AN ABSTRACT OF THE THESIS OF

Sallie Christian Beavers for the degree of Master of Science in Wildlife Science

presented on May 10. 1996. Title: The Ecology the Olive Ridley Turtle

(Lepidochelys olivacea) in the Eastern Tropical Pacific Ocean During the

Breeding/Nesting Season.

Abstract approved:

Bruce R. Mate

The ecology of the olive ridley sea turtle (Lepidochelys olivacea) in the eastern

tropical Pacific (ETP) was studied with satellite telemetry and line transect survey

methods. A satellite-monitored olive ridley turtle was tracked from 26 November

1990 to 18 March 1991. The turtle travelled at least 2,691 km in the open ocean,

averaging 1.28 km/h. The turtle's track suggests that he did not float passively with

the currents. Comparisons with surface current divergence data indicated that the

turtle encountered several large convergence and divergence zones, but did not appear

to favor either. Submergence behaviors were diel. The turtle made more frequent,

shorter-duration submergences during the day, yet also spent more time at the surface.

Nighttime submergence durations were longer and more variable. Daytime

temperatures recorded during transmissions were cooler and more variable and were

correlated with longer submergences. Two types of submergences were recorded,

short duration submergences (<7.7 min) and long dives (8.2 95.5 min). These

Redacted for privacy

behaviors exhibited the same diel pattern. Surfacing durations were long (1-10 min)

but showed no diel differences, possibly due to transmission constraints.

Hard-shelled sea turtles were sighted along shipboard line-transects from July

to December during 1989-1990 in the ETP and were used to estimate relative

densities of turtles for distribution and habitat association analyses. The probability of

animal detection decreases with distance from the trackline, and is further affected by

environmental conditions, observer, and time of day. Therefore, I used detectability

analysis to adjust the effective area surveyed for these variables with least squares

regression analysis and a standard estimator of effective halfwidth surveyed. Sea state

was the primary influence on sea turtle detection. Effective halfwidth was adjusted

for each sea state category, and then effective area surveyed and turtle density were

estimated.

The detectability model was applied to the 1986-1990 data to estimate density

of turtles in daily survey-areas. Maps of relative density distributions showed

concentrations off Baja, Mexico, nearshore from Mexico to Ecuador and out about

700 km, along the equatorial convergence zone, and to a lesser extent along the north

equatorial divergence and the Peru Current boundary. Species distributions were

consistent with reported ranges. El Nifio events appear to influence distribution

patterns. Only identified olive ridleys and unidentified turtles (most likely ridleys)

were used in habitat analyses. Presence of sea turtles was significantly correlated

with surface water density, distance to land, and thermocline depth. Densities of

turtles, where present, were minimally correlated with surface water density, distance

to land, thermocline depth, and chlorophyll concentration (R2 = 0.22).

The Ecology of the Olive Ridley Turtle (Lepidochelys olivacea) in theEastern Tropical Pacific Ocean During the Breeding/Nesting Season

by

Sallie Christian Beavers

A THESIS

submitted to

Oregon State University

in partial fulfillment ofthe requirements for the

degree of

Master of Science

Completed May 10, 1996Commencement June 1996

Master of Science thesis of Sallie Christian Beavers presented on May 10, 1996

APPROVED:

essor, representing Wildlife Science

Chair of Department of Fisheries and Wildlife

I understand that my thesis will become part of the permanent collection of OregonState University libraries. My signature below authorizes release of my thesis to anyreader upon request.

Sallie Christian Beavers, Author

Redacted for privacy

Redacted for privacy

Redacted for privacy

Redacted for privacy

Acknowledgments

I sincerely thank my major Professor, Dr. Bruce R. Mate, for the opportunity

to work with him, for his confidence in my abilities, and for his valuable insights on

my research. I thank the members of my committee, Drs. William G. Pearcy, W.

Daniel Edge, Stephen B. Reilly, and Erik Fritzell, for their helpful discussions and

comments regarding this thesis. In addition to my committee, there are several

individuals who contributed to my academic experience. I am grateful to Dr. Fred L.

Ramsey for the opportunity to work with him and to learn from him. I thank Dr.

Richard A. By les for encouraging me to undertake this project, and for his generosity

and friendship. Dr. Ken Norris taught me to be a field biologist and always had great

faith in my abilities; to him I am especially grateful. This project would not have

been possible without the cooperation of many fine people at the National Marine

Fisheries Service. I am especially indebted to Dr. Stephen B. Reilly, Dr. Paul

Fiedler, and Dr. Tim Gerrodette for generously sharing their environmental data with

me and answering my many questions; to Robert L. Pitman for sharing his sea turtle

data; and to Dr. Doug De Master for allowing me to initiate this project during the

Monitoring of Porpoise Stocks cruises. Robert Holland, Al Jackson, and Valerie

Philbrick cheerfully helped in locating data and providing graphics. Field work could

not have been completed without the support of the officers and crew of the NOAA

research vessels David Starr Jordan and McArthur, and the many oceanographers and

observers who collected data over the years. These dedicated people made the four-

month cruises fun and wonderful experiences. I am grateful to all, and my thanks

especially to Wes Armstrong, Scott Benson, Peter Boveng, Jim Caretta, Edward

Cassano, Jim Cotton, Darlene Everhart, Gary Friedrichsen, Jim Gilpatrick, Bill

Irwin, Sue Kruse, Carrie LeDuc, Rick Le Duc, Scott Leopold, Lisa Lierheimer, Mark

Lowry, Valerie Philbrick, Robert Pitman, Joe Raffetto, Keith Rittmaster, Richard

Rowlette, Scott Sinclair, Brian Smith, Victoria Thayer, and Robin Westlake. Many

people generously provided materials, data, and technical support for this project.

Special thanks are due Don Pacropis and Den-Mat Corp., and Tony Campegna and

3M Corp. for providing the satellite-tag adhesive materials; to Dr. Masa Kamachi of

Florida State University for providing the gridded current data; and to Sea World

San Diego for allowing me to test tag-attachments on their turtles. I am indebted to

Toby Martin and Martha Winsor for their computer programming expertise, advice,

and countless stimulating conversations about telemetry and statistics. I thank Amy

Cutting for her assistance with data entry. Various chapters of this thesis benefited

from reviews by Dr. Paul Crone, Dr. Scott Eckert, Sharon Nieukirk, and Dr. Robert

Steidl. I am grateful to my fellow students Robert Weber, Dr. Robert Steidl, Dr.

Paul Crone, Sharon Nieukirk, Nobuya Suzuki, Mike Lantana, Robert Neely, Lisa

Lierheimer, and Barbara Lagerquist for their support, and for many long discussions

that helped to clarify my thinking and often raised new questions to explore. I extend

my heartfelt gratitude to Lisa Jade Lierheimer for her help as an oceanographer in the

field and especially for her unfailing love and friendship. Finally, special thanks to

Michael Buchal for making the completion of this thesis (nearly) stress-free. This

research was partially funded by National Marine Fisheries Service contract

#40JGNF210366 and was conducted under ESA permit no. 691.

This thesis is dedicated to the memory of my mother, Ode lle Milling Christian,who showed me how to rise to the occasion and to roll with the punches.

Contribution Of Authors

Edward R. Cassano was involved in the planning, testing, and fieldwork of

chapter one. Dr. Fred L. Ramsey developed detectability analysis (Ramsey et a1.

1987), and all mathematical derivations are his. Dr. Ramsey was involved in the

analysis and interpretation of data in chapter two, and in the writing of the procedure

section of that chapter.

Table of Contents

Page

Introduction 1

Movements and Dive Behavior of a Male Sea Turtle (Lepidochelys olivacea) inthe Eastern Tropical Pacific. 6

INTRODUCTION 7

METHODS 7

RESULTS 14

DISCUSSION 20

Detectability Analysis In Transect Surveys 26

INTRODUCTION 27

METHODS 28

RESULTS 37

DISCUSSION 46

Distribution and Habitat Associations of Sea Turtles in the Eastern TropicalPacific Ocean During the Breeding/Nesting Season 50

INTRODUCTION 51

METHODS 52

RESULTS 61

DISCUSSION 76

Table of Contents (Continued)

Page

Summary 85

Bibliography 89

Appendices 100

APPENDIX 1 101

APPENDIX 2 102

Figure

1.1.

List of Figures

Page

Surface currents and movements of a satellite-monitored male oliveridley sea turtle (Lepidochelys olivacea) in the eastern tropical Pacific. . 9

1.2. Data collection and transmission schedule for a satellite-monitored maleolive ridley sea turtle in the eastern tropical Pacific, 1990-91. 11

1.3. Sampled submergence durations (10 s-95.5 min) of a satellite-monitoredmale olive ridley sea turtle in the eastern tropical Pacific 18

1.4. Temperatures recorded at the time of transmission from a satellite-monitored male olive ridley sea turtle in the eastern tropical Pacific. . 19

2.1. Study area and tracklines surveyed by the David Starr Jordan and theMcArthur in the eastern tropical Pacific Ocean, 1989 -90. 33

2.2. Covariates air temperature (ATMP), sea surface temperature (ssT), andBeaufort state (BERT) plotted against ungrouped perpendicular sightingdistances (meters) of sea turtles in the eastern tropicalPacific, 1989-90. 38

2.3. Effective halfwidth (meters) and 95% confidence intervals surveyed atBeaufort sea state levels 0-5 in the eastern tropical Pacific, 1989-90. . . 44

2.4. Density distribution of hard-shelled (cheloniid) sea turtles in the easterntropical Pacific, 1989-90. 45

3.1. Surface water masses and primary surface water circulation of theeastern tropical Pacific Ocean (From Reilly and Fiedler 1994, withpermission). 53

3.2. Combined survey tracklines for the David Starr Jordan and the McArthurin the eastern tropical Pacific during August-November, 1986-90(From Reilly and Fiedler 1994, with permission). 56

3.3. Distribution of identified cheloniid sea turtle species in the easterntropical Pacific, 1986-90. 62

3.4. Relative density distribution of cheloniid sea turtles in the easterntropical Pacific, 1986-90. 64

List of Figures (Continued)

Figure 1e

3.5. Relative density distribution of cheloniid sea turtles in the easterntropical Pacific, 1986. 65

3.6. Relative density distribution of cheloniid sea turtles in the easterntropical Pacific, 1987. 66

3.7. Relative density distribution of cheloniid sea turtles in the easterntropical Pacific, 1988. 67

3.8. Relative density distribution of cheloniid sea turtles in the easterntropical Pacific, 1989. 68

3.9. Relative density distribution of cheloniid sea turtles in the easterntropical Pacific, 1990. 69

3.10. Relative density distribution of cheloniid sea turtles in the easterntropical Pacific from line transect sighting data 1989-90. 70

List of Tables

Table Page

1.1. Mean summary and sampled submergence data for a male olive ridleyin the eastern tropical Pacific per day/night period,29 Nov 1990 18 March 1991. 17

2.1. Forward stepwise selection of variables collected with sea turtle sightingsin the eastern tropical Pacific, 1989-90. 39

2.2. Analysis of variance results for regression model of the log ofperpendicular distance to sea turtles sighted in the ETP, 1989-90,on Beaufort sea state. 41

2.3. Models and parameter number (p) compared by the Mean Square Error(MSE), Cp (Neter et al. 1989), AIC (Akaike 1974), and BIC(Schwarz 1978, Kass and Raftery 1995) selection criteria. 42

3.1. Average ()) ± standard deviation (SD) and associated coefficient ofvariation (CV), and mode of habitat variables used with 1989-90 seaturtle sighting data in the eastern tropical Pacific (ETP). 71

3.2. Logistic regression model parameter estimates for presence of sea turtlesin the eastern tropical Pacific, 1989-90. 73

3.3. Multiple linear regression parameter estimates and analysis of variancefor association of sea turtle density with habitat variables in the easterntropical Pacific, 1989-90. 75

Preface

Chapter one of this thesis was accepted for publication before Dr. Pamela T.

Plotkin's dissertation, Migratory and reproductive behavior of the olive ridley turtle,

Lepidochelys olivacea (Eschscholtz, 1829), in the eastern Pacific Ocean, became

available. I recommend that any reader interested in the movements and behavior of

olive ridley turtles in the eastern tropical Pacific read her excellent work.

The Ecology of the Olive Ridley Turtle (Lepidochelys olivacea) in theEastern Tropical Pacific Ocean During the Breeding/Nesting Season

Introduction

Currently all marine turtle species, except the indigenous Australian flatback

turtle (Natator depressus), are considered threatened or endangered under the

Endangered Species Act of 1973. Despite management programs to protect marine

turtles (primarily in nesting areas), most populations continue to decline (Limpus

1995). It is becoming apparent that intensive protection of female turtles and eggs on

nesting beaches does not necessarily ensure population growth (Crouse et al. 1987).

Frazer (1992:179) eloquently argued for intervention at the sources that threaten sea

turtle populations, rather than attempting "...to release more turtles into a degraded

environment in which their parents have already demonstrated they cannot flourish."

Sea turtles are at increasing risk in the marine environment, where they are threatened

by exposure to fisheries activities, plastic marine debris, and pollutants (Hutchinson

and Simmonds 1992).

Marine turtle management and recovery plans consistently specify studies of

turtle distribution, and quantitative characterization of marine turtle habitats as

research priorities (e.g., Hopkins and Richardson 1984, Thompson et al. 1990).

Correlations of some turtle species with sea surface temperature, water depth, and the

Gulf Stream boundary have been reported off the eastern coast of the United States

(Hoffman and Fritts 1982, Shoop and Kenny 1992, Keinath 1993, Coles et al. 1994,

Epperly et al. 1995a), and the potential influences of ocean currents on the dispersal

2

of young sea turtles has been considered (Collard and Ogren 1990, Witham 1991,

Hall et al. 1992). Aside from these studies, little is known about the biological and

physical influences on turtle distribution and abundance in open water (Carr 1986),

and hypotheses about how pelagic turtles forage, migrate and avoid predators in this

environment are controversial. Hughes and Richard (1974) proposed that sea turtles

migrate by passively riding with ocean currents, but this is not supported by recent

tracking studies (Plotkin 1994). Carr et al. (1978) and Carr (1986, 1987) also

suggested that pelagic turtles use surface currents as transportation aids, and frontal

convergence zones for forage and protection. Witham (1991) argued that

aggregations of organisms at fronts heighten the risk of predation and that food

resources are widely available away from fronts (e.g., Collard 1992).

The lack of information on the pelagic ecology of marine turtles is due to the

difficulties and prohibitive costs of long-term monitoring and extensive dedicated

surveys. The advent of remote tracking-technologies (e.g., Stoneburner 1982, Eckert

et al. 1989, Plotkin et al. 1995), and data collected from "platforms of opportunity"

(e.g., Shoop and Kenny 1992) are beginning to provide new information on turtles in

open water. The 1986-1990 National Marine Fisheries Service (NMFS) Monitoring

of Porpoise Stocks (MOPS) project in the eastern tropical Pacific, ETP (Wade and

Gerrodette 1992) provided an opportunity to apply satellite-telemetry methods to track

an olive ridley turtle (Lepidochelys olivacea) at sea, and to collect data on the

distribution and relative surface density of cheloniid (hard-shelled) turtles, particularly

the olive ridley, in relation to surface features and oceanographic indices (Fiedler et

al. 1990, Reilly and Fiedler 1994).

3

The olive ridley is probably the most abundant and widespread sea turtle in the

ETP (Clifton et al. 1982, Pitman 1990), ranging from Mexico to Peru with primary

breeding populations occurring in Mexico and Costa Rica (Cornelius and Robinson

1986, Marquez 1990). It was the only cheloniid turtle identified in the open ocean in

this study. The olive ridley is listed as threatened, with the exception of endangered

Mexican breeding populations. Populations have declined since the mid-1960's as a

result of intensive take in Mexican and Ecuadoran turtle fisheries, exploitation of

eggs, and incidental take in coastal and offshore fisheries (Clifton et al. 1982,

Cornelius and Robinson 1986). Although commercial harvests have ceased (Frazier

and Salas 1982, Aridjis 1990), olive ridley populations are still at risk from fishery

interactions and poaching.

Three other cheloniid sea turtles inhabit the ETP (described by Marquez 1990)

and are included in a portion of the studies herein. Eastern Pacific green turtles

(Chelonia mydas) range in coastal waters from Mexico to Peru. Nesting populations

are on mainland Mexico (endangered), Costa Rica, Ecuador, and the Galapagos and

Revillagigedos Islands. Loggerhead turtles (Caretta caretta) are concentrated off Baja

California, Mexico, and do not nest in the east Pacific but are likely juveniles from

threatened Japanese breeding stocks (Bowen et al. 1995). The endangered hawksbill

turtle (Eretmochelys imbricata) nests from Mexico to Ecuador and primarily resides in

the coastal littoral zone of the mainland and Galapagos Islands (Green and Ortiz-

Crespo 1982). These three species inhabit primarily continental shelf or island waters

and are infrequently seen at sea (Pitman 1990, 1993, pers. obs.). Although many

nesting beaches are protected and most commercial harvests of adults and eggs have

4

halted, populations are still affected by poaching and incidental take in fisheries

(Marquez 1990).

Steele (1974) noted that in the study of ecosystems, answers to the questions:

where are species distributed, how many species occur there, why do they occur

there, and how do they behave, are unlikely to be gained from a single method and

that any chosen method will emphasize a different pattern or spatial/temporal scale.

The studies reported in this thesis approach the question of the relationship of sea

turtles to the ocean environment of the ETP in two ways: (1) an active, fine-scale

approach of following a single animal over the course of time through its

environment, and (2) a static, large-scale approach of marine turtle distribution in

relation to several environmental variables. Each reveals aspects of the ecology of

sea turtles in open water that would not have been obtainable from the other.

The objectives of the studies reported herein were:

1. To examine the movements and submergence behavior of an olive ridley sea

turtle in relation to geographic surface features and surface currents with

regard to the hypotheses of current-assisted transport and "preference" for

frontal zones.

2. To evaluate and refine data-collection protocols for future satellite-telemetry

studies of olive ridley turtles.

3. To develop a detectability model for cheloniid sea turtles (olive ridley,

loggerhead, and "unidentified" turtles) that adjusts effective area surveyed for

conditions that may affect turtle detection during shipboard surveys, and to use

the model to calculate relative surface-densities by daily area surveyed.

5

4. To compare the combined relative surface-density distribution of cheloniid

species in the ETP for the 1986-1990 breeding/nesting seasons with surface

features and determine if density distribution is concentrated in convergence

zones.

5. To quantitatively correlate olive ridley turtle distribution with habitat variables

and determine whether variables previously used to describe ETP dolphin

habitats (Au and Perryman 1985, Reilly 1990, Reilly and Fiedler 1994) can be

similarly used as descriptors of sea turtle presence or density during the peak

nesting season.

The information gained from these studies provides a foundation for future

research on olive ridley turtles in the open ocean and is important for the conservation

of these turtles in their marine habitat.

6

Chapter 1

Movements and Dive Behavior of a Male Sea Turtle (Lepidochelys olivacea)in the Eastern Tropical Pacific.

Sallie C. Beavers and Edward R. Cassano

Journal of Herpetology, 1996 Vol 3:97-104Reprinted with Permission

7

INTRODUCTION

Most information on sea turtle biology is based on adult females, their eggs,

and hatchlings at nesting beaches. Little is known of the ecology of sea turtles after

they leave the beach, particularly of the adult males, which rarely emerge from the

ocean (Whittow and Balazs 1982). Until recently, logistics and expense limited

researchers to studying the near-shore movements and long-distance, point-to-point

migrations of sea turtles with methods such as aerial survey, conventional radio

telemetry, sonic telemetry, marking, and flipper tagging.

In 1979, the use of satellite-monitored radio transmitters enabled researchers to

follow the movements of turtles into open water (Stoneburner 1982, Timko and Kolz,

1982). Subsequently, the use of satellite telemetry has grown, primarily as a means

to investigate the inter- and postnesting movements of female turtles (Stoneburner

1982, Timko and Kolz 1982, Duron-Dufrenne 1987, Nakamura and Soma 1988, Hays

et al. 1991), juvenile turtles in near-shore areas (By les 1988, Keinath 1993), and

submergence behavior (Nakamura and Soma 1988, Keinath and Musick 1993, Renaud

and Carpenter 1994). We conducted this pilot study to examine the movements and

submergence patterns of a male olive ridley turtle (Lepidochelys olivacea) and to

refine data-collection protocols for future remote-sensing studies of sea turtles in the

open ocean.

METHODS

The eastern tropical Pacific Ocean (ETP) includes the region bounded by

25°N-20°S and the American continents to 140°W. The ETP contains several

8

principal surface currents and fronts (Wyrtki 1965, 1966; Fig. 1.1), and a permanent

shallow thermocline (20-100 m) that is sometimes approximated by the depth of the

20 °C isotherm (Donguy and Meyers 1987, Fiedler 1992). Biologically, the area is

productive and supports a diverse fauna (Wyrtki 1966, Reilly and Fiedler 1994).

We used the Service Argos (Landover, MD) satellite data collection and

location system to monitor the sea turtle (Fancy et al. 1988). In our low-latitude

study area, 7°-18°N, satellites were "visible" to the transmitter eight or nine times

per day for an average of 10 min per pass. We attached a Telonics (Mesa, AZ)

Model ST-3 platform transmitter terminal (PTT; 10 cm x 15 cm x 3 cm; 822 g) to a

large male olive ridley (straight carapace length 67.0 cm, width 59.7 cm, weight 39

kg, tail length: carapace-tip of tail 18.0 cm). The turtle was captured in a breakaway

net taped to a 1.8 m x 0.9 m metal frame that was lowered on a long pole from the

ship's bow.

Before deployment, the PTT was wrapped in Scotch Cast (3M, Saint Paul,

MN), a fiberglass casting material, and painted with anti-fouling paint. The casting

material provided a bonding surface between the PTT and the adhesive, Geri-Stor

(Den-Mat Corp., Santa Maria, CA), a light-cured dental compound. Geri-Stor is

lightweight, generates little heat, and cures in 8 min. Tabs of casting material

extended from the PTT base to increase the attachment surface area. The PT!' was

attached to the carapace over the second neural scute. Flipper tags J0503 and X329

were attached to the front flippers. We released the animal within an hour at

17°08'N, 110°59'W.

0 400 800

20 N

Kilometers

FINAL TRANSMISSION18 MARCH 1991

-10N

MEXICO

Islas Revillagigedo

1 DecI

RELEASE26 NOVEMBER 1990

1 Mar

120W

3 Jan

2 Feb

41m NEC am

Clipperton

110W

NECC im

9

Figure 1.1. Surface currents and movements of a satellite-monitored male oliveridley sea turtle (Lepidochelys olivacea) in the eastern tropical Pacific. The NorthEquatorial Current (NEC) flows westward between 10° and 20°N, and the NorthEquatorial Countercurrent (NECC) flows eastward between 4° and 10°N. Adivergence (upwelling) is between the currents at about 8°30' 10°N. The westerlySouth Equatorial Current (SEC) is between 4°N and 10°S, and a strong convergence(downwelling) is between the NECC and the SEC. The California Current (CC)penetrates the ETP December April at 20°-10°N (Wyrtki 1965, 1966). Large opencircles correspond to the first transmission of each month from release 26 November1990 to final transmission 18 March 1991. Closed circles represent satellite-acquiredlocations 1 (N = 66).

10

The PTT broadcast a 1-watt signal at 401.650 MHz. Transmission uplinks

were made at 61 s intervals and lasted 0.44 s. A saltwater switch, formed by two

external conductivity electrodes on the PTT, allowed transmissions only when the

electrodes were out of the water and counted surfacing events. Because the turtle

passed through two time zones, +7 h (97°30'W-112°30'W) and +8 h (112°30'W-

127°30'W), we report transmission times in Greenwich Mean Time (GMT).

Each transmission consisted of an identification code and 64 bits of encoded

summary and sampled sensor data. Summary data were the number of submergences,

and the average duration of those submergences. For each tracking-day, these data

were summarized by the transmitter into fixed 12-h periods that closely corresponded

to night (0018-1218 GMT) and day (1218-0018 GMT; Fig. 1.2). Percentage of time

at surface was calculated from these data. Sampled data were the duration of the

most recent submergence preceding transmission, and the internal temperature of the

PTT at the time of transmission. To conserve battery power, the PTT was

programmed to transmit the previous period's summary data and the current sampled

data only during the last 6 h of each 12-h period (Fig. 1.2; Night, 0618-1218 GMT;

Day, 1818-0018 GMT). This regimen, or "duty cycle", limited the number of

satellite passes that could receive data to six per day. Satellite coverage during the

study was equal in both periods.

We defined a submergence as lasting at least 10 s. The PTT repeated the

previously transmitted submergence-duration until a new submergence occurred.

Thus, even if the turtle had been on the surface and the PTT transmitting for minutes,

or hours, before the satellite entered reception range, the duration of the most recent

COLLECT

TRANSMIT

11

NIGHT SUMMARY DATA DAY SUMMARY DATA

I

OFF

I

Sampled Data &Previous DaySummary Data

I

OFFSampled Data &Previous NightSummary Data

GMT o 1 2 3 4 5 6 7 8 9 10 11 12 13 11 15 16 17 18 19 20 21 22 23 0 1

+7 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18+8 16 17 18 19

.20

,21 22 23 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

Sunset 27 Nov 15' N 111' W Sunrise

Sunset 17 Mar 10'N 121' W Sunrise

Figure 1.2. Data collection and transmission schedule for a satellite-monitored maleolive ridley sea turtle in the eastern tropical Pacific, 1990-91. Summary data fromthe preceding period were transmitted during (0618-1218, and 1818-0018 GMT).Sampled data were collected and transmitted during (0618-1218, and 1818-0018GMT). Local time zone descriptions are represented by +7 (<112° 30' N) and +8(>112° 30' N). Sunrise and sunset times are given for representative locations anddates for the beginning and end of the study.

12

sampled submergence was still available. We assumed that submergences of equal

duration were unlikely to occur sequentially, and defined transmissions of an identical

submergence-duration repeated at 61 s intervals as "Dry Transmissions." A single

Dry Transmission indicates the turtle was at the surface for 61 s with the PTT

exposed. We used these transmissions to investigate surface bouts. However, by not

clearing this sensor field, some surface information was lost and surface-duration was

likely underestimated.

Herein, we use the general term submergence and reserve "dive" for long

submergences. We used log-survivorship analysis (e.g. Slater 1974, Machlis 1977,

Fagen and Young 1978, Slater and Lester 1982) to examine the sampled

submergence-durations and assign a duration criteria to differentiate between

submergences and "dives". Because the distributions of the submergence data did not

fit the typical log-survivorship pattern (Appendix 1), we modified Machlis' (1977)

method for minimizing misclassification of behavior type by modeling the data as

overlapping negative exponential and normal distributions. We chose the point at

which a submergence had an equal probability of being in either distribution as the

criterion for minimizing the misclassification of a submergence in a behavior type.

Service Argos locates the PIT by Doppler shift of the PTT frequency as the

satellite passes overhead. At the time of our study, Service Argos classified location

reliability (+1 SD) as the following location classes (LC): LC 3 (± 150 m), LC 2 (±

350 m), LC 1 (± 1,000 m), and LC 0 (accuracy unknown). However, accuracy was

poor; Argos (1989) estimated that at least 66% of the computed locations attained the

indicated accuracies. We used only LC 1 or better in analysis. Distance data

13

(calculated with the Great Circle Equation; Bowditch 1977:1258) and transmission

times were used to calculate minimum rates of movement between locations.

We used modeled-current magnitude and direction vectors supplied by the

Mesoscale Air-Sea Interaction Group, Florida State University (FSU), to visually

compare the movements and locations of the turtle with local surface currents and

surface current divergence (au/ax + av/ay, 10' s') to describe the turtle's

movements associated with these features. Current vectors were calculated at 1/4°

grid resolution (Kamachi and O'Brien 1995) forced by ship-reported wind data

(Stricherz et al. 1992). The FSU data were available in six day intervals (N = 22).

Surface currents in 1° grids and surface current divergence in smoothed 2° grids were

centered on each turtle location of the same date or + 1 day. We conservatively used

divergence values -0.1 x 10-6 s' as criteria for determining a location was at a

convergence, and values __0.1 x 10-6 s' as criteria for a location at a divergence.

All data transmissions were screened by a logical-consistency error checking

program (Oregon State University) and rechecked by hand. Summary submergence

data were tested for serial correlation by using partial autocorrelation analysis up to

lag 100. No serial correlation was found and the data were assumed independent.

Submergence frequency data were log-transformed to meet assumptions of parametric

statistical tests, however arithmetic means and standard errors are reported for

convenience. Equality of variances were tested with a variance ratio test (two-tailed).

Student's t-test (two-tailed) was used to test submergence day/night differences, and

Welch's approximate t-test was used when variances were unequal (Zar 1984). Chi-

square goodness of fit analyses with Yates correction for continuity was used to test

14

diel differences in number of satellite passes and transmissions (Zar 1984). The

Mann-Whitney U-test (two-tailed) was used to compare day/night Dry Transmissions

and the temperature data. We excluded temperatures transmitted with Dry

Transmissions from analysis because they represent known warming at the surface

rather than the most recent temperature recorded after a submergence.

RESULTS

We monitored the turtle from 27 November 1990 to 18 March 1991. During

113 days we received 779 transmissions from the PTT (g = 6.9 transmissions/day,

SE = 0.50, range = 0-30 transmissions/day), which provided 121 locations; 66 were

LC 1 ± 1,000 m). More transmissions (61%) were received during the day.

The mean number of transmissions received per pass was equal in day (g = 4.3) and

night (g = 4.2). Eighty-eight transmissions (11.3%) contained errors and were

discarded. Data from summary periods in which only one transmission was received

(N = 14) were also discarded. The PTT recorded 11,012 error-free submergences.

Movements

Following the 27 November 1990 release (Fig. 1.1) the turtle moved 118.4

km northeast to 18°00'N, 110°15'W on 1 December, and then southeast (-140°)

with the California Current (CC). After 4 December, the turtle turned south. He

approached the westerly North Equatorial Current (NEC) at 13°7'N, entering an area

of convergence between the CC and NEC on 16 December (see Appendix 2).

Continuing south, the turtle passed through a convergence on 6 and 7 January, and on

15

16 January 1991 was 64 km (± 350 m) west of Clipperton Island. He then turned

southwest across the NEC and crossed the divergence between the NEC and the

easterly North Equatorial Counter Current (NECC) at 9°30'N on 20 January. The

turtle crossed a convergence on 22 January and maintained a southwesterly heading

across the NECC. The turtle was in a strong convergence on 2 February, and

reached 7°N on 6 February. He moved northwest for the remainder of the month,

and crossed the divergence (8°30'N) back into the NEC between 18 22 February.

During the first two weeks of March, the turtle headed west-northwest with the NEC.

On 15 March, the turtle turned north at 120°W in a divergence area. The final

transmission was on 18 March 1991 in a convergence at 11°05'N, 120°20'W.

The turtle traveled at least 2,691 km during 113 days (Fig. 1.1). Calculated

speeds based on minimum distances between locations ranged from 0.19 km/h to

3.88 km/h (R = 1.28 km/h, SE = 0.10, N = 65). Visual comparison of the turtle's

track with bottom topography showed no observable association with particular depth

contours or sea mounts. Throughout the study, the turtle was in tropical surface

waters historically characterized by high, stable temperatures (__25 °C), low salinity

( 34.00), and a strong, fairly shallow (60-80 m at study area) thermocline (Wyrtki

1966, Fiedler 1992).

Submergence Behavior

All summary and sampled data variables differed between day and night

(P < 0.0001). The turtle made 1.6 times as many submergences during the daytime

summary period as in the night, yet also spent more time at the surface, resulting in

16

shorter average submergence times (Table 1.1). Nighttime submergence durations

were more variable than during the day (F58,51 = 2.7, P = 0.0005).

The distribution of sampled submergence-durations (Table 1.1; range =

10 s-95.5 min, N = 338) was bimodal for both periods (Fig. 1.3), suggesting that

two distinct types of submergence behavior exist. We called these surface-active

submergences and dives. Sampled submergences greater than 7.7 min (the point at

which misclassifying a behavior is minimized, a 0.17% chance of being in either

distribution) were classified as "dives" based on our model of overlapping

distributions. As in the summary data, mean duration of the dives (Table 1) sampled

at night was greater (range = 8.2-95.5 min) than during the day (range = 11.1-47.6

min), and nighttime dive durations were more variable (F44,50 = 4.1; P < 0.0001).

Surface Behavior

We found no diel difference in the percentage of transmissions that were Dry

Transmissions (Night: k = 48.8%, SE = 4.0, N = 52; Day: R = 45.3%, SE = 4.0,

N = 59, P = 0.60). In passes where Dry Transmissions occurred (73.5%), mean

surface-duration calculated from sequential Dry Transmissions did not differ between

periods (Table 1.1, Night: range = 1-7 min; Day: range 1-10 min, P = 0.25).

Temperature

The first temperature transmitted in each pass was consistently lower than, or

equal to, the subsequent temperatures in the pass. We report results from first

transmissions only (N = 149). Daytime temperatures were variable, ranging between

Table 1.1. Mean summary and sampled submergence data for a male olive ridley in the eastern tropical Pacific per day/night

period, 29 Nov. 1990 18 March 1991. Summary Data were summarized during 12-h night (0018-1218 GMT) and day (1218-

0018 GMT) periods. Sampled submergence and temperature data were collected and transmitted during the 6 h night (0618-

1218 MT) and day (1818-0018 GMT) duty cycle. Sequential Dry Transmissions (surface-duration) did not differ between

periods (P = 0.25)a. All other data are significantly different between periods (P < 0.0001).

Summary Data Sampled Data

Submergence % Time at Dive Duration Sequential Dry Temperature (°C)

No. Dives Duration (min) Surface (min) Transmissions(min)a

Period N x SE x SE x SE N SE N x SE N x SE

Night 59 73 6.2 12.3 0.91 11.7 0.65 45 49.8 2.07 59 2.4 0.19 62 31.6 0.19

Day 52 129 15.3 7.15 0.59 17.3 0.65 51 35.5 0.96 67 2.9 0.25 87 26.9 0.43

.a

18

80 -/E20

15

>s101

0C ga) -

Night /

Dives

Day

10

15

20

151 m

I

Dives

152 iII,I.IiiiiiI,I,IiI,I,ILI0 8 16 24 32 40 48 56 64 72 80 88 96

Dive Duration (min)

Figure 1.3. Sampled submergence durations (10 s-95.5 min) of a satellite-monitoredmale olive ridley sea turtle in the eastern tropical Pacific. The means of day/nightdives are significantly different (P < 0.0001). We retained the 95.5 min dive inanalysis because we have no reason to suspect that it is an inaccurate data point, andits removal did not affect the significance level of the t-test.

36

18

J(l'A1 NA ell cp \Co 53 cp \<0 (t) 0A p(k

Date 1990 - 1991

Figure 1.4. Temperatures recorded at the time of transmission from a satellite-monitored male olive ridley sea turtle in the

eastern tropical Pacific, 1990-91. Closed circles represent nighttime temperatures, shaded triangles represent daytime

temperatures.

20

19.9-33.6 °C (Table 1, Fig. 1.4), and cool temperatures were correlated with longer

dive durations (Spearman Rank Correlation, r = 0.76, P < 0.0001). Temperatures

recorded at transmission were warmer at night (Table 1.1). Nighttime temperatures

ranged 24.5-33.1 °C and 93.6% of those were over 30.0 °C. There was no

correlation between temperature and dive durations at night (r = -0.02, P = 0.85).

Patterns of transmitted temperature and sampled submergences changed after 2

February. The transmitted temperatures dropped between 27 January and 2 February,

the subsequent daytime temperatures remained cool (Fig. 1.4), and fewer surface-

active submergences were sampled after 2 February (Before: NB = 208, After: NA

=27; X2 = 89.2, P < 0 .0001). These submergences averaged 31 s (SE = 7.5),

and 74% were _.30 s, suggesting wave action across the saltwater switch. The

frequency of dives increased (NB = 39, NA = 51; X2 = 7.1, P = 0.008), but their

mean duration did not change (R, = 41.3 min, SE = 1.3; t = 0.82, 55 df, P =

0.41), and they were followed by more Dry Transmissions (NB = 150, NA = 153; X2

= 8.8, P = 0.003). Rate of travel did not change (RA = 1.20 km/h, SE = 0.15;

t= 0.23, df = 57, P = 0.81).

DISCUSSION

Movements

Our results (Fig. 1.1), and the observations of others (Cornelius and Robinson

1986, Au 1991, Pitman 1993) support the hypothesis that Pacific olive ridleys spend

much of their time in pelagic waters (Hendrickson 1980). How olive ridleys make

21

use of the pelagic surface currents and surface features during their life cycle is not

known. Carr et al. (1978) and Carr (1986, 1987) suggested that pelagic olive ridleys

utilize surface currents as an aid to transport, and frontal convergence zones for

forage and protection. Witham (1991) countered that food resources are also

available away from convergences and risk of predation is increased at fronts.

The turtle's movement across the prevailing currents suggests some active

swimming rather than continuous passive floating with the currents. The sharp

change in reported temperatures between 27 January and 2 February (Fig. 1.4) and

the cusp-shape of the turtle's track at that time (Fig. 1.1) suggest that the turtle

encountered a temperature front, possibly associated with a mesoscale eddy (Hansen

and Paul 1984), which may represent a site of food concentration as indicated by the

convergence at the 2 February location. Although the turtle in our study encountered

large-scale fronts, we found no evidence to support or refute the hypothesis that sea

turtles specifically concentrate their time in convergences (Appendix 2). However,

fronts occur on many temporal and spatial scales from meters and hours to kilometers

and days (Fedorov 1983), and a turtle may cue on a finer scale than we examined.

Information on the diets of Pacific olive ridleys in pelagic waters would provide

insight into this question, however prey items have only been reported from turtles in

nearshore areas and at oceanic islands (Fritts 1981, Montenegro and Bernal 1982,

Mortimer 1982, Barra* et al. 1992). It is likely that olive ridleys in the open ocean

are opportunistic foragers, feeding at fronts, on epipelagic organisms, or on vertical

migrators when and where food is available rather than consistently targeting a

particular food source.

22

Submergence Behavior

The turtle performed two distinguishable behaviors, surface-active

submergences and dives, which were diel in pattern (Fig. 1.3). The high frequency

of daytime surface-active submergences may reflect either short submergences while

at the surface or wave action across the saltwater switch if the turtle sometimes

floated in a "low basking posture" (Sapsford and van der Reit 1979). Daytime dives

were more often associated with lower temperatures, suggesting that the turtle made

occasional dives to deeper, cooler waters. Nighttime dives were longer and less

frequent, with less time spent in surface-active behavior. Consistently warm

nighttime temperatures suggest that the turtle either made shallow, long dives in the

surface layer or deep dives, and rose to the surface slowly, which allowed the PTT to

warm before surfacing. Renaud and Carpenter (1994) also reported warmer night

temperatures from two loggerhead turtles (Caretta caretta). Although the stratified

thermal structure of the mixed layer in our study area is stable (Wrytki 1966, Fiedler

1992), without concurrent dive profiles it is impossible to be certain if temperatures

reflect dive depths.

Diel dive patterns have been reported for leatherback (Dermochelys coriacea)

(Eckert et al. 1986, 1989, Keinath and Musick 1993) and loggerhead turtles

(Sakamoto et al. 1990, Renaud and Carpenter 1994). Other satellite telemetry studies

describe summary period (Renaud and Carpenter 1994) and sampled data (Keinath

and Musick 1993) similar to ours. Like our turtle, leatherback turtles spent more of

their time at the surface during the day (Eckert et al. 1986, 1989). Further, daytime

dive depths of the leatherbacks were variable, whereas nighttime dives were typically

23

shallow. The temperature data reported from our turtle, more variable during the day

and warmer at night, is consistent with the leatherbacks' dive patterns. Diel

differences in behavior could reflect resting and thermoregulation patterns, foraging

patterns (Eckert et al. 1986, 1989), traveling, avoiding predators, or a combination of

these and other unknown factors.

Surface Behavior

There was no diel difference in the turtle's surface behavior. In the ETP,

olive ridley turtles have been observed in the daytime floating with their carapaces

well exposed, often long enough for seabirds to use them for roosting (Pitman 1993).

We were curious to know whether this behavior indicated longer diurnal surfacings.

Although our turtle spent more time at the surface during the day, it appears to be

due to frequent surfacings rather than a difference in the transmitted behavior while at

the surface. The mean surface durations were similar between day and night, as were

the percentages of Dry Transmissions per period. However, we can not rule out the

possibility that these data were biased by our collection methods. The observation

period of the satellite ( 10 min/pass) may have been too short to measure the length

of surface bouts (e.g. Slater 1974, Fagen and Young 1978), and surface bouts were

underestimated because the submergence sensor did not clear after transmission.

Also, daytime surfacings were frequently transmitted with very short submergences,

which may be a result of wave action rather than deliberate submersion (see Keinath

and Musick 1993).

24

PTT Performance

Our transmission time of 113 days was very good given that the PTT had a

45-day battery life at maximum transmission. The turtle's surfacing behavior

permitted many transmissions per day, indicating that detailed dive data for specific

times of day can be archived and transmitted if the data collection format is planned

appropriately. Summary periods and the duty cycle should be shorter, but not so

short that there is a consistent loss of certain periods from orbit and transmission

constraints. Our long duty cycle limited battery life. The number of bits should be

increased to the maximum 256 bits, to enable transmission of sequences of archived

submergence data and prevent bias against long dives. Long dives are less likely to

be reported during the sampling period unless the turtle surfaces just as a pass is

made. Because our PTT software did not clear the submergence sensor field until a

new submergence was made, this bias was reduced, but not eliminated. Surface-

duration data should also be collected as a summary variable to avoid short sampling

periods. A depth sensor should be included. Because errors occur during

transmission, the data stream should contain either an error detection or correction

code. Sampling rate (Boyd 1993), sample size (Machlis et al. 1985), and the analysis

of time-series data are essential criteria in designing the sensor software, as are

physical tests of the tag housing, batteries, and sensors before deployment.

Satellite telemetry is an effective method for tracking sea turtles; these data

would have been difficult to obtain with conventional tracking methods. Daily

submergence behavior, dive depth, diet, and internal body temperatures from a larger

25

sample of olive ridleys at sea are needed for comparisons among cheloniid sea turtles,

and with leatherback sea turtles.

26

Chapter 2

Detectability Analysis In Transect Surveys

Sallie C. Beavers and Fred L. Ramsey

Department of Fisheries and Wildlife and Department of StatisticsOregon State University, Corvallis, OR

Submitted to Journal of Wildlife Management

27

INTRODUCTION

Line transect sampling relies on the assumption that the probability of

detection of an animal typically decreases with increasing distance from the transect.

Several factors, such as environmental conditions, observer, and time of day, can

further influence the detection of animals (e.g. Scott and Gilbert 1982, Verner 1985,

Holt and Cologne 1987, By les 1988, Marsh and Saalfeld 1989, Buck land et al. 1992)

thus affecting the estimates of effective area surveyed and population density (Ramsey

et al. 1987). Most commonly, these factors are recognized but not included in

analysis (e.g., Scott and Gilbert 1982, By les 1988), mitigated by restricting survey

time to optimum conditions (e.g., Epperly et al. 1995b), or the data set is stratified

relative to the factors (e.g., Holt 1987, Buck land et al. 1992, 1993). However,

restricted survey time and stratification can cause sample size problems, and

conditions can vary within these restrictions. An alternative approach is to include

the influencing factors as covariates in the analysis and adjust the estimates of

effective area surveyed (Ramsey et al. 1987).

We illustrate this method with line transect data, including potential covariates,

collected on sea turtle sightings during 1989-90 of the National Marine Fisheries

Service (NMFS) Monitoring of Porpoise Stocks (MOPS) project in the eastern

tropical Pacific (ETP; Wade and Gerrodette 1992). In this paper we identify sea state

from several covariates as the primary influence on the detection of sea turtles,

incorporate this factor in a model to estimate effective area surveyed, and estimate the

surface density of sea turtles in the ETP.

28

METHODS

Procedure

The theory for transect surveys is presented by Seber (1982) and by Burnham

and Anderson (1976). Buck land et al. (1993) provide a comprehensive treatment.

In a survey, let N be the number of animals present in the target region, R.

Let n be the number of animals detected by the transect survey. The perpendicular

distance from the transect to the ith detected animal is the detection distance, y,. A

detectability function, g(y), equals the conditional probability that an animal is

detected, given that its distance from the transect is y.

The basic results of transect theory are

1. E {n} = Da, where D = N/Area(R) is the population density, and where a is

the effective area surveyed.

2. The effective area surveyed in a line transect of length L is a = 2*L*w, where

w is the effective halfwidth,

= g(y) dy

3. Conditional on n, the set of perpendicular detection distances, fyl, ,y,il are

independent with common probability density function f(y) = g(y)/w.

The following model of varying detectability comes from Ramsey (1979) and

Ramsey et al. (1987). Associated with the ith detected animal is a set of covariates,

xi = (x11, ,x0) describing detectability conditions. Assume that the covariates alter

29

the effective halfwidth multiplicatively so that the effective halfwidth surveyed under

conditions x, is 0.) where

log (6);) = Po EoJ 1.1

:1=1

If it is also assumed that the effective halfwidth is a scale parameter in the

distribution f(y) , then the ordinary least squares regression of {log(yi); i = 1, , n}

on the covariates provides an unbiased estimate of the parameters {01, , (3p). The

least squares estimates of {01, , (3p} are not as efficient as maximum likelihood

estimates (MLE), but obtaining MLEs of WI, , Op} and 00 relies on knowing the

shape of the detectability function. Multiple linear regression will not, however, yield

a sensible estimate for the constant, (30. A maximum likelihood approach is possible,

but also relies on the shape of the detectability function. Likelihood theory indicates

that the estimate of the constant term is approximately independent of the estimates of

the regression parameters if the covariates are centered at their averages (Ramsey et

al. 1987).

This theory suggests the following non-parametric strategy:

1. Estimate the parameters {fib , fip} by least squares regression with log(y)

as the response.

2. Determine average detectability conditions for the covariates, X.

3. Adjust all detection distances to the average conditions according to

= y. x exp(.i= 1

30

4. Select a non-parametric estimator of the effective halfwidth, (.-40, at average

conditions using the adjusted detection distances. Programs DISTANCE

(Laake 1994) or CUMD (Wildman and Ramsey 1985, Scott et al. 1986)

provide an estimate of effective halfwidth.

5. Estimate the constant term with

po = log (60)

This strategy enables the researcher to estimate the effective halfwidth under

the prevailing detectability conditions x(t) = (x, (t), , x, (t)) at any point, t, along a

transect:

6(0 = cbo x exp(ESi(xj(t)i=1

In the multiplicative model, estimation is performed on the logarithmic scale, where

the variance of this estimate is

Var(60) P P\far llog [6) (t)]1 + E E kit)- [xt(t) c'iwoppti.

fri

In this expression, the variance of i.soo comes from the estimate provided by

DISTANCE or CUMD, and the estimated covariances for the regression coefficients

come from least squares analysis.

31

Then, the total effective area surveyed is estimated by

AT = 2 f (t) dt

where T represents the totality of transect segments in the region. Dividing the total

number of detections, nr, by the total effective area surveyed yields a final estimate of

density (DT) .

Factors affecting animal populations usually do so multiplicatively, therefore

statistical inferences concerning densities are also constructed on a logarithmic scale.

The following variances are approximate. First calculate the quantities

2 f (xj(t)-7vi)6(t)dt

B TA T

Then, the variance for the log density is

for j = 1,...,p.

Var(nr) Var(60) P PE E BiBt coy rilj,f3j.

n TVar {log(A)}

226o j=1 k =1

The first term can be approximated as in Burnham et al. (1980:54) by setting

\far (n 7.) = 2(n 7.). The sum of the last two terms is the variance of the log(AT). An

approximate 95% confidence interval for the log density is the log estimate +1.96

times the square root of this variance. The exponentiated endpoints provide the

density interval.

Densities may also be compared in two subregions or from the same region in

different years by estimating their ratio. In doing so, the baseline effective halfwidth

32

estimate, (Jo, cancels. The comparison is therefore unaffected by the choice of

detectability model in DISTANCE, and is equivalent to a comparison of raw counts

adjusted for differing detectability conditions, as recommended by Verner (1985).

For testing and confidence intervals on the ratio of densities in two regions, the

variance of the log-ratio is approximately

\far (ni) Var (112) P Par b2) + E E (B11 B2j)(Bli. B21) czn, oppki

n 21

n22 j = 1 k =1

In this equation, the B-coefficients are calculated separately in the two surveys or

subregions. Inference is then drawn to multiplicative factors affecting density, which

agrees with log-linear analysis in Poisson regression (McCullagh and Nelder 1983).

Survey Methods

The ETP extends from a narrow continental shelf along the west coast of the

Americas out to about 140°W between 25°N and 20°S, and encompasses more than

19 million km2 (Fig. 2.1; Holt and Sexton 1990). The area is biologically productive

(Fiedler et al. 1991) and supports several species of cheloniid (hard-shelled) sea

turtles (Marquez 1990): the olive ridley (Lepidochelys olivacea), the eastern Pacific

green (Chelonia mydas), the hawksbill (Eretmochelys imbricata), and the loggerhead

(Caretta caretta).

Sea turtle sighting data were collected from 23 August to 5 December 1989

and 3 August to 5 December 1990 aboard the National Oceanographic and

Atmospheric Administration (NOAA) research vessel David Starr Jordan, and from 6

3

a

30

20

10

10

20

\ / --,,,,

< --. ----- / -?.-. i .....\ ..,

\\ \ V '''' l (\ /7 / / ,.\ '----. - /'\ I \ I \ \/(/ 1.4f. fr..1.\\ zi -7"--- 3.'" t I /1 ,(,.4.4

: .... x/ ----

---- 1-->---fL- -) -"I\

--- /.--- .-/-

Y 2 /

160 150 140 130 120 110Longitude

100 90 80

Figure 2.1. Study area and tracklines surveyed by the David Starr Jordan and the McArthur in the eastern tropical PacificOcean, 1989-90. Broken lines represent daily search effort.

34

October to 6 December 1989 and 30 July to 5 December 1990 aboard the NOAA R/V

McArthur (Fig. 2.1). Survey dates coincided with the peak nesting season for some

sea turtle populations in the ETP (Richard and Hughes 1972, Marquez 1990). The

ships surveyed in different areas of the ETP, maintained speeds of approximately

18.5 km/h, and remained underway at night to allow greater spatial distribution of the

tracklines.

Each vessel carried two teams of three experienced observers. Port and

starboard-side observers used 25x binoculars mounted on the upper-most deck,

approximately 10.7 m above the sea surface, to search from the trackline out to 90°

on their respective sides. The third observer recorded data and searched the trackline

with the naked eye and with hand-held 7x50 binoculars. Teams rotated equally

through the three duty-stations in 2-h shifts. The bearing and radial distances of

turtles sighted with the 25x binoculars were obtained from 360°-graduated washers

mounted at the binocular base and from distance-calibrated reticles etched in the right

ocular. Perpendicular distance (y) was calculated as radial distance multiplied by the

sine of the sighting angle. When turtles were sighted with 7x binoculars or by naked

eye, observers estimated the perpendicular distance from the transect.

Records of environmental conditions were kept throughout the day's effort.

The sighting time (TIME), the observer team on duty (TEAM, 1-8), the Beaufort-scale

sea state (BFRT; where Beaufort 0 = flat, calm seas, and Beaufort 5 = rough, white-

capped seas with some spray; Bowditch 1977:1267), the absence of sun (NosuN), and

the vertical (vsuN) and horizontal (HsuN) position of the sun (sun glare) in relation to

the trackline were recorded at each sighting. We used a clock-face schematic, with

35

12 o'clock at the trackline, to classify HSUN. The VSUN was classified as a 1/4

clock-face with 3 o'clock at the horizon and 12 o'clock directly overhead. Sea surface

temperature (ssT), collected by thermosalinograph, and dry-bulb air temperature

(ATMP) were recorded hourly. Temperatures used for analysis were within one

half-hour of a sighting.

The number of turtles, their relative size, activity, associations with debris or

other species, and species (if known) were also recorded. Identification of turtle

species at sea is difficult, therefore all hard-shelled (Cheloniidae) species were

combined for these analyses. When possible, species were identified by trained

observers or identified later from photographs.

Daily search-effort continued from sunrise to sunset in BFRT conditions 5.

Transect lengths were unequal and effort was occasionally interrupted when the ship

left the trackline to identify cetacean species.

Analysis

Three assumptions must be met for line transect methods to be reliable: (1)

turtles exactly on the transect are always detected, (2) prior to detection, turtles do

not move in response to the ship, and (3) measurements of turtle position are accurate

with no rounding errors. Some turtles on the transect are likely to be missed because

they are submerged. An adjustment for submerged animals can be applied to

estimates of density (e.g., By les 1988, Hiby and Hammond 1989), but it requires

adequate information on diurnal submergence behavior of each turtle species,

therefore we estimate surface density only. We have not seen evidence of turtles

36

avoiding the ship. Sightings were typically of sea turtles floating motionless at the

surface. Turtles on or near the trackline remained at the surface as the ship passed

by, or dived as the ship's bow wave touched them. Turtles floating off the trackline

did not appear to respond to the ship. Given the speed of the vessels relative to the

motion of the turtles, it is unlikely that a turtle was counted more than once per

transect. Observers have a tendency to round measurements, especially when

estimating distance perpendicular to the transect. Rounding problems were alleviated

prior to adjusting detection distances by grouping data at 20-m intervals.

The influence of environmental and other factors on the detectability of sea

turtles was initially examined with scatter plots of ungrouped perpendicular distance

against each factor. Horizontal sun data were transformed with the equation

Cos(27r(HsuN/12 sighting angle/360)) to examine glare in relation to each sighting.

Vertical sun data were transformed with Cos(27(vsuN/12)). We built a log-linear

multiple regression model of the detection distances on the influencing factors by

stepwise variable selection (F-to-enter = 4), and tested it against several models

containing quadratic and interaction terms. Final model selection was made using the

Bayes Information Criterion (BIC; Schwarz 1978, Kass and Raftery 1995) assisted by

the C, statistic (Neter et al. 1989) and Akaike's Information Criterion (AIC; Akaike

1974). Detection distances were adjusted to average detectability conditions with

parameters from the final model. The adjusted distances, truncated at 1,200 m, were

submitted to programs DISTANCE and CUMD, which provided estimates of the

effective halfwidth surveyed under average conditions. Effective halfwidths under

non-average detectability conditions were then estimated by applying model

37

adjustments to the estimate under average conditions. Adjusted area surveyed was

calculated from distance traveled per sighting condition, and density was estimated

from the number of sightings in the adjusted area.

RESULTS

Detectability Analysis

During the 1989-90 surveys, 346 transects, totaling 50,024 km, were searched

and 579 sea turtles were detected under a variety of conditions. Most (90%) of

survey effort was in BFRT 3, and the average BFRT during sightings was 1.87 (SD=

1.54). Sea surface temperature at sightings ranged 18.3-31.2 °C (average = 28.2

°C, SD = 2.3, n = 577), and ATMP ranged 19.0-30.5 °C (average = 27.7 °C, SD

= 1.8, n = 572). Equipment failure caused the loss of nine temperature values.

Average sighting time was 12:57 (range 05:57-18:44). Sightings were made in all

sun conditions and 55% of the sightings were made while some glare was present in

the search area.

Scatterplots of the variables indicated that BFRT, SST, and ATMP may influence

detection (Fig. 2.2). However, the relationship of SST and ATMP to detection

distances could be an artifact of too few observations at low temperatures and the

influence of wind (BFRT) on temperature; higher winds result in cooling by convection

of SST, and cooler ATMP.

Forward stepwise selection of the possible covariates (Table 2.1) led to a

regression model with BFRT as the only significant covariate (slope: -0.444, SE =

8000

Ea)0CcoU)

64000

U_

Ca)e-a)

0_

i

1 1

0 1 2 3 4Beaufort Sea State

5 18 20 22 24 26 28 30Air Temperature (°C)

I I I

. . .

. ..

-.- ':'-',..: - ., ....;:.....:3'4: -,:;;;..--: . ., 4..,..

" - .:. ........:.;,,..t.....11. wii-r '.1 .1

16 18 20 22 24 26 28 30Sea Surface Temperature (°C)

Figure 2.2. Covariates air temperature (ATMP), sea surface temperature (SST), and Beaufort state (BFRT) plotted againstungrouped perpendicular sighting distances (meters) of sea turtles in the eastern tropical Pacific, 1989-90.

39

Table 2.1. Forward stepwise selection of variables collected with sea turtle sightingsin the eastern tropical Pacific, 1989-90. F-to-enter set as 4.0. BFRT = Beaufort seastate, SST = sea surface temperature, ATMP = air temperature, TIME = time ofsighting, TEAM = observer team, NOSUN = the absence of sun, VSUN = verticalposition of the sun, and HSUN = horizontal position of the sun.

Variable

F to Enter

df Step 1 df Step 2

BFRT 578 129.14

SST 569 44.71 568 0.23

ATMP 569 36.68 568 0.02

TIME 578 12.21 577 1.97

TEAM 572 6.47 570 1.25

NOSUN 578 5.22 577 0.12

HSUN 578 3.64 577 0.13

VSUN 578 2.31 577 0.11

HSUN x VSUN 578 0.37 577 0.89

40

0.039, t = -11.17, P < 0.0001; Table 2.2). We compared the Beaufort model and

several candidate models (Table 2.3). Two more complex models were also

suggested by the data analysis. One, the sea state temperature model, (Table 2.3)

contained an SST x ATMP interaction term that was strong in higher Beaufort

conditions and possibly reflects some decreased detectability in poor weather

associated with either factor. The relationship between environmental temperatures

and sea turtle presence at the surface could be related to thermoregulation behavior

(Heath and McGinnis 1980) and may affect the number of turtles observed at the

surface, but not necessarily the distribution of detection distances. A second model of

the sun variables (Table 2.3) contained the interaction VSUN x HSUN suggesting

decreased detectability for sea turtles in the line of the sun when it is low to the

horizon. However, because vessel speed is fairly slow and the turtles are floating, the

entire field of view is eventually surveyed without sun-glare (Holt 1987) and the

distribution of the perpendicular detection distances should not be affected.

Although these more complex models are suggested by liberal model selection

criteria such as the C, statistic, they are eliminated by the more conservative BIC

(Table 2.3), which emphasizes the best model-fit with the least complexity in

parameters (Lutkepohl 1985, Kass and Raftery 1995).

We adjusted detection distances of sea turtles to average detectability

conditions with the model

log(yi) = po (-0.444)(7,-B,RT BFRT).

41

Table 2.2 Analysis of variance results for regression model of the log ofperpendicular distance to sea turtles sighted in the ETP, 1989-90, on Beaufort seastate.

Sum of MeanSource df Squares Square F-value P Adj R2

Model 1 269.62 269.62 129.14 <0.0001 0.1815

Residual 577 1,204.65 2.09

Total 578 1,474.27

42

Table 2.3. Models and parameter number (p) compared by the Mean Square Error(MSE), Cp (Neter et al. 1989), AIC (Akaike 1974), and BIC (Schwarz 1978, Kassand Raftery 1995) selection criteria. Lowest number in bold type indicates modelselected by that criterion. n = 570 because of seven missing values in ATMP and twomissing values in SST. BFRT = Beaufort sea state, SST = sea surface temperature,ATMP = air temperature, TEAM = observer team, NOSUN = the absence of sun, VSUN= vertical position of the sun, and HSUN = horizontal position of the sun.

Model p MSE CP AIC BIC

All Main Effects 15 2.085 35.7 437.0 471.7

Full 2nd order model 41 1.981 33.4 448.9 514.0

All Interactions 35 2.001 32.6 465.4 617.5

NOSUN, HSUN, VSUN, 5 2.477 136.3 527.0 548.8HSUN x VSUN

BFRT, ATMP, SST, 5SST X ATMP 2.071 22.1 425.0 446.7

BFRT, TEAM 3 2.070 25.7 432.7 471.8

BFRT 2 2.084 23.0 422.6 431.3

43

The hazard-rate model (Ramsey 1979, Buck land 1985) with a 2-term cosine

adjustment of the detection function, g(y), provided the best fit to the standardized

distance data submitted to program DISTANCE (X 2 = 3.83, df = 5, P = 0.52). The

estimated effective halfwidth under average conditions was 192 m (SE = 19.9 m).

The CUMD estimate was also 192 m. Effective halfwidths searched during various

Beaufort conditions (Fig. 2.3) were calculated as

Effective halfwidth = 192 m (0.641 )(BFRT BFRT)

Survey Results

Total effective area surveyed was 8,678 km2 (95% CI 6,879-10,947 km2).

Estimated turtle surface density was 0.067 turtles per km2 (95% CI 0.051-0.086

turtles/km2). Turtles were detected 8.9 to 2,825 km from the nearest mainland shore.

Two main concentrations of density were from the central American coast out to

about 700 km, and along the Equatorial Convergence around 4°N (Wyrtki 1966) out

to 125°W (Fig. 2.4). Other areas were along the southern boundary of the study

area, off Baja California, Mexico, and near the Reviallagigedos Is., Mexico (ca.

19°N, 111°W).

An additional 108 turtles were detected while off survey-effort. Twenty

percent of all sightings were identified to species. Of these, 90% were olive ridley,

8% were loggerhead, and 2% were green turtles. Olive ridleys were near nesting

areas and in open water. All loggerhead turtles were off Baja California, Mexico.

The green turtles were near islands. Our identifications and the known distributions

of turtles in the ETP (Marquez 1990) suggest that most sightings were olive ridley

44

600

500

E_c 400475

.,---

To 300Ia)>

.15 200a)

w100

0

ii

0 1 2 3 4

Beaufort Sea State

5

Figure 2.3. Effective halfwidth (meters) and 95% confidence intervals surveyed atBeaufort sea state levels 0-5 in the eastern tropical Pacific, 1989-90.

ZON,

ION

0

10S

0

0. 0

0

o °

0 0 o0 0 0 r _

00R§)0 0 O 0

0 O. 96C6c6°8.( 00 o ao.

P 0

00 ..0 4 . O. O.. O0 , %9 . 0 0 ° Q . . .0- Q . 0 Po'

0. ° 0 . . o

0

0 0.01-0.05 0 0.05-0.10 0.5-1.0 0 >1.0

110W 100W 90W BOW

45

Figure 2.4. Density distribution of hard-shelled (cheloniid) sea turtles in the easterntropical Pacific, 1989-90. Symbols are centered on the local noon position of eachday's trackline. Small closed points indicate no turtles detected. From smallest tolargest, the open circles represent: 0.01-0.05 turtles/km2, 0.05-0.1 turtles/km2, 0.1-0.5 turtles/km2, 0.5-1.0 turtles/km2, and turtles/km2.

46

turtles. Species identified during effort and used in analyses were olive ridley and

loggerhead turtles.

We compared surface densities between 1989 and 1990. In 1989, surface

density was 0.12 turtles/km2 (95% CI 0.092-0.15 turtles/km2; nT = 396, AT = 3,316

km2, 95% CI 2,653-4,144 km2), and in 1990, surface density was 0.04 turtles/km2

(95% CI 0.029-0.054 turtles/km2; nT = 182, AT = 4,581.9 km2, 95% CI 3,600-5,831

km2). The 1990 area does not include transects west of 125°W (Fig. 2.1) as that area

was not surveyed in 1989, but does include the additional transects surveyed east of

125°W in August and September 1990. The ratio of the two density estimates was

3.0 (95 % CI 2.34-3.86 turtles/km2), indicating a significant difference between years.

Unusually high counts of turtles were recorded along five transects in

September and November 1989 (n = 261) and along only one transect in October

1990 (n = 20) 29 to 440 km off olive ridley nesting beaches in Mexico and Costa

Rica. All identifications (8.8%) were olive ridley turtles. These loose aggregations

were likely precursors to the mass nesting-assemblages characteristic of some

populations of olive ridleys (Richard and Hughes 1972, Plotkin et al. 1995).

DISCUSSION

Survey Data

Of the variables we examined, sea state was the primary factor influencing the

detection of sea turtles. Sighting distances decreased in high Beaufort conditions.

Sea state has also been cited by several authors as a primary factor in affecting sea

47

turtle detections in aerial surveys (Scott and Gilbert 1982, By les 1988, Marsh and

Saalfeld 1989). However, other unexamined variables may bias our density estimates.

Potential detection biases are individual turtle size, and behavioral and morphological

differences among species. Larger turtles are probably seen farther from the

trackline. Size can be entered as a covariate in detectability analysis; however,

although individuals of all sizes were seen in this study, size-data collection was too

inconsistent among observers for inclusion. Species morphology may influence

detectability. For example, although relatively small, olive ridley turtles have domed

carapaces that make them well suited for sighting from ships.

Estimated turtle density in the ETP is a minimum because submerged turtles

were missed. Some authors have calculated correction factors for submerged animals

(e.g., Bayliss 1986, Barlow et al. 1988, By les 1988), but note that activity (i.e.,

feeding, migrating) and species differences may cause variation in dive behavior. A

correction factor for some species might be estimated from the limited telemetry data

available on pelagic turtles in the ETP (Balazs 1994, Plotkin 1994, Beavers and

Cassano 1996), but would be grossly imprecise. Alternatively, Forney et al. (1991)

point out that if the data are to be used as indices of relative abundance and the

proportion of animals detected at the surface in an area does not change over time, a

submergence correction factor is unnecessary. Both alternatives depend on improved

species identification.

Surface density differences between years is likely due to survey location and

timing, and sea turtle reproductive behavior rather than true density differences.

Surface densities can vary widely, at least within 440 km of certain nesting areas,

48

depending on the timing of olive ridley mass nesting-aggregations during peak nesting

season. Also, between nestings, olive ridley turtles that remain in nearshore, shallow

waters spend more time submerged during the day than post-nesting turtles in the

open ocean (Plotkin 1994), indicating that Forney et al.'s (1991) approach would be

more applicable in deeper, offshore regions. The reproductive behavior of olive

ridley turtles in particular (Richard and Hughes 1972, Plotkin 1994) should be

considered in the timing of nearshore transects and analysis of future surveys.

Detectability Analysis

The use of detectability analysis to adjust distances for sighting conditions is

applicable in variable-distance surveys where sighting conditions are systematically

recorded. This analysis relies on the assumptions that effective area surveyed is a

multiplicative scaling parameter for detection area, and is a link to the influencing

covariates (Ramsey et al. 1987). Because detectability conditions vary,

survey-specific adjustments are desirable and can be modelled easily with this

procedure. For factors that vary with distance, this method provides a simple

alternative to stratifying data or restricting survey conditions. However, other factors

that directly influence the animals' movements with respect to the observer, such as

large variations in vessel speed, or height and speed of aerial survey craft, affect the

shape of the detection function and therefore cannot be modelled with this method.

Some researchers have used covariate methods with sighting rate as the

response variable to identify detection biases (Scott and Gilbert 1982), and to adjust

sighting rates accordingly (Gunnlaugsson and Sigurjonsson 1990). However

49

corrections for estimates that are calculated from the same data to be adjusted are not

independent (Scott and Gilbert 1982). Adjusting effective area surveyed rather than

adjusting density is appropriate because sighting and environmental conditions do not

directly affect true density, they alter the area the observer is able to effectively

search. Therefore, by adjusting the area surveyed, the detection bias is eliminated.

50

Chapter 3

Distribution and Habitat Associations of Sea Turtles in the Eastern TropicalPacific Ocean During the Breeding/Nesting Season.

Sallie Christian Beavers

Department of Fisheries and WildlifeOregon State University

Corvallis, Oregon

51

INTRODUCTION

Physical and biological oceanographic features have been used to describe the

distribution and habitats of large pelagic vertebrates such as tuna (e.g., Blackburn

1965, Sund et al. 1981) dolphins (e.g., Reilly 1977, 1990, Au and Perryman 1985,

Polacheck 1987, Fiedler and Reilly 1994, Reilly and Fiedler 1994) and whales (e.g.,

Volkov and Moroz 1977, Reilly and Thayer 1990) in the eastern tropical Pacific

Ocean (ETP). Recent literature records the pelagic distribution of sea turtles,

particularly of olive ridley turtles (Lepidochelys olivacea), in this region (Au 1991,

Pitman 1990, 1993, Plotkin 1994, Beavers and Cassano 1996 (Chapter 1), Chapter 2);

however, marine turtle distribution in relation to the oceanic environment has not

been studied in detail. Obtaining information on the influence of oceanographic

features on the distribution of sea turtles is an important first step towards defining

critical habitat of threatened and endangered marine turtle species (Thompson et al.

1990).

The 1986-1990 National Marine Fisheries Service (NMFS) Monitoring of

Porpoise Stocks (MOPS) project in the ETP (Wade and Gerrodette 1992) provided an

opportunity to investigate the influence of several oceanographic features on sea turtle

distribution in the eastern Pacific. In particular, frontal zones are of interest because

they have been suggested as important oceanic foraging areas for turtles (Carr 1978,

Carr et al. 1986, 1987). The objectives of this chapter were to 1) examine the

relative density distribution of cheloniid (hard-shelled) sea turtles in the ETP with

respect to large-scale oceanographic features, and 2) investigate whether the presence

and the relative density of olive ridley sea turtles in the ETP can be described by

52

simple habitat models developed from the same oceanographic habitat variables used

to describe ETP dolphin habitats (Au and Perryman 1985, Reilly 1990, Reilly and

Fiedler 1994).

METHODS

Study Area

The ETP extends from a narrow continental shelf along the west coast of the

Americas out to about 140°W between 25°N and 20°S and encompasses more than

19 million km' (Holt and Sexton 1990). The regional oceanography is defined by

several water masses and surface currents (Wyrtki 1966, 1967; Fig. 3.1), and a

permanent shallow thermocline (20-100 m) that is sometimes approximated by the

depth of the 20 °C isotherm (Donguy and Meyers 1987, Fiedler 1992). The three

surface-water masses in the ETP are categorized as; (1) tropical surface water (TSW),

centered in the ETP along 10°N and characterized by very warm temperatures and

low salinity; (2) subtropical surface water (SSW), cool, high-salinity water positioned

at the poleward ends of the ETP; and (3) equatorial surface water (ESW), cool water

of intermediate salinity, located between the TSW and the southern SSW masses

extending west along the Equator from the coast of Peru. Upwelling occurs locally in

coastal areas, along 10°N, at the Equator, and at the Costa Rica Dome (a semi-

permanent cyclonic eddy, ca. 9°N and 89°W; Wyrtki 1964).

The two westerly currents, the North Equatorial Current (NEC) between 10°

and 20°N, and the South Equatorial Current (SEC) between 4°N and 10°S are driven

30

20

:;-t; 0

53

10-

20160 150 140 130 120 110 100 90

SUBTROPICAL...SURFACE WATER

North Equatorial Current

MEXICO

TROPICAL

North Equatorial Countercurrent SURFACE WATERCosta Rica Dom.

c992tPri.51Front............. EQUATORIAL ......South Equatorial Current SURFACE WATER Galapagos " ECUADOR

PERU

SUBTROPICAL

SURFACE WATER 91.))

Longitude

80

Figure 3.1. Surface water masses and primary surface water circulation of theeastern tropical Pacific Ocean (From Reilly and Fiedler 1994, with permission).

70

54

by the northeast and southeast trade winds and are fed by the two eastern boundary

currents, the California Current and the Peru Current. The Intertropical Convergence

Zone (ITCZ), where the trade winds meet and weaken, lies between 5° and 10°N.

Because the winds are light there (the doldrums), the North Equatorial Countercurrent

(NECC) is able to flow opposite the SEC and NEC currents, "down" the eastward

horizontal pressure gradient between 4°N and 10°N. The seasonal movement of the

ITCZ north of 10°N in the boreal "summer" half of the year (June December), and

south to near the Equator in "winter" (January June) causes the NECC to strengthen

in September December and to weaken in February April east of 120°W (Fiedler

1992).

The thermocline shoals eastward near the coast (40-60 m) from > 150 m

offshore and is strongest under TSW, but weakens in September November (Fiedler

1992). A divergence (upwelling), and thus "ridging" of the thermocline is between

the NEC and NECC at about 8°30'N to 10°N. A strong convergence (downwelling)

lies between the NECC and the SEC at the equatorial front, causing a "trough" in the

thermocline. These features are intensified during the boreal summer.

Biologically, the area is productive (Fiedler et al. 1991) and supports a diverse

fauna (e.g., Wyrtki 1966, Reilly and Fiedler 1994), including several species of

cheloniid sea turtles (Marquez 1990). The olive ridley is the most abundant sea turtle

in the ETP (Clifton et al. 1982, Marquez 1990) and widely distributed in open water

(Au 1991, Pitman 1990, Pitman 1993, Plotkin 1994, Chapter 2). The eastern Pacific

green turtles (Chelonia mydas) are primarily distributed around islands and coastal

areas (Marquez 1990). Sightings of hawksbill turtles (Eretmochelys imbricata) are

55

rare in the eastern Pacific, particularly offshore (Marquez 1990, Pitman 1993), and

loggerhead turtles (Caretta caretta) are common in the cooler waters off Baja

California, Mexico (Marquez 1990).

Data Collection

Sea turtle sightings and oceanographic data were collected yearly between

1986 and 1990, from 28 July to 6 December aboard the National Oceanographic and

Atmospheric Administration (NOAA) research vessels David Starr Jordan and

McArthur during the MOPS project (Fig. 3.2). Detailed line transect data on sea

turtles were collected only during the final two years of the project (see Chapter 2).

Survey dates coincided with the peak nesting season for some sea turtle populations in

the ETP (Richard and Hughes 1972, Marquez 1990).

The two ships simultaneously surveyed in different areas of the ETP,

maintained speeds of approximately 18.5 km/h during sunrise to sunset survey effort,

and remained underway at night to allow greater spatial distribution of the tracklines.

Each vessel carried two teams of three experienced observers. Port and starboard-

side observers used 25x binoculars to search from the trackline out to 90° on their

respective sides. The third observer recorded sightings, and searched the trackline

with the naked eye and with hand-held 7x50 binoculars. Teams rotated equally

through the three duty-stations in 2-h shifts.

Identification of turtle species at sea is difficult; when possible, species were

identified by trained observers or identified later from photographs. Of the 1,799

turtles counted between 1986-90, 313 (17.4%) were identified. Most (83%) were

19861987198819891990

T

120Longitude

110T I

100 90 80 70

56

Figure 3.2. Combined survey tracklines for the David Starr Jordan and the McArthurin the eastern tropical Pacific during August-November, 1986-90 (From Reilly andFiedler 1994, with permission).

57

olive ridley turtles. Density distributions of species consistently followed described

distribution patterns (Marquez 1990). All identified olive ridley turtles were at sea or

near coastal nesting areas, and other species were in nearshore or island waters.

More than one species was never identified per transect. Therefore, I followed Au

(1991), Pitman (1990), and Pitman (1993), and assumed that oceanic sightings of

unidentified turtles were olive ridley, and excluded all identified sightings of other

species from the habitat analyses. However, the common term "sea turtle" is used in

the discussion.

Habitat Variables

The oceanographic data used in these analyses were originally collected and

summarized by Fiedler et al. (1990) and Reilly and Fiedler (1994) for comparison

with cetacean sightings. Oceanographic data were collected throughout each day's

transect (see Fiedler et al. 1990 and Reilly and Fiedler 1994 for detailed collection

techniques). Sea surface temperature (ssT) and salinity (SAL) were collected

continuously by thermosalinograph and averaged for each transect. A sea-water

density index, a, (SIGMA -T), typically used to describe water masses, was formed by

combining SST and SAL in a standard non-linear function (Pickard and Emery

1990:17-18). Depth of thermocline in meters (z20, defined as the 20 °C isotherm;

Donguy and Meyers 1987) and strength of thermocline (zD), the difference in depth

(m) between the 20 °C and 15 °C isotherms (where a low value indicates a strong

thermocline), were estimated from expendable bathythermographs deployed four to

six times per day. Surface chlorophyll a (cHL) , a standing stock indicator of

58

productivity, was measured by fluorescence and converted to chlorophyll

concentration (mg/m3). Chlorophyll data were collected continuously and calibrated

by at least six discrete surface samples per day, then averaged for each transect. I

calculated distance to land (LAND) as the shortest distance (km) from the approximate

center (noon location) of each transect to the nearest mainland or the nearest island

(Camris® geographic information system software, Ford 1991). Chlorophyll and land

data were log-transformed (LOGC, LOGLAND) prior to analyses to meet assumptions of

parametric statistical models.

Explanatory variables ultimately used in analyses were (1) thermocline depth,

Z20; (2) thermocline strength, zD; (3) chlorophyll, LOGC; (4) distance to nearest land,

LOGLAND; and (5) surface density, SIGMA-T. Surface salinity and surface temperature

were dropped in favor of SIGMA-T as a descriptive variable of water-mass.

Data Analysis

I estimated relative density of cheloniid sea turtles from the daily number of

turtles sighted divided by the effective area searched per day. A model for effective

area searched that linked the area surveyed to prevailing Beaufort-scale sea state

conditions (where Beaufort 0 = calm seas and Beaufort 5 = rough, white-capped seas

with some spray, Bowditch 1977:1267) was constructed by covariate analysis from

the 1989-90 line transect data (Ramsey et al. 1987, Chapter 2). With this model, I

estimated effective area surveyed from distance traveled by Beaufort sea state per day

in 1986-88 (Chapter 2), and then converted daily 1986-88 counts of sea turtles to

density estimates based on the area surveyed. Density estimates represent minimum

59

turtle surface densities, because no correction was made for submerged turtles

(Chapter 2). Relative surface density distribution patterns of all cheloniid species in

the ETP were examined from the combined 1986-90 data.

All transects selected for habitat analyses were surveyed for h in Beaufort

conditions Areas surveyed ranged from 3 km2/d to 98.3 km2/d (average = 25.8

km2/d). Turtle sighting data were analyzed with habitat variables collected during the

same transect. Habitat association analyses were conducted only on the 1989-90

survey data because during those years the collection of sighting data was most

rigorous and there was no detected El Nino event (Fiedler et al. 1992); thus,

interannual variability in the habitat variables was probably low. The relationship of

presence of sea turtles to habitat variables was explored with logistic regression

(Neter et al. 1989), and the relationship of turtle density (turtles/km2), where turtles

were present, to habitat variables was further examined with multiple linear

regression. Full models included all first order, interaction, and quadratic terms.

Logistic regression is an effective method for analyzing wildlife-habitat data

with a binary response function (e.g., y = 1,0; sea turtles present, or sea turtles not

present in an area) and one or more continuous factors (e.g., Ramsey et al. 1994).

The response represents the probability, 7, that y = presence for given levels of the

explanatory variables. The fitted multiple logistic response function is estimated by

exp (bo+E bixi)

1 + exp(bo+E bix)i=1

60

where the response variable, 1, is the estimated probability that a sea turtle is present

in an area, bo is the estimated intercept, and b1 is the estimated regression coefficient

for j = (1, , p) explanatory habitat variables. The practical interpretation of the

response variable, 1, is as an estimated proportion or percent. The odds of turtle

presence, given a set of habitat variables measured within an area, is the estimated

odds ratio, 1/(1-1). The logistic response function is linearized by the logit

transformation of the odds ratio and is referred to as the estimated logit response

function,

loge(1/1-1) = b0 +> bixi.

The estimated logistic response function yields a fitted regression line that is both

sigmoidal and monotonic, is approximately linear where ir is between 0.2 and 0.8,

and is curvilinear at the ends of the range of xi near the asymptotes at 0 and 1 (Neter

et al. 1989). The method of maximum likelihood was used to estimate the parameters

of the logistic response function. The regression coefficients, exp(b), can be

interpreted as the increase in the odds of the response variable (i.e., sea turtle

presence) with a unit increase in a specific habitat variable (xj), keeping other x3

variables constant. The measure of the contribution of the independent variables

towards explaining variation in the response variable in logistic regression is the chi-

square statistic, and can be interpreted analogously to the F-statistic in multiple

regression (Neter et al. 1989).

A model was selected by fitting a logistic model to habitat variables separately

to determine whether a regression relation existed, and then fitting a full model and

performing a likelihood ratio test, or "drop in deviance" test between a full model

61

and successively reduced models (Neter et al. 1989). Variables that do not

significantly reduce the deviance can be dropped from the model. The deviance test

statistic (X2) is distributed approximately as a chi-square statistic (X2dfr_dff) and is

estimated as

X2 = -2 [loge loge Lp],

where Loo is the value of the likelihood function for the reduced model, LIF-) is the

value of the likelihood function for the full model, and dfr and dff are the degrees of

freedom associated with the respective reduced and full models.

The aptness of the logistic regression model was examined graphically and by

informal goodness of fit procedures (Pregibon 1981, Neter et al. 1989). Model

diagnostic procedures based on the R2 statistic were not used in these analyses

because of the limitations of this statistic when applied to binary response variables

(Agresti 1990:112).

Multiple linear regression analyses were based on standard regression

methods (e.g., Neter et al. 1989). The response variable, estimated sea turtle density,

was log-transformed prior to analysis, to meet model assumptions. Final model

selection was made with the Bayes Information Criterion, BIC, (Schwarz 1978).

RESULTS

From 1986 to 1990, 1,799 cheloniid turtles were counted and 313 (17.4%)

were identified to species (Fig 3.3). All loggerhead identifications (12.8%, n = 40

turtles) were off Baja California, Mexico. Green turtles were identified near the

Galapagos Islands, Ecuador (4%, n = 13), and one juvenile hawksbill was sighted

62

20N

lON

0

10S

o°0 ,9 0 o

Oo 0

a (i)0 O o

o. : o 0 44.40 0. 0 00

0 0

0.0 00°

0

Turtles per Square Km

0 ° 0.01-0.05 0 0.05-0.10 0.1-0.5 0 0.5-1.0 0 >1.0

. 0 7 X. 0 4 . ° a 0 o-

0 o. %0 .00 Q. o

. 0 po..v...-

150W 140W 130W 120W 110W 100W 90W 80W

Figure 3.3. Distribution of identified cheloniid sea turtle species in the easterntropical Pacific, 1986-90.

63

with a piece of wood at 14°28'N, 100°7'W. All identified olive ridleys (83%, n =

259) were in oceanic waters or near coastal nesting areas.

The 1986-90 density distribution of all cheloniid species (Fig. 3.4) shows four

primary areas of concentration: (1) off southern Baja California, Mexico; (2) a

nearshore band along the Mexican mainland to Ecuador out to about 700 km; (3)

along the Equatorial Convergence Zone, around 4°N, from 90° to 130°W; and (4)

scattered along 10°N, attenuating at about 110°W where there is a depression in the

thermocline ridge (Fiedler 1992), and increasing from 122°W to 142°W. The

boundary of the Peru current in the southern-most part of the study area was also an

area of light distribution.

When separated by year (Figs. 3.5-3.9), strong interannual variation was

evident in the relative density distribution patterns; suggesting that El Nirio events

(Fiedler et al. 1992) influenced turtle distribution. In "near normal" years, 1986,

1989, and 1990, turtles were distributed along 4°N and near the Central American

coast (Figs. 3.5, 3.8, 3.9). Also in 1989, some turtles were west of 120°W along

10°N (Fig. 3.8). Offshore distribution patterns in 1987, an El Nino year, were more

southerly than in other years (Fig. 3.6). In 1988, an anomalously cool year (La

Nina), turtles were mostly concentrated along the thermocline ridge at 10°N out to

135°W (Fig 3.7).

The 1989 and 1990 line transect data (Fig. 3.10) provided 333 suitable

transects, totaling 49,684 km in length and 8,603 km2 in area for habitat analyses

(Table 3.1). During survey effort, 579 turtles were sighted, and an additional 108

turtles were seen while off survey-effort. Of all turtles counted, 19.6% (n = 135)

64

a20N

.10N 0.. .Q

0

IOS

o 00 ,too(bc,

e .

.

,.0 .00 oo '

0 . '. - ' a % 0.1, go 6.0 0:. ' ' 4 .0- 0 99 d8 09,:,0

. ?. .0 . 2- ccb-o.. .- a 4c,co.0 ,`.3 0

00,

0-''..... :-... ..... caP...°°0.% .07... ..

: 00.. ..toGIN9.Q .. 0 . tr-

O.° .t; *.'0 ...- .0'...: o. Q. @ s.. -.0

O0 0.o 4

Turtles per Square Km, 1986-90

0 0.01-0.05 0 0.05-0.1 0 0.1-0.50 0.5-1.0 0 >1.0 0 >2.0

150W 140W 130W 120W 110W 100W

. .0

0

90W BOW

b

()N

00

. 061117.: 0"C:0 °40

. o o° 6) C/, g)ION o 9. ? 0 c.° Q GS. 000 °° 13 0;9 Cto.

o 0°oOD ata.

o 0 0 Bo 0go 0000 0 a

oo 0 0® o 0.0 g. 0 .00 boa 0

0° °b

00 d 0 0 cog0 0

0 . O 0.0

° o 9 ,S9 0e ° 0 *,5 °

0

10s

0

Turtles per Square Km, 1986-90

0.01-0.05 0 0.05-0.1 0 0.1-0.5 0 0.5-1.00 -1.0 0 >2.0

150W 140W 130W 120W

0 ° 0 00 00 0

0 ° 0 0

0 c)c. 0 0 0

110W 100W 90W BOW

Figure 3.4. Relative density distribution of cheloniid sea turtles in the easterntropical Pacific, 1986-90. All symbols are centered on the local noon position ofeach transect. a) Turtle density distribution, including tracklines surveyed where noturtles were seen (small points). b) Turtle density distribution only where turtleswere seen.

a20N

ION

00

0

o 00 o

° 000°o 0 of

. o o 000

O . 0

o0 0*

0

00 9a 6 v, 0

.

0 o0

Turtles per Square Km, 1986

0 0.01-0.05 0 0.05-0.1 0 0.1-0.50 0.5-1.0 0 >1.0 0 >2.0

150W 140W 130W 120W 110W 100W 90W 80W

65

b20N

ION

0

0 00 0

0

0 0o

° 000°0 0 03

0 000

0 00 o 9 8

0 0

0.01-0.05 0 0.05-0.1 0 0.1-0.50 1.0 0 >2.0

b o

150W 140W 130W 120W 110W 100W 90W BOW

Figure 3.5. Relative density distribution of cheloniid sea turtles in the easterntropical Pacific, 1986. All symbols are centered on the local noon position of eachtransect. a) Turtle density distribution, including tracklines surveyed where noturtles were seen (small points). b) Turtle density distribution only where turtleswere seen.

66

a20N

ION

0

'Os

0

Turtles per Square Km, 1987

0 0 0.01-0.05 0 0.05-0.1o 0.1-0.5 0 0.5-1.0

Oo 0

0

0 °

0

. o0

O

0

o0:0 a a0

150W 140W 130W 120W 110W 100W 90W 80W

0.01-0.05 0 0.05-0.1 0 0.1-0.50 0.5-1.0

Figure 3.6. Relative density distribution of cheloniid sea turtles in the easterntropical Pacific, 1987. All symbols are centered on the local noon position of eachtransect. a) Turtle density distribution, including tracklines surveyed where noturtles were seen (small points). b) Turtle density distribution only where turtleswere seen.

a020N

67

0

JION o o 0 o OD0

0 .0 0 o °

0 0 . ° 0 000 00

° 08 : . 0 - Q. o0

0 b 0 00O . 0 0 bO 0 .!"f

Turtles per Square Km, 1988

0

0 0 0.01-0.05 0 0.05-0.10 0.1-0.5 0 0.5-1.0

150W 140W

O

o0

130W 120W 110W 100W 90W 80W

20N

10N

0

10S

O

0

0

0

00

0c J

0 0 0 0 0 CO

0 0 0 ° ° 0 00 00

0 0 08 . 4

0 o 0O o 0 0

O 0 1.0 o

Turtles per Square Km, 1988

0.01-0.05 0 0.05-0.1 0 0.1-0.5o 0.5-1.0

O

00

0000

150W 140W 130W 120W 110W 100W 90W 80W

Figure 3.7. Relative density distribution of cheloniid sea turtles in the easterntropical Pacific, 1988. All symbols are centered on the local noon position of eachtransect. a) Turtle density distribution, including tracklines surveyed where noturtles were seen (small points). b) Turtle density distribution only where turtleswere seen.

20N

ION0

0

0

68

. 0o 0 0 o

0 0 o- 0 o

° o0 o o 0

0 Qon o

000 00 0o . 0 8

o 0 o °0 0

0 0 CO° o ° o0 0

.0

:El

0 0

Turtles per Square Km, 198910s

0 0 0.01-0.05 0 0.05-0.10 0.1-0.5 0 0.5-1.0 0 '.1.0

0 oo 0

0

150W 140W 130W 120W 110W 100W 90W 80W

ION

I 0 S

00

O

0

60

003

0 0o 0 0 Goo 0 0 00° 00 0

0 0e o 0 0 0

0 0 o 0O00 0 u 0

0 0 80 0 0 0

0 CO0 0 ° 0o 0 0

tsd 0

0 0

Turtles per Square Km, 1989

0 0.01-0.05 0 0.05-0.1 0 0.1-0.50 0.5-1.0 0 -1.0

150W 140W 130W 120W

O0 0

O

'jo0 0

110W 100W 9 0 W 80W

Figure 3.8. Relative density distribution of cheloniid sea turtles in the easterntropical Pacific, 1989. All symbols are centered on the local noon position of eachtransect. a) Turtle density distribution, including tracklines surveyed where noturtles were seen (small points). b) Turtle density distribution only where turtleswere seen.

69

a20N.

ION

0

0

0 0 00 0 0 9

O0

Co 00 0oo 0 ° a .

oo .0. 0 0

° (? 0

Turtles per Square Km, 1990

O

o

O0 0. oo

0.0 o0 o

D

0 0.01-0.05 0 0.05-0.10 0.1-0.5 0 0.5-1.0 0 -LO

150W 140W 130W 120W 110W 100W 90W 80W

20N

ION

0

10S

0 0 0 0 O

0 0 90 0 0 co

0 00 0°

. 000

00 0 0 0 ooo ° 0 0

Turtles per Square Km, 1990

0.01-0.05 0 0.05-0.1 0 0.1-0.5O 0.5-1.0 0 -.1.0

150W 140W 130W 120W

0

0

0

110W 100W 90W 80W

Figure 3.9. Relative density distribution of cheloniid sea turtles in the easterntropical Pacific, 1990. All symbols are centered on the local noon position of eachtransect. a) Turtle density distribution, including tracklines surveyed where noturtles were seen (small points). b) Turtle density distribution only where turtleswere seen.

70

0- 0,

0 o o 0 000

0 0 .

Line Transect Sighting DataTurtles per Square Km, 1989

0 0 0.01-0.05 0

0 0.1-0.5 0 0.5-1.0 00.05-0.1>1.0

140W 130W 120W 110W 100W

20N.

ION

0

10s

o

o

0

o° .00 0 (6 0

0o

Line Transect Sighting DataTurtles per Square Km, 1990

0 0 0.01-0.05 0 0.05-0.10 0.1 -0.5 0 0.5-1.0 0 >1.0

o0 0

0 0

0 80 0

co 00

0

0

00 0

150W 140W 130W 120W 110W 100W 90W BOW

Figure 3.10 Relative density distribution of cheloniid sea turtles in the easterntropical Pacific from line transect sighting data 1989-90. All symbols are centeredon the local noon position of each transect, including tracklines surveyed where noturtles were seen (small points). a) 1989. b) 1990.

Table 3.1 Average (R) ± standard deviation (SD) and associated coefficient of variation (CV), and mode of habitat variablesused with 1989-90 sea turtle sighting data in the eastern tropical Pacific (ETP). Range in parentheses. SIGMA-T (kg /m3) =

surface sea-water density, Z20 (m) = depth of thermocline, LOGC (mg/m3) = chlorophyll concentration, LOGLAND (km) =distance to nearest land, ZD (m) = strength of thermocline, SST (°C) = sea surface temperature, and SAL (psu) = salinity ofsurface water. Arithmetic summary statistics are presented for LOGLAND and LOGC for convenience.

All datan = 333

Presentn = 117

Not presentn = 216

Variable R + SD CV R A- SD CV Mode R SD CV Mode

SIGMA-T 22.09 ± 1.3 6.1 21.66 ± 1.29 6.0 21.04 22.33 ± 1.3 5.9 20.81

(18.79-24.89) (19.23-24.89) (18.79-24.88)

Z20 57.4 ± 26.3 45.8 53.99 ± 22.18 41.1 51.6 59.3 + 28.1 47.4 0.0(0.0-151.9) (0.0-119.8) (0.0-151.9)

LOGC 0.249 ± 0.15 61.1 0.25 ± 0.13 52.6 0.2 0.247 ± 0.16 65.5 0.01

(0.053-1.28) (0.06-1.07) (0.05-1.28)

LOGLAND 815.5 ± 751.0 92.0 600.4 ± 468.4 78.0 474.6 932.0 ± 844.9 90.6 624.8

(6.8-4,044.6) (9.2-2,019) (6.8-4,044.6)

ZD 30.0 ± 15.7 52.3 28.6 ± 13.8 48.3 22.5 30.7 ± 16.6 54.0 23.4

(10.8-115.5) (10.8-91.6) (11.3-115.5)

SST 25.9 ± 2.8 10.8 26.9 + 2.6 9.6 27.8 25.4 + 2.8 10.9 26.1

(17.8-30.6) (19.3-30.4) (17.8-30.6)

SAL 33.77 ± 0.87 2.6 33.58 ± 0.86 2.6 32.76 33.86 ± 0.86 2.5 33.90

(28.95-35.32) (30.64-35.32) (28.95-35.25)

72

were identified and 90% (n = 122) of these were olive ridley. Turtles (n = 573)

were present on 127 transects designated suitable for analyses. Ten transects

exclusively contained species other than the olive ridley (n = 14) and were omitted.

Turtle densities ranged from 0.015 to 1.27 turtles /km2 (average = 0.12 turtles/km2,

SD = 0.16).

In the individual logistic regression analyses, SIGMA-T (P = 0.0001) and

LOGLAND (P = 0.0039) were negatively correlated with turtle presence. Depth of

thermocline, Z20, showed a negative trend with increasing probability of turtle

presence (P = 0.08). During the 1989-90 nesting/breeding seasons, the probability of

sea turtle presence was best described by a shallow thermocline, lower density of

surface water, decreasing distance from nearest point of land, and an interaction of

the latter two variables (-2 loge L = 388.9, X2 = 42.8, df = 4, P = 0.0001;

Table 3.2). The odds of sea turtle presence increased by exp(-0.0128) = 0.987, or

98.7%, with a 1-m decrease in thermocline depth, holding the other variables constant

in the model. Because of the interaction term, exp(b3) for SIGMA-T and LOGLAND

cannot be interpreted as simply the odds ratio of a unit increase in SIGMA-T or in

LOGLAND. To understand the relationship between SIGMA-T and LOGLAND, it is useful

to find the coefficient for SIGMA-T in the estimated response function (Table 3.2):

log-odds of turtle presence = b0 0.128*Z20 -9.419*log(LAND)

[3.097 0.4292*log(LAND)]*SIGMA-T .

Thus a unit increase in SIGMA-T (kg/m') is a change of exp -[3.097

in the odds. An increase in SIGMA-T near land reduces the odds

of presence, but the amount of this reduction decreases farther from land. At

73

Table 3.2. Logistic regression model parameter estimates for presence of sea turtlesin the eastern tropical Pacific, 1989-90. SIGMA-T (kg/m3) = surface sea-waterdensity, Z20 (m) = depth of thermocline, LOGLAND (km) = distance to nearest land.The logistic response function is:

= [1 + exp( -bo b1z20 - b2siGmA-T - b3 LOGLAND - b4 SIGMA-T*LOGLAND)]-1.

Variable Coeff. SEWald's

Chi Square P-value

Constant 68.09 16.68 16.67 0.0001

Z20 -0.0128 0.006 4.91 0.0270

SIGMA-T -3.0968 0.77 14.34 0.0001

LOGLAND -9.4394 2.55 24.57 0.0002

SIGMA-T X 0.4292 0.12 13.29 0.0003LOGLAND

74

distances greater than exp(3.097/0.4292) = 1,360 km (obtained by setting the

coefficient, -[3.097 0.4292*log(LAND)], equal to 0 and solving for LAND), the sign

of the coefficient becomes positive and thus increases in SIGMA-T far from land equal

increases in the odds of turtle presence. For example, the odds ratio for increasing

SIGMA-T at various distances are:

Distance from LAND Odds ratio of 1-unit increase in SIGMA-T

10 km 0.12

100 km 0.33

1,000 km 0.88

4,000 km 1.59

The 1989-90 sea turtle density data were not well explained by the available

habitat variables. The best model, which contained distance to land, surface water

density, chlorophyll, and thermocline depth, accounted for only 22.3% of the

variation in sea turtle density in the ETP (F4,112 = 9.3, P < 0.0001; BIC = -15.6;

Table 3.3). The interaction of SIGMA-T and LOGLAND was a possible candidate for

inclusion, but was marginally significant (P = 0.04, BIC = -14.4) and its addition to

the model increased the adjusted R2 to only 24%. The effect of this interaction is

similar to the one above.

75

Table 3.3. Multiple linear regression parameter estimates and analysis of variance forassociation of sea turtle density with habitat variables in the eastern tropical Pacific,1989-90. SIGMA-T (kg/m3) = surface sea-water density, Z20 (M) = depth ofthermocline, Locc (mg/m3) = chlorophyll concentration, LOGLAND (km) = distanceto nearest land.

Variable Coeff. SE t-value P-value

Constant 2.45 1.24 1.97 0.05

LOGLAND -0.196 0.084 -2.34 0.02

SIGMA-T -0.171 0.065 -2.63 0.009

LOGC 0.514 0.187 2.75 0.007

Z20 0.012 0.004 3.31 0.001

Sum of MeanSource df Squares Square F-value P-value Adj. R2

Model 4 22.48 5.62 9.32 <0.0001 0.223

Residual 112 67.53 0.603

Total 116 90.01

76

DISCUSSION

Geographic distribution

The spatial shifts in yearly relative density distributions of turtles were likely

related to the periodic El Nino/Southern Oscillation (ENSO) events that drive

interannual variability in the ETP. During an El Nino in the eastern Pacific, there is

a warming of surface waters and a reduction in productivity because the trade winds

are weaker and the thermocline deepens, and so upwelled water in the equatorial and

coastal upwelling regions is nutrient-poor (Mann and Lazier 1991). Fiedler et al.

(1992) described the following anomalies in sea surface temperature, SST, thermocline

depth, and chlorophyll concentration data collected during the MOPS study in August-

November 1986-89. In 1986, most of the study area was within normal ranges.

However, around the northern Galapagos and along the equator west of 100°W, the

thermocline was 20-30 m deeper, SST was 1 °C above normal, and chlorophyll was

high south of 5°N; sea turtles were present near these areas. In 1987, warm nutrient-

poor water covered the equator and sea turtles were distributed to the south off Peru,

where chlorophyll was slightly higher than in tropical surface water, TSW. Turtle

densities were lower in the central TSW where chlorophyll was reduced and the

thermocline ridge and trough were flattened. Surface temperatures in 1988 were

colder than normal over most of the study area, but the central band of TSW, where

turtles were mostly distributed, was only 1 °C below normal and the thermocline

ridge at 10°N was 20-40 m shallower than usual. In 1989 and 1990, conditions were

within normal ranges and turtle distribution was similar to that of 1986 with most of

77

the offshore distribution concentrated around the equatorial convergence (2°N-5°N),

with the addition of light distribution at 10°N in 1989. Turtle distribution patterns

show a flexible response to the dynamic oceanographic conditions dominating the

habitat, shifting to areas of higher productivity as conditions change. Shifts in olive

ridley distributions during El Nitio events have also been noted by others (Cornelius

and Robinson 1986, Plotkin 1994).

In addition to the expected nearshore density distribution during the peak

nesting season, major concentrations of sea turtles in 1986-90 coincided with two

primary features of the oceanic eastern tropical Pacific, the equatorial convergence

zone at about 2°- 5°N, and to a lesser extent, the area around the divergence zone at

10°N in 1988 and 1989. These two zones may provide pelagic foraging grounds for

sea turtles. The equatorial convergence is formed by the meeting and sinking of cool,

saline and nitrate-rich waters of equatorial upwelling and Peru Current origin, and the

warmer, less saline TSW (Pak and Zaneveld 1974). As nutrient-rich water moves

away from the area of upwelling towards the front, a community of macroplankton

and nekton forms and is aggregated at the convergence, attracting large predators

(Sette 1955, Vinogradov 1981). This process is strengthened in the summer season.

Likewise, the divergence zone at 10°N is an area of increased biological productivity

(Fiedler et al. 1991) and is underlain by a shallow, strong thermocline, which may

confine potential prey to surface waters (Green 1967, Reilly 1990). Communities of

potential prey may also form some distance laterally away from the divergence zone

by horizontal advection as described above, which accounts for the scattered turtle

distribution to either side of the 10° latitude line (Figs. 3.7, 3.8).

78

The distribution of turtles along the convergence zone in the summer is

complementary to seasonal spatial patterns reported for some delphinid species (Reilly

1990). In the summer, spotted and spinner dolphin (Stenella attenuata and S.

longirostris), and to a lesser extent striped dolphin (S. coeruleoalba), distributions

apparently shift away from their winter centers at 4°N and 6°S, to along 10°N (Reilly

1990). Whereas in this study, sea turtles were more prevalent along 4°N than at

10°N especially in the 1986-87, and 1989-90 summers. Complementary offshore

distributions of turtles and dolphins have been noted in other areas. Primary

concentrations of marine mammals and sea turtles sighted in the waters off the NE

United States rarely overlap, except along the continental shelf edge (Shoop and

Kenny 1992). Sea turtle distribution and the distributions of several dolphin species

in the ETP overlap in the nearshore waters from southern Mexico to northern Peru

(Au and Perryman 1985, Reilly 1990, Reilly and Fiedler 1994), but apparently less so

offshore during the summer months. It is possible that there is seasonal component in

sea turtle distribution as well. Plotkin (1994) tracked post-breeding male and female

olive ridleys in the ETP and they ranged predominantly north of 5°N.

Because survey sampling stopped at the boundary of the Peru Current, it is

unknown if the southern distribution of sea turtles also discontinues south of the

boundary. Certainly turtle densities were lower in the area ( 0.1 turtles/km2) and

were more concentrated there during the 1987 El Niiio. It is likely that the southern

extent of olive ridley distribution is limited by the colder water of the Peru Current.

79

Habitat association

The presence of olive ridley sea turtles in the ETP during August December

1989-90 was significantly associated with nearness to land, low surface-water density

especially when close to land, and moderately shallow depths of the 20 °C isotherm.

These conditions are components of tropical surface water, TSW (Fig. 3.1; Wyrtki

1967), and are consistent with "coastal tropical habitat," as described by Reilly and

Fiedler (1994) for mixed schools of eastern spinner dolphins and spotted dolphins.

Coastal tropical habitat is characterized by a well-stratified surface layer, occasional

and localized upwelling, a relatively shallow thermocline, lower SIGMA-T (higher

surface temperatures, lower salinity), and high chlorophyll concentrations (Reilly and

Fiedler 1994).

The effect of distance to land in describing the presence of sea turtles is

expected, particularly during the peak nesting season when turtles are more likely to

be near certain beaches. Furthermore, the distribution near land is generally

consistent with the trend towards shallower thermocline depths nearshore and lower

surface water density in TSW, where precipitation exceeds evaporation. However

other unmeasured components are included in this variable that may also be

meaningful. For instance, topographically induced fronts, and fronts associated with

wind-driven coastal upwelling (Mann and Lazier 1991, Franks 1992) attract predators

(e.g., Laurs et al. 1977) and may be strong attractors near land for turtles. Fixed

geographic effects were also significant in habitat models of dolphins (Polacheck

1987, Reilly and Fiedler 1994). The increase in odds of turtle presence offshore with

increasing SIGMA-T possibly reflects the distribution of turtles along the equatorial

80

front, where surface water density is higher from input of equatorial surface water,

ESW at the boundary of water-masses.

The inshore (within about 700 km, Fig 3.4) distribution of sea turtles visually

overlaps two types of inshore habitats described by others for Stenella spp. ("coastal

tropical habitat") and common dolphins (Delphinus delphis; "upwelling-modified

habitat") (Au and Perryman 1985, Reilly 1990, Reilly and Fiedler 1994). However,

the overall summary statistics (Table 3.1) of surface water density and thermocline

depth in areas of sea turtle presence are most comparable to those collected in

summer for striped dolphins (xSIGMA-T 21.6, SD = 1.35, n = 341; XZ20 = 57.2 m,

SD = 23.2, n = 341; Reilly 1990) and overlapped those of spinner and spotted

dolphins (isiGmA_T = 21.2, SD = 1.42, n = 401; x Z20 = 67.7 m, SD = 21.3, n =

401; Reilly 1990). These similarities further suggest that sea turtle distribution in the

summer is primarily coastal tropical habitat. Plotkin (1994) characterized distribution

of satellite-acquired locations of olive ridleys as "upwelling-modified." However, the

majority of tracking-days of turtles in that study were during winter months when

coastal upwelling systems intensify due to strong winds from the Gulf of Mexico

funneling through mountain passes into the Gulfs of Tehuantepec, Papagayo, and

Panama (McCreary et al. 1989). Therefore many nearshore areas occupied by sea

turtles in the winter may be more upwelling-modified.

The multiple linear regression model for sea turtle density was similar to the

model for turtle presence, with the addition of chlorophyll concentration as an

explanatory variable. However, the habitat variables were poor descriptors of sea

turtle density in the ETP during 1989-90. Only 22% of the variation in sea turtle

81

surface density was accounted for by SIGMA-T, Z20, LOGLAND, and LOGC, suggesting

these variables are linked indirectly to sea turtle density.

Comparable R2 values were reported for these environmental parameters and

the abundance and distribution patterns of cetaceans, similarly indicating indirect

associations with these variables. Using the same variables as this study and

including geographic locations, Reilly and Fiedler (1994) explained 20.5% of the

variation in the combined relative abundance data of several delphinid species, and

8.8 41.2% of individual species abundance. Cetacean sighting-rate and chlorophyll

concentration were strongly correlated off California (r = 0.92; Smith et al. 1986).

Polacheck (1987) found the highest correlations of several cetacean species'

abundance with sea surface temperature, thermocline depth, 02-minimum layer

thickness, and water depth (a reflection of distance from land). However these

variables also accounted for less than 38% of variation for any species.

The low explanatory ability of the environmental parameters for large

vertebrate abundance and density is not surprising. These variables represent meso-

and large-scale features that are likely linked to the occurrence and density of sea

turtles through finer-scale processes and features. In particular, a lag in time and

space is expected between favorable physical and chemical configurations and the

occurrence of high-level predators. Such a lag can be on the order of months and

hundreds of kilometers (Mann and Lazier 1991). Furthermore, the sampling scale of

this study was large; daily survey areas were often extensive and turtle patchiness

within transect-areas was not examined.

82

Additionally, several other factors likely affect the amount of variation

explained in these data. Because 80% of the sea turtles were unidentified, habitat

preferences and behavior of other species inadvertently included in the sighting data

will increase variation in the results. However, olive ridley turtles are easy to detect

and therefore unidentified sightings in this study were likely biased towards members

of that species (see Chapter 2).

Also, reproductive behaviors influence distribution and density but are not

considered in the model. During the peak nesting season, July December, the

highest inshore densities of olive ridleys (>1 turtle/km2) are likely due to their unique

nesting behavior. The majority of females of this species nest synchronously in large

aggregations (hundreds to thousands of turtles) on certain Central American beaches

(e.g., Richard and Hughes 1972, Plotkin et al. 1995). Turtles loosely aggregate prior

to the mass-nesting emergences, apparently awaiting the appropriate physiological and

environmental cues to go ashore (Plotkin 1994). It is not known how long the

aggregations initiate before the emergence or how far from the beaches they form,

however, the highest density in 1989 (1.27 turtles/km2) was > 100 km offshore.

Finally, an obvious and important omission from the data presented here is the

distribution of prey species of olive ridley sea turtles and its influence on sea turtle

distribution. Unfortunately there are not many studies of the feeding habits of olive

ridleys, particularly in pelagic regions. Available information on prey items suggests

they are generalists, feeding opportunistically in benthic and epipelagic habitats on

mollusks, bryozoans, shrimp, pelagic crabs, fish eggs, cnidareans, salps and tunicates

(Fritts 1981, Montenegro and Bernal 1982, Mortimer 1982, Barragan et al. 1992). A

83

generalist foraging strategy is consistent with the evolution of a oceanic migratory

species in a dynamic habitat like the El Nino/Southern Oscillation dominated ETP.

The presence and density of olive ridley sea turtles in the ETP during 1989-90

was significantly associated with distance to land, water-mass density, depth of

thermocline and chlorophyll concentration. The reader is cautioned, however, that

these analyses are exploratory, rather than predictive. The results presented are valid

only for these data in the ETP during the study period. Even so, the results suggest

the following directions for future research:

1. There may be a seasonal component to sea turtle density distribution

corresponding to seasonal variability in the ETP. Examination of existing data

(e.g., Pitman 1990), or data from fishing vessels in the appropriate seasons

could reveal such changes in distribution.

2. Do sea turtles co-occur spatially and temporally with cetaceans, or are their

distributions complementary (e.g., Shoop and Kenny 1992)? Overlap in

distributions can be determined from comparisons of MOPS cetacean and turtle

daily-sighting data.

3. Interannual variation in the ETP seems to affect turtle distribution. The

relationship of turtle presence and density to annual variation can be studied

statistically with the currently available data.

4. Foraging behavior, prey species availability, and prey distribution are

important elements of sea turtle distribution and are necessary to further study

how marine turtles actively use their habitat.

84

Information on seasonal and interannual distributions, and habitat relationships

is important to conservation and management efforts. Data are especially needed in

areas where sea turtles overlap with, and are incidentally captured in, commercial

fishery operations. Sea turtles occur in local and international waters throughout

Pacific Central America. Management and protection plans will require cooperative

international effort and a strong data base for effective conservation decisions.

85

Summary

Until recently there was little information on the distribution range and habitat

of the olive ridley turtle during the non-nesting portion of its life history. Reports

from tuna vessels (Cornelius and Robinson 1986, Au 1991) and other vessels (Oliver

1946, Pitman 1990, 1993) have documented the presence of olive ridleys in the open

ocean. Recent tracking studies of the olive ridley (Plotkin 1994, Plotkin et al. 1995)

have vastly increased knowledge of their behavior and migrations. However, no

attempts have been made to estimate relative densities or to quantitatively correlate

distribution with oceanographic indices and surface features. The goals of this thesis

were to examine the behavior of an olive ridley with respect to primary surface

currents and features especially convergence zones, evaluate satellite-telemetry as a

method for collecting such data, calculate relative density and compare density

distributions with convergence zones and other primary surface features, and to

quantitatively correlate presence and density of turtles with oceanographic habitat

variables during the peak breeding/nesting season.

The track of the satellite-monitored olive ridley turtle suggests some active

swimming rather than continuous passive floating with the currents, and comparisons

with surface current divergence data indicated that the turtle encountered several large

convergence and divergence zones, but did not appear to favor either. The turtle

travelled at least 2,691 km in the open ocean, averaging 1.28 km/h. Diel patterns

were evident in the turtle's submergence and diving behavior. Daytime dives were

frequent, and of short-duration, yet more time was spent at the surface. Nighttime

86

submergence durations were longer and more variable. Temperatures recorded during

daytime transmissions were cooler and variable, and were correlated with longer

submergences. Nighttime temperatures were consistently warm and there was no

correlation with submergence time. It is possible that temperatures reflect dive

depths, but that remains to be tested by transmitters with depth sensors. Two discrete

submergence behaviors were recorded, short duration submergences (<7.7 min) and

long dives (8.2 95.5 min). These behaviors exhibited the same diel pattern.

Surfacing durations were long (1-10 min) but showed no diel differences, possibly due

to transmission constraints. The surfacing patterns of this olive ridley allowed many

transmissions, indicating that greater amounts of data could be archived and

transmitted in future studies.

Detectability analysis was an effective method for testing the effect of

environmental conditions, observer, and time of day on the detection of turtles, and

for adjusting the effective area surveyed. Sea state was the primary factor influencing

sea turtle detection. Effective halfwidth surveyed was adjusted for each sea state

category. Effective area surveyed under each sea state condition was calculated for

the 1986-1990 data from effective halfwidth and transect length (km/sea state).

Detectability analysis is applicable in any variable distance survey where sighting

conditions are systematically recorded and target animals are not influenced by the

observer. This method provides a simple alternative to stratifying data or restricting

survey conditions.

Turtles were widely distributed throughout the ETP and species distributions

were consistently within reported ranges. All loggerhead turtles were identified off

87

Baja California, all green turtles were identified near the Galapagos, one hawksbill

was sighted off Mexico, and olive ridleys were identified in both nearshore and

offshore waters.

Relative density distributions showed primary concentrations of turtles off

Baja, Mexico, nearshore from Mexico to Ecuador and out about 700 km, along the

equatorial convergence zone, and to a lesser extent along the north equatorial

divergence and the Peru Current boundary. Strong interannual variation was evident

in the yearly relative density distribution patterns, suggesting that turtle distribution

was influenced by the 1987 El Niiio and 1988 La Nina events.

Presence of olive ridley turtles was significantly correlated with surface water

density, distance to land, and thermocline depth. Densities of these turtles, where

present, were minimally correlated with surface water density, distance to land,

thermocline depth and chlorophyll concentration.

Sea turtle density was noticeably concentrated along the equatorial convergence

zone, particularly in 1986, and 1989-90. However the data from these studies are

inconclusive with regard to the relationship of sea turtles to frontal zones. Turtles

were also present in areas of coastal and offshore upwelling divergence (1988, 1989).

Both convergence and divergence features are areas of high productivity (Mann and

Lazier 1991). Whether turtles were directly associated with the divergences, or fronts

related to them (e.g., coastal upwelling fronts), cannot be determined from the data in

these studies. The spatial scales of currents and surface features examined in these

studies are too large to reveal the true association between turtles and these physical

features. The available data (Plotkin 1994, these studies) indicate that olive ridley

88

turtles in the ETP are generalists in their habitat, ranging widely, continuously

moving and probably foraging opportunistically on food when and where it is

available rather than consistently targeting a particular food source.

89

Bibliography

Agresti, A. 1990. Categorical data analysis. John Wiley & Sons. New York.

Akaike, H. 1974. A new look at the statistical model identification. IEEE Trans. onAutomatic Control. 19:716-723.

Argos, 1989. Argos system guide. Service Argos, Landover. Available from:Service Argos 1081 McCormick Dr. Landover, MD 20785.

Aridjis, H. 1990. Mexico proclaims total ban on harvest of turtles and eggs. Mar.Turt. Newsl. 50:1-3.

Au, D. W. 1991. Polyspecific nature of tuna schools: shark, dolphin, and seabirdassociates. Fish. Bull. 89:343-354.

, and W. L. Perryman. 1985. Dolphin habitats in the eastern tropical Pacific.Fish. Bull. 83:623-643.

Balazs, G. H. 1994. Homeward bound: satellite tracking of Hawaiian green turtlesfrom nesting beaches to foraging pastures (Abstract). Pages 205-208. In B. A.Schroeder and B. E. Witherington, compilers. Proc. thirteenth annu. symp.on sea turtle biol. and conserv. NOAA Tech. Memo. NMFS-SEFSC-341.

Barlow, J., C. W. Oliver, T. D. Jackson, and B. L. Taylor. 1988. Harbor porpoise,Phocoena phocoena, abundance estimation for California, Oregon, andWashington: II. aerial surveys. Fish. Bull. 86:433-444.

Barragan, A. R., C. Lopez, M. Mata, A. Quintana, E. L. Santos, and L. Sarti.1992. Estudio de los contenidos estomacales de Lepidochelys olivacea en lacosta sur del estado de Michoacan, Mexico. Pages 39-50. In M. Benabib andL. Sarti (eds.), Memorias del VI Encuentro Interuniversitario Sobre TortugasMarinas, Publ. Soc. Herpetol. Mex. No. 1.

Bayliss, P. 1986. Factors affecting aerial surveys of marine fauna, and theirrelationship to a census of dugongs in the coastal waters of the NorthernTerritory. Aust. Wild. Res. 13:27-37.

Beavers, S. C., and E. R. Cassano. 1996. Movements and dive behavior of a maleturtle (Lepidochelys olivacea) in the eastern tropical Pacific. J. Herpetol.30:97-104.

90

Blackburn, M. 1965. Oceanography and the ecology of tunas. Oceanogr. Mar.Biol. Ann. Rev. 3:299-322.

Bowditch, N. 1977. American practical navigator. Vol. 1. Defense MappingAgency, Hydrographic Center, Washington, D.C.

Bowen, B. W., F. A. Abreu-Grobois, G. H. Balazs, N. Kamezaki, C. J. Limpus,and R. J. Ferl. 1995. Trans-Pacific migrations of the loggerhead turtle(Caretta caretta) demonstrated with mitochondrial DNA markers. Proc. Natl.Acad. Sci. 92:3731-3734.

Boyd, I. L. 1993. Selecting sampling frequency for measuring diving behavior.Mar. Mammal Sci. 9:424-430.

Buck land, S. T. 1985. Perpendicular distance models for line transect sampling.Biometrics 38:469-77.

, and D. R. Anderson. 1976. Mathematical models for nonparametricinferences from line transect data. Biometrics 32:325-336.

, , K. P. Burnham, and J. L. Laake. 1993. Distance sampling;estimating abundance of biological populations. Chapman and Hall, London.

, K. L. Cattanach, and Th. Gunnlaugsson. 1992. Fin whale abundance in theNorth Atlantic, estimated from Icelandic and Faroese NASS-87 and NASS-89data. Rep. Int. Whal. Commn. 42:645-651.

Burnham, K. P., D. R. Anderson, and J. L. Laake. 1980. Estimation of densityfrom line transect sampling of biological populations. Wildl. Monogr. 72:1-202.

By les, R. A. 1988. Behavior and ecology of sea turtles from Chesapeake Bay,Virginia. Unpubl. Ph.D. Dissertation, The College of William and Mary,Virginia.

Carr, A., M. H. Carr, and A. B. Meylan. 1978. The ecology and migrations of seaturtles, 7. The west Caribbean green turtle colony. Bull. Amer. Mus. Nat.Hist. 162:1-46.

. 1986, Rips, FADS, and little loggerheads. BioScience 36:92-100.

. 1987. Impact of nondegradable marine debris on the ecology and survivaloutlook of sea turtles. Mar. Poll. Bull. 18:352-356.

91

Clifton, K., D. 0. Cornejo, and R. S. Felger. 1982. Sea turtles of the Pacific coastof Mexico. Pages 199-209. In K. A. Bjorndal (ed.), Biology andconservation of sea turtles. Smithsonian Institution Press. Washington, D.C.

Coles, W. C., J. A. Keinath, D. E. Barnard, and J. A. Musick. 1994. Sea surfacetemperatures and sea turtle position correlations (Abstract). Pages 211-212.In K. A. Bjorndal, A. B. Bolten, D. A. Johnson, and P. J. Eliazar(compilers), Proceedings of the Fourteenth Annual Workshop on Sea TurtleBiology and Conservation 1 -5 March 1994, Hilton Head, SC. NOAA Tech.Memo. NMFS-SEFSC-351.

Collard, S. B. 1992. Surface ocean fronts and the pelagic habitat of sea turtles: thequestion of food availability in non-frontal zones (Abstract). Page 28. In M.Salmon and J. Wyneken (compilers), Proceedings of the Eleventh AnnualWorkshop on Sea Turtle Biology and Conservation 26 Feb-2 March 1991,Jekyll Island, GA. NOAA Tech. Memo. NMFS-SEFSC-302.

, and L. H. Ogren. 1990. Dispersal scenarios for pelagic post-hatchling seaturtles. Bull. Mar. Sci. 47:233-243.

Cornelius, S. E. and D. C. Robinson. 1986. Post-nesting movements of femaleolive ridley turtles tagged in Costa Rica. Vida Si lv. Neotrop. 1:12-23.

Crouse, D. T., L. B. Crowder, and H. Caswell. 1987. A stage-based populationmodel for loggerhead sea turtles and implications for conservation. Ecology68:1412-1423.

Donguy, J. R., and G. Meyers. 1987. Observed and modelled topography of the 20°C isotherm in the tropical Pacific. Oceanol. Acta 10:41-48.

Duron-Dufrenne, M. 1987. Premier suivi par satellite en Atlantique d'une tortueLuth Dermochelys coriacea. C.R. Acad. Sc. Paris t. 304 Serie 111:399-402.

Eckert, S. A., D. W. Nellis, K. L. Eckert, and G. L. Kooyman. 1986. Divingpatterns of two leatherback sea turtles (Dermochelys coriacea) duringinternesting intervals at Sandy Point, St. Croix, U.S. Virgin Islands.Herpetologica 42:381-388.

, K. L. Eckert, P. Ponganis, and G. L. Kooyman. 1989. Diving and foragingbehavior of leatherback sea turtles (Dermochelys coriacea). Can. J. Zool.67:2834-2840.

Epperly, S. P., J. Braun, A. J. Chester, F. A. Cross, J. V. Merriner, and P. A.Tester. 1995a. Winter distribution of sea turtles in the vicinity of Cape

Hatteras and their interactions with the summer flounder trawl fishery. Bull.Mar. Sci. 56:547-568.

, J. Braun, and A. J. Chester. 1995b. Aerial surveys for sea turtles in NorthCarolina inshore waters. Fish. Bull. 93:254-261.

92

Fagen, R. M., and D. Y. Young. 1978. Temporal patterns of behaviors: Durations,intervals, latencies and sequences. Pages 79-114. In P. W. Colgan (ed.),Quantitative Ethology, John Wiley & Sons, New York.

Fancy, S. G., L. F. Pank, D. C. Douglas, C. H. Curby, G. W. Garner, S. C.Amstrup, and W. L. Regelin. 1988. Satellite telemetry: A new tool forwildlife research and management. U.S. Fish and Wildlife Service ResourcePublication 172.

Fedorov, K. N. 1983. The physical nature and structure of oceanic fronts. M. J.Bowman, R. T. Barber, C. N. K. Mooers, and J. A. Raven (eds.), LectureNotes on Costal and Estuarine Studies, Vol 19. Springer-Verlag, Berlin.

Fiedler, P. C. 1992. Seasonal climatologies and variability of eastern tropicalPacific surface waters. NOAA Tech. Rep., NMFS 109.

, F. P. Chavez, D. W. Behringer, and S. B. Reilly. 1992. Physical andbiological effects of Los Nil los in the eastern tropical Pacific, 1986-1989.Deep-Sea Res. 39:199-219.

, L. J. Lierheimer, S. B. Reilly, S. N. Sexton, R. S. Holt, and D. P.De Master. 1990. Atlas of eastern tropical Pacific oceanographic variabilityand cetacean sightings, 1986-1989. NOAA Tech. Memo., NOAA-TM-NMFS-SWFSC-144.

, V. Philbrick, and F. P. Chavez. 1991. Oceanic upwelling and productivityin the eastern tropical Pacific. Limnol. Oceanogr. 36:1834-1850.

, and S. B. Reilly. 1994. Interannual variability of dolphin habitats in theeastern tropical Pacific. II: Effects on abundances estimated from tuna vesselsightings, 1975-1990. Fish. Bull. 92:451-463.

Ford, R. G. 1991. Computer aided mapping inventory system, CAMRIS® 3.0.Ecological Consulting Inc., Portland, Oregon.

Forney, K. A., D. A. Hanan, and J. Barlow. 1991. Detecting trends in harborporpoise abundance from aerial surveys using analysis of covariance. Fish.Bull. 89:367-377.

93

Franks, P. J. S. 1992. Phytoplankton blooms at fronts: patterns, scales, andphysical forcing mechanisms. Rev. Aquat. Sci. 6:121-137.

Frazer, N. B. 1992. Sea turtle conservation and halfway technology. Cons. Biol.6:179-184.

Frazier, J and S. Salas. 1982. Ecuador closes commercial turtle fishery. Mar. Turt.Newsl. 20:5-6.

Fritts, T. H. 1981. Marine turtles of the Galapagos Islands and adjacent areas of theeastern Pacific on the basis of observations made by J. R. Slevin 1905-1906.J. Herpetol. 15:293-301.

Green, R. E. 1967. Relationship of the thermocline to success of purse seining fortuna. Trans. Am. Fish. Soc. 96:126-130.

Green, D. and F. Ortiz-Crespo. 1982. Status of sea turtle populations in the centraleastern Pacific. Pages 221-233. In K. A. Bjorndal (ed.), Biology andconservation of sea turtles, Smithsonian Institution Press. Washington D. C.

Gunnlaugsson, Th., and J. Sigurjonsson. 1990. NASS-87: Estimation of whaleabundance based on observations made onboard Icelandic and Faroese surveyvessels. Rep. Int. Whal. Commn. 40:571-580.

Hall, M., P. Arenas, M. Garcia, and A. Paros-Sierra. 1992. Drift and dispersalpatterns of sea turtle post-hatchlings in the eastern tropical Pacific: asimulation study (Abstract). Pages 51-52. In M. Salmon and J. Wyneken(compilers), Proceedings of the Eleventh Annual Workshop on Sea TurtleBiology and Conservation 26 Feb-2 March 1991, Jekyll Island, GA. NOAATech. Memo. NMFS-SEFSC-302.

Hansen, D. V., and C. A. Paul. 1984. Genesis and effects of long waves in theequatorial Pacific. J. Geophys. Res. 89:10431-10440.

Hays, G. C., P. I. Webb, J. P. Hayes, I. G. Priede, and J. French. 1991. Satellitetracking of a loggerhead turtle (Caretta caretta) in the Mediterranean. J. Mar.Biol. Ass. U.K. 71:743-746.

Heath, M. E., and S. M. McGinnis. 1980. Body temperature and heat transfer inthe green sea turtle, Chelonia mydas. Copeia 1980:767-773.

Hendrickson, J. R. 1980. The ecological strategies of sea turtles. Amer. Zool.20:597-608.

94

Hiby, A. R., and P. S. Hammond. 1989. Survey techniques for estimatingabundance of Cetaceans. Pages 47- 80. In G. P. Donovan, ed. Thecomprehensive assessment of whale stocks: the early years. Rep. Int. Whal.Commn. Spec. Issue 11.

Hoffman, W. and T. H. Fritts. 1982. Sea turtle distribution along the boundary ofthe Gulf Stream Current off eastern Florida. Herpetologica 38:405-409.

Holt, R. S. 1987. Estimating density of dolphin schools in the eastern tropicalPacific Ocean by line transect methods. Fish. Bull. 85:419-434.

, and J. Cologne. 1987. Factors affecting line transect estimates of dolphinschool density. J. Wildl. Manage. 51:836-843.

, and S. Sexton. 1990. Monitoring trends in dolphin abundance in the easterntropical Pacific using research vessels over a long sampling period: analysesof 1986 data, the first year. Fish. Bull. 88:105-111.

Hopkins, S. R. and J. I. Richardson (ed.$). 1984. A recovery plan for marineturtles. Available from NTIS.

Hughes, D. A. and J. D. Richard. 1974. The nesting of the Pacific ridley turtleLepidochelys olivacea on Playa Nancite, Costa Rica. Mar. Biol. 24:97-107.

Hutchinson, J. and M. Simmonds. 1992. Escalation of threats to marine turtles.Oryx. 26:95-102.

Kamachi, M. and J. J. O'Brien. 1995. Continuous data assimilation of drifting buoytrajectory into an equatorial Pacific Ocean model. J. Mar. Sys. 6:159-178.

Kass, R. E., and A. E. Raftery. 1995. Bayes factors. J. Am. Stat. Assoc. 90:773-795.

Keinath, J. A. 1993. Movements and behavior of wild and head-started sea turtles(Caretta caretta, Lepidochelys kempii). Unpubl. Ph.D. Dissertation, TheCollage of William and Mary, Virginia.

, and J. A. Musick. 1993. Movements and diving behavior of a leatherbackturtle, Dermochelys coriacea. Copeia 1993:1010-1017.

Laake, J. L., S. T. Buck land, D. R. Anderson, and K. P. Burnham. 1994.DISTANCE user's guide V2.1. Colorado Cooperative Fish & WildlifeResearch Unit Colorado State University, Fort Collins, Colo.

95

Laurs, R. M., H. S. H. Yuen, and J. H. Johnson. 1977. Small-scale movements ofalbacore, Thunnus alalunga, in relation to ocean features as indicated byultrasonic tracking and oceanographic sampling. Fish. Bull. 75:347-355.

Limpus, C. 1995. Global overview of the status of marine turtles: a 1995 viewpoint.Pages 605-609. In K. A. Bjorndal (ed.), Biology and Conservation of seaturtles, revised edition. Smithsonian Institution Press, Washington D.C.

Lutkepohl, H. 1985. Comparison of criteria for estimating the order of a vectorautoregressive process. J. Time Ser. Ana ly. 6:35-52.

Machlis, L. 1977. An analysis of the temporal patterning of pecking in chicks.Behaviour 63:1-70.

, P. W. D. Dodd, and J. C. Fentress. 1985. The pooling fallacy: Problemsarising when individuals contribute more than one observation to the data set.Z. Tierpsychol. 68:201-214.

Mann, K. H., and J. R. N. Lazier. 1991. Dynamics of marine ecosystems,biological-physical interactions in the oceans. Blackwell ScientificPublications, Oxford.

Marquez, R. M. 1990. FAO species catalogue. Vol. 11: Sea turtles of the world.An annotated and illustrated catalogue of sea turtle species known to date.FAO Fish. Synop. No. 125, Vol 11.

Marsh, H., and W. K. Saalfeld. 1989. Aerial surveys of sea turtles in the northernGreat Barrier Reef Marine Park. Aust. Wildl. Res. 16:239-249.

McCreary Jr, J. P., H. S. Lee, and D. B. Enfield. 1989. The response of thecoastal ocean to strong offshore winds: with application to circulations in theGulfs of Tehuantepec and Papagayo. J. Mar. Res. 47:81-109.

McCullagh, P., and J. A. Nelder. 1983. Generalized linear models. Chapman andHall, London.

Montenegro, B. and N. Bernal. 1982. Analisis del contenido estomacal deLepidochelys olivacea. Unpublished Thesis Profesional. ENEP Iztacala,Universidad Nacional Autonoma de Mexico.

Mortimer, J. A. 1982. Feeding ecology of sea turtles. Pages 103-109. In K. A.Bjorndal (ed.), Biology and Conservation of Sea Turtles. SmithsonianInstitution Press, Washington, D.C.

96

Nakamura, A., and M. Soma. 1988. Biological telemetry of sea turtles utilizing theARGOS system. Pages 91-100. In C. J. Amlaner Jr. (ed.), Biotelemetry X:Proceedings of the Tenth International Symposium on Biotelemetry, July 31August 5, 1988, Fayetteville, Arkansas. University of Arkansas Press,Fayetteville.

Neter, J., W. Wasserman, and M. H. Kutner. 1989. Applied Linear RegressionModels. Second ed. Irwin, Inc., Homewood, Ill.

Oliver, J. A. 1946. An aggregation of Pacific sea turtles. Copeia 1946:103.

Pak, H. and J. R. V. Zaneveld. 1974. Equatorial front in the eastern Pacific Ocean.J. Phys. Oceanogr. 4:570-578.

Pickard, G. L. and W. J. Emery. 1990. Descriptive physical oceanography, anintroduction. Fifth (SI) ed. Pergamon Press, Oxford.

Pitman, R. L. 1990. Pelagic distribution and biology of sea turtles in the easterntropical Pacific (Abstract). Pages. 143-148. In T. H. Richardson, J. I.Richardson, and M. Donnelly (Compilers). Proceedings of the Tenth Ann.Workshop on Sea Turtle Biol. and Conserv. NOAA Tech. Memo. NMFS-SEFC-278.

. 1993. Seabird associations with marine turtles in the eastern Pacific Ocean.Col. Waterbirds. 16:194-201.

Plotkin, P. T. 1994. Migratory and reproductive behavior of the olive ridley turtle,Lepidochelys olivacea (Eschscholtz, 1829), in the eastern Pacific Ocean.Unpubl. Ph.D. Dissertation, Texas A&M Univ., College Station.

, R. A. By les, D. C. Rostal, and D. W. Owens. 1995. Independent versussocially facilitated oceanic migrations of the olive ridley, Lepidochelysolivacea. Mar. Biol. 122:137-143.

Polacheck, T. 1987. Relative abundance, distribution and inter-specific relationshipof cetacean schools in the eastern tropical Pacific. Mar. Mammal Sci.3:54-77.

Pregibon, D. 1981. Logistic regression diagnostics. Ann. Stat. 9:705-724.

Ramsey, F. L. 1979. Parametric models for line transect surveys. Biometrika66:505-512.

, V. Wildman, and J. Engbring. 1987. Covariate adjustments to effective areain variable-area wildlife surveys. Biometrics 43:1-11.

97

, M. McCracken, J. A. Crawford, M. S. Drut, and W. J. Ripple. 1994.Habitat association studies of the northern spotted owl, sage grouse, andflammulated owl. Pages 189-209. In Case studies in biometry. N. Lange, L.Ryan, L. Billard, D. Brillinger, L. Conquest and J. Greenhouse (ed.$). JohnWiley & Sons, Inc. New York.

Reilly, S. B. 1977. The distribution of pilot whales, Globicephala macrorhyncusGray, 1896, in the eastern tropical Pacific. M.S. Thesis, Calif. Polytech. St.Univ., San Luis Obispo, California.

. 1990. Seasonal changes in distribution and habitat differences amongdolphins in the eastern tropical Pacific. Mar. Ecol. Prog. Ser. 66:1-11.

, and P.C. Fiedler. 1994. Interannual variability of dolphin habitats in theeastern tropical Pacific, I: Research vessel surveys, 1986-1990. Fish. Bull.92:434-450.

, and V. G. Thayer. 1990. Blue whale (Balaenoptera musculus) distribution inthe eastern tropical Pacific. Mar. Mammal. Sci. 6:265-277.

Renaud, M. L. and J. A. Carpenter. 1994. Movements and submergence patterns ofloggerhead turtles (Caretta caretta) in the Gulf of Mexico determined throughsatellite telemetry. Bull. Mar. Sci. 55:1-15.

Richard, J. D., and D. A. Hughes. 1972. Some observations of sea turtle nestingactivity in Costa Rica. Mar. Biol. 16:297-309.

Sakamoto, W., Y. Naito, I. Uchida, and K. Kureha. 1990. Circadian rhythm ondiving motion of the loggerhead turtle Caretta caretta during inter-nesting andits fluctuations induced by the oceanic environmental events. Nippon SuisanGakk. 56:263-272.

Sapsford, C. W., and M. van der Riet. 1979. Uptake of solar radiation by the seaturtle, Caretta caretta, during voluntary surface basking. Comp. Biochem.Physiol. 63A:471-474.

Schwarz, G. 1978. Estimating the dimension of a model. Ann. Stat. 6:461-464.

Scott, G. P., and J. R. Gilbert. 1982. Problems and progress in the US BLM-sponsored CETAP surveys. Rep. Int. Whal. Commn. 32:587-600.

Scott, J. M., Ramsey, F. L. and C. B. Kepler. 1986. Forest bird communities ofthe Hawaiian Islands: their dynamics, ecology and conservation. Studies inAvian Biology, No. 9.

98

Seber, G. A. F. 1982. The estimation of animal abundance. Second ed. Griffin,London.

Sette, 0. 1955. Consideration of midocean fish production as related to oceaniccirculatory systems. J. Mar. Res. 14:398-414.

Shoop, C. R. and R. D. Kenny. 1992. Seasonal distributions and abundances ofloggerhead and leatherback sea turtles in waters of the northeastern UnitedStates. Herp. Monogr. 6:43-67.

Slater, P. J. B. 1974. Bouts and gaps in the behaviour of zebra finches, with specialreference to preening. Rev. Comp. Anim. 8:47-61.

, and N. P. Lester. 1982. Minimizing errors in splitting behaviour into bouts.Behaviour 79:153-161.

Smith, R. C., P. Dustan, D. Au, K. S. Baker, and E. A. Dunlap. 1986.Distribution of cetaceans and sea-surface chlorophyll concentrations in theCalifornia Current. Mar. Biol. 91:385-402.

Steele, J. H. 1974. The structure of marine ecosystems. Harvard University Press.Cambridge, MA.

Stoneburner, D. L. 1982. Satellite telemetry of loggerhead sea turtle movement inthe Georgia Bight. Copeia 1982:400-408.

Stricherz, J. N., J. J. O'Brien, and D. M. Leg ler. 1992. Atlas of Florida StateUniversity Tropical Pacific Winds for TOGA 1966-1985. A Mesoscale Air-SeaInteraction Group Technical Report. Florida State University, Tallahassee, FL.Available from: MASIG, Florida State University, Tallahassee, FL.

Sund, P. N., M. Blackburn, and F. Williams. 1981. Tunas and their environment inthe Pacific Ocean: a review. Oceanogr. Mar. Biol. Ann. Rev. 19:443-512.

Thompson, N., T. Henwood, S. Epperly, R. Lohoefener, G. Gitschlag, L. Ogren, J.Mysing, and M. Renaud. 1990. Marine turtle habitat plan. NOAA Tech.Memo. NMFS-SEFC-255.

Timko, R. E., and A. L. Kolz. 1982. Satellite sea turtle tracking. Mar. Fish. Rev.44:19-24.

Verner, J. 1985. Assessment of counting techniques. Pages 247-302 in R. F.Johnston, ed. Current Ornithology. Vol. 2. Plenum Press, New York.

99

Vinogradov, M. E. 1981. Ecosystems of equatorial upwellings. Pages 69-93. In A.R. Longhurst (ed.) Analysis of marine ecosystems. Academic Press. NewYork.

Volkov. A. F. and I. F. Moroz. 1977. Oceanological conditions of the distributionof Cetacea in the eastern tropical part of the Pacific Ocean. Rep. Int. Whal.Commn. 27:186-188.

Wade, P. R., and T. Gerrodette. 1992. Estimates of dolphin abundance in theeastern tropical Pacific: preliminary analysis of five years of data. Rep. Int.Whal. Commn. 42:533-539.

Whittow, G. C., and G. H. Balazs. 1982. Basking behavior of the Hawaiian greenturtle (Chelonia mydas). Pac. Science 36:129-139.

Wildman, V. J., and F. L. Ramsey. 1985. Estimating effective area surveyed withthe Cumulative distribution function. Tech. Rep. 106, Dept. Statistics, ORState Univ., Corvallis.

Witham, R. 1991. On the ecology of young sea turtles. Fla. Sci. 54:179-190.

Wyrtki, K. 1964. Upwelling in the Costa Rica Dome. Fish. Bull. U.S.63:355-372.

. 1965. Surface currents of the eastern tropical Pacific Ocean. Bull.Inter-Amer. Trop. Tuna Commn. 9:271-294.

. 1966. Oceanography of the eastern equatorial Pacific Ocean. Oceangr. Mar.Biol. Ann. Rev. 4:33-68.

. 1967. Circulation and water masses in the eastern equatorial Pacific Ocean.Int. J. Oceanol. Limnol. 1:117-147.

Zar, J. H. 1984. Biostatistical analysis. 2nd Edition. Prentice-Hall, Inc., EnglewoodCliffs, NJ.

100

Appendices

1000

4-0

Aco

o 100cv

0a)

004- 10

-0

1

APPENDIX 1

1'1'1'1'1'1Nightn = 274

o Dayn = 403

11 T III 1 1 I I I

Long Dives

I 1111111 II0 8 16 24 32 40 48 56 64 72 80 88 96

Dive Duration t (min)

101

Log-survivorship plot of day and night sampled submergence durations. The slopesof the curves are proportional to the probability of a submergence longer than aparticular duration (t) occurring. A change in the slope is used to discriminatebetween types of submergence behavior (Fagan and Young 1978).

0.30.A'

w

a;oc 0.00a)EP

u)

-0.30

APPENDIX 2

Same Day1 Day

A + 1 Day A

A

El

IA

6 I

AA A

A

11111111111111111111111CO CO CO N CO CD t" Co CO '' CO Ps. CO 1. CO Q)

JAG1Z C) (le qi C) ) CV CV C) C) 1/ il, 0 0 '. t.

---\ 0 0 0 0 0 042)(2)C42(424 CO T

C C Q0040 40 40 40 O0O TTTTTT 44OkOkO4 4<\T T T

Date

102

Dates of satellite-acquired locations of an olive ridley turtle in the eastern tropicalPacific plotted against modeled surface current divergence (10-6 s'). Plotted locationswere one day before, after, and on the same day as available divergence data.Surface divergence data was available in six day intervals. The criterion fordetermining a location was at a convergence (shown as dotted lines on the graph) wasvalues __ -0.1 x 10-6 s-1. The criterion for a location at a divergence was values>0.1 x 10' s'.