Temporal patterns of wildfire probability and intensity in Northern Sardinia, Italy

The prediction of burn probability and intensity by the use of wildfire simulators is a valuable method to capture the large variability of fire spread and behavior produced by weather and fuel conditions, especially when extreme weather conditions, and their effects on fuel characteristics, leads to large fires. The planning of fireprone landscapes is the main area when the evaluation of the burn probability and intensity can provide additional information for the decision makers. In addition, policy makers and management agencies could use wildfire simulators at landscape level to monitor and evaluate the impact of prevention plans. In this work we estimated the past variations of burn probability and intensity using the scenarios of the key wildfire factors (wind field, fuel moisture, fuel characteristics, and ignition patterns) historically observed during severe and extreme conditions. Simulations of fire spread were performed using the FlamMap fire mapping and analysis system, based on the fire spread model of Rothermel and the fire growth method of minimum travel time. The set of simulations were conducted using the landscape characteristics existing in the year 2000, and were replicated for two time steps in the past: 1954 and 1977. The wind data required by the simulator were developed by the analysis of the weather records collected near the experimental area, and using the wind data to simulate the spatial distribution of wind by a mass-consistent model. The sets of ignition points were generated by sampling algorithms based on random sampling process, and alternatively considering the historical probability of fire to stratify the random sample. The fuel moisture condition typical of very dry and extreme days were estimated by the analysis of weather data and considering local representative data on fuel moisture conditions. Both graphical and analytical methods were used to highlight the main variations of burn probability and severity, and to find the relationships with the variations of the main landscape characteristics. The experimental data provided by the simulator shows a clear increase of both burn probability magnitude and spatial extension induced by the extreme conditions, in particular by the extreme fuel moisture scenarios. This trend is more evident for the data simulated using the historical ignition scenario. The experimental data showed also that the land use change observed throughout the study period (from 1954 to 2000) determined a reduction of burn probability in the last time step (2000), while limited differences were observed from 1954 to 1977. The recent variation of burn probability was more evident when the historical ignition scenario was used to perform the simulations, and can be attributed to the modifications of the landscape generated by the increase of the human activities in the coastal areas, where regional Forest Service statistics show the highest number of ignitions. In addition, in the last time step we observed a general decrease of the areas characterized by herbaceous vegetation, more prone to large fire sizes and able to produce large values of burn probability.