Wildfire Modeling

3. Efficient first-order algorithms for large-scale, non-smooth maximum entropy models with application to wildfire science

We propose a new algorithm to solve the Maxent optimization algorithm that outperforms the state of the art by one order of magnitude. As an example application, we implement our Maxent algorithm to model monthly wildfire probabilities as a function of fire-related environmental features.

G. P. Langlois, J. Buch, & J. Darbon (2024), Efficient first-order algorithms for large-scale, non-smooth maximum entropy models with application to wildfire science, Entropy, https://doi.org/10.3390/e26080691 || preprint

2. SEASFire: Seasonal and subseasonal-to-seasonal forecasts of fire frequency and sizes

Github: https://github.com/jtbuch/SEASFire

Manuscript in progress

1. Stochastic machine learning (SML) model of wildfire activity in the western US (SMLFire1.0)

Github: https://github.com/jtbuch/smlfire1.0

We predict the fire frequency and sizes in each grid cell using a pair of Mixture Density Networks (MDNs) trained on climate, vegetation, and human predictors.

J. Buch, A. Park Williams, C.S. Juang, W.D. Hansen, P. Gentine (2023), SMLFire1.0: a Stochastic Machine Learning Model for Wildfire Activity in the Western United States, Geosci. Model Dev., https://doi.org/10.5194/gmd-16-3407-2023