Title
Improved variance estimates of biomass for stream-dwelling fish calculated using removal estimators
Author(s)
Shepard, B. B.;Taper, M. L.;Zale, A. V.
Published
2013
Publisher
Transactions of the American Fisheries Society
Abstract
Total biomass of fish populations has traditionally been computed by multiplying the mean weight of captured fish by the estimated abundance of fish. The variance estimator for this method overestimates variance when a moderate to large portion of the population is sampled. We developed and evaluated finite population correction (FPC) methods for estimating fish biomass in small streams (<5m wetted width) in conjunction with removal population estimators. Using simulated data we found that removal estimates could deviate by more than 50% from true population abundances and be biased when the ratio of total captured fish to the estimated abundance was less than 0.6. These deviations were generally not biased and within 25% of the true population when the ratio was 0.6 or higher. To estimate biomass using FPC methods we investigated a classical a priori sample design and an a posteriori modeled estimator. Both methods take advantage of the fact that relatively high proportions of total populations are typically captured (and can be weighed) in small streams during removal estimates. The a posteriori model-based method can be used to partition estimates of variance and incorporate measurement error, but the a priori design cannot. Estimates of biomass using all three methods were identical. However, precision of FPC methods was significantly better (Wilcoxon paired sign-ranked test: P < 0.001) than the traditional method. Coverage by 95% confidence intervals for the model-based FPC method were much closer to the 95% nominal level than for the traditional method, especially when capture probabilities were higher than 0.5. We recommend using the model-based FPC method because it reduces estimates of SE as the sample size (n) approaches the population size (N) and it provides the most reliable 95% confidence intervals. Received September 20, 2012; accepted January 18, 2013

Access Full Text

A full-text copy of this article may be available. Please email the WCS Library to request.



Back

PUB14740