Non-separable spatio-temporal Poisson point process models for fire occurrences

Abstract Our study addresses the analysis of environmental concerns through point process theory. Among those, Sicily faced an escalating issue of uncontrolled fires in recent years, necessitating a thorough investigation into their spatio-temporal dynamics. Each fire is treated as a unique point in both space and time, allowing us to assess the influence of environmental and anthropogenic factors. A non-separable spatio-temporal Poisson model is applied to investigate the influence of land use types on fire distribution, controlling for other environmental covariates. The results highlight the significant effect of human activities, altitude, and slope on spatio-temporal fire occurrences, also confirming their dependence on various environmental variables, including the maximum daily temperature, wind speed, surface pressure, and total precipitation. As a model with constant parameters in space and time may be too restrictive, a local version of the proposed model is also fitted. This allows us to obtain better performance and more valuable insight into the estimated effects of the different environmental covariates on the occurrence of fires, which we find to vary both in time and space. This research work’s relevance lies in the analysis of an important environmental problem through complex point process models, yet easily interpretable, given their resemblance to regression-type models. We also provide reference to newly available open-source software for estimating such models. Finally, we contribute to the framework of spatio-temporal point process modelling by integrating data with different spatio-temporal resolutions from very diverse sources.