Application of Logistic Regression to Salmon Species Apportionment Using Sonar Length Measurements

The migration timing of adult pink salmon and other salmon species, like sockeye salmon, are similar in some rivers of Alaska.  Fish wheel catch rates have been used to apportion salmon counts from the sonars.  However, pink salmon are smaller and tend to swim closer to shore to avoid the strong current, there is a concern that the fish wheel has a higher selectivity of pink salmon, and therefore results in negatively biased for the other salmon species.  In this study we used sonar length measurements and applied a logistic regression model to separate the abundant pink salmon from other salmon species. With particular fish lengths from the sonars, the probabilities of being pink salmon can be estimated.  We then simulated draws of fish (pink or non-pink salmon) with the probabilities. This way, we were able to estimate not only the proportion of pink salmon passing through the sonar site, but also precision of the proportion.