Sampling designs for estimating the total number of fish in small streams /

Abstract

A common objective of fisheries research is estimating the total number of fish in small streams. The conventional approach involves (1) selecting a small sample of equal-length sections of stream, and (2) estimating the total number of fish in each section using removal method or mark-recapture estimators. Error of estimation of the total number of fish in a stream arises from two sources: (1) extrapolation from the small number of sampled sections to the entire stream, and (2) errors of estimation of fish numbers within sampled sections. This report shows that errors arising from the first source will usually be larger than those arising from the second source. Total errors of estimation can be reduced by making sampled sections equivalent to natural habitat units. Entire pools or riffles should be sampled rather than fixed-length sections of streams. The relative performances of three alternative sampling designs, which can be used when sampled sections are equivalent to natural habitat units, are contrasted in terms of accuracy and cost-effectiveness. Accuracy of estimation can be dramatically improved if sampling designs account for the usually strong, positive correlation between fish numbers and habitat unit sizes.