Probability of Detecting Marbled Murrelets at Sea: Effects of Single versus Paired Observers

Abstract

Density estimates for marbled murrelets (Brachyramphus marmoratus) are obtained from at-sea surveys, but survey methods vary among regions. We compared the performance of a single observer with that of an observer operating in a paired-observer team in detecting marbled murrelets during marine surveys in the San Juan Islands, Washington, USA. Performance was measured against an independent observer (IO) who selected a sample of birds; we used this sample to determine the proportion detected by the observer(s). To represent probability of success, we used a function that was the product of a half-normal detection function modified to incorporate scale covariates and a logistic function to represent detection probability on the line, g(0). We used Akaike's Information Criterion (AIC) to select the set of covariates (murrelet group size and behavior, observer, wave height, cloud cover, and the number of primary observers) that best explained variability in g(0) and scale (effect of distance on detection probability). Single observers detected 80% of 274 targets, and paired observers detected 84% of 343 targets selected by the IO. Detection probability was affected by observer, murrelet behavior, and group size; wave height influenced the effect of distance on detection probability. Estimates of detection probability on the line [g(0)] ranged from 0.78 to 0.95 with a single observer; average estimates from paired observers were similarly biased (0.84 to 0.93), but less variable. Options for surveying include retaining single- or paired-observer surveys and accepting bias and variability from g(0) < 1, attempting to reduce bias by using 2 or more observers in a different configuration with a different search pattern, or using an independent observer to remove bias.