottlenose dolphin distribution and foraging activity in Barataria Bay, Louisiana. However, their analysis showed water depth, distance to shore, and salinity as additional significant factors influencing foraging choice, and noted dissolved oxygen had a positive correlation with bottlenose dolphin habitat suitability. Hornsby et al. (2017) found bottlenose dolphins tended to avoid waters less than ~5 ppt salinity in Barataria Bay and suggested the ~8 ppt isohaline is a reasonable predictive boundary for dolphin habitat selection there. Not all studies find reason to associate abiotic environmental variables with dolphin distribution.

In their three-year study of bottlenose dolphins in the Indian River Lagoon, Florida,

Mazzoil et al. (2008) found no statistically significant correlation between dolphin density and salinity or pollutant loads for the 615 individual dolphins identified in their study.

I modeled the relationship between dolphin distribution and density with archived hydrologic data and physiographic factors to determine the best predictors for bottlenose dolphins in the embayments of this thesis. My results are consistent with prior research indicating dolphins tend to aggregate in deep navigational channels and suggest water depth is a good predictor for estuarine dolphin distribution in areas where there are deep engineered waterways. The importance of these results is not only in the provision of baseline data supporting continued monitoring of this protected species, but also in the implication for potential human-induced concentration of dolphin density when navigational channels are dredged and the requisite management responsibilities that are incurred due to their construction and modification.

Figure 1-1. Northwestern Gulf of Mexico estuaries surveyed in this study: West Bay, Galveston Bay, and Sabine Lake.

CHAPTER 2 BACKGROUND AND METHODS

Survey Area Background

The geographic span of this study encompassed an ~1800 km2 area in the nwGoMx along coastal northern Texas and western Louisiana, spanning ~161 km from

Christmas Bay, Texas, to approximately 10 km east of the Texas / Louisiana border

(Figure 2-1). Regional climate has been characterized as humid and wet sub-humid with a mean annual average air temperature of ~21°C (Anderson et al., 2014; Fisher et al., 1972). Bay salinities vary seasonally, with lower salinities occurring in spring during periods of increased river inflow, and overall salinity ranges from ~0.5 to 30 ppt, with salinity generally decreasing with increasing distance away from each pass to the nwGoMx. The upper Texas coast is home to more than 7 million people and highly concentrated industries related to petroleum and chemical production. Shrimp, finfish, and shellfish are among the natural resources harvested from upper Texas waters and in total during 2017, some 36 million pounds of commercial fisheries landings were recorded at just two of the largest ports in the region [Galveston and Port Arthur,

(NMFS, 2019)]. The Gulf Intracoastal Waterway (GIWW), a dredged canal spanning much of the Gulf of Mexico coast, connects coastal bodies of water to facilitate protected inland commercial shipping interests. Industrial interests in the regions are dominated by dense clusters of chemical and petroleum production facilities, with 50% of total U.S. chemical production and 30% of petroleum production located around

Galveston Bay, alone (USEPA, 1999). Commercial and recreational fishing, cruise ship terminals, intermodal facilities, stormwater and wastewater runoff, and large container shipping vessels add to the list of potential anthropogenic threats to wildlife in the

region. Extensive human modification to the topography of the bays has created an artificial connection between the bays and barriers that have altered freshwater circulation and saltwater intrusion. As a result, the deep navigation channels associated with each Gulf pass have provided a mechanism for developing salt wedges that reach into the estuary via density currents. A density current is the “mean current directed from the mouth to the head of the estuary forced by the seaward gradient in salinity” and the difference in water density creates a horizontal pressure gradient that drives the higher salinity water into the estuary [(Ward, 1980), p. 197]. Density currents increase in proportion to the cube of water depth (Ward, 1980), therefore, deep-draft channels are ideal mechanisms for the transport of density currents and results in a “tongue” of salinity bisecting adjacent waters that otherwise do not have a vertical salinity gradient

(Orlando et al., 1993; Ward, 1980). Anthropogenic threats (e.g., ship strikes, oil spills, cumulative industrial pollution, fisheries interactions) exist for dolphins inhabiting the upper Texas coast, as reviewed by Phillips and Rosel (2014). All three research areas of this study—West Bay, Galveston Bay, and Sabine Lake, were scored as high research priorities for common bottlenose dolphins due to the anthropogenic threats

(Phillips & Rosel, 2014).

West Bay

West Bay (~150 km2, mean depth 1.5 m) is oriented southwest to northeast along

Galveston Island, and is ~30 km in length by 3.5–6 km wide. West Bay is part of the

Galveston Bay system but is nearly separated from Galveston Bay by the Texas City

Dike–an 8.5 km engineered structure designed to protect industrial infrastructure on the mainland shoreline of Galveston Bay from storm surges. A complex of small dredge spoil islands and oyster reefs on the northern end of West Bay provides additional

impediments to water currents. The GIWW passes through these topographical restrictions, however, freshwater from the Galveston Bay system tends to bypass West

Bay resulting in a marked difference in West Bay salinity that is chiefly moderated by tidal movement through the natural inlet of San Luis Pass (Orlando et al., 1993). San

Luis Pass is at the southeastern point of West Bay and is the primary point of sea water exchange with the nwGoMx. In West Bay the prevailing winds and local runoff tend to be the primary factors influencing salinity (Orlando et al., 1993).

Galveston Bay

Galveston Bay (~1400 km2, mean depth 2.5 m) is the largest bay on the upper

Texas coast (USEPA, 1999). The primary sources of freshwater input to Galveston Bay are the San Jacinto and Trinity rivers. The Trinity River provides approximately 83% of the freshwater inflow into Galveston Bay and is the main determinant for the bay salinity, which is generally lowest in spring and increases through the month of June

(Orlando et al., 1993). However, wind patterns and density currents are important secondary factors (Orlando et al., 1993). Bolivar Roads serves as a vessel anchorage and key pass from the Gulf of Mexico into Galveston Bay and intersects the Houston

Ship Channel (HSC), Galveston Ship Channel, and the GIWW. Density currents move through Bolivar Roads and the HSC, resulting in a salt wedge in the lower 40 km of

Galveston Bay (Orlando et al., 1993). Bolivar Roads is approximately 3 km wide and features granite rock jetties extending approximately 3 km into the nwGoMx from the nearest point of land; it is dredged to approximately 15 m to accommodate large vessels in transit to the Port of Houston through the HSC. The Galveston Ship Channel, a 365- m wide, 15-m deep engineered inlet, branches away from Bolivar Roads and leads to the Port of Galveston.

Sabine Lake

Sabine Lake (~240 km2, mean depth 2 m) is a brackish estuarine embayment bisected by the Texas–Louisiana border (USEPA, 1999). Sabine Lake receives the highest freshwater input per unit volume of any Gulf of Mexico embayment, primarily from the Sabine and Neches rivers flowing into its northern end (Ward, 1980). Orlando et al. (1993) attributed the dominant seasonal influence on salinity patterns in Sabine

Lake to the volume and duration of water flow in Sabine and Neches rivers. The river flow of the Sabine-Neches watershed is influenced by the Toledo Bend (Sabine River) and Sam Rayburn (Neches River) reservoirs that were built for hydroelectric power generation. Generally, increased inflows into the reservoirs occur during warmer months to meet increased energy demands, thus reducing the volume of freshwater reaching Sabine Lake with a resulting seasonal increase in salinity. Sabine Lake salinity is lower in winter and spring (December through May) and higher in summer and fall [June through November, (Longley, 1994; Orlando et al., 1993)]. The lake is adjacent to engineered channels on its southern and western boundaries: the Port

Arthur Ship Canal, Sabine-Neches Waterway, GIWW, and the Sabine Pass Channel–an energetic tidal inlet and the only nwGoMx access point into Sabine Lake. Sabine Lake was not always a brackish environment. Salt-water intrusion likely began during the late

1800’s when the Sabine Pass Channel was initially dredged for navigation through the offshore sand bar (Ward, 1980). By the 1960’s, saline water reached 65 km upstream in the Neches River, requiring construction of salt water barriers in the Neches River to protect the fresh water supply for nearby communities (TTWP, 1998; Ward, 1980). The

Sabine Pass Channel features granite jetties that extend its entrance to the nwGoMx

approximately 4 km from the Texas / Louisiana shorelines and has a total length of ~15 km as measured from the nwGoMx entrance to the southern boundary of Sabine Lake.

Visual Surveys

Survey Procedure

Photo-ID mark-recapture surveys were conducted from 7 m center-console boats equipped with one or two four-stroke outboard engines. Each survey team consisted of three individuals. All members visually scanned for dolphin groups 180° ahead of the research vessel, and individual operational duties included driving, photography, and recording data. Surveys were typically conducted in a Beaufort sea state 3 or less on predetermined survey routes at a speed of 28–32 km/h. A survey effort log (e.g., start and end time, weather conditions, effort status) was completed for each survey and the vessel track lines were recorded on a handheld global positioning system (GPS) receiver unit. The survey team was recorded in “on-effort” status when following the prescribed visual survey methodology. If those conditions were not met (e.g., transiting at higher speeds to or from the dock) the team was working in “off-effort” status.

When a dolphin, or group of dolphins was sighted on-effort, it was approached for data collection consistent with photo-ID methods described by Melancon et al.

(2011); data included date, start and end time, start and end GPS locations, environmental conditions, group size and composition (e.g., observations of calves), behavioral observations, and general notes. Photographs of dorsal fins were collected with digital single lens reflex (DSLR) cameras equipped with 100-400 mm telephoto zoom lenses. Photo-ID teams attempted to photograph every dolphin in each group regardless of dorsal fin distinctiveness, with passes on both left and right sides.

Occasionally, the team stopped to photograph dolphin groups observed while off-effort using similar protocols; images and data collected were recorded as such.

Survey Design

Small-boat photo-ID mark-recapture surveys were conducted in West Bay in

December 2014 and June 2015; Galveston Bay, January and July 2016; and in Sabine

Lake, February and June 2017 (Figure 2-1). Additional scouting or monitoring surveys were also conducted to ground-truth navigable survey routes or for periodic monitoring efforts from October 2014–April 2018. However, only data from the structured mark- recapture surveys during 2014-2017 were used to estimate abundance. Survey design followed Pollock’s “robust design” structure for mark-recapture surveys with “primary” seasonal sessions (i.e., winter and summer) and multiple “secondary” survey occasions within each season (Pollock, 1982). Within each primary session, the population is assumed to be “closed.” When closure assumptions are valid, it can be inferred that the data collected in each survey area are representative of each bay’s population during the study period. Two primary sessions were conducted in presumed non-transitional meteorological seasons [e.g., December-February in winter; June-August in summer

(Trenberth, 1983)]; however, transitional periods can vary geographically. A primary session consisted of three replicate secondary occasions, where each consisted of complete coverage of the survey area completed in as few days as possible, weather permitting. One “mixing day” was allotted between secondary occasions to allow time for marked and unmarked dolphins to mix spatially, as suggested in Rosel et al. (2011).

The duration of each primary session was kept as short as possible to help improve the likelihood that the closure assumption was met. The survey transect-line design (Figure

2-1) was non-randomly spatially stratified, based on several factors: maximizing

coverage of the study area, minimizing heterogeneity in capture probability, dolphin group density noted during exploratory surveys, anecdotal reports of dolphin sightings, and the presumed salinity gradient (i.e., salinity decreases as distance from the nwGoMx increases). The survey design for each embayment also included two parallel coastal transects. The first transect line was followed along the beach contour 300–500m from the beach and the second transect was placed ~2 km offshore; each transect extended

~5km or more in each direction away from the inlet. These coastal transects were included in the study design because telemetry tag and photo-identification data from other embayments within the Gulf of Mexico have demonstrated that resident estuarine animals may regularly utilize nearshore coastal waters (Balmer et al., 2008; Wells et al., 2017).

Survey route schedules were not temporally fixed, in that due to inclement weather or differences in the number of dolphin groups encountered on a given day, routes did not begin and end at the same time or at the same location. Thus, the detection of dolphin groups was not temporally correlated to a specific time or location. Sub-areas within each survey area were designated for fine-scale classification of dolphin group locations

(e.g., coastal waters, open bay waters, and sub-embayments). Topographic features were used to delineate the boundaries of the pass or channel entrance to each embayment separating bay and coastal waters. For example, the San Luis Pass sub- area was delineated as the waters between Cold Pass and a natural sandbar restricting ingress / egress to the nwGoMx. In the Galveston Bay and Sabine Lake survey areas, the Bolivar Roads and Sabine Pass Channel sub-areas were designated as the waters between the end of the channel weir jetty structures meeting the nwGoMx and the inner topographical opening to each embayment.

West Bay

Other ecological studies of the Galveston Bay system include West Bay as a sub-embayment of Galveston Bay. However, the NMFS has delineated the West Bay stock as a separate management unit from the rest of the Galveston Bay system

(Hayes et al., 2017) and previous research in West Bay suggests its population is comprised of a core resident group of animals with long-term site fidelity (Irwin &

Würsig, 2004; Maze & Würsig, 1999). West Bay surveys were conducted with one research vessel in December 2014 and June 2015 and included Chocolate Bay,

Bastrop Bay, Christmas Bay, Cold Pass, San Luis Pass, and the nwGoMx coast adjacent to San Luis Pass along Galveston and Follet’s islands. Previous bottlenose dolphin studies identified the southwestern end of West Bay as the area most likely inhabited by dolphins (Henderson & Würsig, 2007; Irwin & Würsig, 2004; Maze &

Würsig, 1999), however, I included the entirety of West Bay for a geographically comprehensive assessment. Survey transects across the middle of the bay were separated by approximately 3 km, with a single line or loop in the adjacent bay areas and connecting water ways. On the initial December 2014 survey, Christmas Bay and

Bastrop Bay were found to be too shallow for the research vessel. However, the water depth was acceptable in Christmas Bay during June 2015, so it was added for the summer sampling session.

Galveston Bay

Galveston Bay surveys were conducted with two research vessels in January and July of 2016, and included upper Galveston Bay, Trinity Bay, Galveston Bay, East

Bay, “Back Bay”, the Galveston Ship Channel, and Bolivar Roads. Galveston Bay bottlenose dolphin sightings and salinity levels (Ward & Armstrong, 1992) are generally

higher in the western and southern waters of the Galveston Bay system. Bottlenose dolphins are frequently observed in Galveston Bay’s deep channels and research suggests such navigational channels are important components of dolphin habitat

(Moreno & Mathews, 2018; Piwetz, 2019). Therefore, the highest concentration of animals was anticipated in the southern area of the bay. Survey transect lines were drawn to a finer scale in Bolivar Roads to maximize the capture probability for dolphin groups. Survey transect line spacing ranged from 1–1.25 km in the southern area of the bay. In the central section of the bay, transect line spacing ranged from 1.5 km in

Galveston Bay out to 5 km in East Bay. In Upper Galveston Bay, transect lines were spaced 1.5 km apart. Few dolphin encounters were expected in Trinity Bay due to historically low salinity levels and the low salinity (~1–2 ppt) found during preliminary scouting surveys. Due to the low expectation of dolphin encounters and resource limitations, the perimeter and inner transect lies were surveyed only once per season on the first sampling occasion; thereafter only the western-most lines were surveyed.

Sabine Lake

Sabine Lake surveys were conducted with one research vessel in February and

June of 2017. The Texas Marine Mammal Stranding Network has performed multiple out-of-habitat interventions or stranding responses for dolphins in fresh water in the outer reaches of Sabine Lake (Whitehead & Ronje, 2017) and reports archived from the general public indicate groups of dolphins use the navigation channels and rivers.

Therefore, survey routes were planned in the Sabine Pass Channel, Port Arthur Ship

Canal, Sabine-Neches Waterway, and Neches and Sabine rivers. Survey line spacing in the open water of Sabine Lake was graduated from ~1–2 km to concentrate survey effort with the presumption of encountering more dolphin groups closer to the Sabine

Pass Channel, where surface salinity would likely be greater (Longley, 1994). Upper lake lines were surveyed in alternate sessions, so that each upper lake line was surveyed by the end of the primary session.

Photo Analysis

Dorsal fin marks were assumed to be unique and not mismatched or misread during subsequent photo analysis. Bottlenose dolphins accumulate nicks and notches over time and the long-lasting nature of the marks is well established (Würsig & Würsig,

1977). Primary sessions were completed in approximately two weeks, and while new marks could have been acquired between primary sessions, or during the multi-year period over which these surveys were conducted, it was unlikely changes in marks would alter the fin beyond recognition (Urian et al., 2015; Williams et al., 2002). Photos were processed for primary matching and verification similar to the protocols outlined in

Melancon et al. (2011) and combined with their associated data to create a catalog of bottlenose dolphin dorsal fins for northern Texas / western Louisiana embayments

(NorTex catalog). Photo analysis was done with the aid of FinBase, a software package designed for bottlenose dolphin photo-ID that consolidates data tracking, image analysis, and a multiple-attribute catalog sorting algorithm to expedite photo analysis and generate encounter histories for use in mark-recapture studies (Adams et al.,

2006). Primary matching and verification were conducted by at least two experienced technicians. Photo quality scores and dorsal fin distinctiveness grades were assigned to each photo to determine their suitability for analysis and ranged from a photo quality

(PQ) score of PQ 1–3, with PQ 1 and 2 considered acceptable. Photo quality is cumulatively scored for each image in FinBase using five categories of quality: focus, contrast, angle, partial (e.g., a dorsal fin partially submerged), and distance.

Distinctiveness levels assigned to each dorsal fin were classified similar to Urian et al.

(2015), as follows: D4 (not distinct)–no useful mark information (≤ 1 small nick / tiny notch); D3 (low)–few and / or small marks (e.g., 2 small nicks); D2 (average)–1 or more permanent marks or notches; D1 (high)–major, prominent feature(s) unlikely to be mistaken even in poor quality photos. Only those individuals classified as D1 or D2 and

PQ1 or PQ2 were used for mark-recapture abundance estimation. After photo analysis, discovery curves for each survey area were constructed to visualize the newly marked individuals encountered, the recaptured individuals, and the cumulative total of individuals for each survey area by as they were added to the NorTex catalog.

Discovery curves were representative of the total number of distinctive (D1 and D2) individuals encountered in mark-recapture surveys.

Figure 2-1. Northwestern Gulf of Mexico survey areas. Bold black lines represent predetermined photo-ID mark-recapture survey routes in West Bay (2014-2015), Galveston Bay (2016), and Sabine Lake (2017).

CHAPTER 3 ABUNDANCE ESTIMATION

Estimating Abundance and Identifying Potential Transients

Photo-ID studies of bottlenose dolphins indicate some dolphins exhibit long-term site fidelity of specific embayments (Wells, 2014) and individual movements may range from within an estuary to adjacent passes and coastal waters (Fazioli et al., 2006; Irvine et al., 1981; Laska et al., 2011). Migratory dolphins from other stocks may travel along the coast and spend time within the stock boundaries of a different population (Maze &

Würsig, 1999; Speakman et al., 2010; Urian et al., 2018; Wells et al., 1987). Including seasonal transient individuals in the abundance estimate may over-estimate the size of a given stock, with potential implications for management decisions. Therefore, for each of the three embayments in this study, I used a simplified approach to potentially identify transient animals by parsing encounter histories of individual dolphins by spatiotemporal parameters before estimating the abundances of each survey area.

Encounter histories were a subset for each survey area as follows:

A. Includes all animals and their re-sights during the mark-recapture surveys and the spatial boundaries of the survey area (bay and coastal waters)

B. Includes only individuals encountered within the boundary of the embayment and channel connecting to the nwGoMx (i.e., excludes all initial observations and re- sights of animals encountered in coastal waters)

C. Includes groups encountered within the boundary of the embayment and channel connecting to the nwGoMx, individuals encountered in coastal waters that were observed in both seasons, and excludes individuals found to use more than one stock management area (inter-bay matches). The data for subset C were derived by using the NorTex FinBase catalog to query inter-bay dorsal fin matches (e.g., between Sabine Lake and Galveston Bay ~90 km to the southwest) from all group encounters. If a match was confirmed, the fin ID was removed from the dataset used to estimate the abundance. To exclude additional transient animals, data were filtered to exclude all individuals observed only in coastal waters and only in one season using a custom python script in ESRI ArcGIS Pro 2.4.1 (ESRI, 2019). The delineation of the coastal waters

began at the mouth of the engineered channel (or natural pass in the case of West Bay) where the coastal waters enter the estuary.

Model Selection

Estimating abundance may also be described as an estimation of the parameters of capture (p) and recapture (c) probability in a population, from which an estimate of abundance (N̂ ) is derived. After parsing the data into the type A–C datasets, encounter histories for each primary session (winter or summer) were analyzed in program MARK

9.0 (White & Burnham, 1999) using the closed capture “Huggins’ p and c” conditional likelihood approach (Huggins, 1991). The Huggins’ p and c approach is conditional only on the number of animals encountered, requiring only two parameters to estimate and is suitable for sparse data. The “full-likelihood” approach [an approach that accounts for animals in the population that were not encountered by including their probabilities of capture in the likelihood (Darroch, 1958)] was also explored but models did not consistently converge for the smaller data sets. Using the Otis et al. (1978) notation, the closed capture models considered under each approach to estimate abundance were:

• Mo: p(.) = c(.)–time constant capture probability

• Mb: p(.), c(.)–behavioral response influences capture probability and therefore initial capture probability (p) may differ from subsequent recapture probability (c)

• Mt: p(t) = c(t)–time variable capture probability allowing differences in recapture probability between secondary sampling occasions

Individual heterogeneity [Mh (Chao et al., 1992)] was not modeled due to the low number of sampling occasions in this study [3–4 sampling occasions per primary in this study, minimum of 5 or 6 recommended (Conn et al., 2006; Lukacs, 2001)]. Model results were evaluated using Akaike’s Information Criterion (Akaike, 1973), corrected for

small sample sizes (AICc) within Program MARK, and the most parsimonious model was selected as the model with the lowest AICc score. If the difference (∆AICc) in one or more models was ≤ 4 AICc points of the minimum score, those model outputs were combined using the weighted model averaging function within program MARK.

Correction for Unmarked Fins

Closed capture models only estimate the number of marked individuals and correction is needed to account for the number of unmarked individuals in each sample.

A corrected estimate of abundance (N̂ c) to account for both marked and unmarked individuals was calculated by dividing N̂ by the proportion of marked fins. Because the proportion of distinctive fins can change between survey areas and within survey areas across seasons (Balmer et al., 2019), seasonal θ values were calculated for each of the three survey areas. As with previous studies, theta (θ) was derived by calculating the proportion of distinctive fins (D1 + D2) out of all fins (D1+D2+D3) that met photo quality criteria (i.e., PQ1 and PQ2) and that were collected during on-effort group encounters

(Read et al., 2003). Subsequently using θ, I calculated a corrected estimate for total abundance: N̂ c = N̂ / θ for each winter / summer primary. The variance and confidence intervals of the corrected estimates were calculated using the delta method (Speakman et al., 2010; Wilson et al., 1999).

Results

Survey Effort and Dolphin Group Encounters

In total, 91 small-boat photographic identification surveys covering ~1,800 km2 comprising ~11,000 km of track-line were conducted in ~ 608 hours during summer and winter seasons from 2014–2018 in West Bay (WB, n = 25), the Galveston Bay system

(GB, n = 49), Sabine Lake (SL, n = 17), and including the adjacent nwGoMx coastal

waters (Table 3-1). Overall, photos were collected from 404 dolphin groups: WB (n =

52), GB (n = 260), and SL (n = 92). Generally, bottlenose dolphin groups were encountered more often in or near to the deep channels and passes to the nwGoMx, particularly during the winter (Figures 3-1 to 3-3).

Photo Analysis

Approximately 36,000 photos of bottlenose dolphin dorsal fins were collected and sorted for the identification of individuals. In total, 1,271 individuals were identified during mark-recapture surveys. Of those, 1,113 (n = 88 West Bay, n = 736 Galveston

Bay, n = 289 Sabine Lake) were well-marked (D1 + D2, >PQ3) individuals used for abundance estimation. By survey area and season (winter / summer), the number of individuals (and proportion distinct) were: West Bay–40 (0.82) / 72 (0.86), Galveston

Bay–336 (0.76) / 539 (0.73), Sabine Lake–130 (0.61) / 185 (0.84) (Table 2). Mean catalogued group sizes (SE) were 10.2 (1.4), 8.4 (0.5), and 7.9 (1.0) for West Bay,

Galveston Bay, and Sabine Lake, respectively. The discovery curves indicate a greater number of marked individuals were initially encountered for all three survey areas during the summer surveys (Figure 3-4).

In total, 40 inter-bay matches were found, all observed in the Galveston Bay survey area: 15 / 40 individuals for West Bay and 25 / 40 individuals for Sabine Lake.

Inter-bay matches represented 15.2% (15 / 99) of individuals in West Bay, 4.7% (40 /

843) in Galveston Bay, and 7.6% (25 / 329) in Sabine Lake. No inter-bay matches were found between the West Bay and Sabine Lake survey areas, yet individuals from both populations were plotted well into the Galveston Bay stock area (Figure 3-5). For example, dolphin #2 (Table A-1) was observed in West Bay in the winter / summer survey seasons 2014-2015, observed in Galveston Bay during the summer of 2016,

then re-sighted in West Bay during the spring of 2017 and 2018. Similar movement patterns were observed for the other inter-bay matching individuals. In total, 150 observations of inter-bay matching individuals were recorded: 42 (28%) in coastal waters, 49 (33%) in the pass or channel entrance to each bay (San Luis Pass, Bolivar

Roads, Sabine Pass Channel), and 59 (39%) inside the interior of each embayment (not including the pass or channel entrance).

Abundance Estimates

Model selection varied by survey area and data subset. The Mt model was consistently supported by West Bay data, while a combination of the Mt model and model averaging was used for Galveston Bay and Sabine Lake, depending on the season. The AICc results consistently supported the Mt model for winter Galveston Bay data subsets and model averaging of all three models (Mo, Mb, Mt) for the summer data subsets. Sabine Lake AICc results indicated support for the Mt model for all data subsets except for the winter type B data subset which was model averaged between all three models (Mo, Mb, Mt). The Mb model failed with certain datasets (West Bay winter data, Sabine Lake type C winter and all summer data). For all survey areas, the estimates based on the type C data subset are preferred because potential migratory and transient animals are removed from consideration, yet animals with at least dual- season site fidelity to the coast contribute to the population estimate. Hereafter, results will describe the “best” abundance and parameter estimates based on the model selected for the type C data subset. Table 3-2 summarizes the number of encounter histories, abundance estimates, models selected, and distinctiveness proportions for all approaches and data subsets. Figure 3-6 illustrates the estimate and 95% CI for each

survey area using the best estimate, and Tables B-1 and C-1 contain the AICc results and parameter estimates of the models selected, respectively.

West Bay

Capture probabilities for each sampling occasion ranged from 0.16–0.67 in winter and 0.13–0.89 in summer. The best estimates corrected for unmarked individuals were

37.9 (95% CI = 28.7–47.1) and 36.5 (95% CI = 32.7–40.3) for the winter / summer primary sampling sessions. Respectively, for winter and summer: applying the type B constraint (removing all coastal observations) to the type A data (all individuals encountered during each mark-recapture survey period) resulted in ~18% (33 from 40) and ~54% (33 from 72) fewer individuals. Applying the type C constraint (removing individuals observed only on the coast and in only one season, and individuals matched to more than one survey area) to the type A data resulted in ~33% (27 from 40) and

~57% (31 from 72) fewer individuals. Of those individuals removed with the type C constraint, inter-bay matches to Galveston Bay were 9 and 5 (winter / summer), with 2 of those present in both winter and summer mark-recapture primaries in West Bay.

Galveston Bay

Capture probabilities for each sampling occasion ranged from 0.11–0.19 in winter and 0.22–0.23 in summer. The best estimates corrected for unmarked individuals were

841.9 (95% CI = 693.5–990.4) and 1131.5 (95% CI = 846.3–1416.7) for the winter / summer primary sampling sessions. Respectively, for winter and summer: applying the type B constraint to the type A data resulted in ~10% (301 from 336) and ~14% (464 from 539) fewer individuals. Applying the type C constraint to the type A data resulted in ~7% (311 from 336) and ~%18 (440 from 539) fewer individuals. Of those individuals

removed with the type C constraint, inter-bay matches to Galveston Bay were 11 and 28

(winter / summer), with 3 of those present in both mark-recapture primaries in Galveston

Bay.

Sabine Lake

Capture probabilities for each sampling occasion ranged from 0.15–0.39 in winter and 0.15–0.37 in summer. The best estimates corrected for unmarked individuals were

121.6 (95% CI = 73.0–170.3) and 162.2 (95% CI = 114.3–210.2) for the winter / summer primary sampling sessions, respectively. Respectively, for winter and summer abundance estimation: applying the type B constraint to the type A data resulted in

~85% (20 from 130) and ~69% (58 from 185) fewer individuals. Applying the type C constraint to the type A data resulted in ~66% (44 from 130) and ~57% (79 from 185) fewer individuals. Of those individuals removed with the type C constraint, inter-bay matches to Galveston Bay were 16 and 3 (winter / summer), with 2 of those present in both mark-recapture primaries in Sabine Lake.

Discussion

These results provide updated population estimates for each of three nwGoMx embayments. I estimated abundances for three bottlenose dolphin stocks in the nwGoMx using the Huggins’ p and c approach with three different models and found the

Mt model or a combination of averaged models best applied across these sparse data.

The Mb model failed for several of the datasets, particularly for winter West Bay and summer Sabine Lake data, likely due to a combination of low capture probability (e.g., p

= 0.06–0.15 for the summer Sabine Lake type A subset, Table C-1), few sampling occasions (<5), and heterogeneity in individual capture probabilities (White & Cooch,

2017). Ideally, more sampling occasions would be planned for future studies of these

areas to provide enough data to justify modeling additional parameters (e.g., individual heterogeneity and survival). The large spatial range of my survey areas (particularly

Galveston Bay) made additional sampling occasions impractical, however, the distribution of dolphin groups presented here and in other studies, particularly for West

Bay, may guide future photo-ID mark-recapture studies to be directed at areas where dolphin encounters are more likely, allowing for more survey iterations within each primary [5 + iterations (Conn et al., 2011; Lukacs, 2001)].

The “type C” parsed dataset reduces the probability of including non-resident animals in my estimates for each embayment because it excludes any animal seen in more than one embayment and animals seen primarily in coastal waters. It is possible the same individual transient dolphins were observed in coastal waters in both primaries, however, if so, those individuals were using the nearshore coastal waters in a minimum of two non-contiguous seasons, showing at least some site fidelity to a single survey area. Except for Sabine Lake, population estimates between the type A, B, and

C subsets were relatively similar during winter. Sabine Lake was exceptionally different from other survey areas, in that not only was the total population estimate much higher than the estimate based on the type B and C data subsets in both seasons, but the type

B subset estimate was lower than the estimate based on type C parsed data—less

~53% / 30 %, winter / summer, respectively. This suggests that there are dolphins

(potentially the majority of dolphins in winter) with site-fidelity to the Sabine Lake survey area that prefer to use those nwGoMx waters near the end of and adjacent to the

Sabine Pass Channel, which is not surprising given that Sabine Lake salinity is typically

lowest in winter due to increased freshwater riverine flow bypassing upstream hydroelectric reservoirs draining into Sabine Lake (Orlando et al., 1993).

From 1997–2000, photo-ID surveys were conducted in West Bay, where the population was estimated at 28–34 resident dolphins or 28–38 including nonresident dolphins (Irwin & Würsig, 2004). The research conducted by Irwin and Würsig (2004) was supported by long-term studies of West Bay dating to 1990, leading them to conclude the carrying capacity of the West Bay habitat to be at approximately 30 dolphins. Litz et al. (2019) estimated abundance for West Bay using an earlier variation of the West Bay study data used here. The method Litz et al. (2019) used to parse potential nonresidents from the presumed resident population involved a thorough review of individual sighting histories and locations, but without inter-bay match information. The best West Bay estimate from Litz et al. (2019) was 50.6 / 44.4 in winter / summer, respectively, and the differences between those estimates and my

West Bay abundance estimates are likely due to my exclusion of probable transient dolphins observed using the combination of West Bay and Galveston Bay data. This study also excluded “off-effort” sighting data for abundance estimation (a difference of 1 winter and 6 summer group encounters), resulting in different proportions of marked animals (e.g., Litz et al. (2019) winter / summer θ = 0.90 / 0.95 to this study’s 0.82 /

0.86). The best corrected estimate of the West Bay population presented here (N̂ c =

37.9 / 36.5, winter / summer, respectively) is consistent with the range of estimates from

2004, suggesting the West Bay population remains relatively stable close to the carrying capacity (30) estimated by Irwin and Würsig (2004).

The NMFS conducted line transect aerial surveys from 1983–1984 and 1992–

1993 during all seasons over Sabine Lake and the Galveston Bay system resulting in calculations of only 0–2 dolphins for Sabine Lake and 152 dolphins for Galveston Bay

(Blaylock & Hoggard, 1994; Scott et al., 1989). Henningsen and Würsig (1991) conducted visual small-boat surveys in the general Galveston Bay area including coastal waters and identified 1,002 individuals, however, only 135 were re-sighted during their study (April-October 1990), primarily in the bay. Bräger (1993) estimated about 200 dolphins to use Galveston Bay year-round, based on 97 surveys conducted from June through November 1991. The count of 1,002 distinct individuals in Galveston

Bay (Henningsen & Würsig, 1991) is within the range of best estimates calculated in this study (N̂ c = 841.9 / 1131.5, winter / summer, respectively) and clearly much higher than the estimate of Galveston Bay “resident” dolphins from previous estimates. This study has several important differences relative to earlier Galveston Bay studies that may be responsible for the large differences. For example, the estimates of “resident” individuals provided by Henningsen and Würsig (1991) and Bräger (1993) were a reflection of the distinctive individuals present in their respective catalogs observed in all seasons, while this study focused on the estimate of animals using the habitat of each survey area during each of the winter and summer seasons without comparisons of site fidelity across seasons. Additionally, the abundance estimates were adjusted to account for unmarked dolphins, thus further contributing to the higher estimates of abundance. Most dolphin groups sighted in the Sabine Lake survey area during this study were in the Sabine Pass Channel and nearshore coastal waters and those factors are likely responsible for the large difference in my abundance estimates (N̂ c = 121.6 /

162.2, winter / summer, respectively), where NMFS aerial survey crews were primarily assessing Sabine Lake proper and not the Sabine Pass Channel and adjacent coastal waters.

The distribution of the inter-bay matches identified here also corresponds in-part to natural and engineered deep channels of San Luis Pass and the GIWW on the mainland side of West Bay, the Galveston Ship Channel, Houston Ship Channel,

Bolivar Roads, and the Sabine Pass Channel. The presence of dolphins, a federally protected species, in high-traffic areas may have implications for future coastal engineering projects, particularly in winter, when dolphin group density in the coastal inlets of this study appeared greater. For example, in recent years, Bolivar Roads (the channel between Galveston Bay and the nwGoMx) has been a focal point in a broader plan by the state and federal government to protect the economic interests of the community and industry making use of Texas shorelines (USACE, 2018). The tentatively selected plan (TSP) includes a sea gate that will potentially alter or close

Bolivar Roads as a storm-surge protection device for oil refineries and communities near Galveston Bay (USACE, 2018). The impacts of large scale human modifications to Bolivar Roads (e.g., the TSP) or other bay access points on bottlenose dolphin should be considered (Ronje et al., 2018). The matches identified in different survey areas in this study all had the Galveston Bay survey area in common, and notably, there was no crossover in matches between West Bay and Sabine Lake. Tyson et al. (2011) and Urian et al. (2013) posited community boundaries for adjacent dolphin populations in their study areas by identifying a geographical boundary line on which the fewest number of overlapping dorsal fin matches were found. It is possible Bolivar Roads is

the location of a similar type of community boundary, yet also serving as a potential mixing area or functioning as a “hub” for the populations of West Bay, Galveston Bay,

Sabine Lake, and potentially other populations occurring primarily in nearshore coastal waters.

Long-distance movements of individual dolphins or groups of dolphins to multiple embayments have been noted in the north-central Gulf of Mexico (Balmer et al., 2016;

Ronje et al., 2017), and it is possible such dolphins are part of three Gulf of Mexico bottlenose dolphin coastal stocks, (i.e., Northern, Western, and Eastern Coastal stocks) delineated by the NMFS (Waring et al., 2016). The Western Coastal Stock range extends from the Mississippi River Delta to the Texas-Mexico border from the coastal shoreline out to the 20-m isobath in the Gulf of Mexico. Few studies have identified bottlenose dolphins belonging specifically to the Western Coastal Stock, however, the animals found to match in more than one survey area in this study are potential members of the Western Coastal Stock, or some other population. The importance of transient individuals from other stocks mixing with the three discrete population areas studied here is in the potential for genetic exchange or transmission of communicable disease between these populations. If functioning in this manner, the three population areas studied here may fit the description of a metapopulation, defined as “a system of discrete local populations, each of which determines its own internal dynamics to a large extent, but with a degree of identifiable and nontrivial demographic influence from other local populations through dispersal of individuals” (Kritzer & Sale, 2004). Photo-ID mark-recapture techniques can provide insight into metapopulations of dolphins occupying discrete geographic areas. For example, Chabanne et al. (2017) applied

multistate capture-recapture robust design to characterize the metapopulation structure of Indo-Pacific (T. aduncus) subpopulations in Western Australia occupying habitats in three distinct geographic sites. Using four years of year-round mark-recapture survey data, Chabanne et al. (2017) estimated site-specific capture probabilities, abundances and survival rates, and transition probabilities to characterize the metapopulation dynamic of the dolphins in each site. Similar studies in the nwGoMx could further elucidate the movement patterns of dolphins in this study region and provide insight into the significance of inter-population demographic influence between the three embayments studied here and potentially the Western Coastal Stock. The methods of

Chabanne et al. (2017) may be challenging (resource-intensive) to duplicate for the larger study area of this thesis, because the multistate mark-recapture survey design required that each of their three sites were sampled within the same primary session.

However, if the survey routes for the embayments of this thesis were limited to areas where dolphin group encounters were more likely (e.g., a route limited to the areal nwGoMx pass to each embayment), significantly less resources would be required to complete a primary sampling session in each survey area, and would potentially open up the possibility for three boats (one in each survey area) to sample each stock area simultaneously to collect the requisite data.

Limitations

In Galveston Bay it was not uncommon for research crews to “lose” dolphins near the HSC during a group encounter to cargo or tanker vessels when dolphins would leave the research vessel to bow-ride on the passing vessel (Würsig, 2009). Likewise, it was possible to “gain” dolphins that would quit bow-riding on a vessel passing our research vessel and merge with a group near us. A method to account for the

heterogeneity resulting from such behavior was not determined. It was often not clear if, or when, dolphins would drop off a vessel and join a sighting in-progress, and if it was noticed, the number of dolphins involved was usually unknown. In some cases, it was very clear our research vessel had lost a group of dolphins to a passing vessel before adequate photo-ID of individuals was collected. Some researchers have suggested dolphins are preferentially seeking out bow-riding opportunities to gain an advantage in the energetic cost of traveling long distances (Williams et al., 1992), but others feel this is unlikely and suggest bow-riding activity is merely for amusement (Würsig, 2009). In any case, if dolphins are traveling long distances across Galveston Bay (perhaps even entering or leaving the bay system temporarily on the bow of exiting vessels), such travel back and forth through Galveston Bay and Bolivar Roads and the resulting gains and losses of dolphins from group encounters may be worth considering in assessing how to approach future abundance and habitat assessments.

Table 3-1. Summary of 2014–2018 common bottlenose dolphin photo-ID survey effort. “Total catalogued dolphins” includes images of all distinctive and marginally distinctive (D1 + D2 + D3) dorsal fin images meeting minimum photo quality criteria (PQ1 + PQ2). Survey Mean Total Survey Survey Area Effort Surveys Groups Group Catalogued Hours (km) Size Dolphins West Bay 130 2,127 25 52 10.2 99 Galveston Bay 348 6,557 49 260 8.4 843 Sabine Lake 130 2,524 17 92 7.9 329 Total 608 11,208 91 404 8.8 1,271

Table 3-2. Bottlenose dolphin abundance estimates for West Bay, Galveston Bay, and Sabine Lake. Set = Parsed encounter histories, n = number of marked individuals in NorTex catalog for each Set, N̂ = estimated abundance, W = winter, S = summer, N̂ c = corrected abundance estimate. Multiple models indicate the estimate was derived using model averaging. Uncertainty was not estimated for the “% Marked.”

Model n N̂ (SE) % Marked N̂ c (95% CI) Set W S W S W S W S W S West Bay

A (Mt) (Mt) 40 72 44.5 (3.0) 79.7 (4.2) 54.3 (43.6-65.0) 92.7 (79.7-105.7)

B (Mt) (Mt) 33 33 38.1 (3.5) 33.6 (0.9) 0.82 0.86 46.5 (35.7-57.3) 39.1 (34.9-43.3)

C (Mt) (Mt) 27 31 31.1 (3.1) 31.4 (0.7) 37.9 (28.7-47.1) 36.5 (32.7-40.3) Galveston Bay

A (Mt) (Mo, Mb, Mt) 336 539 686.3 (55.5) 1101.4 (111.9) 903.0 (749.9-1056.1) 1508.8 (1198.5-1819.1)

B (Mt) (Mo, Mb, Mt) 301 464 646.8 (58.6) 918.3 (134.1) 0.76 0.73 851.1 (691.5-892.2) 1258.0 (892.2-1623.8)

C (Mt) (Mo, Mb, Mt) 311 440 639.9 (54.1) 826.0 (104.0) 841.9 (693.5-990.4) 1131.5 (846.3-1416.7) Sabine Lake

A (Mt) (Mt) 130 185 288.8 (48.9) 613.1 (113.7) 473.4 (303.4-643.4) 729.9 (460.7-999.1)

B (Mo, Mb, Mt) (Mt) 20 58 34.8 (16.7) 94.7 (14.9) 0.61 0.84 57.0 (2.8-111.2) 112.7 (77.2-148.2)

C (Mt) (Mt) 44 79 74.2 (14.2) 136.3 (20.1) 121.6 (73.0-170.3) 162.2 (114.3-210.2)

Figure 3-1. West Bay seasonal bottlenose dolphin group observations and survey track-line.

Figure 3-2. Galveston Bay seasonal bottlenose dolphin group observations and survey track-line.

Figure 3-3. Sabine Lake seasonal bottlenose dolphin group observations and survey track-line.

West Bay

New Recaptured Cumulative

100

Distinctive Distinctive Individuals 0 s1 s2 s3 s1 s2 s3 A

Galveston Bay 800 700 600 500 400 300 200 100 0 s1 s2 s3 s4 s1 s2 s3 B

Sabine Lake 350 300 250 200 150 100 50 0 s1 s2 s3 s1 s2 s3 C

Figure 3-4. Discovery curves of estimated individuals (D1+D2, PQ1 + PQ2). Secondary sampling occasions designated with “s1” at the beginning of each seasonal primary. A) West Bay, B) Galveston Bay, C) Sabine Lake.

Figure 3-5. Sighting locations for bottlenose dolphin inter-bay matches.

Figure 3-6. Best estimated abundance corrected for unmarked animals and 95% CI for each survey area by seasonal primary session.

CHAPTER 4 HABITAT ANALYSIS

Modeling Dolphin Distribution and Abiotic Factors

In this chapter, summer survey effort and dolphin count data were examined in context with water quality and physiographic variables with a geographic information system (GIS), a dorsal fin catalog, and statistical models. The aims were to describe dolphin group spatial distribution, assess the density of bottlenose dolphin groups relative to habitat characteristics, and provide maps of the observed and model- predicted bottlenose dolphin density. Habitat covariates included water quality data

(temperature, pH, dissolved oxygen, salinity) and bathymetry (water depth and sea floor slope) and the proximity of dolphin group locations to the entrance of the Gulf pass or barrier island of each embayment.

Water Quality

Water quality data was not available for the coastal waters, so the habitat analysis was limited to the survey data collected in the interior of each embayment.

Insufficient data were available from the period of my surveys, so older data were used as proxies. Water quality data were obtained from the Texas Water Development

Board (TWDB, 2019) and consisted of data collected in June 1992 for West Bay, May

1989 for Galveston, and June of 1990 and 1996 for Sabine Lake during intensive hydrographical surveys. Water quality data were a combination of manual boat-based collections at pre-determined stations and automated fixed datasondes (Figure 4-1A and 4-2A, Table 4-1 and 4-2). Data were collected approximately hourly and at 4–5 different depths in the water column for a vertical water quality profile, ranging from

approximately 0.3 m beneath the surface to a depth of up to 20 m in dredged navigational channels (e.g., Table 4-3).

I attempted to assess whether previously measured water quality data could be expected to be relevant for the 2015–2017 period of my surveys. Orlando et al. (1993) suggested an analysis of salinity (and presumably other water quality data) for an estuary be considered in the context of recent changes to the estuary or antecedent events influencing the system during the study period (e.g., increased water flow from upstream dams or storms) and suggested a representative time period of 3 months for analysis. To determine if the mean water quality parameters for each survey area used in this analysis were temporally consistent with long-term data of the same season

(May-July), data from 2 fixed datasondes (station IDs: BOLI, Bolivar Roads, Galveston

Bay; SAB2, upper Sabine Pass Channel, Sabine Lake) were temporally split in the R statistical software [openair package (Carslaw & Ropkins, 2012)], and the mean value of each covariate was tested for statistically significant differences between data from

May-July 1990 (used as a proxy to represent the water quality for each survey area) and data from May-July during 1991–2005. Additionally, estimated BOLI salinity for

May-July 1981–2015 obtained from the TWDB “TXBlend” salinity model (TWDB, 2019) was tested against the 1990 BOLI data. A Shapiro-Wilk statistical test (R Core Team,

2018) was applied to the data of each covariate (temperature, pH, dissolved oxygen, and salinity) for each survey area to test for normality and results indicated non-normal distributions (p < 0.05), therefore, I used a non-parametric Mann-Whitney test (α = 0.05) to determine if there were significant differences in means.

GIS Data Preparation

GIS data preparation and analysis was completed in ESRI ArcGIS Pro 2.4.1

(ESRI, 2019). Water quality data were imported into with each station as a point feature and water quality parameters from each station were interpolated between stations and over the survey areas using an inverse distance weighted (IDW) interpolation. The IDW interpolation calculates a weighted mean; weights are assigned to the given point data as a function of distance, with less weight applied toward more distant points. Mean predicted values are calculated in between the locations of water quality stations using a line-of-sight method and cannot be less than the minimum or more than the maximum observed data (O'Sullivan & Unwin, 2010). The interpolated water quality values were estimated with respect to the topographic restrictions to water circulation. Polyline barriers were traced to mirror the topography of the embayment and provided as input to the IDW tool, preventing measured values that would otherwise not have line-of-sight from being interpolated with each other across the dredge spoil islands. Exceptions allowing line-of-sight include the eastern side of Sabine Lake that could “see” those in the Sabine Pass Channel to provide more data to improve the estimates (see polyline boundaries in Figure 4-1A and 4-2A). Bathymetry data were extracted from a 2014 ninth arc second topobathy digital elevation model (DEM) obtained from the NOAA

Office for Coastal Management (NOAA, 2019b). Distance from the location of each dolphin group was measured as the geodesic distance (m) between the centroid of the grid cell in which each group was first sighted and a fixed point at the end of the Gulf pass (for Sabine Lake) or to the coastal barrier (West Bay and Galveston Bay) using a custom polyline and the Near Tool (geodesic distance was measured “as the crow flies”). The Zonal Statistics as Table tool was used to extract mean water quality values

and degree of slope from their respective rasters. Maximum depth was used instead of the mean, because the narrow width of the navigational channels and the proximity of some dolphin groups to the shoreline resulted in some land included, thus invalidating a calculation of mean water depth. See Figures 4-1B–4-1G and 4-2B–4-2G for the visual representation of the interpolated covariates and bathymetry.

All data were projected to the spatial reference of NAD 1983 UTM Zone 15N.

The spatial extent of survey effort track-line and bottlenose dolphin group data were examined in context with the geographic locations of water quality stations. Those data

(track-line and dolphin group locations) not within the spatial boundaries of the water quality stations were removed from the analysis as habitat information could not be accurately extrapolated to geographic space beyond the perimeter formed by the interpolation. Survey effort during visual surveys was denoted as “on-effort” if the survey team was actively searching for dolphin groups while surveying transects. Track distance recorded while transiting to and from the dock or between survey areas was considered “off-effort” and contributed significantly to the total recorded track distance

(30% Sabine Lake, 21% West Bay, and 28% Galveston Bay). Therefore, to avoid diluting the density estimates, on-effort track-line was intersected with the fishnet grid.

The number of photo-identified dolphins in the NorTex catalog from each grid cell were divided by the summer distinctiveness proportion determined in the abundance estimation (West Bay = 0.86, Galveston Bay = 0.73, Sabine Lake = 0.84) to account for unmarked dolphins. Mean values of interpolated water quality parameters

(temperature, pH, dissolved oxygen, salinity), mean slope, and maximum depth were

joined to the fishnet grid containing the survey track-line effort and dolphin group data, similar to McBride (2016), (Figures 4-3 and 4-4).

Habitat Modeling

GIS data were exported into R-3.5.1 statistical software for analysis (R Core

Team, 2018)]. Predictor variables in ecological data sets are often highly correlated

(e.g., water temperature and depth), and when collinear can skew parameter estimates and bias the results of a regression analysis. Therefore, collinearity was assessed using a variance inflation factor (VIF) test [corvif function, (Zuur et al., 2009)]. In an iterative process, the covariate with the highest VIF was sequentially dropped followed by a recalculation of VIF, until all covariates were approximately at a conservative value of 3 or less (Zuur et al., 2010). A second check for collinearity on the remaining covariates was done with a Pearson’s product-moment correlation test [cor.test function, (R Core Team, 2018)] using a threshold of 0.7 for removing covariates from the analysis (Dormann et al., 2013).

Species distribution models may be fit using various methods but may perform best when built on a foundation of presence-absence data, rather than presence data alone. Presence-only data may be defined as information for which there are no (or insufficient) data for where species are absent. Presence-only data are often obtained from archived records, situations where organized surveys were not conducted (e.g., some citizen science data), or the methods of data collection are not known, so absences cannot be inferred reliably. Marine mammals may be difficult to detect, depending on the species, survey platform, and survey conditions. Some researchers incorporate pseudo-absences into the estimated visual survey strip-width distance of presumed detection (Derville et al., 2018), or randomly generate pseudo-absences in

study areas where no dolphins were sighted to balance the presence values for modelling presence / absence (Torres et al., 2008). Others used the GPS coordinates of a point where their research vessel paused for collecting environmental data to represent “true” absences, under the assumption that if dolphins were present, observers would see them while the vessel is stopped (Heinrich et al., 2019). I have used the continuous stream of absences (on-effort track-line) within each grid cell to represent true absences where observers did not find dolphins. I reason that the survey design for each embayment was topographically and geographically comprehensive with 3–4 iterations of survey effort even in areas where we presumed a priori dolphins would not be encountered. Survey transects were specifically designed based on knowledge of bottlenose dolphin biology, water quality reports from each embayment, and habitat characteristics previously associated with dolphin groups in estuarine environments (e.g., distance to coast, salinity records, deep channels). Additionally, based on prior knowledge, I incorporated increased effort in the survey design where the densest dolphin distribution was anticipated, and arranged transect-line spacing for maximum potential visual detection of dolphin groups. As the majority (93%) of dolphin encounters were sighted within 500 m of the research vessel, I chose a 500 x 500 m fishnet grid under the assumption of high confidence of dolphin group detection within the 250 m transect strip width on each side of the research vessel. This is similar to the approach of Macleod et al. (2008) where harbor porpoise (Phocoena phocoena) absences were assigned to grid cells if the cell had been surveyed at least three times without a group sighting to reduce the likelihood of false negative detection. After the

data described above were prepared, observed bottlenose dolphin density (Dobs) was calculated as:

퐶표푢푛푡푐 퐷 = 표푏푠 퐸푓푓표푟푡 × 푆 × 푃 where for each grid square, Countc is the number of individual catalogued dolphins

(corrected for unmarked fins), Effort is the summed track distance (km), S is the visual survey strip width (0.5 km), and P is the probability of dolphin group detection (P assumed = 1 in this thesis).

Generalized additive models [GAMs, (Hastie & Tibshirani, 1986)] were used to model dolphin distribution in R-3.5.1 [“mgcv” package, (R Core Team, 2018; Wood,

2017)]. A GAM is an extension of a generalized linear model with a linear predictor composed of additive “smooth” functions for each covariate (Wood, 2017). Generalized additive models have proven useful to examine non-parametric relationships between the environment and a wide variety of species distributions (Guisan et al., 2002), and have been used to model the distribution of cetaceans when sightings are scarce (Keller et al., 2012; Virgili et al., 2017). As the response variable of individual dolphin density consisted of 96–97% zeroes due to the lack of dolphin encounters in all three survey areas, the GAMs were specified with a Tweedie distribution, which is suitable for modeling a combination of continuous data (e.g., individual dolphin density values) with a spike at zero [for grid cells where dolphins were not encountered, (Mannocci et al.,

2015)]. Predictor variables in GAMs are fit with smoothing functions, each with a specified number of “knots” that determine the number of functions which are added to create the smoothed overall model fit. An examination of the adjusted-R2 (adj.-R2) values during exploratory modeling for each survey area indicated increasing the

number of knots for each function provided negligible increases in explanatory power.

Therefore, GAM smooth parameters were automatically set using the restricted maximum likelihood (REML) option in the mgcv package with the default thin plate regression spline. Estimated degrees of freedom (edf) for each covariate are provided in a diagnostic check function of the mgcv package. An edf ~1.0 indicates a linear relationship between the covariate and the response variable, negating the need for a smooth function on that covariate. Where the edf was ~1, the smooth function was removed and the covariate was left in the GAM as a parametric term to examine the results of removing the smoother. The GAM terms included an offset to correct for different amounts of survey effort in each grid cell. The offset term has a coefficient of 1 and was quantified as the log of the summed distance of effort in each grid cell, multiplied by the area of the cell. For example, the full GAM took the form of:

Countc ~1+s(Depth)+s(Slope)+s(Dist.Coast)+s(Salinity)+s(Temperature)+

s(Diss.Oxy)+s(pH)+offset(log(Effortc)) where Countc is the number of marked catalogued dolphins corrected with each survey area’s distinctiveness proportion, “s” designates the smooth for each covariate, and

Effortc is the summed kilometers of visual survey effort in each grid cell multiplied by the area of the cell (0.5 km).

Model selection took place in an iterative process of backward covariate elimination, whereby the least significant covariate was sequentially removed until all remaining covariates were statistically significant at an α = 0.05 (Crawley, 2005).

Models were evaluated using a combination of the adj.-R2, percent of deviance explained, and Akaike’s information criterium [AIC, (Akaike, 1973)] calculated for each

model iteration. The adj.-R2 is a weighted measure of model fit that accounts for the number of parameters in the model (and corresponding degrees of freedom) relative to the amount of variation in the response variable explained by the model. The AIC measures the lack of model fit to the observed data using the negative log-likelihood estimated by the model, and corrects for bias by increasing the AIC score with increasing parameters (Johnson & Omland, 2004). The lowest AIC score may represent the best fitting model, but models with AIC scores within 2 AIC points of each other merit further consideration to determine the most parsimonious model (Symonds

& Moussalli, 2011). If two or more models scored ≤ 2 points of the minimum AIC score, the simplest model (fewest parameters) was chosen. The mean squared error (the mean squared difference between the observed and predicted values) was calculated for each best model. Finally, the dolphin group density values predicted by the selected model of each approach were joined with the fishnet grid to visually compare the observed group density with the predicted group density. The symbology used to represent the number of individuals in the maps was manually classified in four or five number breaks for all maps: ≤ 1.0, ≤ 2.5, ≤ 5.0, ≤7.5, and ≤15.0. For numbers > 15.0, a

6th classification break was added to represent the maximum value.

Results

West Bay and Galveston Bay

In total, 27 surveys comprising ~293 hours and ~2,509 km of on-effort track-lines were conducted in West Bay and Galveston Bay, during June 2015 and July 2016, respectively (Table 4-4). Overall, 134 groups of dolphins were sighted (including coastal waters), 80 of which were included in the habitat analysis. Mean group size was

10.7 ± 8.8 dolphins and ranged from 1–34. Water quality values (푥̅ ±SD) were

interpolated for temperature (25.5 ± 0.7 °C), pH (8.1 ± 0.2), dissolved oxygen (6.8 ±

0.4 mg/l), and salinity (18.5 ± 4.5 ppt). Water depth and mean slope values extracted from the DEM were 3.6 ± 3.8 m and 0.6 ± 0.6 degrees, respectively. Mean distance from the centroid of each grid cell to the designated coastal point were 13.4 ± 10.2 km.

Of the 2,059 grid cells, 69 contained dolphin density values.

The results of the Mann-Whitney test of Galveston Bay water quality values indicated no statistically significant difference in means between the distributions of the

1990 dataset and of the 1991–2005 test data (temperature: p = 0.73; pH: p = 0.06; dissolved oxygen: p = 0.23; salinity: p = 0.74) or the 1981–2015 TXBlend estimated salinity data (p = 0.37) . Collinearity was indicated for the distance to coast covariate

(VIF = 4.8), so it was removed. The remaining covariates had VIF scores < 3. Results of the Pearson correlation test on the remaining covariates indicated a significant positive correlation for depth and slope (r = 0.73, p-value < .001). The DEMs for both survey areas indicate that significant changes in slope are spatially limited to the dredged navigation channels (Figure 4-1G and 4-2G); very little indication of slope variation is observed elsewhere. However, in Galveston Bay, bottlenose dolphin groups were encountered in open water > 5 km from dredged channels. Given the natural seafloor of the region is relatively smooth, I removed slope from the model under the assumption it would not be a suitable predictor of dolphin presence in open water away from dredged navigation channels.

The GAM indicated depth alone (p < 0.001) was a statistically significant predictor (AIC = 2481.4, adj.-R2 = 0.20, deviance explained = 28.2%, Table 4-5). The

GAM smooth plot indicates dolphin density is positively correlated with depth above ~

2.5 m, increasing to 12.5 m (Figure 4-5). Observed mean dolphin density was 0.14 ±

1.2 km-2, ranging from 0–33.8 km-2, with the highest observed dolphin density in the

HSC and Bolivar Roads areas. Mean GAM predicted dolphin density was 0.26 ± 0.7 km-2 and fitted values ranged from 0–7.2 km-2. The MSE was 2.6 dolphins km-2 (Table

4-7, Figure 4-6).

Sabine Lake

In total, 7 surveys comprising 62.0 hours and covering 471 km of on-effort track- lines were conducted in Sabine Lake during June 2017 (Table 4-4). Overall, 47 groups of dolphins were sighted (including coastal waters), 25 of which were included in the habitat analysis. Mean group size was 6.5 ± 7.3 dolphins and ranged from 1 – 31.

Water quality values (푥̅ ±SD) for were interpolated for temperature (29.9 ± 0.5 °C), pH

(7.7 ± 0.3), dissolved oxygen (5.6 ± 0.6 mg/l), and salinity (12.4 ± 2.4 ppt). Water depth and mean slope values extracted from the DEM were 5.0 ± 5.4 m and 0.9 ± 1.2 degrees. Mean distance from the centroid of each grid cell to the designated coastal point were 24.8 ± 7.1 km. Of the 566 grid cells, 21 contained dolphin density values.

The results of the Mann-Whitney test of Sabine Lake water quality values indicated no statistically significant difference between the distribution of the 1990 dataset and that of the test data (temperature: p = 0.44; pH: p = 0.95; dissolved oxygen: p = 0.11; salinity: p = 0.41). Collinearity was indicated for covariates of mean slope (VIF

= 8.6) and pH (VIF=3.1). These were removed from the modeling process. The remaining covariates had VIF scores < 3. Results of the Pearson correlation test indicated no significant correlations ≥ 0.7.

The best GAM indicated depth and distance from coast as statistically significant predictors (p < .005, AIC = 686.9, adj.-R2 = .08, deviance explained 48.1%, Table 4-6).

The GAM smooth plot for depth suggests a nearly linear relationship between depth and dolphin density (Figure 4-7). A diagnostic check indicated the estimated degrees of freedom for depth was close to 1 (edf = 1.24), but removing the smooth function decreased the explanatory power of the model by half (adj.-R2 = .08 with the smooth, adj.-R2 = .04 without the smooth) and resulted in a negligible decrease in the AIC score, so the smooth was left in place for the depth covariate. The smooth plot for distance

(edf = 3.16) indicates a positive correlation between dolphin density and distance that increases until approximately 22.5 km away from the end of the Sabine Pass jetties

(Figure 4-7). Observed mean dolphin density was 0.12 ± 0.85 km-2, ranging from 0–

14.7 km-2, with the highest observed dolphin density in the Sabine Pass Channel. Mean

GAM predicted dolphin density was 0.17 ± 0.70 km-2 and fitted values ranged from 0–

7.4 km-2 and was greatest in the Sabine Pass channel in an area within approximately 5 km of the southern boundary of Sabine Lake. The MSE was 1.8 dolphins km-2 (Table 4-

7, Figure 4-8).

Model validation. Cross validation of modeling results can be done by splitting the data before analysis into training and test data sets. After analysis, the model fit can be validated using the independent data of the test data set. Too few samples were available to split the summer data, however, given the GAM indicated only physiographic covariates as statistically significant predictors, water quality data

(including temperature) was not required to test the models. GAM models for each

summer survey area were fitted to the winter survey effort data (groups and on-effort track-line) to test the fit of the model.

Respectively, West Bay–Galveston Bay and Sabine Lake observed mean winter

-2 -2 density was 0.08 ± 0.55 km and 0.06 ± 0.75 km , mean predicted winter density was

0.17 ± 0.40 km-2 and 0.08 ± 0.25 km-2. The MSE of the winter models was 0.75 and

0.65 dolphins km-2 for West Bay–Galveston Bay and Sabine Lake, respectively (Table

4-7).

Discussion

The observed seasonal density pattern was consistent with the seasonal difference in abundance estimates, with greater observed density in summer. Summer mean predicted density for Sabine Lake (0.17) was higher than what was observed

(MSE = 1.8), while the predicted mean density for Galveston Bay (0.26) was nearly twice as high as observed (MSE = 2.6). However, the MSE for the winter density predicted by the summer models was lower for both survey areas (MSE = 0.75 and

0.65, West Bay–Galveston Bay and Sabine Lake, respectively). Fewer open water sightings were detected during the winter surveys, particularly in Galveston Bay, and less prediction error for the winter season may be a result of higher concentrations of bottlenose dolphin groups in deep channels nearer to the coast during winter, which is a distribution more aligned with the models indicating depth and distance as significant predictors. Both summer models predicted a maximum of 7–8 dolphins km-2 for any given grid square and the adj.-R2 for the two best models were relatively low (ranged from 4–20%). A visual evaluation of the mapped predictions indicates the models performed well in terms of where their highest relative density was spatially predicted but failed to predict dolphins in open water away from deep channels. While that result

is consistent with what was observed in West Bay and Sabine Lake, dolphins were observed in Galveston Bay open water > 5 km from a deep channel. Future population assessments may be improved if they are informed by a spatial distribution model of fish-prey which was demonstrated to have high predictive performance of dolphin habitat use (Torres et al., 2008). Another potential reason for the difference between what was observed and predicted in Galveston Bay is the observed dolphin interactions with shrimp trawlers. During this study, shrimp trawlers encountered in Galveston Bay

(but not in Sabine Lake or West Bay) were associated with foraging dolphins in a wide range of depths and locations. Observations from other studies conducted in Galveston

Bay confirm the high correlation of dolphins with shrimp trawler activity [e.g., (Fertl,

1994; Moreno & Mathews, 2018; Piwetz, 2019). Moreno and Mathews (2018) found statistically significant clusters of foraging dolphin groups in connection with shrimp trawlers and seabird activity and estimated deep channels in Bolivar Roads and the

Galveston Ship Channel to include 91% of observed foraging groups yet comprised only

1 / 5 of their survey area. Likewise, Piwetz (2019) estimated 57% of observed dolphin groups used the Galveston Ship Channel as foraging habitat and noted dolphin reorientation rate to significantly increase in the presence of shrimp trawlers. However, shrimp trawler activity is common in Galveston Bay away from the deep channels, and the presence of shrimp trawlers is likely a powerful determinant of dolphin distribution in open water as well. It is not clear the extent of heterogeneity that foraging behavior introduced into my density estimates, however, future bottlenose dolphin species distribution models for Galveston Bay might be improved if shrimp trawler interaction is included as a factor.

In Sabine Lake, I observed small groups (< 6-8 individuals) traveling upstream and away from the coast into the freshwater Neches and Sabine rivers, where dolphin density was predicted to be < 1.0. Reports from the public suggest bottlenose dolphins are venturing farther west in the Neches River as far as the city of Beaumont, ~20 km upstream of Port Neches1* and the Texas Marine Mammal Stranding Network has removed several live dolphins from so-called out-of-habitat environments in the outer reaches of Sabine Lake (Whitehead & Ronje, 2017). With a few exceptions, most dolphins observed in the Sabine Lake study appeared in good physical condition.

However, it is not clear that dolphins using the Sabine Lake area tolerate a lower salinity threshold, relative to other studied populations. Notably, my survey results suggest the relatively shallow Sabine Lake proper [which does not have stratified salinity (Ward,

1980)], may not be used frequently for travel to the northern end of the lake. Dolphins might exploit the stratified vertical salinity gradient in the deeper waters of the engineered Sabine-Neches Waterway. Ward (1980) described an unusual phenomenon peculiar to Sabine Lake, whereby the Sabine-Neches Waterway conducts a vein of higher salinity water relative to Sabine Lake on the other side of its channel boundary. The wedge of saline water is apparently so persistent there, it can result in the salinity of northern Sabine Lake measuring higher than the southern end, creating a reverse salinity gradient across the lake. As previously discussed, density currents altered by the engineering of deep waterways (e.g., the Sabine-Neches Waterway and the HSC) facilitate salt-water intrusion into the estuaries and this may have the unintended effect of increasing estuarine capacity for fish and dolphin habitat.

1 *Personal communication: W.Cascio-Smith and D. Shaeffer, 9 July 2018

However, salinity was not a statistically significant factor in the best model for either survey area, and it was the least significant term for Sabine Lake (first covariate removed). The water quality data I used were collected from a vertical profile at each station, providing a range of salinity values. I used the mean salinity of the vertical profile in my models, but perhaps there could be significant influence of one or more of the different levels of salinity in the vertical profile with dolphin density. Likely there is some interaction between depth and salinity that explains how dolphins are interacting with the bay habitats, but this study clearly establishes the importance of deep channels and Gulf passes to estuarine bottlenose dolphins in the nwGoMx.

Limitations

A critical part of any SDM focused on environmental factors is high-quality data at the temporal and spatial resolution required by the study. After an exhaustive search for data contemporary to my studies, I found the TWDB water quality dataset was the best publicly-available one. It is possible the age of the water quality data and the temporal shift to the timing of my studies (May – June water quality data and June –

July survey data) introduced bias in analyzing the spatial distribution of dolphins in the respective bays. Yet, obtaining enough high-quality data to create a water quality profile consistent with the spatial and temporal parameters (and vertical stratifications) of a population study site remains a challenge in estuarine environments. It is worth mentioning that the removal of slope from each model, due to the collinearity between slope and depth, probably does not invalidate slope as an important factor. The bathymetry of these estuaries is such that the bay bottoms are naturally smooth, and slope is paired with the navigational channels or Gulf passes so that a consideration of slope without depth may not be meaningful for these survey areas.

Table 4-1. West Bay–Galveston Bay water quality station locations and mean water quality values for temperature (Temp), pH, dissolved oxygen (DO), and salinity collected by the Texas Water Development Board in 1989 and 1992. NA represents no data collected for the covariate at a station. Station Location Latitude Longitude Date Temp pH DO Salinity 1 San Luis Pass 29.07861 -95.1261 May-89 25.7 7.7 6.9 25.6 1A Cold Pass 29.07722 -95.1367 May-89 25.7 7.7 6.8 25.7 GIWW at South 2 Chocolate Bay 29.14583 -95.1583 May-89 25.8 7.7 7.5 23.4 GIWW at I45 3 Causeway 29.29722 -94.8858 May-89 25.5 8.5 7.4 NA Galveston Channel 4 at A&M Docks 29.31278 -94.8214 May-89 25.2 8.1 7.0 NA GIWW at Pelican 5 Island 29.35139 -94.8158 May-89 24.9 NA NA 19.9 Texas City Dike 6 side of HSC 29.36528 -94.8039 May-89 24.9 8.2 8.0 NA 6A Texas City Channel 29.36222 -94.8158 May-89 25.0 8.2 7.6 NA Bolivar Peninsula 7 side of HSC 29.37083 -94.7869 May-89 25.1 8.4 7.4 22.0 GIWW at Bolivar 8 Peninsula 29.38000 -94.7750 May-89 25.3 8.2 8.1 19.0 9 Rollover Pass 29.50833 -94.5000 May-89 25.5 NA NA 16.0 GIWW near Oyster 10 Bayou 29.57167 -94.4856 May-89 25.8 NA NA 7.2 11 Trinity River 29.77000 -94.6861 May-89 23.6 8.0 8.8 NA 11A Long Island Bayou 29.80139 -94.7431 May-89 24.3 8.0 8.8 NA HSC at Baytown 12 Tunnel 29.70417 -95.0194 May-89 25.4 7.6 5.9 9.9 Houston Entrance 13 Channel 29.34167 -94.7417 May-89 24.7 8.3 6.3 24.0 Lower Carancahua 14A Reef 29.20778 -94.9903 May-89 26.0 NA 6.4 25.1 Mid Carancahua 14B Reef 29.21472 -95.0022 May-89 25.9 NA 6.4 25.0 Upper Carancahua 14C Reef 29.22056 -95.0147 May-89 25.9 NA 6.5 24.8 15A Eagle Point 29.49611 -94.8819 May-89 25.9 NA NA 18.2 15B Smith Point 29.52028 -94.8422 May-89 25.2 NA NA 14.3 Trinity Bay near D1 Double Bayou 29.66111 -94.7458 May-89 24.3 7.9 6.9 6.5 Upper Galveston D2 near Red Bluff 29.58056 -94.9417 May-89 25.1 7.8 6.1 12.9 Redfish Reef East D3 of HSC 29.51667 -94.8583 May-89 25.3 8.2 6.9 13.4 HSC off Dollar D4 Point 29.47111 -94.8489 May-89 25.9 8.1 6.5 12.7 West Bay, D5 Carancahua Reef 29.25278 -94.9792 May-89 25.9 8.0 7.1 24.4 D6 Bolivar Roads 29.34167 -94.7833 May-89 24.7 8.2 6.3 21.7

Table 4-1. Continued Station Location Latitude Longitude Date Temp pH DO Salinity D7 East Bay at Marsh Point 29.53194 -94.5764 May-89 24.7 8.2 6.3 21.7 10-CB Cold Pass 29.07000 -95.1431 Jun-92 26.7 8.0 5.7 27.9 1-CB San Luis Pass 29.08194 -95.1247 Jun-92 25.6 8.4 4.5 29.9 2B-CB Titlum Tatlum 29.07694 -95.1428 Jun-92 27.0 8.0 6.3 30 3-CB Guyton Cut 29.10167 -95.1594 Jun-92 27.5 8.2 7.2 23.3 4-CB Christmas Point 29.08056 -95.1714 Jun-92 27.1 7.7 5.9 27.1 5-Cb ICWW E of Oyster Lake 29.13000 -95.1761 Jun-92 28.0 8.0 7.1 NA 6A-CB ICWW N of Bastrop Bayou 29.10556 -95.2019 Jun-92 28.3 8.0 5.9 19.6 6B-CB Bastrop Bayou 29.09722 -95.2069 Jun-92 28.7 7.9 5.9 16.2 7A-CB Port Morris Cut 29.09250 -95.2022 Jun-92 28.5 7.5 6.2 18.9 7B-CB ICWW S of Port Morris Cut 29.09000 -95.2028 Jun-92 27.9 7.2 5.9 24.4 8-CB Nicks Cut 29.02806 -95.2278 Jun-92 28.4 NA NA 20.9 9A-CB ICWW near Surfside 28.98917 -95.2586 Jun-92 25.6 7.9 5.7 NA 9B-CB Old ICWW near Surfside 28.98806 -95.2561 Jun-92 26.7 7.9 6.4 NA 10-CB Rattlesnake Point 29.02833 -95.2100 Jun-92 28.7 NA NA 18.4 D1-CB Cold Pass 29.07583 -95.1344 Jun-92 26.0 8.0 5.6 23.1 D2-CB Christmas Bay 29.04167 -95.1750 Jun-92 26.2 8.0 3.5 20.2 D3-CB Swan Lake Boat Basin 28.98000 -95.2697 Jun-92 25.8 7.8 5.2 14.8

Table 4-2. Sabine Lake water quality station locations and mean water quality values for temperature (Temp), pH, dissolved oxygen (DO), and salinity collected by the Texas Water Development Board in 1990 and 1996 in Sabine Lake. NA represents no data collected for the covariate at a station. Station Location Latitude Longitude Date Temp pH DO Salinity 1 Sabine Pass 29.69167 -93.8389 Jun-90 30.0 7.7 3.2 20.1 2 Sabine Lake at Hwy. 82 29.76667 -93.8958 Jun-90 30.2 7.6 4.7 NA 3 Port Arthur Canal 29.76389 -93.8958 Jun-90 30.1 7.7 3.4 16.0 4 Keith Lake Fish Pass 29.77500 -93.9417 Jun-90 30.8 7.4 4.6 9.1 GIWW at W. Port Arthur 5 Bridge 29.82222 -93.9653 Jun-90 30.4 7.3 5.1 NA 6 GIWW North of Port A 29.93056 -93.8750 Jun-90 30.0 6.8 3.7 6.8 7 Rainbow Bridge 29.97917 -93.8694 Jun-90 29.9 6.6 3.6 NA 8A Sabine River 29.99861 -93.7694 Jun-90 29.8 6.8 5.0 1.5 8B Black's Bayou 29.99306 -93.7639 Jun-90 30.2 6.9 5.5 0.7 9A Johnsons Bayou 29.84944 -93.7847 Jun-90 32.0 7.4 5.9 NA 9B Willow Bayou 29.86389 -93.7806 Jun-90 32.4 7.5 6.5 NA D1 Sabine Causeway 29.76528 -93.8958 Jun-90 29.4 7.4 6.5 5.5 D2 Mid Sabine Lake 29.93750 -93.8083 Jun-90 29.8 NA 9.2 0.7 1 Sabine Pass 29.73389 -93.8722 Jun-96 28.8 8.2 6.1 17.7 2 Sabine Lake at Hwy. 82 29.76667 -93.8958 Jun-96 29.0 8.1 6.1 17.8

Table 4-2. Continued Station Location Latitude Longitude Date Temp pH DO Salinity 3 Port Arthur Canal 29.76389 -93.9125 Jun-96 28.9 8.1 5.9 17.7 5 GIWW at Highway 166 29.82222 -93.9653 Jun-96 29.2 7.7 6.1 16.5 6 Sabine-Canal 29.93833 -93.8694 Jun-96 29.4 7.2 5.2 13.8 7 Rainbow Bridge 29.97694 -93.8636 Jun-96 29.7 7.5 5.2 12.4 8 GIWW at Stewts Island 29.97278 -93.8483 Jun-96 29.5 7.5 5.4 12.7 9 Sabine River 30.00417 -93.7667 Jun-96 29.0 7.4 7.3 10.9 10 Black Bayou 29.99306 -93.7639 Jun-96 29.0 7.5 7.8 12.3 11 Three Bayou 29.91389 -93.7611 Jun-96 29.7 7.3 6.8 10.1 13 Platform "A" 29.95056 -93.8258 Jun-96 29.2 7.8 8.9 13.9 D1 Keith Lake Fish Pass 29.94167 -93.7750 Jun-96 28.9 8.1 5.0 19.0 D2 GIWW at Highway 87 29.82222 -93.9653 Jun-96 29.2 7.7 6.3 15.2 D3 Platform "A" 29.94444 -93.8286 Jun-96 28.8 NA 3.0 18.4 Sabine River Platform D4 (shallow) 30.00417 -93.7667 Jun-96 28.9 7.5 4.5 8.5 Sabine River Platform D5 (deep) 30.00417 -93.7667 Jun-96 28.8 7.3 4.2 8.3 D6 Black Bayou 29.99361 -93.7600 Jun-96 29.0 NA NA 8.3 D7 Johnson Bayou 29.84556 -93.7792 Jun-96 29.1 NA NA 11.7

Table 4-3. Example of raw water quality data recorded by the Texas Water Development Board in Sabine Lake during June 1990. TD = Total depth at sampling location. Station Date Time TD Depth Temperature pH DO Salinity 1 900620 911 53 1.0 30.1 7.96 4.50 16.7 1 900620 911 53 12.5 30.0 7.94 2.51 22.9 1 900620 911 53 25.0 28.8 7.79 1.01 26.6 1 900620 911 53 37.0 28.0 7.67 0.21 29.0 1 900620 911 53 50.0 27.9 7.66 0.09 29.3

Table 4-4. Summary results of West Bay–Galveston Bay and Sabine Lake surveys, mean interpolated water quality values, and fishnet grid information. Note that water quality parameters were measured decades before the dolphin surveys were conducted. West Bay-Galveston Bay Sabine Lake Year 2015 - 2016 2017 Surveys 27 7 Hours 293 62 Effort (km) 2,509 471 Total Groups 134 47 Groups Analyzed 80 25 Mean Group size 10.7 6.5 Temperature °C (archived data) 25.5 29.9 pH (from archived data) 8.1 7.7 Dissolved Oxygen (mg/l) (archived data) 6.8 5.6 Salinity ppt (archived data) 18.5 12.4 Depth (m) 3.6 5.0 Slope (degrees) 0.6 0.9 Distance (km) 13.4 24.8 Total grid cells 2,059 566 Grid cells with density 69 21

Table 4-5. Generalized additive models attempted for West Bay–Galveston Bay. % adj. Dev. Model AIC R2 Exp. Countc ~1+s(Depth)+s(Temp)+s(Diss.Oxy)+s(Salinity)+s(pH) +offset(log(Effortc) 2487.4 0.22 30.1

Countc ~1+s(Depth)+s(DissO2)+s(Salinity)+s(pH)+offset(log(Effortc) 2485.2 0.22 30.1

Countc ~1+s(Depth)+s(Salinity)+s(pH)+offset(log(Effortc) 2483.2 0.22 30.0

Countc ~1+s(Depth)+s(pH)+offset(log(Effortc) 2482.6 0.20 29.4

Countc ~1+s(Depth)+offset(log(Effortc) 2481.4 0.20 28.2

Table 4-6. Generalized additive models attempted for Sabine Lake. % Dev. Model AIC adj. R2 Exp. Countc ~1+s(Depth)+s(Salinity)+s(Dist.Coast)+s(Temp)+s(Diss.Oxy) +offset(log(Effortc) 692.3 0.13 48.9

Countc ~1+s(Depth)+ s(Dist.Coast)+s(Temp)+s(Diss.Oxy)+offset(log(Effortc)) 689.7 0.05 49.2

Countc ~1+s(Depth)+ s(Dist.Coast)+s(Temp)+offset(log(Effortc)) 687.9 0.02 48.1

Countc ~1+s(Depth)+ s(Dist.Coast)+offset(log(Effortc)) 686.9 0.08 48.1

Countc ~1+Depth+ s(Dist.Coast)+offset(log(Effortc)) 686.3 0.04 47.9

Table 4-7. Summary of mean observed and predicted dolphin density from selected models of each approach. GAM = generalized additive model, MSE = mean squared error West Bay – Galveston Bay Sabine Lake

Season Observed (SD) GAM predicted (SD) MSE Observed (SD) GAM predicted (SD) MSE

Summer 0.14 (1.2) 0.26 (0.70) 2.6 0.12 (0.85) 0.17 (0.70) 1.8

Winter 0.08 (0.55) 0.17 (0.40) 0.75 0.06 (0.75) 0.08 (0.25) 0.65

Figure 4-1. Water quality stations, dolphin groups, and mapped environmental profiles for West Bay and Galveston Bay. 2A) Water quality stations in West Bay and Galveston from May 1989 (triangles) and June 1992 (squares). Dolphin groups analyzed are in grey, white circles are dolphin groups excluded from the habitat analysis. The blue line represents the boundary of the interpolation, 2B–2E) respectively, are interpolated water quality measurements collected at each station: temperature, pH, dissolved oxygen, and salinity (data were not available for a full interpolation of salinity for Trinity Bay), 2F–2G) Elevation (m) and slope (degrees) extracted from the digital elevation map

Figure 4-1. Continued

Figure 4-2. Water quality stations, dolphin groups, and mapped environmental profiles Sabine Lake. 3A) Water quality stations in Sabine Lake from 1990 (squares) and 1996 (triangles). Dolphin groups analyzed are in grey, hollow circles are dolphin groups excluded from the habitat analysis. The blue line represents the boundary of the interpolation, 3B–3E): respectively, are interpolated water quality measurements collected at each station: temperature, pH, dissolved oxygen, and salinity., 3F–3G): Elevation (m) and slope (degrees) extracted from the digital elevation map

Figure 4-2. Continued

Figure 4-3. Fishnet grid for West Bay and Galveston Bay with on-effort track-line and bottlenose dolphin group locations. Only grid cells with the full suite of environmental and survey effort data are included.

Figure 4-4. Fishnet grid for Sabine Lake with on-effort track-line and bottlenose dolphin group locations. Only grid cells with the full suite of environmental and survey effort data are included.

Figure 4-5. Smooth plot of West Bay–Galveston Bay GAM. The x-axis of the smooth plots represents the data points of the covariate and the solid line of the y- axis represents the correlation of the smooth function to the response variable. Mean values are at zero, and values above and below zero represent values above and below the mean. The dashed line on each side represent the bounds of the 95% confidence interval and changes in response to the number of observations (narrows with higher number of observations).

Figure 4-6. West Bay–Galveston Bay observed and predicted bottlenose dolphin density A) Observed density, B) Generalized Additive Model predicted density C) Barplot of mean observed and predicted density in summer and winter.

Figure 4-6. Continued

Figure 4-7. Smooth plots of the Sabine Lake GAM. The x-axis of the smooth plots represents the data points of the covariate and the solid line of the y-axis represents the correlation of the smooth function to the response variable. Mean values are at zero, and values above and below zero represent values above and below the mean. The dashed line on each side represent the bounds of the 95% confidence interval and changes in response to the number of observations (narrows with higher number of observations).

Figure 4-8. Sabine Lake observed and predicted bottlenose dolphin density A) Observed density, B) Generalized Additive Model predicted density C) Barplot of mean observed and predicted density in summer and winter.

Figure 4-8. Continued

CHAPTER 5 CONCLUSIONS

Some challenges I encountered during field work and analysis prompt a consideration of how to address methodological limitations. Options for modeling the smaller datasets parsed for potential transient dolphins were few. Additional modeling of abundance and survival could have been completed with additional mark-recapture primaries, and future studies may be improved if at least one primary for each of the four seasons is included. This thesis provided both observed and predicted dolphin spatial distributions that should be useful in locating dolphin groups in each bay. Future mark-recapture surveys planned to estimate abundance may find it helpful to direct their limited resources towards those areas where dolphins are more likely to be encountered, thus reserving valuable resources to conduct surveys more frequently.

In Galveston Bay, vessel traffic is frequent in the ship channels where dolphins are most dense, and passing vessels were a frequent behavioral “attractant” for dolphin groups. It is well-known that bottlenose dolphins (and some other cetacean species) are sometimes attracted to the bows of larger vessels, but a literature search of other cetacean photo-ID articles found no method described to compensate for the “loss” of dolphins to passing vessels during photo-ID as discussed in this thesis. If researchers suspected dolphins from a passing vessel joined the group of focus, then every effort was made to photograph the “new” dolphins and they were included with the group by the research vessel. However, if dolphins were “lost” to a passing vessel before their photographs were collected, there seemed little we could do to compensate in the field.

It was possible, on occasion, to pace the passing vessel and photograph dolphins while they rode the bow, but photos obtained in that way were rarely useful for analysis. It

100

seems likely there is individual heterogeneity in how certain dolphins respond to passing vessels. Piwetz (2019) suggested the physiological limitations of bottlenose dolphin calves likely hinders the maneuverability of mother-calf pairs in response to passing vessels. It may be that estimating abundance with models including parameters for individual heterogeneity would partially account for certain dolphins lost to passing vessels. However, as mentioned in the discussion, modeling individual heterogeneity requires more observations than what was possible in this thesis.

Additional improvements to the collection of environmental data should also be considered for future surveys. Although environmental data were collected during scouting trips and at each dolphin group location (e.g., water temperature, salinity, wind speed), data points were few and relatively concentrated to where dolphins were encountered. Future surveys might be improved if the collection of environmental data were prioritized in the survey design. For example, incorporating a grid of GPS locations for environmental data collection stations to provide additional water quality data to improve the interpolations. It may also be possible to mount a thermosalinograph on a survey vessel to continuously record water quality data.

However, even with temporally and spatially consistent environmental data collected at the surface, it is not clear how well those data would represent the actual relationship of dolphin group locations to water quality parameters because dolphin behavior is not necessarily synchronized to the observed environmental conditions. For example, in

Sabine Lake during June 2017, dolphins were observed traveling up the Neches River in a surface salinity of 0.2 ppt, as measured at the surface with a hand-held sonde, however the range of vertical stratification in surface salinity can be significant (see

101

Table 4-3, where full-depth salinity was twice as high as the surface) and it would be necessary to collect water quality parameters across the range of depths to determine if density currents were influencing dolphin habitat use. Additionally, even if vertically stratified water quality data was systematically collected, it is not clear how informative those data would be when limited to the short-term seasonal surveys of this thesis.

During both summer primaries, strong tropical storm systems moved through the West

Bay and Sabine Lake survey areas [Tropical Stormy Cindy, 20 June 2017; and Tropical

Storm Bill, 16 June 2015, (NOAA, 2019a)]. Weather events can quickly change the salinity of embayments and it is possible that temporally consistent water quality data would not represent the true influence of those factors on dolphin distribution. For example, Mazzoil et al. (2008) observed bottlenose dolphins in the Indian River Lagoon did not avoid waters during high freshwater intrusion events. In West Bay-Galveston

Bay, dolphins were observed to remain in waters polluted by oil spills and harmful algal blooms [reviewed in (Irwin, 2005)]. If bottlenose dolphin behavior is not correlated with short-term changes in environmental parameters, then I recommend only long-term, spatially comprehensive, and vertically stratified water quality data be used to investigate distributional trends, however, such data are difficult to obtain. In the absence of such a comprehensive data set, physiographic features may serve as effective predictors for dolphin distribution in the estuaries of this thesis.

Overall dolphins abundance was greater during summer in this thesis, however, distributions appear to remain concentrated in deep navigational channels in all survey areas in summer and winter, consistent with previous observations reported in bottlenose dolphin population studies on the Texas coast (Gruber, 1981; Maze &

102

Würsig, 1999; Shane, 1980). Navigational channels may serve to funnel fish into concentrated prey aggregations, and the steep sides of such channels are likely advantageous for prey capture (Wells, 2019). Potential density currents (Ward, 1980) in the navigational channels may be related to the concentration of fishes, which serve as an attractant for dolphin groups. Consistent support for depth was strong in both models, but the proportional relationship of depth with the density current described by

(Ward, 1980) are difficult to tease apart so certainly salinity should not be dismissed as non-influential, particularly as it is important to dolphin biology and their prey.

Regardless of the specific characteristic that brings bottlenose dolphins to engineered channels, the results from this thesis clearly point to the channels and Gulf passes as the most important factor driving dolphin distribution in these estuaries. It follows that an unintended consequence of dredging new canals may be creating additional management opportunities for this protected species in areas where before there were few dolphins. Modifications to existing points of bay access from the Gulf may have been a factor in the death of 23 bottlenose dolphins in East Matagorda Bay during 1990 (Ronje et al., 2018). It is important that dredging activity be paired with a plan for monitoring potential ingress / egress of bottlenose dolphins into newly dredged areas, as well as develop a management plan to address the potential for increased numbers of individual dolphins exploiting new deep channels. Likewise, if modifications to existing navigation channels are planned (e.g., the TSP), such plans need to include mitigation to avoid anthropogenic disturbance in what may be the primary habitat selection for this protected species.

103

The results of the photo-ID analysis demonstrated that the ranging patterns of some individuals include Galveston Bay, the adjacent estuaries, and nearshore coastal waters. Bolivar Roads may be an important mixing area for these dolphin populations and serve as a type of dolphin community boundary line. However, the magnitude of this connectivity, and its role in the dynamics of each of these populations, requires additional study likely including multi-year photo-identification studies, tag telemetry studies, and assessment of the genetic independence of the defined stocks. Given the high degree of commercial activity, history of industrial pollution in the region, and anthropogenic modification to the environment, future research should be directed to collect more information on the degree to which bottlenose dolphin populations or subpopulations are mixing in Bolivar Roads to inform the assessments of future coastal engineering projects impacting habitats for bottlenose dolphins in the northwestern Gulf of Mexico.

104

APPENDIX A TABLE OF INTER-BAY BOTTLENOSE DOLPHIN MATCHES

Table A-1. Sighting histories for inter-bay bottlenose dolphin matches by survey area, year, and season. Each cell represents one photo-ID survey. WB, GB, and SL = West Bay, Galveston Bay, Sabine Lake, respectively.

2015 2016 2017 2018 Dolphin Winter Summer Winter Summer Winter Spring Summer Spring 1 GB GB GB GB SL 2 WB WB WB WB WB WB GB GB WB WB 3 GB SL 4 GB SL 5 GB SL 6 GB SL SL 7 WB WB WB GB GB WB 8 WB GB GB GB 9 WB WB WB GB WB 10 WB GB GB GB 11 GB GB SL SL SL 12 GB GB GB GB SL 13 GB SL 14 GB SL 15 GB GB SL 16 GB SL 17 GB SL 18 GB WB WB GB 19 WB WB WB WB WB WB GB GB WB 20 WB WB GB GB WB 21 WB GB GB GB GB 22 WB WB WB GB WB 23 WB WB WB GB 24 WB GB 25 WB WB GB 26 WB GB

105

Table A-1. Continued

2015 2016 2017 2018 Dolphin Winter Summer Winter Summer Winter Spring Summer Spring 27 GB GB GB SL 28 GB GB SL SL 29 GB SL 30 GB GB SL 31 GB SL 32 GB GB GB SL SL 33 GB SL 34 GB SL 35 GB GB SL 36 GB SL 37 GB SL 38 GB SL 39 GB WB 40 GB SL

106

APPENDIX B ABUNDANCE MODEL RANK

Table B-1. Ranked scores of abundance models in program MARK for each season and survey area using Akaike’s Information Criterium adjusted for small sample sizes (AICc). Winter Model Results West Bay Delta AICc Model Num. Model AICc AICc Weights Likelihood Par. Deviance -2log(L) Type A [Mt] 122.04 0.00 0.99885 1.00 3 224.25 115.83 Check Par. Cnt. [Mb] 135.57 13.53 0.00115 0.00 2 239.89 131.46 [M0] 156.61 34.57 0.00000 0.00 1 263.00 154.58 Type B [Mt] 104.55 0.00 0.62885 1.00 3 176.42 98.30 Check Par. Cnt. [Mb] 105.61 1.05 0.37114 0.59 2 179.61 101.48 [M0] 127.65 23.10 0.00001 0.00 1 203.73 125.61 Type C Check Par. Cnt. [Mb] 91.00 0.00 0.59699 1.00 2 135.73 86.85 [Mt] 91.79 0.79 0.40262 0.67 3 134.36 85.48 [M0] 105.68 14.68 0.00039 0.00 1 152.51 103.63

Galveston Bay Type A [Mt] 1392.53 0.00 0.70522 1.00 4 3257.12 1384.50 [Mb] 1410.22 17.68 0.00010 0.00 2 3278.82 1406.21 [M0] 1410.27 17.73 0.00010 0.00 1 3280.88 1408.26 Type B [Mt] 1223.03 0.00 0.73184 1.00 4 2859.62 1214.99 [Mb] 1240.63 17.60 0.00011 0.00 2 2881.24 1236.62 [M0] 1242.67 19.64 0.00004 0.00 1 2885.29 1240.66 Type C [Mt] 1291.81 0.00 0.99625 1.00 4 2974.57 1283.78 [M0] 1303.67 11.86 0.00265 0.00 1 2992.46 1301.67 [Mb] 1305.42 13.61 0.00110 0.00 2 2992.20 1301.41

107

Table B-1. Continued Winter Model Results Sabine Lake Delta AICc Model Num. Model AICc AICc Weights Likelihood Par. Deviance -2log(L) Type A [Mt] 338.76 1.07 0.36988 0.59 3 1030.42 332.70 [Mb] 380.78 43.08 0.00000 0.00 2 1074.47 376.75 [M0] 401.44 63.75 0.00000 0.00 1 1097.16 399.43 Type B [M0] 71.80 1.27 0.24610 0.53 1 93.53 69.73 [Mt] 72.18 1.65 0.20320 0.44 3 89.55 65.75 [Mb] 73.88 3.35 0.08699 0.19 2 93.47 69.67 Type C [Mt] 144.85 0.00 0.86931 1.00 3 258.44 138.66 Check Par. Cnt. [Mb] 148.92 4.07 0.11356 0.13 2 264.60 144.82 [M0] 152.70 7.85 0.01713 0.02 1 270.45 150.67 Summer Model Results West Bay Type A [Mt] 228.06 0.0 0.98217 1 3 489.46 221.94 [Mb] 236.07 8.0 0.01783 0.0182 2 499.53 232.02 [M0] 279.03 51.0 0.00000 0 1 544.53 277.01 Type B [Mt] 80.93 0.0 1.00000 1 3 174.89 74.68 [Mb] 122.47 41.5 0.00000 0 2 218.55 118.34 [M0] 129.29 48.4 0.00000 0 1 227.45 127.25 Type C [Mt] 72.92 0.0 1.00000 1 3 162.47 66.65 [Mb] 111.97 39.1 0.00000 0 2 203.66 107.83 [M0] 120.51 47.6 0.00000 0 1 214.29 118.46

108

Table B-1. Continued Summer Model Results Galveston Bay Delta AICc Model Num. Model AICc AICc Weights Likelihood Par. Deviance -2log(L) Type A [M0] 1791.23 0.0 0.51695 1 1 5714.36 1789.23 [Mb] 1793.22 2.0 0.19140 0.3702 2 5714.35 1789.21 [Mt] 1793.33 2.1 0.18132 0.3507 3 5712.45 1787.31 Type B [M0] 1572.31 0.0 0.51677 1 1 4780.35 1570.31 [Mb] 1573.65 1.3 0.26509 0.513 2 4779.68 1569.64 [Mt] 1574.78 2.5 0.15050 0.2912 3 4778.80 1568.76 Type C [M0] 1495.57 0.0 0.57239 1 1 4452.29 1493.57 [Mb] 1497.13 1.6 0.26247 0.4585 2 4451.85 1493.12 [Mt] 1498.06 2.5 0.16514 0.2885 3 4450.76 1492.04

Sabine Lake Type A [Mt] 514.88 0.0 0.72810 1 3 1573.71 508.84 Check Par. Cnt. [Mb] 530.46 15.6 0.00030 0.0004 2 1591.31 526.44 [M0] 541.05 26.2 0.00000 0 1 1603.92 539.05 Type B [Mt] 194.04 0.0 0.54630 1 3 366.13 187.90 Check Par. Cnt. [Mb] 195.61 1.6 0.24890 0.4556 2 369.77 191.54 [M0] 203.36 9.3 0.00517 0.0095 1 379.57 201.34 Type C [Mt] 255.62 0.0 0.81849 1 3 545.20 249.52 Check Par. Cnt. [Mb] 258.64 3.0 0.18117 0.2213 2 550.27 254.58 [M0] 271.22 15.6 0.00034 0.0004 1 564.88 269.20

109

APPENDIX C CAPTURE PROBABILITIES

Table C-1. Estimated capture probability parameters for selected models. Winter Parameters West Bay Index Label Estimate SE LCI UCI Type A, [Mt] 1 p 0.11 0.05 0.05 0.25 2 p 0.65 0.08 0.48 0.79 3 p 0.67 0.08 0.50 0.81 4 c 0.65 0.08 0.48 0.79 5 c 0.67 0.08 0.50 0.81 Type B, [Mt] 1 p 0.13 0.06 0.05 0.28 2 p 0.47 0.09 0.30 0.65 3 p 0.71 0.10 0.49 0.86 4 c 0.47 0.09 0.30 0.65 5 c 0.71 0.10 0.49 0.86 Type C, [Mt] 1 p 0.16 0.07 0.07 0.34 2 p 0.51 0.10 0.32 0.70 3 p 0.67 0.11 0.44 0.84 4 c 0.51 0.10 0.32 0.70 5 c 0.67 0.11 0.44 0.84

Galveston Bay Type A, [Mt] 1 p 0.20 0.02 0.16 0.25 2 p 0.11 0.02 0.09 0.15 3 p 0.13 0.02 0.10 0.17 4 p 0.17 0.02 0.13 0.21 5 c 0.11 0.02 0.09 0.15 6 c 0.13 0.02 0.10 0.17 7 c 0.17 0.02 0.13 0.21

110

Table C-1. Continued Winter Parameters Galveston Bay Index Label Estimate SE LCI UCI Type B, [Mt] 1 p 0.20 0.02 0.15 0.25 2 p 0.10 0.02 0.08 0.14 3 p 0.12 0.02 0.09 0.16 4 p 0.15 0.02 0.12 0.20 5 c 0.10 0.02 0.08 0.14 6 c 0.12 0.02 0.09 0.16 7 c 0.15 0.02 0.12 0.20 Type C, [Mt] 1 p 0.19 0.02 0.15 0.24 2 p 0.11 0.02 0.09 0.15 3 p 0.13 0.02 0.10 0.17 4 p 0.17 0.02 0.14 0.22 5 c 0.11 0.02 0.09 0.15 6 c 0.13 0.02 0.10 0.17 7 c 0.17 0.02 0.14 0.22

Sabine Lake Type A, [Mt] 1 p 0.11 0.03 0.07 0.18 2 p 0.08 0.02 0.05 0.13 3 p 0.33 0.06 0.22 0.46 4 c 0.08 0.02 0.05 0.13 5 c 0.33 0.06 0.22 0.46 Type B, [Mo, Mb, Mt] 1 p 0.25 0.13 0.08 0.56 2 p 0.21 0.12 0.06 0.53 3 p 0.29 0.15 0.09 0.62 4 c 0.22 0.09 0.08 0.45 5 c 0.30 0.12 0.12 0.56 Type C, [Mt] 1 p 0.22 0.06 0.12 0.36 2 p 0.15 0.05 0.07 0.27 3 p 0.39 0.09 0.23 0.58 4 c 0.15 0.05 0.07 0.27 5 c 0.39 0.09 0.23 0.58

111

Table C-1. Continued Summer Parameters West Bay Index Label Estimate SE LCI UCI Type A, [Mt] 1 p 0.79 0.06 0.65 0.89 2 p 0.41 0.06 0.30 0.53 3 p 0.21 0.05 0.14 0.32 4 c 0.41 0.06 0.30 0.53 5 c 0.21 0.05 0.14 0.32 Type B, [Mt] 1 p 0.80 0.07 0.63 0.91 2 p 0.89 0.06 0.71 0.96 3 p 0.12 0.06 0.05 0.28 4 c 0.89 0.06 0.71 0.96 5 c 0.12 0.06 0.05 0.28 Type C, [Mt] 1 p 0.86 0.07 0.68 0.95 2 p 0.89 0.06 0.71 0.96 3 p 0.13 0.06 0.05 0.29 4 c 0.89 0.06 0.71 0.96 5 c 0.13 0.06 0.05 0.29

Galveston Bay Type A, [Mo, Mb, Mt] 1 p 0.20 0.02 0.15 0.25 2 p 0.20 0.03 0.16 0.26 3 p 0.20 0.02 0.16 0.25 4 c 0.20 0.02 0.17 0.24 5 c 0.20 0.02 0.17 0.23 Type B, [Mo, Mb, Mt] 1 p 0.21 0.03 0.15 0.28 2 p 0.21 0.03 0.15 0.29 3 p 0.21 0.03 0.15 0.28 4 c 0.22 0.02 0.19 0.26 5 c 0.22 0.02 0.19 0.26

112

Table C-1. Continued Summer Parameters Galveston Bay Index Label Estimate SE LCI Type C, [Mo, Mb, Mt] 1 p 0.22 0.03 0.17 0.29 2 p 0.22 0.03 0.17 0.29 3 p 0.22 0.03 0.17 0.29 4 c 0.23 0.02 0.20 0.27 5 c 0.23 0.02 0.20 0.27

Sabine Lake Type A, [Mt] 1 p 0.06 0.01 0.04 0.09 2 p 0.13 0.03 0.08 0.19 3 p 0.15 0.03 0.10 0.22 4 c 0.13 0.03 0.08 0.19 5 c 0.15 0.03 0.10 0.22 Type B, [Mt] 1 p 0.18 0.05 0.10 0.29 2 p 0.21 0.05 0.12 0.33 3 p 0.40 0.08 0.26 0.56 4 c 0.21 0.05 0.12 0.33 5 c 0.40 0.08 0.26 0.56 Type C, [Mt] 1 p 0.15 0.04 0.09 0.24 2 p 0.21 0.05 0.13 0.32 3 p 0.37 0.07 0.25 0.52 4 c 0.21 0.05 0.13 0.32 5 c 0.37 0.07 0.25 0.52

113

LIST OF REFERENCES

Adams, J. D., Speakman, T., Zolman, E., & Schwacke, L. H. (2006). Automating image matching, cataloging, and analysis for photo-identification research. Aquatic Mammals, 32(3), 374. http://doi.org/10.1578/AM.32.3.2006.374

Akaike, H. (1973). Maximum likelihood identification of Gaussian autoregressive moving average models. Biometrika, 60(2), 255-265. https://doi.org/10.1093/biomet/60.2.255

Allen, M. C., Read, A. J., Gaudet, J., & Sayigh, L. S. (2001). Fine-scale habitat selection of foraging bottlenose dolphins (Tursiops truncatus) near Clearwater, Florida. Marine Ecology Progress Series, 222, 253-264. https://doi.org/10.3354/meps222253

Anderson, J. B., Wallace, D. J., Simms, A. R., Rodriguez, A. B., & Milliken, K. T. (2014). Variable response of coastal environments of the northwestern Gulf of Mexico to sea-level rise and climate change: Implications for future change. Marine Geology, 352, 348-366. https://doi.org/10.1016/j.margeo.2013.12.008

Balmer, B., Sinclair, C., Speakman, T., Quigley, B., Barry, K., Cush, C., Hendon, M., Mullin, K., Ronje, E., Rosel, P., Schwacke, L., Wells, R., & Zolman, E. (2016). Extended movements of common bottlenose dolphins (Tursiops truncatus) along the northern Gulf of Mexico's central coast. Gulf of Mexico Science, 1, 93-97. https://doi.org/10.18785/goms.3301.08

Balmer, B., Watwood, S., Quigley, B., Speakman, T., Barry, K., Mullin, K., Rosel, P., Sinclair, C., Zolman, E., & Schwacke, L. (2019). Common bottlenose dolphin (Tursiops truncatus) abundance and distribution patterns in St. Andrew Bay, Florida, USA. Aquatic Conservation: Marine and Freshwater Ecosystems, 29(3), 486-498. https://doi.org/10.1002/aqc.3001

Balmer, B., Wells, R., Nowacek, S., Nowacek, D., Schwacke, L., McLellan, W., & Scharf, F. (2008). Seasonal abundance and distribution patterns of common bottlenose dolphins (Tursiops truncatus) near St. Joseph Bay, Florida, USA. Journal of Cetacean Research Management, 10(2), 157-167.

Blaylock, R. A., & Hoggard, W. (1994). Preliminary estimates of bottlenose dolphin abundance in southern US Atlantic and Gulf of Mexico continential shelf waters. (Technical Memorandum NMFS-SEFSC-356, 10 p). Miami, FL, USA: National Oceanic and Atmospheric Administration.

Bosch, S., Tyberghein, L., Deneudt, K., Hernandez, F., & De Clerck, O. (2018). In search of relevant predictors for marine species distribution modelling using the MarineSPEED benchmark dataset. Diversity and Distributions, 24(2), 144-157. https://doi.org/10.1111/ddi.12668

114

Bouchet, P. J., Meeuwig, J. J., Salgado Kent, C. P., Letessier, T. B., & Jenner, C. K. (2015). Topographic determinants of mobile vertebrate predator hotspots: current knowledge and future directions. Biological Reviews, 90(3), 699-728. https://doi.org/10.1111/brv.12130

Bräger, S. (1993). Diurnal and seasonal behavior patterns of bottlenose dolphins (Tursiops truncatus). Marine Mammal Science, 9(4), 434-438. https://doi.org/10.1111/j.1748-7692.1993.tb00477.x

Bräger, S., Würsig, B., Acevedo, A., & Henningsen, T. (1994). Association patterns of bottle-nosed dolphins (Tursiops-truncatus) in Galveston Bay, Texas. Journal of Mammalogy, 75(2), 431-437. http://doi.org/10.2307/1382564

Carslaw, D. C., & Ropkins, K. (2012). openair --- an R package for air quality data analysis. Environmental Modelling & Software, 27-28, 52-61. https://doi.org/10.1016/j.envsoft.2011.09.008

Chabanne, D. B., Pollock, K. H., Finn, H., & Bejder, L. (2017). Applying the multistate capture‐recapture robust design to characterize metapopulation structure. Methods in Ecology and Evolution. https://doi.org/10.1111/2041-210X.12792

Chao, A., Lee, S., & Jeng, S. (1992). Estimating population size for capture-recapture data when capture probabilities vary by time and individual animal. Biometrics, 48(1), 201-216. https://doi.org/10.2307/2532750

Conn, P. B., Arthur, A. D., Bailey, L. L., & Singleton, G. R. (2006). Estimating the abundance of mouse populations of known size: Promises and pitfalls of new methods. Ecological Applications, 16(2), 829-837. https://doi.org/10.1890/1051- 0761(2006)016[0829:ETAOMP]2.0.CO;2

Conn, P. B., Gorgone, A. M., Jugovich, A. R., Byrd, B. L., & Hansen, L. J. (2011). Accounting for transients when estimating abundance of bottlenose dolphins in Choctawhatchee Bay, Florida. The Journal of Wildlife Management, 75(3), 569- 579. https://doi.org/10.1002/jwmg.94

Crawley, M. J. (2005). Statistics An Introduction Using R. Chichester, England: John Wiley & Sons, Ltd.

Darroch, J. N. (1958). The multiple-recapture census: I. Estimation of a closed population. Biometrika, 45(3/4), 343-359. https://doi.org/10.2307/2333183

Derville, S., Torres, L. G., Iovan, C., & Garrigue, C. (2018). Finding the right fit: Comparative cetacean distribution models using multiple data sources and statistical approaches. Diversity and Distributions, 24(11), 1657-1673. https://doi.org/10.1111/ddi.12782

115

Dormann, C. F., Elith, J., Bacher, S., Buchmann, C., Carl, G., Carré, G., Marquéz, J. R. G., Gruber, B., Lafourcade, B., & Leitão, P. J. (2013). Collinearity: a review of methods to deal with it and a simulation study evaluating their performance. Ecography, 36(1), 27-46. https://doi.org/10.1111/j.1600-0587.2012.07348.x

Elith, J., & Leathwick, J. R. (2009). Species distribution models: ecological explanation and prediction across space and time. Annual Review of Ecology, Evolution, and Systematics, 40, 677-697. https://doi.org/10.1146/annurev.ecolsys.110308.120159

ESRI. (2019). Environmental Research Systems Institute. ArcGIS Pro (Version 2.4.1).

Fazioli, K. L., Hofmann, S., & Wells, R. S. (2006). Use of Gulf of Mexico coastal waters by distinct assemblages of bottlenose dolphins (Tursiops truncatus). Aquatic Mammals, 32(2), 212-222. https://doi.org/10.1578/am.32.2.2006.212

Fertl, D. (1994). Occurrence patterns and behavior of bottlenose dolphins (Tursiops truncatus) in the Galveston ship channel, Texas. Texas Journal of Science, 46(4), 299-318.

Fisher, W. L., McGowen, J. H., Brown Jr., L. F., & Groat, C. G. (1972). Environmental geologic atlas of the Texas coastal zone - Galveston - Houston area. Austin, TX: The University of Texas at Austin.

Friedrich, H. (1969). Marine Biology, an Introduction to its Problems and Results. London: Sidgwick and Jackson Ltd.

Gruber, J. A. (1981). Ecology of the Atlantic bottlenosed dolphin (Tursiops truncatus) in the Pass Cavallo area of Matagorda Bay, Texas. (MSc. Thesis), Texas A&M University, 182 p,

Guisan, A., Edwards Jr, T. C., & Hastie, T. (2002). Generalized linear and generalized additive models in studies of species distributions: setting the scene. Ecological Modelling, 157(2-3), 89-100. https://doi.org/10.1016/S0304-3800(02)00204-1

Gunter, G. (1942). Contributions to the natural history of the bottle-nose dolphin Tursiops truncatus (Montague) on the Texas coast. Journal of Mammalogy, 23(3), 267-276. https://doi.org/10.2307/1374993

Harvey, G. K., Nelson, T. A., Fox, C. H., & Paquet, P. C. (2017). Quantifying marine mammal hotspots in British Columbia, Canada. Ecosphere 8(7):e1884, 1-22 p. https://doi.org/10.1002/ecs2.1884

Hastie, G. D., Wilson, B., & Thompson, P. M. (2003). Fine-scale habitat selection by coastal bottlenose dolphins: application of a new land-based video-montage technique. Canadian Journal of Zoology-Revue Canadienne De Zoologie, 81(3), 469-478. https://doi.org/10.1139/Z03-028

116

Hastie, T., & Tibshirani, R. (1986). Generalized Additive Models. Statistical Science, 1(3), 297-318.

Hayes, S. A., Josephson, E., Maze-Foley, K., Rosel, P. E., Byrd, B. L., Cole, T. V., Engleby, L., Garrison, L. P., Hatch, J. M., & Henry, A. (2017). US Atlantic and Gulf of Mexico marine mammal stock assessments - 2016. (Technical Memorandum NMFS-NE-241, 282). Woods Hole, MA: National Oceanic and Atmospheric Administration, US Department of Commerce.

Heinrich, S., Genov, T., Fuentes Riquelme, M., & Hammond, P. S. (2019). Fine‐scale habitat partitioning of Chilean and Peale's dolphins and their overlap with aquaculture. Aquatic Conservation: Marine and Freshwater Ecosystems, 29, 212-226. https://doi.org/10.1002/aqc.3153

Henderson, E. E., & Würsig, B. (2007). Behavior patterns of bottlenose dolphins in San Luis Pass, Texas. Gulf of Mexico Science, 2, 153-161. https://doi.org/10.18785/goms.2502.06

Henningsen, T., & Würsig, B. (1991). Bottle-nosed dolphins in Galveston Bay, Texas: numbers and activities. Proceedings of the Fifth Annual Conference of the European Cetacean Society, Sandefjord, Norway, December 1991.

Hornsby, F. E., McDonald, T. L., Balmer, B. C., Speakman, T. R., Mullin, K. D., Rosel, P. E., Wells, R. S., Telander, A. C., Marcy, P. W., & Klaphake, K. C. (2017). Using salinity to identify common bottlenose dolphin habitat in Barataria Bay, Louisiana, USA. Endangered Species Research, 33, 181-192. https://doi.org/10.3354/esr00807

Huggins, R. (1991). Some practical aspects of a conditional likelihood approach to capture experiments. Biometrics, 47(2), 725-732. https://doi.org/10.2307/2532158

Hutchinson, G. E. (1957). Concluding remarks. Cold Spring Harbor Symposium Quantitative Biology, 22, 415-427.

Ingram, S. N., & Rogan, E. (2002). Identifying critical areas and habitat preferences of bottlenose dolphins Tursiops truncatus. Marine Ecology Progress Series, 244, 247-255. https://doi.org/10.3354/meps244247

Irvine, A. B., Scott, M. D., Wells, R. S., & Kaufmann, J. H. (1981). Movements and activities of the Atlantic bottlenose dolphin, Tursiops truncatus, near Sarasota, Florida. Fishery Bulletin, 79(4), 671-688.

Irwin, L.-J. (2005). Marine toxins: Adverse health effects and biomonitoring with resident coastal dolphins. Aquatic Mammals, 31(2), 195. https://doi.org/DOI:10.1578/AM.31.2.2005.195

117

Irwin, L.-J., & Würsig, B. (2004). A small resident community of bottlenose dolphins, Tursiops truncatus, in Texas: monitoring recommendations. Gulf of Mexico Science, 2004(1), 13-21. https://doi.org/10.18785/goms.2201.02

Jefferson, T. A., Webber, M. A., & Pitman, R. L. (2008). Marine Mammals of the World: A Comprehensive Guide to their Identification. Elsevier, Canada, 592 p.

Johnson, J. B., & Omland, K. S. (2004). Model selection in ecology and evolution. Trends in Ecology & Evolution, 19(2), 101-108. https://doi.org/10.1016/j.tree.2003.10.013

Keller, C. A., Garrison, L. P., Baumstark, R., Ward-Geiger, L. I., & Hines, E. (2012). Application of a habitat model to define calving habitat of the North Atlantic right whale in the southeastern United States. Endangered Species Research, 18(1), 73-87. https://doi.org/10.3354/esr00413

Kritzer, J. P., & Sale, P. F. (2004). Metapopulation ecology in the sea: from Levins' model to marine ecology and fisheries science. Fish and Fisheries, 5(2), 131- 140. https://doi.org/10.1111/j.1467-2979.2004.00131.x

Lambert, C., Mannocci, L., Lehodey, P., & Ridoux, V. (2014). Predicting cetacean habitats from their energetic needs and the distribution of their prey in two contrasted tropical regions. PLoS ONE, 9(8), e105958. https://doi.org/10.1371/journal.pone.0105958

Laska, D., Speakman, T., & Fair, P. A. (2011). Community overlap of bottlenose dolphins (Tursiops truncatus) found in coastal waters near Charleston, South Carolina. Journal of Marine Animals and Their Ecology, 4, 10-18.

Litz, J. A., Ronje, E. I., Whitehead, H., & Garrison, L. P. (2019). Updated abundance estimate for common bottlenose dolphins (Tursiops truncatus) inhabiting West Bay, Texas. Aquatic Conservation: Marine and Freshwater Ecosystems, 12. https://doi.org/10.1002.aqc.3195

Longley, W. L. (1994). Freshwater inflows to Texas bays and estuaries: ecological relationships and methods for determination of needs. 426 p). Austin, TX: Texas Water Development Board and Texas Parks and Wildlife Department.

Lukacs, P. (2001). Closed population capture-recapture models. In E. G. Cooch & G. C. White (Eds.), Using MARK—a gentle introduction, 19th edition. Volume II. Advanced Methods.

Lynn, S. K., & Würsig, B. (2002). Summer movement patterns of bottlenose dolphins in a Texas bay. Gulf of Mexico Science, 20(1), 25-37. https://doi.org/10.18785/goms.2001.03

118

Macleod, C. D., Mandleberg, L., Schweder, C., Bannon, S. M., & Pierce, G. J. (2008). A comparison of approaches for modelling the occurrence of marine animals. Hydrobiologia, 612(1), 21-32. https://doi.org/10.1007/s10750-008-9491-0

Mannocci, L., Monestiez, P., Spitz, J., & Ridoux, V. (2015). Extrapolating cetacean densities beyond surveyed regions: habitat‐based predictions in the circumtropical belt. Journal of Biogeography, 42(7), 1267-1280. https://doi.org/10.1111/jbi.12530

Maze, K. S., & Würsig, B. (1999). Bottlenosed dolphins of San Luis Pass, Texas: Occurrence patterns, site-fidelity, and habitat use. Aquatic Mammals, 25, 91-104.

Mazzoil, M., Reif, J. S., Youngbluth, M., Murdoch, M. E., Bechdel, S. E., Howells, E., McCulloch, S. D., Hansen, L. J., & Bossart, G. D. (2008). Home ranges of bottlenose dolphins (Tursiops truncatus) in the Indian River Lagoon, Florida: Environmental correlates and implications for management strategies. EcoHealth, 5(3), 278-288.

McArthur, M., Brooke, B., Przeslawaki, R., Ryan, D., Lucieer, V., Nicol, S., McCallum, A., Mellin, C., Cresswell, I., & Radke, L. (2010). A review of surrogates for marine benthic biodiversity. GeoScience Australia, 42 p(2009). https://doi.org/10.1016/j.ecss.2010.03.003

McBride, S. M. (2016). Habitat Use by Bottlenose Dolphins, Tursiops truncatus, in Roanoke Sound, North Carolina. (Ph.D. Dissertation), University of Southern Mississippi, Department of Psychology (Dissertations. 886) https://aquila.usm.edu/dissertations/886

Melancon, R., Lane, S., Speakman, T., Hart, L., Sinclair, C., Adams, J., Rosel, P., & Schwacke, L. (2011). Photo-identification field and laboratory protocols utilizing Finbase version 2. (Technical Memorandum NMFS-SEFSC-627, 52 p). Lafayette, LA: National Oceanic and Atmospheric Administration

Miller, C. E., & Baltz, D. M. (2010). Environmental characterization of seasonal trends and foraging habitat of bottlenose dolphins (Tursiops truncatus) in northern Gulf of Mexico bays. Fishery Bulletin, 108(1), 79-86.

Miller, J. (2010). Species distribution modeling. Geography Compass, 4(6), 490-509. http://doi.org/10.1111/j.1749-8198.2010.00351.x

Moreno, P., & Mathews, M. (2018). Identifying foraging hotspots of bottlenose dolphins in a highly dynamic system: A method to enhance conservation in estuaries. Aquatic Mammals, 44(6), 694-710. https://doi.org/10.1578/AM.44.6.2018.694

Mullin, K. D., Rosel, P. E., Hohn, A. A., & Garrison, L. P. (2007). Bottlenose dolphin stock structure research plan for the central northern Gulf of Mexico. 27 p). Pascagoula, MS, USA: US Department of Commerce, National Oceanic and Atmospheric Administration, NOAA Technical Memorandum NMFS-SEFSC-563.

119

NMFS. (2019). National Marine Fisheries Service, National Oceanic and Atmospheric Administration, U.S. Department of Commerce. Total commercial fishery landings at an individual U. S. port for all years after 1980, https://www.st.nmfs.noaa.gov/commercial-fisheries/commercial-landings/other- specialized-programs/total-commercial-fishery-landings-at-an-individual-u-s-port- for-all-years-after-1980/index, accessed May 2019.

NOAA. (2019a). National Centers for Environmental Information, Hurricanes and Tropical Storms, National Oceanic and Atmospheric Administration, U.S. Department of Commerce. Retrieved from https://www.ncdc.noaa.gov/sotc/tropical-cyclones/ 3 October 2019

NOAA. (2019b). Office of Coast Survey, National Oceanic and Atmospheric Administration, U.S. Department of Commerce. Retrieved from https://www.nauticalcharts.noaa.gov/ 3 October 2019

O'Sullivan, D., & Unwin, D. (2010). Geographic Information Analysis (2nd ed.). Hoboken, NJ, USA: John Wiley & Sons, Inc.

Orlando, S. P. J., Rozas, L. P., Ward, G. H., & Klein, C. J. (1993). Salinity characteristics of Gulf of Mexico estuaries. 209 p). Silver Spring, MD: U.S. Department of Commerce, National Oceanic and Atmospheric Administration, Office of Ocean Resources Conservation and Assessment, NOAA's National Estuary Inventory Series.

Otis, D. L., Burnham, K. P., White, G. C., & Anderson, D. R. (1978). Statistical inference from capture data on closed animal populations. Wildlife Monographs(62), 3-135.

Phillips, N. M., & Rosel, P. E. (2014). A method for prioritizing research on common bottlenose dolphin stocks through evaluating threats and data availability: Development and application to bay, sound and estuary stocks in Texas (Technical Memorandum NMFS-SEFSC-665, 154 p). Lafayette, LA: National Oceanic and Atmospheric Administration, National Marine Fisheries Service.

Pitchford, J. L., Howard, V. A., Shelley, J. K., Serafin, B. J. S., Coleman, A. T., & Solangi, M. (2014). Predictive spatial modelling of seasonal bottlenose dolphin (Tursiops truncatus) distributions in the Mississippi Sound. Aquatic Conservation: Marine and Freshwater Ecosystems, 26(2), 289-306. https://doi.org/10.1002/aqc.2547

Piwetz, S. (2019). Common bottlenose dolphin (Tursiops truncatus) behavior in an active narrow seaport. PLoS ONE, 14(2), e0211971. https://doi.org/10.1371/journal.pone.0211971

Pollock, K. H. (1982). A capture-recapture design robust to unequal probability of capture. The Journal of Wildlife Management, 46(3), 752-757. https://doi.org/10.2307/3808568

120

R Core Team. (2018). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/.

Read, A. J., Urian, K. W., Wilson, B., & Waples, D. M. (2003). Abundance of bottlenose dolphins in the bays, sounds, and estuaries of North Carolina. Marine Mammal Science, 19(1), 59-73. https://doi.org/10.1111/j.1748-7692.2003.tb01092.x

Redfern, J., Ferguson, M., Becker, E., Hyrenbach, K., Good, C., Barlow, J., Kaschner, K., Baumgartner, M. F., Forney, K., & Ballance, L. (2006). Techniques for cetacean–habitat modeling. Marine Ecology Progress Series, 310, 271-295. https://doi.org/10.3354/meps310271

Robinson, L., Elith, J., Hobday, A., Pearson, R., Kendall, B., Possingham, H., & Richardson, A. (2011). Pushing the limits in marine species distribution modelling: lessons from the land present challenges and opportunities. Global Ecology and Biogeography, 20(6), 789-802. https://doi.org/10.1111/j.1466- 8238.2010.00636.x

Rogan, E., Cañadas, A., Macleod, K., Santos, M. B., Mikkelsen, B., Uriarte, A., Van Canneyt, O., Vázquez, J. A., & Hammond, P. S. (2017). Distribution, abundance and habitat use of deep diving cetaceans in the North-East Atlantic. Deep Sea Research Part II: Topical Studies in Oceanography, 141, 8-19. https://doi.org/10.1016/j.dsr2.2017.03.015

Ronje, E., Whitehead, H., & Mullin, K. (2018). The 1990 common bottlenose dolphin (Tursiops truncatus) mass die-off in East Matagorda Bay, Texas -- New insight into a cold case. Southeastern Naturalist, 17(3), 411-422. https://doi.org/10.1656/058.017.0306

Ronje, E. I., Barry, K. P., Sinclair, C., Grace, M. A., Barros, N., Allen, J., Balmer, B., Panike, A., Toms, C., & Mullin, K. D. (2017). A common bottlenose dolphin (Tursiops truncatus) prey handling technique for marine catfish (Ariidae) in the northern Gulf of Mexico. PLoS ONE, 12(7), e0181179. https://doi.org/10.1371/journal.pone.0181179

Rosel, P., Mullin, K., Garrison, L., Schwacke, L., Adams, J., Balmer, B., Conn, P., Conroy, M., Eguchi, T., Gorgone, A., Hohn, A., Mazzoil, M., Schwartz, C., Sinclair, C., Speakman, T., Urian, K., Vollmer, N., Wade, P., Wells, R., & Zolman, E. (2011). Photo-identification capture-mark-recapture techniques for estimating abundance of bay, sound and estuary populations of bottlenose dolphins along the U.S. east coast and Gulf of Mexico: A workshop report. (Technical Memorandum NMFS-SEFSC-621, 38 p). Lafayette, LA: National Oceanic and Atmospheric Administration.

Scott, G. P., Burn, D. M., Hansen, L. J., & Owen, R. E. (1989). Estimates of bottlenose dolphin abundance in the Gulf of Mexico from regional aerial surveys. (CRD 88/89-07, 71 p). Miami, FL: National Marine Fisheries Service.

121

Shane, S. (1977). The population biology of the Atlantic bottlenose dolphin, Tursiops truncatus, in the Aransas Pass area of Texas. Master's Thesis. In Texas A&M University, Port Aransas, TX, USA. 257 p.

Shane, S. H. (1980). Occurrence, movements, and distribution of bottlenose dolphin, Tursiops truncatus, in southern Texas. Fishery Bulletin, 78(3), 593-601.

Speakman, T. R., Lane, S. M., Schwacke, L. H., Fair, P. A., & Zolman, E. S. (2010). Mark-recapture estimates of seasonal abundance and survivorship for bottlenose dolphins (Tursiops truncatus) near Charleston, South Carolina, USA. Journal of Cetacean Research and Management, 11(2), 153-162.

Symonds, M. R., & Moussalli, A. (2011). A brief guide to model selection, multimodel inference and model averaging in behavioural ecology using Akaike’s information criterion. Behavioral Ecology and Sociobiology, 65(1), 13-21. https://doi.org/10.1007/s00265-010-1037-6

Torres, L. G., Read, A. J., & Halpin, P. (2008). Fine-scale habitat modeling of a top marine predator: do prey data improve predictive capacity. Ecological Applications, 18(7), 1702-1717. https://doi.org/10.1890/07-1455.1

Trenberth, K. E. (1983). What are the seasons? Bulletin of the American Meteorological Society, 64(11), 1276-1282. https://doi.org/10.1175/1520- 0477(1983)064<1276:WATS>2.0.CO;2

TTWP. (1998). Trans-Texas Water Program, Southeast Area. Environmental Analysis for the Neches Salt Water Barrier, Beaumont, Texas. Memorandum Report. 63 p.

TWDB. (2019). Texas Water Development Board. Estuarine Hydrographic Surveys. Retrieved from http://www.twdb.texas.gov/surfacewater/bays/surveys/index.asp August 2019

Tyson, R. B., Nowacek, S. M., & Nowacek, D. P. (2011). Community structure and abundance of bottlenose dolphins (Tursiops truncatus) in coastal waters of the northeast Gulf of Mexico. Marine Ecology Progress Series, 438, 253-265. https://doi.org/10.3354/meps09292

Urian, K., Gorgone, A., Read, A., Balmer, B., Wells, R. S., Berggren, P., Durban, J., Eguchi, T., Rayment, W., & Hammond, P. S. (2015). Recommendations for photo-identification methods used in capture-recapture models with cetaceans. Marine Mammal Science, 31(1), 298-321. https://doi.org/10.1111/mms.12141

Urian, K. W., Kaufmann, R., Waples, D. M., & Read, A. J. (2018). The prevalence of ectoparasitic barnacles discriminates stocks of Atlantic common bottlenose dolphins (Tursiops truncatus) at risk of entanglement in coastal gill net fisheries. Marine Mammal Science, 35(1), 290-299. https://doi.org/10.1111/mms.12522

122

Urian, K. W., Waples, D. M., Tyson, R. B., Hodge, L. E., & Read, A. J. (2013). Abundance of bottlenose dolphins (Tursiops truncatus) in estuarine and nearshore waters of North Carolina, USA. Journal of North Carolina Academy of Science, 129(4), 165-171. https://doi.org/10.7572/2167-5880-129.4.165

USACE. (2018). U.S. Army Corps of Engineers, Galveston District Southwestern Division. Coastal Texas protection and restoration feasibility study. Draft integrated feasibility report and environmental impact statement. 442 p.

USEPA. (1999). Ecological condition of estuaries in the Gulf of Mexico. (EPA 620-R-98- 004, 80 p). Gulf Breeze, FL: U.S. Environmental Protection Agency, Office of Research and Development, National Health and Environmental Effects Research Laboratory, Gulf Ecology Division.

Virgili, A., Racine, M., Authier, M., Monestiez, P., & Ridoux, V. (2017). Comparison of habitat models for scarcely detected species. Ecological Modelling, 346, 88-98. https://doi.org/10.1016/j.ecolmodel.2016.12.013

Wade, P. R., & Angliss, R. P. (1997). Guidelines for assessing marine mammal stocks: Report of the GAMMS Workshop April 3-5, 1996, Seattle, Washington. (NOAA Technical Memorandum NMFS-OPR-12, 97 p). Silver Springs, MD: Office of Protected Resources, National Marine Fisheries Service.

Ward, G. H. (1980). Hydrography and circulation processes of Gulf estuaries. In P. Hamilton & K. B. Macdonald (Eds.), Estuarine and Wetland Processes, Marine Science Book Series, Volume 11 (pp. 183-215). New York, NY: Springer Science+Business Media. http://doi.org/10.1007/978-1-4757-5177-2

Ward, G. H., & Armstrong, N. E. (1992). Ambient water and sediment quality of Galveston Bay: present status and historical trends. University of Texas at Austin 19-20

Waring, G. T., Josephson, E., Maze-Foley, K., & Rosel, P. E. (2016). US Atlantic and Gulf of Mexico marine mammal stock assessments - 2015. (NOAA Technical Memorandum NMFS NE 238, 501 p). Available from: National Marine Fisheries Service, 166 Water Street, Woods Hole, MA 02543-1026, or online at http://www.nefsc.noaa.gov/publications/.

Wells, R. S. (2014). Social structure and life history of bottlenose dolphins near Sarasota Bay, Florida: Insights from four decades and five generations. In J. Yamagiwa & L. Karczmarski (Eds.), Primates and Cetaceans. Primatology Monographs (pp. 149-172). Tokyo: Springer. https://doi.org/10.1007/978-4-431- 54523-1_8

Wells, R. S. (2018). Identification Methods. In B. Würsig, J. G. M. Thewissen, & K. M. Kovacs (Eds.), Encyclopedia of Marine Mammals. 3rd ed. San Diego, CA: Academic Press/Elsevier. http://dx.doi.org/10.1016/B978-0-12-804327-1.00009- 1

123

Wells, R. S. (2019). Common bottlenose dolphin foraging: behavioral solutions that incorporate habitat features and social associates. In B. Würsig (Ed.), Ethology and behavioral ecology of odontocetes. Ethology and behavioral ecology of marine mammals (pp. 331-344): Springer, Cham. https://doi.org/10.1007/978-3- 030-16663-2_15

Wells, R. S., Schwacke, L. H., Rowles, T. K., Balmer, B. C., Zolman, E., Speakman, T., Townsend, F. I., Tumlin, M. C., Barleycorn, A., & Wilkinson, K. A. (2017). Ranging patterns of common bottlenose dolphins Tursiops truncatus in Barataria Bay, Louisiana, following the Deepwater Horizon oil spill. Endangered Species Research, 33, 159-180. https://doi.org/10.3354/esr00732

Wells, R. S., & Scott, M. D. (2018). Bottlenose dolphin, Tursiops truncatus, common bottlenose dolphin. In I. Würsig, J. G. M. Thewissen, & K. M. Kovacs (Eds.), Encyclopedia of Marine Mammals. 3rd ed. San Diego, CA: Academic Press/Elsevier. https://doi.org/10.1016/B978-0-12-804327-1.00072-8

Wells, R. S., Scott, M. D., & Irvine, A. B. (1987). The social structure of free-ranging bottlenose dolphins. In H. H. Genoways (Ed.), Current Mammalogy (pp. 247- 305). Boston, MA: Springer. https://doi.org/10.1007/978-1-4757-9909-5_7

White, G. C., & Burnham, K. P. (1999). Program MARK: survival estimation from populations of marked animals. Bird Study, 46, S120-S139. https://doi.org/10.1080/00063659909477239

White, G. C., & Cooch, E. G. (2017). Population abundance estimation with heterogeneous encounter probabilities using numerical integration. The Journal of Wildlife Management, 81(2), 322-336. https://doi.org/10.1002/jwmg.21199

Whitehead, H. R., & Ronje, E. I. (2017). Case reports of out-of-habit marine mammals: common bottlenose dolphins (Tursiops truncatus) and north Atlantic right Whales (Eubalaena glacialis) in the western Gulf of Mexico. Poster presented at: 22nd Biennial Conference. 2017 October 22-28, Halifax, Nova Scotia, Canada.

Williams, B. K., Nichols, J. D., & Conroy, M. J. (2002). Analysis and management of animal populations. San Diego, CA: Academic Press.

Williams, T. M., Friedl, W. A., Fong, M. L., Yamada, R. M., Sedivy, P., & Haun, J. E. (1992). Travel at Low Energetic Cost by Swimming and Wave-Riding Bottle- Nosed Dolphins. Nature, 355(6363), 821-823. https://doi.org/10.1038/355821a0

Wilson, B., Hammond, P. S., & Thompson, P. M. (1999). Estimating size and assessing trends in a coastal bottlenose dolphin population. Ecological Applications, 9(1), 288-300. https://doi.org/10.2307/2641186

Wood, S. N. (2017). Generalized additive models: an introduction with R: Chapman and Hall/CRC. https://doi.org/10.1201/9781315370279

124

Würsig, B. (2009). Bow-riding. In Encyclopedia of marine mammals (pp. 133-134): Elsevier. https://doi.org/10.1016/B978-0-12-373553-9.00037-7

Würsig, B., & Lynn, S. K. (1996). Movements, site fidelity, and respiration patterns of bottlenose dolphins on the central Texas coast. (Technical Memorandum NMFS- SEFSC-383, 128). Miami, FL: National Oceanic and Atmospheric Administration.

Würsig, B., & Würsig, M. (1977). The photographic determination of group size, composition, and stability of coastal porpoises (Tursiops truncatus). Science, 198(4318), 755-756. http://doi.org/10.1126/science.198.4318.755

Zuur, A., Ieno, E. N., Walker, N., Saveliev, A. A., & Smith, G. M. (2009). Mixed effects models and extensions in ecology with R: Springer Science & Business Media.

Zuur, A. F., Ieno, E. N., & Elphick, C. S. (2010). A protocol for data exploration to avoid common statistical problems. Methods in Ecology and Evolution, 1(1), 3-14. https://doi.org/10.1111/j.2041-210X.2009.00001.x

125

BIOGRAPHICAL SKETCH

Errol’s major was in fisheries and aquatic sciences. A native Texan, he grew up on the Gulf Coast. While simultaneously conducting protected species research throughout the northern Gulf of Mexico for the National Marine Fisheries Service, he divided his time between the study of fish and marine mammal population ecology, abundance estimation, habitat modeling, and geographic information systems, graduating with a Master of Science degree in December of 2019.

126