We consider the problem of prepositioning a set of firefighting resources for wildfire suppression. As wildfires are highly affected by uncertainty we devise a two-stage model where the prepositioning of resources are first-stage decisions and the movement of these resources after the fire ignitions are known are the second-stage decisions. We use the minimum travel time principle and mixed integer programming to model the fire spread in the landscape and decisions related to the prepositioning of resources and their movement to attack positions. To model uncertainty we use robust optimization based on a discrete set of scenarios which represent ignition locations, wind speed and directions. As the size of the model grows considerably with the number of scenarios, a row-and-column decomposition algorithm is proposed. Computational experiments based on an actual landscape are reported showing the efficiency of the decomposition algorithm.