Railway mortality for several mammal species increases with train speed, proximity to water, and track curvature

Wildlife-vehicle collisions, particularly on roads and railways, pose a significant threat to various large mammal populations worldwide, potentially impacting their viability. Mitigation strategies are crucial for species conservation, and these require careful planning to balance cost-effectiveness with ecological impact. There's a critical need to decide between extensive (widespread, continuous mitigation like fencing) and intensive, site- or time-specific approaches. Extensive mitigation is often preferred where affordability allows, aiming to prevent population fragmentation and widespread mortality, especially in protected areas with threatened species. However, intensive, specific interventions are more suitable when resources are limited or when specific hotspots of collisions endanger certain species in predictable locations or seasons. The choice depends on factors like the type of transportation corridor (roads vs. railways), the cost of mitigation, and the societal impact of collisions (human injury). Railways, unlike roads, often lack the societal pressure for expensive, extensive mitigation due to a lower likelihood of human injury. This study addresses the lack of comprehensive research on railway-wildlife collisions by focusing on the spatial and temporal factors affecting mortality risk and their variability among different mammal taxa.

Existing literature highlights the significant impact of road and railway collisions on large mammal populations. Studies have examined mitigation measures for roads, with extensive fencing and crossing structures emerging as a gold standard in affluent areas. However, these measures are not always cost-effective or feasible, particularly for railways where human injury is less frequent. Research on railway mitigation is relatively limited compared to road mitigation, but some advocate a similar extensive approach. However, limitations such as increased mortality at fence ends and reduced societal demand for expensive mitigation due to the lower risk of human injury, make site-specific mitigation strategies more appealing for railways. The literature also lacks a generalized understanding of the spatial and temporal factors influencing railway-caused wildlife mortality, particularly regarding the variation across different taxa.

This study analyzed a 24-year (1995-2018) database of 646 confirmed wildlife mortality events from the Parks Canada Agency in Banff and Yoho National Parks, Canada. Eleven mammal species were categorized into three guilds: bears, other carnivores, and ungulates. The researchers compared the spatial attributes (topography, land cover, train operation) of mortality sites with paired random locations at four spatial scales (study area, 2.5km, 5km, 10km radii). Logistic regression models were used to determine which landscape variables best explained mortality locations, considering factors related to train detection, animal movement, forage availability, and human security. The explanatory variables included track curvature, topographic complexity, distance to water and water cover, distance to roads and sidings, shrub cover, canopy closure, seasonal factors (winter/summer, spring/autumn), snow cover, precipitation, and maximum train speed. Conditional logistic regression was used to compare mortality locations with paired random locations within specified distances. Model selection was based on the Bayesian Information Criterion (BIC). A separate logistic regression analysis investigated the influence of season, precipitation, and time of day on mortality risk for each guild. The analyses were conducted using R software and the survival package.

The study revealed 59 bear, 27 other carnivore, and 560 ungulate mortalities. Maximum train speed was the strongest predictor of mortality locations across all guilds and spatial scales, especially at the largest spatial scale. Proximity to and amount of water were also strong predictors, suggesting that water bodies constrain animal movement and increase collision risk. Track curvature also positively correlated with mortality for bears and ungulates at all scales. Other variables showed less consistent effects across guilds and scales: ungulates had higher mortality near sidings, whereas carnivores had higher mortality further from sidings; shrub cover increased mortality for ungulates and bears; topographical complexity showed mixed effects; and proximity to roads increased mortality for bears. Seasonally, ungulate mortality peaked in winter, while bear mortality was slightly higher in late spring. More mortalities occurred during the day than at night. Logistic regression models had moderate to poor accuracy in differentiating mortality sites from random locations, with AUC values of 0.735 (ungulates), 0.634 (bears), and 0.683 (other carnivores).

The study's findings strongly suggest that train speed and track curvature significantly impact wildlife mortality by reducing animals' ability to detect approaching trains. The consistent effect of high train speed across guilds supports previous research. The influence of water proximity and amount reflects limitations on animal escape routes. Differences in responses to other variables across guilds indicate the complexity of factors influencing mortality. For example, the attraction of ungulates to sidings, potentially due to spilled grain, contrasts with carnivores' avoidance of human-influenced areas. The seasonal variation in mortality highlights the importance of time-specific mitigation efforts. The moderate model accuracy indicates the need for additional variables to improve predictions. Integrating animal movement data with collision data could offer better insights into risk areas.

This study provides valuable insights for developing cost-effective mitigation strategies for railway-wildlife collisions. Prioritizing mitigation in spatial and temporal hotspots could be more efficient than extensive, widespread mitigation. Reducing train speed, especially in areas with high curvature and water proximity, is a key recommendation. Future research should explore additional variables, refine spatial scale analysis, and incorporate animal movement data to improve risk prediction and guide targeted interventions. Understanding indirect effects on habitat loss and fragmentation is also important.

The study's limitations include unequal sample sizes among guilds, which affected model robustness and power, particularly for less abundant species. Using mortality data alone might not accurately reflect future risk, as past collisions could lead to avoidance behavior. The study did not include all potentially relevant variables (e.g., tributaries, human use). The analysis used a guild-based approach, overlooking potential species interactions. The spatial scales considered for predictor variables might not encompass the full range of biological relevance.