Detectability analysis in transect surveys

We discuss and illustrate an analysis strategy that adjusts for the influence of variables such as weather, time of day, and observer on the detectability of animals in line transect surveys. The strategy employs ordinary least squares regression analysis followed dy use of a standard estimator of effective area. No new computer software is required because multiple linear regression is available ill all statistical software packages, and a variety of suitable estimators for effective half-width are available in the program DISTANCE. The strategy is applicable to all wildlife surveys that take systematic records of detectability conditions. We apply this approach to a shipboard survey of cheloniid sea turtles in the eastern tropical Pacific Ocean. Sea state (calm seas, white-capped seas) was the primary variable influencing detection of sea turtles. Effective half-width was adjusted for each category of sea state, and effective area surveyed was calculated. Surface density of turtles in 1989-90 was 0.067 turtles/km(2) (95% CI = 0.033-0.084).