Estimating Fish Abundance in Stream Surveys by Using Double-Pass Removal Sampling

Abstract

Double-pass removal sampling is frequently used to quantify absolute abundance of fish in streams. The method can fail, however, if catch on the second pass equals or exceeds that on the first pass. One way to avoid this problem is to pool catch data from several sample sites, but this strategy introduces bias from unequal probability of capture across sample sites. Here we present an analytical approach for pooling that nearly eliminates the bias if probability of capture exceeds 50% and that reduces bias at lower capture probabilities. We also present an application of this approach to the Maryland Biological Stream Survey program, which used double-pass sampling at 270 randomly selected sites throughout Maryland in 1995. Although double-pass sampling failed for 24 of 26 species, it only failed for 2 species with pooling. Bias introduced by pooling was generally less than 10% and was corrected for by our analytical methodology; the estimated probability of capture exceeded 50% for 20 of the 26 species examined.