Collision risk models (CRMs) are potentially useful, albeit crude, tools in predicting the potential avian collision mortality rates which rotating turbine blades at operational wind farms may incur. The 'Band' CRM (Band et al. (2007) has probably received the most utility in practice and therefore, attention, concerning potential deficiencies.
The most obvious barrier to the Band CRM in realising its potential to produce realistic mortality predictions is that it assumes that birds do not avoid collision, whereas it is apparent that avoidance rates are very high at operational wind farms. Deriving estimates of avoidance rates has thus become an urgent research field.
The present study calculated estimates of avoidance rates for the golden eagle Aquila chrysaetos under the Band CRM using results from monitoring studies at four USA wind farms, attempting to account for the many biases which are inherent in avoidance rate calculations.
Avoidance rate estimates for golden eagles varied between 98.64 % and 99.89 % depending on site and uncertainty associated with observed mortality rates before and after adjustment for potential biases. An overall 'worst case' estimate weighted by the scale of study was 99.33 % and the mean unweighted 'worst case' (lowest) avoidance rate for the four wind farms was 99.19 %. A precautionary value of 99.0 % is therefore recommended for use in predictive assessments of wind farm proposals. Other recommendations include the need for further research which avoids the biases inherent in many existing studies of wind farm effects on birds.
The estimated avoidance rates, and the means of their derivation, documented by the present study, are contrasted with those calculated for golden eagle by Fernley (2008), which are higher. Several discrepancies are identified which would lead to elevated estimates of avoidance rates by Fernley (2008), such as not accounting for some eagle deaths or relatively high inactivity of turbines at some sites, or using inflated measures of eagle activity.