Introduction

Fire is a powerful force of nature and one of the most important disturbances in the Earth’s system. A recent scientific paper involving SPECIAL group PI Sandy Harrison, sheds light on how well global Earth System Models (ESMs) simulate fire dynamics, focusing on the latest generation of models—those in CMIP6 (Coupled Model Intercomparison Project Phase 6).

Fire is influenced by a complex web of climatic, ecological, and human factors, making it challenging to simulate accurately. While previous model generations in CMIP5 included fire to a limited extent, CMIP6 has seen a significant leap forward. Firstly, the scope of models including fire in CMIP6 has expanded to 19 EMSs. Within these models, nine simulate burned area (BA) and 18 produce estimates of fire emissions.

Process of Comparison

The ESMs in CMIP6 which produce fire outputs use either the GlobFIRM, Li scheme, or SPITFIRE fire models. To assess the CMIP6 model simulations of fire properties the authors in Li et al. (2024) utilised a large set of benchmarking data. This included benchmarks for BA, emissions and the paleo fire record. For burnt area GFED5, ESACC1 5.1 and MODIS C6 were used. For emissions, GFED4, GFAS 1.2 and FEER-G1.2 were used. Lastly for the paleo fire record the Reading Paleofire Database was used, the database was developed here at Reading University within the SPECIAL research group.

Finally, to simulate the fire drivers including precipitation, sea surface temperature and population density, data was taken from CMIP6 for the climate variables and population density was taken from HYDE v3.2.

Key Findings

The study compared CMIP6 fire simulations to observations over time revealing both strengths of the CMIP6 models as well as areas needing improvement.

1.Performance Improvements Over CMIP5

    • Burned Area (BA): Six out of nine CMIP6 models simulate BA within the range of satellite products, a clear improvement over CMIP5. Enhanced seasonal variability and better representation of high BA regions, such as Africa, are notable advancements.
    • Fire Emissions: Eleven of the 18 CMIP6 models produced emissions estimates within benchmark ranges, outperforming CMIP5 in accuracy.
    • Larger spread in CMIP6 vs CMIP5: The increase in spread in CMIP6 fire model outputs is driven, for BA largely by the GlobFIRM model, and in terms of fire emissions is due to how combustion completeness is treated differently between the three fire models.

Figure 1: The 2001–2014 spatial distribution of the annual burned area fraction (% yr−1) for (a–c) benchmarks and (d–l) CMIP6 models. The spatial correlations of simulations with three benchmarks are also given in parentheses. Figure and Caption taken from Li et al. (2024).

2. Alignment with Observations

    • CMIP6 models better capture the spatial patterns and seasonal cycles of fire. For instance, they simulate peak fire seasons during the dry tropics and warm extra-tropics well. Additionally, the correlation coefficient of modelled outputs and observations spatially improved from 0.15-0.34 in CMIP5 to 0.28-0.7 in CMIP6. However, they miss the spring fire peak in the Northern Hemisphere, which is driven by crop fires.
    • There is greater alignment between CMIP6 models and charcoal-based reconstructions (RPD), especially compared to CMIP5. However, discrepancies remain, particularly in regions like southern South America and eastern North America.
    • CMIP6 models capture response to El Niño in the fire emissions except those using SPITFIRE

3. Ongoing Challenges

    • Human Fire Suppression: CMIP6 models struggle to reproduce the observed global decline in BA and emissions over the past two decades, partly due to the absence of fire management dynamics.
    • Crop Fires: Fires associated with agriculture remain underestimated, affecting seasonal fire peaks.
    • Interannual Variability: Large discrepancies exist among CMIP6 models in simulating year-to-year fire changes.

What Makes CMIP6 Better?

Several factors contribute to the improved performance of CMIP6 models:

  • Advanced Fire Schemes: A shift from GlobFIRM to the Li fire model in many ESMs has enhanced BA simulations and spatial agreement with observations.
  • Improved Climate Simulations: Developments in climate modelling have had flow on effects to more realistic fire property simulation.

Future Directions

While CMIP6 represents a leap forward, the study highlights key areas requiring attention:

  • Global Fire Trends: Models must better simulate the recent decline in BA and fire emissions which requires the incorporation of human management strategies into fire models. Additionally, continuing to improve the total estimate of BA sum is still important.
  • Crop Fires and Seasonal Peaks: Addressing agricultural fires is essential for improving spring fire season estimates.
  • Fire Sensitivity to Precipitation: Reparametrizing fuel buildup and understanding precipitation effects will enhance model reliability.

Research collectives that members of the SPECIAL research group are part of, such as the Leverhulme Centre, FIRE-ADAPT and LEMONTREE projects, are already working on elements of these challenges.

Conclusion

CMIP6 marks a significant step forward in simulating fire dynamics, narrowing the gap between model predictions and real-world observations. However, the persistent challenges highlight the intricate nature of fire systems and the pressing need for continued refinement. A key insight from the evaluation of CMIP6 models is that complexity alone does not guarantee better performance; even the most sophisticated models, like SPITFIRE, do not always deliver the most accurate results.

Looking ahead, advancements in fire modelling will depend on integrating critical factors such as human fire management and agricultural practices, which are currently underrepresented. While adding these elements may enhance realism, the findings also emphasize that striking the right balance between complexity and accuracy is essential. By focusing on targeted improvements and leveraging collaborative efforts across research communities, we can build models that not only capture the intricacies of fire dynamics but also provide valuable insights for managing fire in a changing world.

The SPECIAL group has many plans to continue working on developments in global fire models in 2025 so stay tuned!

You can read the full paper here: Li, F., Song, X., Harrison, S.P., Marlon, J.R., Lin, Z., Leung, L.R., Schwinger, J., Marécal, V., Wang, S., Ward, D.S., Dong, X., Lee, H., Nieradzik, L., Rabin, S.S., Séférian, R. 2024. Evaluation of global fire simulations in CMIP6 Earth system models. Geoscientific Model Development,  https://doi.org/10.5194/gmd-17-8751-2024