Estimation results of RPLM model.

Animal-vehicle crashes (AVC) pose risks in rural areas, often leading to casualties and injuries. Despite their infrequent occurrence, AVC can have significant consequences, especially when larger animals are involved. This study investigates factors contributing to fatalities and injuries resulting from animal-involved collisions. It examines 24 variables using 1403 animal-vehicle crash observations on intercity and major intra-city roads from 2016–2021. The study employs a random parameters logit model (RPLM) and ensemble machine learning approaches to explore the contributory factors in crashes. The RPLM accounts for unobserved heterogeneity, identifying significant variables. Meanwhile, the ensemble learner and Shapley Additive exPlanations (SHAP) provide further insights. Key findings show that expressways, roads with one or two lanes per direction, horizontal curvature, and structurally poor pavement surfaces increase the risk of severe crashes, i.e., fatalities and injuries. Side fence barriers and speed bumps also impact crash severity. The absence of side fencing and damaged fencing both positively influence severe crashes, while the presence of speed bumps is likely to increase severe crashes. Camel exposure, vacation-period crashes, and adverse weather also play positive roles. However, heavy truck involvement is negatively associated with severe crashes. Policymakers and road safety authorities can use these findings to implement effective countermeasures to prevent such collisions.