Line transect sampling for murrelets: Accounting for incomplete detection and identification

Abstract

We assessed boat-based line transect sampling for monitoring population status and trend of the Kittlitz's Murrelet Brachyramphus brevirostris in Glacier Bay National Park and Preserve, Alaska. We used field experiments to compare efficiency of one versus two observers and to test the assumption that detection near the transect center line was 100%. Because coexisting Kittlitz's Murrelets and Marbled Murrelets B. marmoratus cannot always be distinguished on sight, we developed analytic methods to account for unidentified murrelets in density estimates. Relative to one observer, two observers had 56% higher encounter rates, a >20% higher probability of species identification, and better met the criteria for robust estimation of detection probability. More encounters also increase precision of estimated detection probability and group size. We estimated detection probability near the transect center line to be 0.94 (SE 0.03) and considered methods to relax the assumption of complete detection near the transect center line when estimating density. Relative to methods that exclude unidentified birds (53% of observations), analytic methods incorporating unidentified murrelets increased density estimates for both Kittlitz's and Marbled murrelets by >100% and reduced coefficients of variation by 9% and 15%, respectively. Failure to account for unidentified murrelets and for incomplete detection near the transect center line creates substantial and variable bias and error in density estimates, lessening the ability to assess population status and trend. We recommend the use of two observers, periodic calibration of detection near the transect center line and its incorporation into density estimates, and the use of skilled observers coupled with analytic methods to account for unidentified murrelets.