Understanding the distribution and abundance of plant and animal populations is fundamental to ecology, yet abundance is one of the most difficult parameters to estimate, even for a closed population. At wind power facilities we are concerned with estimating the abundance of the population of animals killed after colliding with turbine blades, yet the unique characteristics of this problem preclude the use of many currently available tools: (1) the population is not mobile; (2) the population is not closed; (3) the parameter of interest is the super population, i.e., the total number of animals entering the population; (4) the probability of detection of members of the population<1; and (5) the probability of detection is not equal among members of the population and may be unique to each. Numerous abundance estimators (of fatality) have been developed to simultaneously address these constraints but inherent differences in their assumptions can lead to radically different estimates of fatality, resulting in confusion and poor inferential capacity.
Here we present a case study highlighting this problem and suggest a solution to alleviate the confusion surrounding which assumptions are most likely met and which estimator is the most appropriate to use. Statisticians who developed several of the estimators in current use at wind power facilities, have recognized the commonalities among their approaches and are working together to develop software that combines these approaches under a single generalized estimator. The generalized estimator (GenEst) has several advantages over previous estimators in that it addresses all the constraints listed above, allows the user to test assumptions regarding input parameters, and select the approach that best reflects the situation and data. This flexibility allows GenEst to yield statistically valid results across a wide spectrum of study designs with greatly reduced potential for user error. The goals of this endeavor include 1) providing guidance on study design to increase efficiency and reduce costs of fatality studies, 2) standardizing carcass searches and data analyses, and 3) reducing bias and thereby improve accuracy of fatality estimates generated from carcass searches. The applicability of the approach is not confined to wind power facilities, but can be used in any situation in which the objective is an estimate of a dead super-population, e.g., oil spills, power-line or fence-line fatality rates, etc.