Wildfire Suppression: Complexity, Models, and Instances

Comparative Analysis of Wildfire Spread Models Under Differing Environmental Conditions in Central Europe

Wildfires are an increasing threat in Central Europe and pose challenges for protective forests and areas at the wildland–urban interface (WUI). Understanding, describing and predicting fire behaviour is therefore becoming more relevant for fire management. This work aims to reconstruct the fire spread behaviour of past fire events occurred under differing environmental conditions with selected fire spread models. The three fire spread models Farsite, SimtableTM and Prometheus were selected according to a list of predefined properties they were expected to fulfil. Subsequently, they were tested under different environmental conditions and evaluated against documented perimeter of past fire events. The focus of the analysis was on the spatial perimeter to quantify metrices such as over- and underestimated areas in percent, Sørensen–Dice coefficient and the Jaccard similarity coefficient. Farsite showed the best overall results in both regions. Simtable performed well in steep and complex terrain but produced underestimations in flat terrain. Prometheus lagged, likely due to inadequate parametrization of fuel data, which is a key input parameter in fire spread modelling. As Farsite is readily accessible, it has the greatest potential for further application and more in-depth research. For higher reliability, additional empirical data on fire behaviour are needed to develop custom fuel models or refine current adjustments used to simulate fire spread.

Scalability of Parallel Shared-Memory Isogeometric Alternating-Direction Solver for Wildfires Simulations

Modelling Wildfire Burn Severity in Canadian Megafires

Abstract. Wildfire activity in Canada has increased significantly in recent decades, shifting to larger, more frequent fires and the emergence of megafires (>10,000 ha) across various ecozones. These events typically exhibit complex spatial patterns of burn severity, including larger and more homogeneous patches of high severity. The burn severity patterns and their drivers in megafires remain unclear, in particular, across diverse ecozones. Remote sensing indices such as the Relativized Burn Ratio (RBR) provide an effective means of quantifying burn severity at large spatial scales. This study uses RBR to evaluate nine megafires (each >50,000 ha) representing the 95th percentile and above of fire size within varying ecozones between 2016 and 2022. These fires were used to develop two random forest models: one predicting RBR and another predicting the within-fire z-score of RBR. Within-fire standardization of RBR was conducted to see whether it alters the relative importance of environmental drivers. In the RBR model (OOB R² = 0.75), regional variables such as ecozone and fire ID, along with drought code, were dominant predictors. In contrast, the z-score model (OOB R² = 0.68) emphasized fuel characteristics, including biomass and canopy closure, with additional contributions from elevation and drought-related variables. These results suggest that broad regional and fire-regime controls exert a stronger influence on burn severity than local fuel conditions at the megafire scale. Standardizing burn severity within fires reduces this regional signal but does not improve predictive performance, highlighting the importance of accounting for regional variability in large-fire dynamics.

A reduced PDE model for wildfire spread with probabilistic interpretation

Recently, a stochastic forest fire model was proposed in which the forest state is represented by sub-probability densities of firing, green, and burnt trees, whose evolution is governed by a master equation formulated as a system of integro-differential equations. Within this framework, the present paper restricts attention to the case of short-range fire propagation and derives a reduced model whose dynamics is described by a reaction–diffusion–convection equation for the firing-tree sub-probability density, coupled with an ordinary differential equation for the green-tree sub-probability density. The reduced system is interpreted as a short-range approximation of the original integro-differential model, while retaining the same probabilistic meaning of the evolving densities. Numerical simulations show that it reproduces the main qualitative features of the full model in the parameter regimes considered, while quantitative comparisons based on Hellinger and total variation distances indicate that the two dynamics remain close over simulation times corresponding to real times of the order of a few tens of hours. At the same time, the reduced formulation yields a substantial reduction in computational cost, making it particularly suitable for repeated simulations and for the early stages of wildfire propagation, where stochastic effects are relevant and fire spotting can still be neglected. When used together with the original integro-differential model, the reduced model contributes to a stochastic modelling framework that complements approaches based on cellular automata. Finally, some mathematical properties of the reduced model, including nonnegativity of solutions and long-time asymptotic behaviour, are also discussed.

Explaining building damage from wildfires in California

A physically interpretable, data-driven framework was developed to elucidate causal interactions, model, and predict wildfire-induced building damage across California. More than 100,000 damage inspection records (2013 to 2024) were used to model building damage from static environmental variables (topography, vegetation, and human footprint), dynamic weather inputs, and a proposed Composite Building Flammability Rating (CBFR). Three model configurations were tested: (i) a comprehensive model integrating all variables, (ii) an enviro-weather hybrid excluding CBFR, and (iii) an environmental exposure model excluding both weather and CBFR. A strict 200-meter spatial dead-zone constraint was applied to eliminate local autocorrelation, and the comprehensive model achieved 88% (±0.4%) accuracy, which dropped to 82.9% (±0.6%) without CBFR and to 74.5% (±0.5%) without both weather and CBFR. Spatial grid-based cross-validation demonstrated a diverse accuracy of 68.0 (±17%), 66.0 (±16%), and 62.0 (±13%), respectively. Building flammability, dew point temperature, and near-surface wind speed were identified as the most important predictors of damage. Vapor pressure deficit had the strongest causal effect on damage probability, though spatial variability was observed in the causal effects of climate and geographic variables. A 100-meter-resolution Wildfire Building Damage Risk Index was also developed to highlight high-risk damage zones. Findings emphasize that wildfire impacts in the wildland-urban interface result from a confluence of structural vulnerability, atmospheric dryness, and fuel exposure, offering scalable tools for risk forecasting, defensible space planning, and climate-resilient infrastructure development.

Hybrid Two-Dimensional Wildfire Growth Model for High Resolution Environments