1Understanding wildlife-vehicle collision risk is critical to mitigating its negative impacts on 2 wildlife conservation, human health and economy. Research often focuses on collisions between 3 wildlife and road vehicles, but collision risk factors for other types of vehicles, less examined in 4 the literature, may be informative. 5We studied spatial and temporal variation in wildlife-train collision risk in the State of Victoria, 6Australia. We quantified train movements in space and time, and mapped species occurrence 7 likelihood, across the railway network. Using spatially-and temporally-resolved collision data, 8 we fitted a model to analyse collisions between trains and kangaroos; accounting for time of day, 9 train frequency and speed, and kangaroo occurrence. We then predicted collision rates on the 10 passenger railway network under three management scenarios relating to train speed and 11 occurrence of kangaroos near the railway lines. 12Temporal variation in animal activity was the strongest predictor of collision risk. Train speed 13 was the second most influential variable, followed by spatial variation in likelihood of species 14 occurrence. Reducing speeds in areas of high predicted species occurrence and during periods of 15 peak animal activity (early morning and evening for kangaroos) was predicted to reduce collision 16 risk the most. 17Our results suggest mechanisms that might improve existing wildlife-transport collision 18 analyses. The model can help managers decide where, when and how best to mitigate collisions 19 between animals and transport. It can also be used to predict high-risk locations or times for (a) 20 timetable/schedule changes (b) proposals for new routes or (c) disused routes considered for re-21 opening. 22 3 Keywords 23 crepuscular; railway; risk; species distribution model; temporal; WTC 24 9Kangaroos are widespread (Dawson, 2012) and abundant in many parts of Victoria, and Eastern 126 Grey Kangaroos, in particular, are known to occur throughout the entire extent of the regional 127 train network, yet comprehensive distribution records are lacking in many areas. To represent 128 risk of collision by exposure to threat, we required continuous distributional data across the 129 entire study area and used species distribution modelling to predict relative likelihood of Eastern 130 Grey Kangaroo occurrence. Habitat predictors -often explicitly included in wildlife-vehicle 131 collision models -were alternatively used in a model to determine the relative occurrence of the 132 species across the train network. We emulated methods by Elith et al. (2008) to model and 133