Wildfire Propagation Modeling using Satellite-Derived Parameters and Generalized Elliptical Frames
Hengameh R. Dehkordi
TL;DR
The paper addresses wildfire spread prediction in data-limited settings by integrating satellite-derived thermal, atmospheric, and vegetation information into a geometric front propagation framework. It introduces two complementary strategies: (i) a Huygens-based expansion using an anisotropic directional rate of spread $r(\theta,x,y,t)$ and (ii) a generalized elliptical-frame method that encloses burned regions, both grounded in a unified mathematical description of front evolution. A data-driven correction links thermal-derived rates $R_H^T$, $R_B^T$ to regional rates $R_H$, $R_B$ via wind and fuel moisture, employing an exponential wind-moisture coupling $R = \mathbf{A} U^{\alpha} \exp(-\beta M)$ and seasonal LFMC adjustments. Case studies on the Eaton Fire and data-limited scenarios demonstrate qualitative agreement with observed spread and emphasize the framework's robustness, scalability, and applicability to regional and global wildfire monitoring.
Abstract
Wildfires pose significant threats to ecosystems and communities, yet accurately modeling fire spread remains challenging, particularly in regions where environmental and fuel data are scarce or unavailable. This study introduces an innovative conceptual and methodological framework for simulating wildfire propagation and estimating the rate of spread using a hybrid geometric and data-driven approach that relies exclusively on multi-source satellite observations. The framework integrates thermal fire-front detections, atmospheric conditions, and vegetation indices using two complementary geometric modeling strategies. The first strategy applies the Huygens principle, where generalized elliptical frames are expanded locally at every point along the fire perimeter, and their combined envelope forms the evolving wavefront. This method is best suited for situations in which environmental variables are available and can be incorporated to refine the anisotropic spread function. The second strategy relies solely on the generalized elliptical frames themselves; for each time step, an elliptical frame is constructed from the inferred head and back rates of spread and wind, and the burned area is obtained by enclosing the region determined by these curves. Together, these two methods provide a flexible toolkit that adapts to both data-rich and data-limited conditions while retaining a unified geometric interpretation of wildfire spread. To demonstrate the applicability of the method, we present a case study based on the Eaton Fire, January 2025, using publicly available multi-day satellite imagery. Despite the absence of complete operational datasets for that event, the model driven only by satellite-derived parameters reproduces key qualitative features of the observed propagation pattern, underscoring the flexibility and robustness of the proposed approach in data-limited contexts.