Pattern recognition and modelling of virulent wildfires in Spain

Background Driven by global warming and land use, fire dynamics are changing worldwide, increasing fire activity and its impacts on ecosystems, livelihoods and human settlements. In southern European countries, large forest fires are increasing annually and a high spatio-temporal concentration has been shown to overwhelm fire suppression systems. Aims Propose sound statistical criteria for grouping fire ignitions based on auxiliary information. Methods Our methodology combines pattern recognition and complex statistical modelling that, by capturing underlying dependencies and adapting to the nature of the data, provides a reliable statistical basis for formulating firefighting strategies. Key results Weekly and provincial data from 2007 to 2015 are used to illustrate the performance of the procedures. Conclusions Research indicates that weather conditions, the simultaneous occurrence of events and the proximity of urban areas contribute to making the Spanish landscape highly susceptible to wildfires in summer. Implications Risk mapping is key to fire management and efficient resource allocation.