Prediction of forest fires with Poisson models

This paper presents a general statistical methodology for the prediction of forest fires in the context of Poisson models. It also gives quantitative and qualitative tools for the assessment of different models, as well as some decision theoretical and cost considerations. Case studies from France, Italy, Portugal, and Switzerland illustrate the theory. It turns out that the Poisson models incorporating a fire danger index with other highly important explanatory variables are always superior to empirical use of a single fire hazard index.