Underestimation of Trout Population Size by Maximum-Likelihood Removal Estimates in Small Streams

Abstract

We conducted four-pass electrofishing in seven sections of five small Colorado mountain streams to estimate the frequency that two- and three-pass removal estimates underestimated trout abundance, Results for 23 four-pass removal estimates suggest that underestimation by two- and three-pass estimates occurred more than 50% of the time, and this underestimation was most often related to decreasing capture probabilities but was not related to the amount of deep or complex habitat in the streams. Monte Carlo simulations indicated that removal estimates with high capture probabilities and a greater number of capture occasions show the least bias under conditions of decreasing capture probability. When at least three passes are completed, a chi-square test can be used to detect decreasing capture probability, but the power of this test is low below a population size of 200. We therefore recommend that biologists perform at least three passes when using the removal method.