Neural-Parameterized Cellular Automata for Wildfire Spread
Abstract
Traditional wildfire models rely on rigid, low-dimensional parameters and static fuel maps, frequently underpredicting fire spread. To address this weakness, we introduce a hybrid deep-learning parameterized Probabilistic Cellular Automata (CA) framework implemented in JAX. Our approach employs a Multi-Scale Convolutional Neural Network to dynamically generate spatially varying parameters that govern fire-spread probability, wind alignment, and slope influence. This hybrid design captures complex, nonlinear environmental interactions while preserving the physical interpretability of the underlying three-state CA. The JAX implementation enables hardware acceleration and gradient-based parameter calibration. Evaluated on six large-scale wildfires in the western United States, the model maintains IoU > 0.6 over 72-hour forecast horizons after a 10-day data assimilation window during which the model is fitted incrementally to observed perimeters; the resulting forecast is a conditional projection of fire growth under the suppression regime already ncoded in those observations.
Versions
Related
BibTeX
@misc{zhenirovskyy2026neuralparameterized,
title = {{Neural-Parameterized Cellular Automata for Wildfire Spread}},
author = {Maksym Zhenirovskyy and Ion Matei and Rohit Vuppala and Takuya Kurihana and Hon Yung Wonga},
howpublished = {arXiv},
year = {2026},
month = jun,
doi = {10.48550/arxiv.2606.11676},
eprint = {2606.11676},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2606.11676},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Neural-Parameterized Cellular Automata for Wildfire Spread
AU - Maksym Zhenirovskyy
AU - Ion Matei
AU - Rohit Vuppala
AU - Takuya Kurihana
AU - Hon Yung Wonga
PY - 2026
DA - 2026/06/10
PB - arXiv
DO - 10.48550/arxiv.2606.11676
UR - https://arxiv.org/abs/2606.11676
AB - Traditional wildfire models rely on rigid, low-dimensional parameters and static fuel maps, frequently underpredicting fire spread. To address this weakness, we introduce a hybrid deep-learning parameterized Probabilistic Cellular Automata (CA) framework implemented in JAX. Our approach employs a Multi-Scale Convolutional Neural Network to dynamically generate spatially varying parameters that govern fire-spread probability, wind alignment, and slope influence. This hybrid design captures complex, nonlinear environmental interactions while preserving the physical interpretability of the underlying three-state CA. The JAX implementation enables hardware acceleration and gradient-based parameter calibration. Evaluated on six large-scale wildfires in the western United States, the model maintains IoU > 0.6 over 72-hour forecast horizons after a 10-day data assimilation window during which the model is fitted incrementally to observed perimeters; the resulting forecast is a conditional projection of fire growth under the suppression regime already ncoded in those observations.
ER -