Road network connectivity explains wildfire probability in Southern Italy

Abstract

Wildfire ignition and spread are strongly influenced by anthropogenic factors, particularly road networks. This study investigated the relationship between road network connectivity and wildfire probability in the Apulia region of southern Italy. A wildfire probability map was developed using an Artificial Neural Network (ANN) model, incorporating wildfire occurrence data from 2000 to 2021, along with landscape, climatic, and topographic variables. The resulting probability map was classified into three categories: High Wildfire Probability (HWP), Medium Wildfire Probability (MWP), and Low Wildfire Probability (LWP). Cochran's formula was used to calculate the required sample size, and Neyman allocation was applied to ensure optimal stratified sampling across probability categories. Connectivity indices were then estimated for each sampled area and statistically analyzed using the Kruskal-Wallis test. Results indicated that the majority of the study area fell within the LWP category (65.69 %), followed by MWP (21.48 %) and HWP (12.83 %). Statistically significant differences were observed across all categories for key connectivity indices. HWP areas exhibited higher road density, node counts, and number of links, reflecting more complex and interconnected road networks. In contrast, LWP areas showed lower connectivity and simpler network structures. The Alpha, Gamma, and Eta indices in particular displayed marked variation, with HWP zones exhibiting significantly higher values. These findings highlight the dual role of road networks: while they are vital for wildfire suppression and evacuation, their increased presence and complexity may elevate ignition risk due to greater human activity in densely connected areas.