The Terai ecoregion of the Himalayan foothills is among the most fire-affected ecosystems in the Indian sub-continent. Although most of the Terai has already been lost to agriculture and urbanization, the few remaining native habitats are strictly protected due to their high biodiversity and ecological importance. The use of fires to maintain vegetation and wildlife habitat in these protected areas is an integral part of forest management. Although fires are initiated by forest managers or local people, their eventual spread and behaviour are not controlled. We hypothesize that distributions of fires are determined by several direct and indirect drivers like fuel load, fuel moisture content, presence of natural or artificial fire breaks, and climatic attributes of precipitation and temperature. Using the moderate-resolution (375m) satellite sensor-based data we studied the environmental influence on the spatial-temporal patterns of fire events over 18 years (2000-2018) in a 519 km2 protected area in northeastern India. The park has a mosaic of vegetation formations – including dry and swampy alluvial grassland, early successional woodland, and moist tropical forest. Despite high rainfall, there is an intense dry season that renders the herbaceous vegetation susceptible to fires. Using spatial and spatial-temporal Poisson regression models in a rigorous conditional autoregressive Bayesian framework, we found that net primary productivity (a proxy of vegetation type and fuel load), distance to roads (a measure of human influence), elevation (through its influence on floods and vegetation type), and river area extent (by determining the area under vegetation) had a significant influence on the spatial distributions of fires. The climatic signal on interannual variation in fires was weak, but dry season rainfall reduced fire incidence. The disproportionate distribution of fires adjacent to roads compared to the interior, and the repeated burning of some grassland patches need to be addressed in fire management.