Confirmation that eagle fatalities can be reduced by automated curtailment of wind turbines

Abstract

Automated curtailment is potentially a powerful technique to reduce collision mortality of wildlife with wind turbines. Previously, we used a before–after–control–impact framework to demonstrate that eagle fatalities declined after automated curtailment was implemented with the IdentiFlight system at a wind power facility in Wyoming, USA. We received substantial interest and feedback regarding our study and, here, we implement several analytical suggestions and include more recent data that strengthen the inference we draw from our results. The five main analytical suggestions we received were to (1) exclude from analysis data that were collected during the period when automated curtailment was only partially implemented; (2) only analyse data from a single make and model of turbine; (3) evaluate changes in the rate of fatality, instead of the yearly numbers of fatalities that result from fluctuations around that rate; (4) calculate a standard measure determining effects of a treatment in a before–after–control–impact study and (5) examine yearly fluctuations of the fatality rate during the before period. After incorporating these suggestions and including additional data collected since the prior paper was published, our results confirm prior work. We demonstrate that eagle fatalities were reduced by 85% (95% highest density interval = 12%, 100%) after implementation of automated curtailment. Rate of fatalities declined by 2.85 eagles per year (−0.67, 5.70) between before and after periods at the treatment site and increased by 2.26 eagles per year (−1.77, 7.37) at the control site. Overall, the fatality rate declined by 4.91 (−0.27, 11.27) more eagles per year at the treatment site than at the control site. The probability that the fatality rate declined at the treatment site relative to the control site was 0.97. Our re‐analysis strengthens our inference by using more robust analyses and data to support the conclusions of the prior study suggesting that automated curtailment was effective at reducing eagle fatalities at our treatment site. Because of the site‐ and species‐specific nature of our work, future research should examine the efficacy of automated curtailment at other sites, with other species, and under different curtailment regimes.

Reanalysis ignores pertinent data, includes inappropriate observations, and disregards realities of applied ecology: Response to Huso and Dalthorp (2023)

“…Their critique has several flaws and, as we explain below, our original results remain informative. Further, we have now published a reanalysis of our original work that includes additional data, incorporates some suggestions included in Huso and Dalthorp's (2023) critique, and which confirms the results of our original analysis (McClure et al, 2022). Huso and Dalthorp (2023) describe several aspects of our study as purported errors.…”

“…Surely, concurrent fatality rates at a site with similar habitat and so nearby (~15 km) as to be well within the daily movement capacity of eagles, are useful to provide inference into changes in fatalities at our treatment site. In our updated analysis (McClure et al, 2022), we demonstrate that the fatality rates at the control and treatment sites appeared to fluctuate similarly during the before period, with the treatment site being deadlier in every year than the control site. Yet, only after the implementation of automated curtailment did the control site become deadlier than the treatment site.…”

“…It is almost universally the case that post hoc manipulations of a dataset can produce novel results and, thus, we stand by our demarcation date as a reasonable a priori choice in our original paper. In a subsequent reanalysis that validates our original conclusions (McClure et al, 2022), we were able to take an even more conservative approach, ignoring all data from the transition period during which automated curtailment was being incrementally implemented.…”

Effectiveness of an artificial intelligence-based system to curtail wind turbines to reduce eagle collisions

“…The time required to curtail a wind turbine could influence effectiveness by increasing the distance at which curtailments must be issued to be effective. At another North American wind project, where turbines could be curtailed as quickly as 20 s, this system was shown to reduce eagle collisions by 82–85% [ 7 , 8 ]. However, the average time required to curtail turbines at Manzana Wind Power Project is approximately 60 s (A.E.P.…”

“…One AI-based system, IdentiFlight, has been installed at locations in North America, Europe, and Australia and has been associated with reduced mortalities of intended targets (eagles) in one location where it was assessed [ 7 , 8 ]. This system incorporates a network of tower-mounted cameras and computers to detect, identify, and track avian species up to 1000 m from the cameras (see S1 File ).…”