The Human Factor in Wildfires: Towards a Long-Term Fire Risk Estimation Model for Central Banking

Climate change elevated the relevance of fire risk projections for the future, which are essential for estimating the potential changes to expected economic losses, including the changing risk exposure of banking assets and insurance policies. And while the use of empirical models is required to lend credibility to this undertaking, human-induced fires, particularly stubble (or crop residue) fires, distort the observed relationship between weather and fire risk. This paper presents an empirical-statistical model to estimate future fire risk using MODIS satellite data for fire occurrences, the Keetch-Byram Drought Index (KBDI) as weather-based input, and CORINE Land Cover information. The model is developed for Hungary at a 5 km resolution over the period 2001-2024. In addition, we account for the Human Factor (HF) in fire risk by introducing a Google Trends-based proxy reflecting legislation and public awareness of fire. Alternative definitions of this variable are also tested. Incorporating the HF mitigates omitted variable bias, improves the explanation of past fire occurrences, and reverses the overall KBDI-fire risk relationship to the expected positive sign. Under the pessimistic RCP8.5 scenario, current fire occurrence risk is projected to double by the end of the century, with a potential of reaching a fivefold increase in extreme years, assuming the present HF in fire prevention and land cover flammability remain constant.