Abstract
We implemented an experiment testing the effectiveness of changing turbine cut-in speed on reducing bat fatality at wind turbines at the Casselman Wind Project in Somerset County, Pennsylvania in 2008 and 2009. Our objectives were to 1) determine the difference in bat fatalities at turbines with different cut-in-speeds relative to fully operational turbines, and 2) determine the economic costs of the experiment and estimated costs for the entire project area under different curtailment prescriptions and timeframes.
Twelve of the 23 turbines at the study site were randomly selected for the experiment and we employed three treatments at each turbine: 1) fully operational, 2) cut-in speed at 5.0 m/s (C5 turbines), and 3) cut-in speed at 6.5 m/s (C6 turbines), with four replicates on each night of the experiment. We used a completely randomized design and treatments were randomly assigned to turbines each night of the experiment, with the night when treatments were applied as the experimental unit. We re-randomized these treatments during the second year of the study. We conducted daily searches at the 12 turbines from 27 July to 9 October 2008, and 26 July to 8 October 2009. During this same period, we also conducted daily searches at 10 different turbines that were part of a complementary study to determine if bat activity data collected prior to construction with acoustic detectors can be used to predict post-construction fatalities, and to meet permitting requirements of the Pennsylvania Game Commission‟s (PGC) voluntary agreement for wind energy (herein referred to as “PGC” turbines). These 10 turbines formed an alternative „control‟ to the curtailed turbines. We performed two different analyses to evaluate the effectiveness of changing turbine cut-in speed to reduce bat fatalities; for one we used 12 turbines to determine differences in fatality between curtailment levels and for another we used 22 turbines to determine differences in fatalities between curtailment and fully operational turbines. The experimental unit in the first analysis was the turbine-night and turbines were considered a random blocking factor within which all treatments were applied. In our first analysis, the total number of fatalities estimated to have been killed the previous night, herein referred to as “fresh” fatalities, in each treatment at each turbine was modeled as a Poisson random variable with an offset of the number of days a treatment occurred within a turbine (due to the slight imbalance of the design). For our second analysis, the turbine was the experimental unit, with 12 turbines receiving the curtailment treatment, 10 the control (fully operational at all times). We used all carcasses found at a turbine to estimate the total number of bat fatalities that occurred at each turbine between 27 July to 9 October 2008 and 26 July to 8 October 2009 and compared fatalities using one-way ANOVA.
In 2008, we found a total of 32 fresh bat fatalities at the 12 treatment turbines. At least one fresh fatality was found at each turbine, and 10 of the 12 turbines had at least 1 fatality during a fully operational night, indicating that fatalities did not occur disproportionately at only some turbines, but were well distributed among all turbines. There was strong evidence that the estimated number of fatalities differed among turbine treatments (F2,33 = 7.36, p = 0.004). There was no difference between the number of fatalities for C5 and C6 turbines (χ1 2 = 0.68, p = 0.41). Total fatalities at fully operational turbines were estimated to be 5.4 times greater on average than at curtailed turbines (C5 and C6 combined; χ1 2 = 14.11, p = 0.0005, 95% CI: 2.08, 14.11); in other words, 82% (95% CI: 52–93%) of all fatalities at curtailment turbines likely occurred when the turbines were fully operational. Estimated total bat fatalities per turbine (i.e., all carcasses found and corrected for field bias) were 1.48–5.09 times greater (mean = 2.57) at PGC turbines relative to curtailed turbines, further supporting the contention that reducing operational hours during low wind periods reduces bat fatalities.
In 2009, we found a total of 39 fresh bat fatalities at the 12 treatment turbines. Similar to 2008, we found at least one fresh fatality at each turbine, and 11 of the 12 turbines had at least 1 fatality during a fully operational night, indicating that fatalities did not occur disproportionately at only some turbines and were well distributed among all turbines. We found strong evidence that the estimated number of fatalities over 25 nights differed among turbine treatments in 2009 (F2,33 = 6.94, p = 0.005). There was no difference between the number of fatalities for C5 and C6 turbines (χ1 2 = 0.24, p = 0.616). Total fatalities at fully operational turbines were estimated to be 3.6 times greater on average than at curtailed turbines (C5 and C6 combined; χ1 2 = 12.93, p = 0.0003, 95% CI: 1.79, 7.26); in other words, 72% (95% CI: 44–86%) fewer fatalities occurred when the turbines were curtailed than when the turbines were fully operational. Estimated total bat fatalities per turbine (i.e., all carcasses found and corrected for field bias) were 1.23–2.58 times greater (mean = 1.80) at PGC turbines relative to curtailed turbines, again providing further support for the contention that reducing operational hours during low wind periods reduces bat fatalities. Our comparisons between PGC and curtailed turbines in both years of the study are conservative estimates of the difference because treatment turbines were fully operational one-third of the time during the study.
The lost power output resulting from the experiment amounted to approximately 2% of total project output during the 75-day study period for the 12 turbines. Hypothetically, if the experimental changes in cut-in speed had been applied to all 23 turbines at the Casselman site for the study period (0.5 hour before sunset to 0.5 hour after sunrise for the 75 days we studied), the 5.0 m/s curtailment used would have resulted in lost output equaling 3% of output during the study period and only 0.3 % of total annual output. If the 6.5 m/s curtailment were applied to all 23 turbines during the study period, the lost output would have amounted to 11% of total output for the period and 1% of total annual output. In addition to the lost power revenues, the company also incurred costs for staff time to set up the processes and controls and to implement the curtailment from the company‟s offsite 24-hour operations center.
Our study demonstrated nightly reductions in bat fatality ranging from 44–93% with marginal annual power loss. Given the magnitude and extent of bat fatalities worldwide, the conservation implications of our findings are critically important. While more studies are needed to test changes in turbine cut-in speed among different sizes and types of turbines, wind regimes, and habitat conditions, we believe changing cut-in speeds to the levels we tested offers an effective mitigation strategy for reducing bat fatalities at wind facilities.