Publication Date
Summer 2026
Degree Type
Thesis
Degree Name
Master of Science (MS)
Department
Meteorology and Climate Science
Advisor
Adam Kochanski; Craig Clements; Kyle Hilburn
Abstract
Wildfire simulations rely heavily on near-surface atmospheric conditions and accurate fire-state initialization, and uncertainty in either can result in significant forecast errors. Using observational data from the California Canyon Fire Experiment (CCFE), this study investigates how assimilating atmospheric and fire behavior observations reduces these uncertainties and improves wildfire simulations using a high-resolution WRF–SFIRE model. Atmospheric uncertainty is addressed using a cycling 3DVAR framework within WRFDA, which assimilates SoDAR profiles at 5-minute intervals to continuously improve representation of the evolving canyon-scale flow. Fire-state uncertainty is addressed through a perimeter-constrained replay approach that prescribes fire progression until the first observed perimeter, effectively acting as fire-state assimilation. Simulations employing atmospheric-only, fire-only, and combined assimilation are compared against simulations initialized from estimated ignition lines. Fire spread is evaluated using the Jaccard index, while near-surface winds are evaluated using statistical metrics. Results show that atmospheric and fire-state uncertainties both strongly influence wildfire simulation accuracy. Errors in early fire progression are cumulative and can persist throughout the simulation, while wind errors affect the direction, structure, and rate of fire spread. Fire-state assimilation reduces errors associated with early fire growth, and atmospheric assimilation improves representation of canyon-scale flow, but neither approach alone fully resolves forecast errors. Combined assimilation produces the strongest agreement with observed fire behavior, demonstrating the benefits of joint assimilation.
Recommended Citation
Moser, Daniel, "Improving Fire Simulations in Complex Terrain by Assimilating Fire and Weather Data" (2026). Master's Theses. 5826.
DOI: https://doi.org/10.31979/etd.wb7h-mh8a
https://scholarworks.sjsu.edu/etd_theses/5826
