A data-driven approach for wildfire risk modelling in Southern Europe

Wildfires are one of the most devastating natural disasters with wide impacts across economic sectors and society in Europe. Modeling wildfire risk remains a complex challenge, particularly as predictions of fire risk often rely on weather-based operational indices, such as the Fire Weather Index (FWI) which exclude key factors such as human activities and other ignition sources. In addition, it is unclear how wildfire risk will evolve in the future under different climate scenarios.

The study shows how data driven models can be used to combine human, topographic, land cover and weather data to quantify wildfire risk in Southern Europe on a 10 x 10 km grid at a daily resolution. Four classification machine learning models are trained on a historical fire record from 2008 to 2023, obtained from the European Forest Fire Information System (EFFIS). The best performing model is a Random Forest (RF) model with an AUC of 0.95 and F1 score of 0.89.  

The RF model is first validated by comparing the model output run using weather variables from the ERA5-land reanalysis to the historical fire record from EFFIS. The results show that the RF model can successfully identify high-risk fire regions both seasonally and daily and provides a more accurate representation of fire risk compared to the FWI. The model is subsequently run using future climate data (2081 – 2100) from the ClimEx2 regional climate model under the Shared Socioeconomic Pathways (SSP) 1 and SSP3 scenarios, showcasing its potential to evaluate fire risk under evolving climate conditions.