A Conceptual Guide to Detection Probability for Point Counts and Other Count-based Survey Methods

Accurate and precise estimates of numbers of animals are vitally needed both to assess population status and to evaluate management decisions. Various methods exist for counting birds, but most of those used with territorial landbirds yield only indices, not true estimates of population size. The need for valid density estimates has spawned a number of models for estimating p, the ratio of birds detected to those present. Wildlife biologists can be assisted in evaluating these methods by appreciating several subtleties of p: (1) p depends upon the duration of the count, (2) p has two independent components, availability of cues and detectability of cues, (3) detectability is a function of both the conspicuousness of cues (e.g., vocalizations, conspicuous movements) and the abundance of cues, and (4) discontinuous production of cues lowers availability, which has a direct and sometimes profound effect on p. Two recently-updated methods of estimating p, doubleobserver sampling and distance-sampling, are better suited to estimating detectability, while two others, double sampling and removal sampling, are better suited to estimating availability. While none of the four offers a complete solution at present, hybrids are under development, and a technique that can yield estimates of availability and detectability from survey data may be available in the near future.