Fire Spread Modelling Viability using FARSITE in Montane Environments of Veneto, Italy

In the coming decades, the impact of forest fires in Alpine environments is likely to increase and may further compound with local natural disturbances due to the advent of a warming climate. The effects from these compounding disturbances may impede the use of local ecosystem services and damage community infrastructure in the Alpine areas of the Italian region of Veneto. The ability to accurately and rapidly develop models of wildfire perimeters is an important tool in avoiding the risk of and mitigating the damage from natural disturbances in the region. The viability of doing this is dependent on the quality of regionally available data like local fuel maps and weather data. This study utilizes the fire modelling program FARSITE to output 26 different fire perimeter simulations attempting to emulate 26 relatively large, observed fires that occurred in the region in the years 2015-2024. A custom fuel model developed for the province of Veneto is utilized alongside publicly available topographical and meteorological data as inputs to attempt to emulate the conditions of the observed fires on the day of their ignition. Measurable physical properties of the simulated wildfire perimeters, such as rate of spread and fire intensity, are used as metrics to attempt to understand the spatial behaviour of the fires and diagnose spatial inconsistencies when compared to observed perimeters. An alternative moisture scenario developed to emulate a drier fuel bed as a consequence of climate change is presented alongside this analysis to spatially demonstrate what an increase in fuel dryness could mean for future wildfires in the coming decades. The simulated perimeters are tested for their similarity to their real-life observed counterparts through creating confusion matrixes from their spatial overlaps and calculating Sørensen-Dice coefficients from these values as a comparative statistic. With a few exceptions, the majority of the simulated perimeters in the study had extremely low spatial similarities with the observed perimeters. It is likely that this spatial disconnect comes from a multitude of input variables. Weather data, specifically related to wind, differences between observed and actual fuels in the regional fuel map, and the discrepancies between observed and actual burn periods of selected observed fires are posited as likely influences. Improvements to quality of data to enhance the viability of spatial modeling of wildfire in the region are suggested based on international research and documented practice as the influence that wind and temporal data have on model accuracy is well demonstrated.