Simulating wildfire propagation in Central Europe - Evaluation of input data variation and parameter adjustment for the PhyFire model

Wildfires pose an increasing risk in many parts of the world. In Central Europe and especially in the Alpine region, uncontrolled wildfires are becoming a relevant issue due to climatic and socio-economic changes. Fire modelling in these areas requires tools that can predict fire behavior and spread under the specific environmental conditions. This study aims to apply the PhyFire wildfire simulation model-originally developed for Mediterranean environments-to Central European conditions, refine its parameters and assess its applicability for a case study in Austria. PhyFire, a 2D model with partial 3D capabilities including wind modeling, is based on the principles of mass and energy conservation. A variety of input datasets, fuel-type maps (EFFIS, firEUrisk, and a manually derived high‑resolution map), meteorological and adjusted fuel‑type‑dependent parameters (including newly specified values for forest fuel-type classes) were tested. Additionally, the parameter pfRad was defined for each fuel type to compute flame radiation and energy flux to surrounding fuels. Therewith ignition and spread representation under Central European conditions were improved. Model outputs were compared against observed temporal and spatial spread patterns based on operational records. The simulation accuracy was evaluated by calculating over- and underestimation and computing the Sørensen-Dice coefficient for each input-data combination. The model performed well overall but was overly sensitive to slope and wind direction. The study confirms PhyFire's potential for Central European wildfire modeling but highlights the need for a careful selection of the input data and the required parameter adjustments for increasing predictive reliability. For a broader application more recent high-resolution fuel data are essential to enhance predictive power and model accuracy.