I analyzed moose (Alces alces)-vehicle collisions (MVCs) in Maine from 1992-2005 using spatial statistics and Geographic Information Systems (GIS). My objectives were to describe temporal and spatial distributions of MVCs and to develop predictive models based on landscape characteristics. MVCs were most frequent from June-October and clustered spatially at local and regional scales. Logistic regression modeling showed that the predicted probability of MVC increased by 57% for each 500-vehicle/day increase in traffic volume, by 35% for each 8-km/hour increase in speed limit, and by 36% for each 5% increase in cutover forest cover. Land cover covariates were most explanatory at spatial extents (2.5-5 km) that approximated the spatial requirements of moose. Where the reduction of timber harvesting, conifer cover, and wetlands over large areas is not feasible, lowering driving speeds during high-risk times of day and year and in high risk areas may be most effective for reducing MVCs.