Benson et al,: Abundance, distribution, and habitat of Dermochelys coriaceo off California 



339 



40"N- 



38 N" 



Analytical methods 42 N- 



For analysis of regional patterns of leatherback 

 turtle density and distribution, the study area was 

 divided into five geographic strata, near prominent 

 features of the coastline, to capture variation in 

 bathymetric and oceanographic characteristics (Fig. 

 1); north coast (3765 km^), Pt. Arena (772 km-). 

 Gulf of the Farallones (4189 km-'), Monterey Bay 

 (908 km-), and south central California (1849 km^). 



Leatherback turtle sighting rates were evalu- 

 ated for potential effects of sea state, glare, and 

 cloud cover, by using a two-way extension of the 

 Kruskal-Wallis nonparametric analysis of vari- 

 ance (Scheirer et al., 1976), because these factors 

 can influence one's ability to detect marine ani- 

 mals. Glare conditions were categorized as either 

 optimal, when the sun position was behind the 

 aircraft or directly ahead and did not affect the 

 primary field of view, or marginal, when the sun 

 position was just ahead of or perpendicular to the 

 aircraft's travel direction. Cloud cover was divid- 

 ed into four categories: clear (<25% cloud cover), 

 partly cloudy (26-50%), mostly cloudy (51-75%), 

 and overcast (76-100%). Data collected in sea 

 states greater than Beaufort 3 were excluded from 

 the analysis; cloud cover and glare did not appear 

 to exhibit any effect (see "Results" section) and 

 were not considered further. 



The detection function of leatherback turtles 

 was estimated from the pooled perpendicular dis- 

 tances by using DISTANCE software (Thomas 

 et al.2). Truncation of the 5-10% most distant 

 sightings was investigated but it did not improve 

 precision or model fit, and the final models in- 

 cluded all data without truncation. Hazard, half- 

 normal, and uniform models with and without 

 cosine adjustment terms were fit to the ungrouped 

 perpendicular distance data. The best model was 

 selected according to Akaike's information crite- 

 rion, AIC (Akaike, 1973), and visual inspection of 

 goodness-of-fit. 



The density (D) and abundance (A'^) of leather- 

 back turtles within each geographic stratum, j, were 

 estimated by using standard line-transect formulae 

 (Buckland et al., 2001): 



36"N- 



34 N- 





North 

 Coast 



Monterey 

 Bay 



North Pacific 

 Ocean 



South 



Central 



California 



Pt. Conception 



126°W 



124°W 



— r— 

 122W 



120°W 





D.= 



N,=D^A^, 



(1) 



(2) 



Figure 1 



Cahfornia study area with survey transects and geographic 

 strata. Open squares represent locations of leatherback 

 turtle t Dermochelys coriacea) sightings during systematic 

 surveys. Thin gray line denotes the 90-m isobath. 



where n = the total number of turtles seen during 

 systematic surveys; 

 /(O) = the probability density function evaluated 

 at zero perpendicular distance; 

 L = the linear distance surveyed in km; 

 ^(0) = the probability of detection at zero perpen- 

 dicular distance, estimated from leather- 

 back turtle dive data (see below); and 

 A, = the area size in km-. 



2 Thomas, L., J. L. Laake, S. Strindberg, F. F. C. Marques. S. 

 T. Buckland. D. L. Borchers, D. R. Anderson, K. P. Burnham, 

 S. L. Hedley, J. H. Pollard, and J. R. B. Bishop. 2004. Dis- 

 tance 4.1. Release 2. Research Unit for Wildlife Population 

 Assessment, University of St. Andrews, UK. Website: http:// 

 www.ruwpa.st-and.ac.uk/distance/ (accessed 19 November 

 2006). 



Although we attempted to complete each transect the 

 same number of times, weather conditions often resulted 

 in uneven coverage. To avoid this potential within- 

 stratum source of bias, encounter rates (rij/Lj) for each 

 geographic stratum were calculated from the individual 

 transect encounter rates, weighted according to the 

 proportional contribution of each transect: 



