Fig 1.
Locations of the Juniper Canyon, Big Horn I, and Big Horn II wind energy facilities in Washington, USA, and the Shiloh I wind energy facility in California, USA.
Table 1.
Field method details for the four study periods.
We conducted scanning searches at the Juniper Canyon (JC), Big Horn I (BH I), Big Horn II (BH II), and Shiloh I wind energy facilities. Due to the exploratory nature of our study, we made minor protocol adjustments (e.g., expanded search radii, size of search team reduced) after encouraging results from early trials.
Table 2.
Visibility class definitions and percent coverages within searched areasa during searcher efficiency trials at the Juniper Canyon (JC), Big Horn I (BH I), Big Horn II (BH II), and Shiloh I wind energy facilities.
Fig 2.
Diagram of eagle scan positions around a turbine (not to scale).
Fig 3.
Turkey decoys with feather harnesses were used as a proxy for eagle carcasses during searcher efficiency trials.
Fig 4.
Examples of visibility classes present with search plots.
(a) easy, (b) moderate, (c) difficult, (d) obstructed.
Table 3.
Searcher efficiency trial results as a function of year and visibility class at the Juniper Canyon (JC), Big Horn I (BH I), Big Horn II (BH II), and Shiloh I wind energy facilities.
Overall, searchers detected 76% of all decoy placements.
Table 4.
Searcher efficiency models with delta (Δ) sample-size Akaike’s information criterion (AICc) ≤2.
Top models for decoy detection all included the main effects of distance, visibility, facility, and year. We selected the third-ranked model because a distance by visibility interaction was more plausible than a distance by year interaction and because the information criterion differed little between the second- and third-ranked models.
Table 5.
Model parameters for logistic regression of searcher efficiency (response was the log-odds of a detection) at the Juniper Canyon (JC), Big Horn I (BH I), Big Horn II (BH II), and Shiloh I wind energy facilities.
The reference parameter (i.e., the intercept) predicts detection probability at zero distance in difficult visibility during the first year at the Big Horn I (BH I) and Big Horn II (BH II) wind energy facilities. Note that this reference condition does not occur in the data because difficult visibility areas were not searched at the Big Horn I (BH I) and Big Horn II (BH II) wind energy facilities.
Fig 5.
Decoy detection within visibility classes at the Juniper Canyon, Big Horn I/Big Horn II, and Shiloh I wind facilities.
Table 6.
Searcher efficienciesa for the different studies at the Juniper Canyon (JC), Big Horn I (BH I), Big Horn II (BH II), and Shiloh I wind energy facilities.
Table 7.
Tested carcass density distribution models with corrected Akaike’s information criterion (AICc) and delta (Δ) AICc, and fitted parameters.
Table 8.
Proportion of large raptor carcasses expected within each visibility classa at the Juniper Canyon (JC), Big Horn I (BH I), Big Horn II (BH II), and Shiloh I wind energy facilities.
Fig 6.
Fitted and observed density distribution of raptor carcasses.
Table 9.
Carcass persistence modelsa with corrected Akaike’s information criterion (AICc) and delta (Δ) AICc by facility and year for the Juniper Canyon (JC), Big Horn I (BH I), Big Horn II (BH II), and Shiloh I wind energy facilities.
Table 10.
Carcass persistence distributions, model parameters and probabilities of persistencea by facility, bird type, and search interval for the Juniper Canyon (JC), Big Horn I (BH I), Big Horn II (BH II), and Shiloh I wind energy facilities.
Fig 7.
Empirical carcass persistence for raptors and game birds, by facility.
Table 11.
Overall probability of detectiona by facility, bird type, and search interval for Juniper Canyon (JC), Big Horn I (BH I), Big Horn II (BH II), and Shiloh I.
Table 12.
Evidence of Absence results using a range of overall probability of detection estimates and carcass counts.
Higher detection probabilities result in lower take estimates for a fixed carcass count.