his article is a tutorial for the R-package carcass. It starts with a short overview of common methods used to estimate mortality based on carcass searches. hen, it guides step by step through a simple example. First, the proportion of animals that fall into the search area is estimated. Second, carcass persistence time is estimated based on experimental data. hird, searcher efficiency is estimated. Fourth, these three estimated parameters are combined to obtain the probability that an animal killed is found by an observer. Finally, this probability is used together with the observed number of carcasses found to obtain an estimate for the total number of killed animals together with a credible interval.
Many wind-power facilities in the United States have established effective monitoring programs to determine turbine-caused fatality rates of birds and bats, but estimating the number of fatalities of rare species poses special difficulties. The loss of even small numbers of individuals may adversely affect fragile populations, but typically, few (if any) carcasses are observed during monitoring. If monitoring design results in only a small proportion of carcasses detected, then finding zero carcasses may give little assurance that the number of actual fatalities is small. Fatality monitoring at wind-power facilities commonly involves conducting experiments to estimate the probability (g) an individual will be observed, accounting for the possibilities that it falls in an unsearched area, is scavenged prior to detection, or remains undetected even when present. When g < 1, the total carcass count (X) underestimates the total number of fatalities (M). Total counts can be 0 when M is small or when M is large and g<< 1. Distinguishing these two cases is critical when estimating fatality of a rare species. Observing no individuals during searches may erroneously be interpreted as evidence of absence. We present an approach that uses Bayes' theorem to construct a posterior distribution for M, i.e., P(M \ X, ĝ), reflecting the observed carcass count and previously estimated g. From this distribution, we calculate two values important to conservation: the probability that M is below a predetermined limit and the upper bound (M*) of the 100(1 - α)% credible interval for M. We investigate the dependence of M* on α, g, and the prior distribution of M, asking what value of g is required to attain a desired M for a given α. We found that when g < -0.15, M* was clearly influenced by the mean and variance of ĝ and the choice of prior distribution for M, but the influence of these factors is minimal when g > -0.45. Further, we develop extensions for temporal replication that can inform prior distributions of M and methods for combining information across several areas or time periods. We apply the method to data collected at a wind-power facility where scheduled searches yielded X = 0 raptor carcasses.
This is an open access article under the terms of the Creat ive Commo ns Attri bution-NonCo mmerc ial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.
Estimating bird and bat mortality at wind facilities typically involves searching for carcasses on the ground near turbines. Some fraction of carcasses inevitably lie outside the search plots, and accurate mortality estimation requires accounting for those carcasses using models to extrapolate from searched to unsearched areas. Such models should account for variation in carcass density with distance, and ideally also for variation with direction (anisotropy). We compare five methods of accounting for carcasses that land outside the searched area (ratio, weighted distribution, non-parametric, and two generalized linear models (glm)) by simulating spatial arrival patterns and the detection process to mimic observations which result from surveying only, or primarily, roads and pads (R&P) and applying the five methods. Simulations vary R&P configurations, spatial carcass distributions (isotropic and anisotropic), and per turbine fatality rates. Our results suggest that the ratio method is less accurate with higher variation relative to the other four methods which all perform similarly under isotropy. All methods were biased under anisotropy; however, including direction covariates in the glm method substantially reduced bias. In addition to comparing methods of accounting for unsearched areas, we suggest a semiparametric bootstrap to produce confidence-based bounds for the proportion of carcasses that land in the searched area.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.