Wildfires are an increasingly urgent concern as climate change alters fire regimes worldwide. While historical fire records provide valuable insights, they are too short to capture the full range of extreme wildfire years that may occur. To better understand possible wildfires in a given year, in both recent and future climate conditions, a new study published in Frontiers applies a large ensemble (LE) modelling approach to wildfire occurrence.

Led by LEMONTREE researcher Theo Keeping, the study models 1,600 years of wildfires in the United States to understand the extent to which stochastic climate variability can drive differences in wildfire outcomes between years. By comparing a modern reference period (2000–2009) to a future scenario with 2°C of global warming, the research highlights how fire regimes will shift across the contiguous U.S., revealing increased fire occurrence, longer fire seasons, and stronger geographic patterns in interannual wildfire variability.

Why Large Ensemble Modelling is Crucial for Wildfire Projections

Unlike traditional wildfire models driven by historical observations, LE modelling provides a probabilistic approach to understanding wildfire variability. This method captures a wider range of possible fire years, including years with extreme fire events that have not yet been observed.

Key findings from the study show that:

  • Regions with a high mean number of wildfires also exhibit greater variability in fire occurrence. However, certain areas—particularly the Great Plains and southwestern U.S.—experience proportionally higher variability than others.
  • Fire occurrence is controlled primarily by two key factors: fuel production (linked to vegetation growth) and atmospheric moisture availability (which affects fuel drying). The dominant factor varies by region.
  • The distribution of fire season length is highly variable, with some areas—especially the Southwest—experiencing extreme fire seasons that are much longer than the historical record suggests.

“Fire models are increasingly being used to project future trends. These predictions are often quite specific, but to truly understand present and future fire regimes we need to somehow characterise the aleatoric spread of present and future fire outcomes. In my view large ensembles are the most sensible way of doing this.”

Theo Keeping – University of Reading

How Will Wildfires Change with 2°C of Warming?

The study’s future scenario, which examines conditions under an additional 2°C of warming, reveals several significant changes in wildfire risk:

  • The mean number of wildfires increases across most of the contiguous U.S., but the spread of fire occurrence increases even more dramatically. This means that extreme fire years will become much more common.
  • Regions that were previously less fire-prone, such as parts of the northern U.S., become increasingly susceptible to extreme wildfire seasons.
  • Fire-prone areas in the western U.S. experience a strong amplifying effect, with both the annual number of fires and fire season length increasing significantly.
  • Fuel availability becomes an even stronger driver of fire occurrence, meaning that vegetation growth—rather than just dry conditions—plays a larger role in determining fire risk.

These findings highlight a fundamental shift in fire dynamics: Climate warming not only increases the average likelihood of wildfire but also amplifies the likelihood of extreme fire years, particularly in already fire-prone areas.

Figure 1: Observed and modelled annual wildfire occurrence across the contiguous U.S. Panels show (a) observed mean fire occurrence (1992–2020), (b) reanalysis-based model mean (1990–2019), (c) modern ensemble mean (2000–2009), (d) modern ensemble spread, (e) ratio of spread to mean for the modern ensemble, (f) +2°C ensemble mean, (g) +2°C ensemble spread, and (h) ratio of spread to mean for the +2°C ensemble.

 

A More Effective Approach to Fire Management

By incorporating climate stochasticity and fire randomness, large ensemble modelling provides a more robust framework for wildfire risk assessment than traditional models. This approach is valuable for:

  • Estimating the probability of extreme fire years—helping fire agencies plan for worst-case scenarios.
  • Identifying regions with the potential for very long fire seasons, which is critical for managing fire suppression resources.
  • Understanding emerging wildfire risks in areas experiencing rapid shifts in fire regimes, such as the Great Plains and northern U.S.
  • Informing climate adaptation and mitigation strategies by providing a clearer picture of how wildfire regimes will evolve.

Given the increasing costs and impacts of wildfires in the U.S., this research underscores the need for proactive fire management and policy planning. The combination of rising global temperatures and the stochastic nature of wildfire events makes extreme fire years more likely and more unpredictable. Tools like large ensemble modelling can help improve wildfire forecasting, allowing communities and policymakers to better prepare for the challenges ahead.

📖 Read the full paper here:

Keeping, T.R., Zhou, B., Cai, W., Shepherd, T.G., Prentice, I.C., van der Wiel, K. & Harrison, S.P. 2025. Present and future interannual variability in wildfire occurrence: A large ensemble application to the United State. Frontiers,8, https://doi.org/10.3389/ffgc.2025.1519836