Using Machine Learning Models to Evaluate Raptor Relocation for Red-Tailed Hawks at Portland International Airport (PDX)

Portland International Airport

As part of their effort to reduce risks from wildlife at Portland International Airport (PDX), the Port of Portland implements a raptor management program that includes relocating red-tailed hawks from  PDX to distant locations. This strategy is intended to reduce the chance for human-wildlife conflicts and serious safety risks such as wildlife-aircraft collisions. Relocating raptors is a non-lethal alternative to the more common practice of lethal removal of raptors at airports. PDX is one of a few airports that is investigating whether raptor relocation is a viable alternative to lethal methods. To help address this matter, the Port is interested in answering the following question: Are there variables such as bird age and sex, the time of year when birds are released, and distance of release sites that increase the likelihood of a bird not returning such that would indicate the success of the relocation program?

Anchor QEA leveraged advanced machine learning techniques such as Classification and Regression Trees with Random Forest and Generalized Logistic Regression with exhaustive model screening, to develop various models and evaluate the relative significance each variable has on the operation’s effectiveness to predict the relocation success rate. The best model was selected by assessing the goodness of fit, variance versus bias tradeoff, and predictive accuracy.

Compared to previous studies using models such as Generalized Linear Regression, the implementation of machine learning models provided better predictive accuracy on the success rate of the relocation operation and a clearer picture of what variables really impact successful relocation. This enabled our team to provide the Port with a robust evaluation of the current raptor relocation program and offer advice on how to increase the success rate of the relocation operation.