Implementing a probabilistic fire modeling system at the pan-European level

Abstract

This research shows the potential use of cross-boundary fire modeling systems at the pan-European level. Despite the growing interest in building fire-resilient cultural landscapes, European Union (EU) level efforts have been reactive and focused on early detection, fire propagation monitoring, and perimeter mapping rather than predicting where the disaster can potentially occur to develop a comprehensive wildfire management strategy. We propose a modeling system that integrates wildfire occurrence models and observed fire-weather scenarios with a fire spread model to generate the probabilistic risk components, i.e., wildfire likelihood and hazard estimates. We selected four NUTS-2 level administrative division pilot sites from different fire-prone countries to implement the modeling system. The European-level pyromes were first delineated based on ecoregions and historical wildfire activity. We then generated human and lightning wildfire occurrence models to display the ignition points. Remote sensing products were used to derive fire spread modeling spatial input data such as surface fuels and canopy metrics. Global atmospheric products were used to calculate the fuel moisture content with physical models and determine the most frequent wind scenarios for each pyrome. We then used the Minimum Travel Time algorithm to model the fire footprints that correspond to 10,000 years or synthetic iterations. This modeling approach accounted for the spatial variation of ignition locations and the changing weather conditions across the different pyromes within each pilot site. The modeling results include the annual burn probability and fire intensity rasters, and fire perimeter vector outputs. Modeled burn patterns showed a close agreement with observed fire size distributions. We compared this modeling system with previous works to explain why stochastic fire modeling is essential to assess wildfire exposure of natural values at risk and human communities. Our results may help predict future catastrophic fires and provide quantitative estimates to identify high-priority management areas within vast regions. The probabilistic predictions generated in this work represent the foundation for developing long-term adaptation strategies to better coexist with fire. This work is also a demonstration of how this modeling system is replicable in any European country.