Wind energy developers commonly assess collision mortality of birds and bats at wind facilities by conducting post-construction monitoring (PCM) studies. However, it is difficult to detect every instance of a bird or bat collision fatality and, therefore, accurately estimate mortality from collisions. Sources of detection error can include searchers not finding every carcass, scavengers removing carcasses, or carcasses falling outside of the search area. To estimate collision mortality as accurately as possible, investigators need to estimate the correction needed for sources of detection error, often through field study, or what are also referred to as bias trials. Mortality estimators use the results of the trials to calculate probability that a carcass will be detected and combine these with observed counts to ultimately get an estimate of total mortality from collisions.
In recent years, the two estimators most often used to correct for sources of detection error and estimate mortality in North America are the Huso and the Shoenfeld estimators. GenEst is the newest estimator and was designed to improve on the Huso and Shoenfeld estimators by generalizing the key assumptions in both estimators to produce more accurate mortality estimates and to improve comparability among new PCM studies.
This study compared the performance of GenEst to the Huso and Shoenfeld estimators by simulating mortality and search conditions that might occur in in the field and evaluating each estimator’s ability to accurately estimate a known carcass count. The estimators were compared on three metrics:
- Tendency of an estimator to over- or under-estimate mortality
- Ability of an estimator to constrain an estimate and its uncertainty to a narrow range
- Accuracy of the confidence intervals
The study showed that GenEst consistently outperformed the other estimators on all three metrics. Practitioners switching from the Huso estimator to GenEst should see a comparable magnitude in estimates under many circumstances, while those switching from the Shoenfeld estimator should expect to see mortality estimates remain the same or increase.