Evaluation of Removal Sampling for Basinwide Assessment of Atlantic Salmon

Abstract

Removal estimators for stream fish abundance are widely used but can result in biased population estimates at the site level. We conducted computer simulations to examine how the Carle and Strub (1978) estimator, coupled with variation in catchability, influences the accuracy of population estimates at the site level. Site-level population estimates were then used to examine what effect potential bias in the population estimate at a site had on basinwide abundance estimates. Historic electrofishing data collected from Atlantic salmon Salmo salar in the Narraguagus River, Maine, were used as the baseline for construction of these simulations. At the site level, mean percent bias of population estimates was −23% when catchability was low (0.30–0.40) and when the true population was low (1–20 fish). Bias was reduced as the true population size increased and catchability increased. The negative bias at the site level affected total population estimates for the entire river basin. Under current sampling methodology in the Narraguagus River, basinwide population estimates are probably 11–17% lower than the true population. Confidence intervals (95%) would be expected to cover the true population between 65% and 79% of the time. Increasing the amount of sampling had little effect on the negative bias of basinwide population estimates but did reduce the error around the estimate, as expected. These results should serve as a reference point for gauging the effectiveness of current sampling efforts in providing reliable estimates of Atlantic salmon parr in the Narraguagus River. The methodology employed by this simulation study can also be applied to other Atlantic salmon rivers to evaluate current sampling programs.