16 Nov 2022
16 Nov 2022
Status: this preprint is open for discussion.

SMLFire1.0: a stochastic machine learning (SML) model for wildfire activity in the western United States

Jatan Buch1, A. Park Williams2, Caroline S. Juang1,3, Winslow D. Hansen4, and Pierre Gentine5 Jatan Buch et al.
  • 1Lamont-Doherty Earth Observatory, Columbia University, Palisades, NY, USA
  • 2Department of Geography, University of California, Los Angeles, CA, USA
  • 3Department of Earth and Environmental Sciences, Columbia University, New York, NY, USA
  • 4Cary Institute of Ecosystem Studies, Millbrook, NY, USA
  • 5Department of Earth and Environmental Engineering, Columbia University, New York, NY, USA

Abstract. The annual area burned due to wildfires in the western United States (WUS) increased by more than 300 % between 1984 and 2020. However, accounting for the nonlinear, spatially heterogeneous interactions between climate, vegetation, and human predictors driving the trends in fire frequency and sizes at different spatial scales remains a challenging problem for statistical fire models. Here we introduce a novel stochastic machine learning (SML) framework, SMLFire1.0, to model observed fire frequencies and sizes in 12 km x 12 km grid cells across the WUS. This framework is implemented using Mixture Density Networks trained on a wide suite of input predictors. The modeled WUS fire frequency corresponds well with observations at both monthly (r = 0.94) and annual (r = 0.85) timescales, as do the monthly (r = 0.90) and annual (r = 0.88) area burned. Moreover, the annual time series of both fire variables exhibit strong correlations (r ≥ 0.6) in 16 out of 18 ecoregions. Our ML model captures the interannual variability and the distinct multidecade increases in annual area burned for both forested and non-forested ecoregions. Evaluating predictor importance with Shapley additive explanations, we find that fire month vapor pressure deficit (VPD) is the dominant driver of fire frequencies and sizes across the WUS, followed by 1000-hour dead fuel moisture (FM1000), total monthly precipitation (Prec), mean daily maximum temperature (Tmax), and fraction of grassland cover in a grid cell. Our findings serve as a promising use case of ML techniques for wildfire prediction in particular and extreme event modeling more broadly. They also highlight the power of ML driven parameterizations for potential implementation in the fire modules of Dynamic Global Vegetation Models (DGVMs) and Earth System Models (ESMs).

Jatan Buch et al.

Status: open (until 11 Jan 2023)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse
  • RC1: 'Comment on egusphere-2022-1148', Anonymous Referee #1, 05 Dec 2022 reply

Jatan Buch et al.

Data sets

Western US MTBS-Interagency (WUMI) wildfire dataset Caroline Juang, A. Park Williams

Model code and software

SMLFire1.0: a stochastic machine learning (SML) model for wildfire activity in the western United States Jatan Buch

Jatan Buch et al.


Total article views: 276 (including HTML, PDF, and XML)
HTML PDF XML Total Supplement BibTeX EndNote
181 87 8 276 21 2 3
  • HTML: 181
  • PDF: 87
  • XML: 8
  • Total: 276
  • Supplement: 21
  • BibTeX: 2
  • EndNote: 3
Views and downloads (calculated since 16 Nov 2022)
Cumulative views and downloads (calculated since 16 Nov 2022)

Viewed (geographical distribution)

Total article views: 269 (including HTML, PDF, and XML) Thereof 269 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
Latest update: 06 Dec 2022
Short summary
We leverage machine learning techniques to construct a statistical model of grid-scale fire frequencies and sizes using climate, vegetation, and human predictors. Our model reproduces the observed trends in fire activity across multiple regions and timescales. We provide uncertainty estimates to inform resource allocation plans for fuel treatment and fire management. Altogether the accuracy and efficiency of our model makes it ideal for coupled use with large scale dynamical vegetation models.