This paper integrates a spatial fire-behavior model and a stochastic dynamic-optimization model to determine the optimal spatial pattern of fuel management and timber harvest. Each year's fire season causes the loss of forest values and lives in the western United States. We use a multi-plot analysis and incorporate uncertainty about fire ignition locations and weather conditions to inform policy by examining the role of spatial endogenous risk-where management actions on one stand affect fire risk in that and adjacent stands. The results support two current strategies, but question two other strategies, for managing forests with fire risk.