Bayesian Classification of a Meteorological Risk Index for Forest Fires: DSR

Abstract

The Daily Severity Rating (DSR), a meteorological rating for assessing the risk of fires by using the forest fire index, is calculated on a daily basis in two sequential steps: first it is forecast for the day of interest, within a day of advance, and for a limited set of control points, using the Canadian Forest Fire Weather Index System (CFFWIS). Afterwards the forecast values are interpolated (ordinary kriging) into a regular grid of points covering Portugal. In this chapter we propose a model for fire risk assessment in Portugal that is based on the conditional probability of fire, I(x), as given by the class of DSR predicted for that specific period of time - p(I(x)∣R(x)). The evaluation of this a posteriori p(I(x)∣R(x)) is based on the update for marginal local probability of fire in each chosen county. Mapping of the risk of fire is obtained for the entire country by kriging. As the definition of DSR classes should be dependent on specific conditions for each county, in the last part of this study the thresholds of DSR classes that can lead to a high risk of fires are calculated for each county. This regional DSR threshold is an indirect measure of other factors that are the cause of fires. [ABSTRACT FROM AUTHOR]