Epperly et al.: Aerial surveys for sea turtles 



255 



Cape Hatieras 



ATLANTIC OCEAN 



Cape Lookout 



Figure 1 



Core Sound and subareas of Pamlico Sound flown in aerial 

 surveys for sea turtles in North Carolina inshore waters, 

 1989-91. 



the surveys monthly beginning in spring 1989 and 

 bimonthly May-November 1990 and 1991. Because 

 it was difficult to obtain military airspace clearance 

 over Pamlico Sound and because the results of the 

 1989-90 surveys indicated that our effort was best 

 expended in Core Sound (greater sighting rates), the 

 only area surveyed in 1991 was Core Sound. 



We employed a systematic sampling design. The 

 underlying assumption was that the systematic 

 sample could be treated as a random sample. There 

 was no reason to assume that the number of turtles 

 sighted per transect would be autocorrelated (i.e. we 

 assumed no areal trend in density or correlations 

 between neighboring transect values). As recom- 

 mended by Cochran (1977) and Eberhardt et al. 

 (1979) in order to avoid potential selection biases of 

 systematic sampling, the starting transect for each 

 survey was chosen at random from all possible 

 transects in the survey. Transect lines ran east-west 

 and were spaced equi-distant from the starting 

 transect. On the basis of the maximum known swim- 

 ming speed of a loggerhead turtle (6 km/h, Keinath, 

 1993), transects were spaced far enough apart so that 

 a turtle could not be sighted twice during any one 

 survey. LORAN was used to maintain position on the 

 prescribed transects. Beginning and ending longitu- 

 dinal coordinates and time were recorded for each 

 transect flown. Two observers on opposite sides of 

 the plane scanned the waters, recording the time 



(with synchronized watches) and perpendicular angle 

 to each turtle sighted (with handheld clinometers). 

 On the assumption that groundspeed within a 

 transect was constant, turtle positions were calcu- 

 lated by interpolating time and longitudinal coordi- 

 nates and by converting sighting angle and survey 

 altitude to perpendicular distance from the transect. 

 We used both strip- and line-transect theory to 

 analyze the data. First, a histogram of all perpen- 

 dicular sighting distances was constructed, one for 

 Core Sound and one for Pamlico Sound (Fig. 2). From 

 these histograms we empirically determined the strip 

 width over which the probability of sighting a turtle 

 was not reduced by nearness to the plane ( acute view- 

 ing angle; turtles diving to avoid the plane) or by 

 distance from the plane (reduced detection) to be 

 0.15-0.30 km from the flight line. Observations 



