Modelling wildfires occurrence in Italy using Generalized Additive Models

The primary objective of this thesis is to use Generalized Additive Models (GAMs) to compute wildfire occurrence probability in the Italian region Calabria, by knowing the past wildfires, occurred from 2000 to 2024. For each year, the area of interest is divided in a 250 x 250 meters grid, where each pixel is addressed with a set of morphologic and climatic variables. The first ones are derived from the Digital Elevation Model (DEM), and they are elevation, slope, aspect, profile curvature and tangential curvature. On the other hand, the climatic variables are temperature, rainfall, wind speed and near-surface specific humidity, and they are collected from NEX-GDDP-CMIP6 dataset. The model is trained over the period from 2000 to 2014, during which observed climatic variables are available. In this way, the model learns the conditions under which wildfires occur. Its performance is then evaluated over the period 2015–2024 using predicted climatic variables. The structure of this thesis comprehends an introduction, presented in chapter 1, followed by chapter 2, which describes how data have been collected and chapter 3 focused on the methodology. Chapter 4 shows the results obtained from the model, together with some comments that could be helpful to lead the discussion. Finally Chapter 5 is the conclusion.