Human-caused Wildfire Ignition Assessment - A Comparison of Different Regions in Europe, Using Remote Sensing and Geodata

Kurzfassung

Wildfires are of increasing global concern due to their devastating effects and their role in climate change through positive feedback loops from greenhouse gas emissions. While between 95-97 % of wildfires in Europe are triggered by human-related factors, to date, cross-regional fire prediction research has mostly focused on weather data and fuel characteristics. The presented work addresses this gap by proposing a predictive spatial modelling approach for human-caused fire ignition risk and by analysing the contributing factors. By using openly available geospatial and remote sensing data, the built random forest models are transferable to any study area in Europe when trained with local data. The models were applied to and tested in four distinct study areas: northern Portugal, north-western Spain, the Athens metropolitan region (Greece), and Brandenburg, Germany. For each study area, eight models were trained and evaluated, incorporating different combinations of up to seven explanatory variables that try to depict human-caused fire ignition. These variables are mainly related to human presence and activities in wildland areas, primarily measured by the distance to different kinds of infrastructure, and in one model, also the distance to agriculture and forested areas. The results revealed substantial regional variations in performance, with exceptional performance in Brandenburg (F1-score: 0.97), high accuracy in Greece (F1: 0.86) and moderate performances in Spain and Portugal (F1s: 0.65 & 0.59). The predominant variables contributing to human-caused ignition risk in the Mediterranean regions are distance to railways and the wildland-urban interface. In Brandenburg, distance to footpaths was predicted to be the primary factor. Interestingly, military training Areas showed a strong spatial correlation with fire ignitions, although they were not included as a variable in the analysis. When used in conjunction with live fuel and weather maps, the results can provide policymakers and stakeholders with valuable tools for implementing targeted localized fire risk reduction measures and optimizing resource allocation for fire management. The transferability of the methodology and the identification of region-specific risk factors can help to develop locally tailored fire prevention strategies across Europe.