Forest fires are not spatially uniform events. They result in a complicated mosaic of burned and unburned vegetation. To manage fuel loads and the associated fire hazard it is essential to improve our understanding of the spatial patterns of the potential effects of future fires. However, the extremely complex interplay of biological, meteorological, topographic, and social factors make accurate fire modeling and prediction very difficult. Traditional fire modeling efforts quantify and map the individual variables that might effect fire spread and behavior, and recombine these variables in complex mathematical propagation models. The study presented here, instead, models fire hazard through the direct interrogation of the relationship between the spatial patterns of previous fires and the remotely sensed spectral and terrain patterns observable in the pre-fire landscape. The recent Cerro Grande Fire in Northern New Mexico was started as a prescribed fire, but eventually burned into the town of Los Alamos. Such projects are undertaken because of the recognized risks associated with continued accumulation of fuels. This fire offers a chilling indication of the complex problems faced by a growing number of other western communities with urban-wildland interface areas and similar fuel accumulation. This paper uses the Cerro Grande to explore the effectiveness of a technique for modeling the internal severity patterns of future wildfires directly from remotely sensed data of nearby historic fires. Such a technique offers a valuable tool for communities wishing to understand the likely outcome of their next big fire. The Cerro Grande was similar to another recent fire in the area. In 1996 the Dome Fire burned 16,000 acres directly south of the Cerro Grande. TM data from before and after the Dome Fire are used to map the internal heterogeneity of the Dome*s severity. These maps are used to identify areas that underwent complete canopy mortality, and areas that underwent no significant canopy mortality. Once identified, these areas are digitized to extract pre-fire spectral and terrain patterns associated with these observed fire effects. Once extracted, these signatures are used as the basis of a classification of the entire Jemez Mountains, including the area of the Cerro Grande Fire. If this technique continues to perform well in other areas it will offer a fast and efficient tool for mapping fire hazards across large areas.