Effects of population variability on the accuracy of detection probability estimates

Abstract

Observing a constant fraction of the population over time, locations, or species is virtually impossible. Hence, quantifying this proportion (i.e. detection probability) is an important task in quantitative population ecology. In this study we determined, via computer simulations, the effect of population characteristics on estimates of detection probability. Simulation results showed a consistent and significant over-prediction of detection probability across sampling methodologies and estimators. Comparisons between estimators and sampling methods showed significant differences amongst them (estimators accounting for heterogeneity are the most accurate). Population size was the most important factor influencing the uncertainty and accuracy of the estimates. Our results show the need to include population characteristics (either as a correction function or as factors in the observation model) when estimating detection probabilities. The study concludes by suggesting two methods for using population characteristics to estimate detection probabilities.