Investigating forest fire causes through an integrated Bayesian network and geographic information system approach

Forests are among the most critical natural resources worldwide and are essential for a sustainable ecosystem. In recent years, climate change and human activities have increased their impact on forests, and forest fires-especially in summer- have caused significant forest losses globally. This paper integrates Bayesian Network (BN) and Geographic Information Systems (GIS) to pinpoint the possible causes of forest fires and analyze their complex interactions. The study used data from 1465 fires between 2017 and 2022 in Mu & gbreve;la province in southwestern T & uuml;rkiye. Within the scope of the study, 11 variables were chosen, such as elevation, slope, aspect, wind speed, population density, and road network, to build a BN model that combines physical and human geographical features. These variables overlapped using the pairwise intersect tool from ArcGIS Pro during the BN model setup, and probability values were calculated. The overall probability of fire in the BN model was determined to be 0.81, with probabilities ranging from 0.81 to 0.56 in low-altitude, moderately sloped, and south-facing areas. Scenario analyses examined fire risk under different conditions, highlighting the most influential variables for fire prevention efforts. The study identified fire-prone areas as spatial data, revealing that densely forested coastal and certain inland regions are at higher risk, whereas bare high-altitude areas with steep slopes pose lower risk. The BN model can be further enhanced by incorporating additional variables, making it a valuable tool for future fire risk assessment and mitigation strategies research.