Epperly et al.: Aerial surveys for sea turtles 



259 



pattern (Table 1; Fig. 3). Volunteer commercial fish- 

 ermen and the general public reported turtles in in- 

 shore waters April-December. Turtles were also 

 sighted in the sounds during April-December aerial 

 surveys. Spring aerial surveys (March-May) indi- 

 cated that turtles were distributed mainly in Core 

 Sound and along the eastern edge of southern 

 Pamlico Sound. Summer (June-August) and fall 

 (September-November) aerial surveys demonstrated 

 that turtles were distributed throughout the sounds. 

 No sea turtles were sighted during fall 1990 aerial 

 surveys, but turtles were reported in the area by the 

 public and by commercial fishermen (Epperly et al., 

 in press, a). Turtles were still present in Core Sound 

 in December 1989, but none was sighted during Janu- 

 ary or March 1990 aerial surveys. 



Species were generally indistinguishable from the 

 air because of their small size, except for leather- 

 back sea turtles, Dermochelys coriacea, which were 

 sighted only during the December 1989 survey (three 

 individuals). The loggerhead turtle, with a reddish- 

 brown carapace, was the species most often identi- 

 fied. Data from commercial fishermen indicated that 

 the species composition in Pamlico and Core Sounds 

 was 80% loggerhead, 159?- green, and 5% Kemp's rid- 

 ley sea turtles; leatherback turtles infrequently en- 

 ter inshore waters, and hawksbills, Eretmochelys 

 imbricata, are very rare (Epperly et al., in press, a). 

 Nearly all of the turtles measured by fishermen were 

 greater than 30 cm carapace length (measured over 

 the curve) — the smallest model tested and successfully 

 detected in the aerial survey experiment. 



Density estimates from line- and strip-transect 

 analysis are given in Table 1. The Fourier series es- 

 timator fit the sighting distance data from both 

 sounds well <x 2 goodness-of-fit test, P>0.05). Values 

 of/10) differed substantially between the two sounds: 

 /tO) Con? =8.13 (SE=0.75) and/lO) Pam/;co =5.99 (SE=0.52). 

 Confidence intervals for the estimates of /10) s over- 

 lapped at the 95% confidence level but not at the 90% 

 confidence level. The cause of the difference in our 

 ability to sight turtles between the two sounds was 

 not obvious. Observer fatigue could have been a fac- 

 tor. Transects in Core Sound were short (2.7-14.9 

 km) and observers were able to take frequent breaks 

 between them. Pamlico Sound transects were long 

 ( 13.9-57.1 km in northern Pamlico Sound; 37.5-94.1 

 km in southern Pamlico Sound), and breaks occurred 

 infrequently. Homogeneity of background could have 

 been another factor. Core Sound waters were rela- 

 tively clear, and bottom structures (channels, 

 seagrass beds, etc.) were usually visible. This het- 

 erogeneous background served to attract observers 

 and to intensify the observers' searches in order to 

 detect turtles. Consequently, the visual sweep of the 



observers was confined to an area near the flight 

 path. Conversely, the majority of Pamlico Sound 

 waters usually were turbid and presented a homo- 

 geneous background, except for the easternmost por- 

 tion of the sound which was very similar to Core Sound. 



Line-transect estimates of density in Core Sound 

 averaged 40% greater than estimates derived from 

 strip-transect theory (Table 1). Line-transect esti- 

 mates of density in Pamlico Sound were, on average, 

 14% greater than strip-transect estimates. Coefficients 

 of variation of strip- and line-transect estimates of den- 

 sity were nearly identical within each sound (67% for 

 strip- and 66% for line-transect estimates for Pamlico 

 Sound; 47% and 54% for strip- and line-transect esti- 

 mates, respectively, for Core Sound) (Table 1). 



Application of line-transect and strip-transect 

 analyses to the North Carolina aerial survey data 

 requires several assumptions. Strip-transect analy- 

 sis assumes that 1) transect lines are randomly lo- 

 cated, 2) the strip over which all turtles are assumed 

 to be seen and counted, 0.15-0.30 km, remains con- 

 stant during a single survey and from survey to sur- 

 vey, i.e. sighting conditions (distance from plane, size 

 of turtles, sun position and glare, sea state, weather, 

 etc.) do not affect the ability to sight turtles, 3) no 

 turtle is counted more than once in a given survey, 

 and 4) sightings are independent events. In addition 

 to the first, third, and fourth assumptions above, line- 

 transect analysis requires that 1) all turtles on the 

 line (defined as 0.15 km from the flight line) are seen 

 with certainty, 2 ) turtles do not move prior to sight- 

 ing or before distance measurements are made, 3) 

 measurements are taken without error, and 4) the 

 probability density function remains constant dur- 

 ing a single survey and from survey to survey (i.e. 

 the ability to sight turtles does not change). 



The underlying assumptions of both methodologies 

 are violated in important ways, primarily with re- 

 spect to the ability to sight turtles. For strip-transect 

 analysis, conditions are such that probabilities of 

 sighting individual turtles within the strip are less 

 than one. In addition, these probabilities vary within 

 and among surveys. For line-transect theory, we do 

 not know that all turtles on the line (x=0) are seen. 

 The histogram of sighting distances (Fig. 2) indicates 

 avoidance behavior in response to the aircraft in com- 

 bination with poor downward visibility near the air- 

 plane, but we cannot be sure that locating the line 

 Cr=0) at 0.15 km from the flight line eliminated this 

 effect entirely. In addition, the use of a pooled pdf 

 may not be completely valid, because factors affecting 

 the ability to sight turtles varied over the course of the 

 study. As applied, strip-transect methods assuredly 

 underestimate the density of turtles on the surface of 

 the water. Line-transect methods, however, may over- 



