Modelling Fire Regimes in Social-Ecological Systems through Geospatial Analyses | Modélisation des régimes d'incendie dans les systèmes socio-écologiques au moyen d'analyses géospatiales

English. Fires in the Anthropocene are a complex process driven by a multitude of interacting factors. Understanding how these factors affect fire and its features, such as frequency, seasonality, and severity – collectively referred to the concept of fire regime - is essential to develop landscape strategies aimed at reducing hazards and restoring the ecological role of fire. In this thesis, we investigated large scale fire behaviour within Social-Ecological Systems, a framework that extends the concept of ecosystem by integrating human and ecological factors. Specifically, we focused on clarifying the relationships between complex land-use dynamics and land governance models, and fire regimes, while accounting for climate, landscape, and social variables. Additionally, we aimed to assess the impacts of fires driven by such processes on key dimensions of interest, such as forest vegetation. Our methodology primarily relied on geospatial analysis and statistical regression modelling, using both traditional and machine learning approaches, to produce quantifiable and spatially explicit results. Our three case studies were selected to represent contrasting contexts and varying geographical scales. In the first research chapter, we analysed the effects of devegetation (i.e., native vegetation loss due to anthropogenic land conversion) on fire occurrence, size distribution, and seasonality across five states in the Brazilian Cerrado. Additionally, we quantified the loss of native vegetation caused by fires linked to devegetation and identified land tenure types that effectively limit these impacts. Our findings indicate that approximately 12% of all the burned area in the case study is attributable to the devegetation process, which is responsible for burning one fourth of the native vegetation surface in the study area, with altered seasonality and size. Protected Areas and Indigenous Lands emerged as highly effective tools in limiting these impacts. In the second research chapter, we evaluated the effects of land abandonment on fire occurrence within the forest landscapes of Italy's Piedmont region. We also assessed the potential of protection and rewilding policies to mitigate fire-related impacts. Our results show that land abandonment increases fire occurrence, regardless of forest category or landscape features, especially within protected areas. In the third research chapter, we examined fire occurrence and impacts distribution across Italy. Our results showed that, net of various climatic, landscape, and socioeconomic drivers, fire impacts are significantly prevented and mitigated in areas that implement active governance through sustainable agropastoral and forest management systems, protected areas, and biodiversity conservation and rural development strategies. Overall, this thesis shows that fire regimes are heavily influenced by complex processes shaping Social-Ecological Systems, providing quantitative insights. It also identifies the most suitable strategies for addressing altered fire regimes and their associated impacts, according to specific context.

French. Les incendies dans l'Anthropocène sont un processus complexe régi par une multitude de facteurs en interaction. Il est essentiel de comprendre comment ces facteurs affectent les incendies et leurs caractéristiques, telles que la fréquence, la saisonnalité et la sévérité - collectivement désignées par le concept de régime des incendies - pour développer des stratégies paysagères visant à réduire les risques et à restaurer le rôle écologique des incendies. Dans cette thèse, nous avons étudié le comportement des incendies à grande échelle au sein des systèmes socio-écologiques, un cadre qui étend le concept d'écosystème en intégrant les facteurs humains et écologiques. Plus précisément, nous nous sommes attachés à clarifier les relations entre les dynamiques complexes d'utilisation des terres, les modèles de gouvernance des terres et les régimes d'incendie, tout en tenant compte des variables climatiques, paysagères et sociales. En outre, nous avons cherché à évaluer l'impact des incendies provoqués par ces processus sur des dimensions clés, telles que la végétation forestière. Notre méthodologie repose principalement sur l'analyse géospatiale et la modélisation de la régression statistique, en utilisant des approches traditionnelles et d'apprentissage automatique, afin de produire des résultats quantifiables et spatialement explicites.