Spatial point process modeling applied to the assessment of risk factors associated with forest wildfires incidence in Castellón, Spain

During the last decades the Mediterranean zone in Europe has experienced an increment in the incidence of forest wildfires. This increase is partly explained by higher mean temperature and lower relative humidity, while socioeconomic change has lead to the abandonment of farms, resulting in an increase in an unusual accumulation of forest fuels, increasing the risk of wildfires. Mapping wildfire risk is highly important because wildfires are known to potentially lead to landscape changes and to modify fire regime by inducing potential changes in vegetation composition. Also, they pose a hazard to human property and life. Maps of wildfire risk based on statistical models provide a measure of uncertainty for the inferences derived from such risk maps, leaving a quantitative error margin for managers and decision takers. Further, some of the model parameters often have a physical or a biological interpretation which can give ecologists and forest engineers answers about scientific questions of interest. In this paper, we analyze the incidence of wildfires in the province of Castellon in Spain in order to identify risk factors associated with wildfire incidences during the years 2001–2006. We used the discrete nature of wildfire events to build such models using point process theory and methods and included information about elevation, slope, aspect, land use and distance to nearest road as covariates in our modeling process. Our results show that wildfire risk in Castellon is associated with all the covariates considered and that three land-use categories have the highest risk of wildfire incidence. Also, wildfire incidences are not independent and some degree of interaction exists, which indicates that the commonly used Poisson point process models are not applicable in this case, but instead area-interaction models should be considered.