A logistic regression model for estimating turbine mortality at hydroelectric generating stations

We present a method that allows separation of fish mortality caused by handling and capture techniques from that caused by passage through a turbine. Fish that are naturally entrained into the turbine tube are captured with nets deployed in the turbine tailrace for varying lengths of time. The live or dead status of captured fish is modeled as a binomial response that is a function of the duration of net deployment. Within this model, the intercept is an estimate of the mortality of fish that have spent zero time in the net. For species that do not suffer high mortality from other components of the capture process (such as removal from the net), this intercept may be interpreted as an estimate of turbine mortality. If mortality from other components is high, the intercept cannot be interpreted as turbine mortality without correction for mortality from the other sources. We suggest a modification to the model that allows estimation of mortality from these components. We demonstrate the method with data for 12 species of fish captured at the Annapolis Tidal Generating Station, Nova Scotia, Canada. Acute turbine mortality estimates ranged from 0.0% for sea lamprey Petromyzon marinus to 23.4% for American shad Alosa sapidissima.