Using model-based geostatistics to predict lightning-caused wildfires

The probability of fire in a particular area depends on a range of environmental and geographic variables. Fire prevention planning can be assisted by the construction of models to identify the variables that have a significant influence on the occurrence of fires and by building maps showing the spatial probability distribution for fires occurring in specific geographic areas. We used generalized spatial linear models to predict spatially distributed probabilities for fire occurrence in locations where storms featuring lightning occurred, on the basis of a set of variables related to climatology, orography, vegetation and lightning characteristics, and to assess the relative importance of these variables. A comparison of this model with simple logistic regression models used by other researchers to resolve similar problems demonstrates the importance of bearing in mind spatial correlation between variables.