Abstract
Studies at renewable energy sites often attempt to estimate avian and bat fatalities or fatality rates (e.g., fatalities/megawatt [MW]/year). However, searcher efficiency and level of survey effort among sites are variable. We evaluated how searcher efficiency and proportion of area surveyed affected precision of fatality estimates and probability of detecting rare fatalities (e.g., fatalities of an endangered species). We measured searcher-efficiency rates of human and dog-handler teams for 3 fatality types: feather spots, small carcasses, and large carcasses. We also created high and low searcher-efficiency scenarios to evaluate effects of searcher efficiency across a broader range. Model selection on a set of logistic regression models for our empirical trials supported search team and fatality type as important predictors of searcher-efficiency rates, and demonstrated that dog-handler teams had higher searcher efficiency for all fatality types than human search teams. We used model simulations to evaluate fatality estimates obtained across the 4 searcher-efficiency scenarios and 10 levels of survey coverage: 10–100%, in increments of 10%. At each level of survey coverage, width of confidence intervals decreased as searcher efficiency increased; however, as survey coverage increased, width of the confidence intervals stabilized, and further increases in coverage did not lead to increases in precision. Average fatality estimates were biased positively and this bias increased with lesser searcher efficiency. The survey coverage necessary to achieve 80% detection probability for rare fatalities (defined as 1, 5, or 15 fatalities) decreased as searcher efficiency increased for all fatality types. When a rare fatality was defined as a single fatality out of 1,000, an 80% detection probability was achieved only by dog-handler teams and the high searcher-efficiency scenario, and only for large carcasses. Our results emphasize the need for managers and agencies to consider study objectives and site characteristics when selecting level of survey coverage, type of search team, and other options to maximize searcher efficiency.