The spatial pattern of forest fire locations is important in the study of the dynamics of fire disturbance. In this article we used a spatial point process modeling approach to quantitatively study the effects of land cover, topography, roads, municipalities, ownership, and population density on fire occurrence reported between 1970 and 2002 in the Missouri Ozark Highland forests, where more than 90% of fires are human-caused. We used the AIC (Akaike information criterion) method to select an appropriate inhomogeneous Poisson process model to best fit to the data. The fitted model was diagnosed using residual analysis as well. Our results showed that fire locations were spatially clustered, and high fire occurrence probability was found in areas that (1) were public land, (2) within 6 km to 17 km of municipalities, and (3) <500 m from roads where forests are accessible to humans. In addition, fire occurrence probability was higher in pine-oak forests on moderate (<25 degree) slopes and xeric aspects and at higher (>270 m) elevations, reflecting the effects of natural factors on fire occurrence. The results serve as a provisional hypothesis for expanding fire risk estimation to surrounding areas. The spatial scale of analysis (approximately 1 ha) provides new information to guide planning and risk reduction efforts.