Keeping et al. (2025) Influence of global climate modes on wildfire occurrence in the contiguous United States under recent and future climates

Identification

• Journal: Climate Dynamics
• Year: 2025
• Date: 2025-12-19
• Authors: Theodore Keeping, Theodore G. Shepherd, I. Colin Prentice, Karin van der Wiel, Sandy P. Harrison
• DOI: 10.1007/s00382-025-07998-w

Research Groups

• Centre for Environmental Policy, Imperial College London, London, UK
• Geography & Environmental Science, University of Reading, Reading, UK
• Leverhulme Centre for Wildfires, Environment and Society, Imperial College London, London, UK
• Department of Meteorology, University of Reading, Reading, UK
• Georgina Mace Centre for the Living Planet, Department of Life Sciences, Imperial College London, Ascot, UK
• Royal Netherlands Meteorological Institute (KNMI), De Bilt, The Netherlands

Short Summary

This study uses a large ensemble climate model and a wildfire occurrence model to characterize the spatial patterns and magnitude of global climate mode influence on wildfire occurrence in the contiguous United States under recent and future (+2 °C) climates. It finds that ENSO, IOD, and lagged TNA are the primary drivers, with their influence intensifying and other modes becoming significant under future warming.

Objective

• To characterize the spatial patterns, magnitude, and seasonal timing of the influence of global climate modes on wildfire occurrence in the contiguous United States under recent (2000–2009) and future (+2 °C global warming) climates, and to assess how these relationships change with warming.

Study Configuration

• Spatial Scale: Contiguous United States (CONUS), with wildfire occurrence modeled at 0.1° spatial resolution and aggregated to 0.5° for analysis.
• Temporal Scale: Two decade-long time slices: recent (2000–2009) and future (2075–2084, corresponding to approximately +2 °C global warming). A large ensemble of 1600 simulated years (160 ensemble members per time slice, each 10 years long) was used, with daily wildfire probability modeled.

Methodology and Data

• Models used:
  • Wildfire occurrence model (Keeping et al. 2024): Generalized linear model trained on wildfire occurrence data, predicting daily probability of wildfires greater than 0.1 hectares. Predictors include climate (diurnal temperature range, precipitation, wind speed, snow cover fraction, vapor pressure deficit), vegetation (cropland, needleleaf, shrub fractions, gross primary production), and human factors (rural population density).
  • Climate model: EC-Earth3 (Döscher et al. 2022) within the KNMI-LENTIS large ensemble (Muntjewerf et al. 2023).
  • Gross Primary Production (GPP) model: P model (Wang et al. 2017; Stocker et al. 2020) for simulating photosynthesis.
• Data sources:
  • Wildfire occurrence data: Short et al. (2022) (satellite data after 1984, aggregated state/federal records after 1992).
  • Reanalysis data: ERA5-Land (Muñoz-Sabater et al. 2021) for bias correction and downscaling of climate predictors.
  • Climate model outputs: KNMI-LENTIS (EC-Earth3) for sea-level pressure (SLP) and sea surface temperature (SST) fields to derive climate mode indices, and for climate predictors.

Main Results

• ENSO, the Indian Ocean Dipole (IOD), and the 1-year lagged Tropical North Atlantic (TNA+1) are the primary global climate modes influencing interannual wildfire variability in the contiguous US under recent conditions.
• The La Niña phase of ENSO positively influences wildfire probability over 91% of the contiguous US, with the greatest relative increases in Mediterranean California, the central Great Plains, and southern Florida.
• Negative IOD and positive TNA+1 show similar associations with wildfire to La Niña, affecting 90% and 85% of the CONUS, respectively. While correlated with ENSO, TNA+1 retains a substantial independent influence.
• Contrary to previous expectations, El Niño is not significantly associated with wildfire occurrence in the northwestern US in this 1600-year simulation.
• ENSO's influence on wildfire is linked to its effects on vapor pressure deficit (VPD), particularly strong in the Great Plains, and precipitation, especially in the southwestern US.
• El Niño (La Niña) is associated with a later (earlier) wildfire season peak in the southwestern US and an earlier (later) peak in the southeastern US. La Niña increases fire season length in Mediterranean California, the Arizonan Mountains, the Great Plains, and southern Florida.
• Under an additional +2 °C global warming:
  • The Atlantic Multidecadal Oscillation (AMO+1) and Pacific/North American (PNA) mode show the greatest expansion in their area of significant influence on wildfire (AMO+1 from 38% to 68%, PNA from 16% to 56%).
  • The magnitude of impact of ENSO and TNA+1 on annual wildfire occurrences increases strongly, particularly over the Great Plains and in the West.
  • The Great Plains is identified as the region most sensitive to climate variability and change, showing a greater than 25% increase in annual wildfire occurrences under specific climate mode phases (e.g., La Niña, positive TNA+1, negative IOD in recent climate, and additionally AMO+1 in the future climate).

Contributions

• This study is the first to leverage a large ensemble (1600 simulated years) with a probabilistic wildfire occurrence model to provide high-resolution geographical patterns of climate mode impact on wildfire across the contiguous US, overcoming limitations of short observational records.
• It quantifies the strength and geographical variation of relationships between key climate modes and wildfire occurrence, including their influence on fire season timing and length, under both recent and future climate scenarios.
• It clarifies the independent and correlated influences of ENSO, IOD, and TNA+1, identifying them as the principal drivers of US wildfire variability.
• The research challenges existing assumptions by finding no significant relationship between El Niño and wildfire occurrences in the northwestern US.
• It highlights the Great Plains as a critical region highly sensitive to climate variability and change regarding wildfire probability.
• The findings provide robust, statistically significant evidence for the utility of incorporating multiple climate modes into long-range fire season forecasts, enhancing risk management capabilities.

Funding

• Schmidt Sciences, LLC
• LEMONTREE (Land Ecosystem Models based On New Theory, obseRvations and ExperimEnts) project
• SCENARIO NERC Doctoral Training Programme (NE/S007261/1)

Citation

@article{Keeping2025Influence,
  author = {Keeping, Theodore and Shepherd, Theodore G. and Prentice, I. Colin and Wiel, Karin van der and Harrison, Sandy P.},
  title = {Influence of global climate modes on wildfire occurrence in the contiguous United States under recent and future climates},
  journal = {Climate Dynamics},
  year = {2025},
  doi = {10.1007/s00382-025-07998-w},
  url = {https://doi.org/10.1007/s00382-025-07998-w}
}