Estimating wildlife strike costs at US airports: A machine learning approach

• DOI: 10.1016/j.trd.2021.102907
• Publication Date: 2021-05-27T14:23:00Z
• Authors: Levi Altringer; Jordan Navin; Michael J. Begier; Stephanie A. Shwiff; Aaron Anderson
• Keywords: Veterinary Medicine; Cost imputation; Population Biology; Terrestrial and Aquatic Ecology; 330; Epidemiology; 05 social sciences; Natural Resources Management and Policy; 0211 other engineering and technologies; Life Sciences; 02 engineering and technology; Other Veterinary Medicine; Veterinary Microbiology and Immunobiology; Wildlife strikes; Animal Sciences; Natural Resources and Conservation; Machine learning; 0502 economics and business; and Public Health; Other Environmental Sciences; Veterinary Infectious Diseases; Veterinary Preventive Medicine; Zoology; Environmental Sciences

Abstract Current lower bound estimates of the economic burden of wildlife strikes make use of mean cost assignment to impute missing values in the National Wildlife Strike Database (NWSD). The accuracy of these estimates, however, are undermined by the skewed nature of reported cost data and fail to account for differences in observed strike characteristics—e.g., type of aircraft, size of aircraft, type of damage, size of animal struck, etc. This paper makes use of modern machine learning techniques to provide a more accurate measure of the strike-related costs that accrue to the US civil aviation industry . We estimate that wildlife strikes costed the US civil aviation industry a minimum average of $54.3 million in total losses annually over the 1990–2018 period. If one assumes that wildlife strikes were underreported by as much as a factor of 3 over the same period, our estimates still fall below previous lower bound estimates.