| 16 Sep 2026
Global fuel and burned area reconstructions and projections for 1950–2100
Abstract. Fire is a key component of the Earth system, influencing ecosystems, atmospheric composition, and carbon–climate feedbacks. Although observations, process-based models, and data-driven approaches have advanced understanding of historical global fire activity, projecting fire regimes under concurrent changes in climate, fuels, land cover, and socioeconomic conditions remains challenging. Here, we present Sparky-Climate, a data-driven framework providing a global reconstruction and scenario-based projections of burned area and eight fuel characteristics from 1950 to 2100 at monthly approximately 25 km resolution. The framework emulates fuel load, vegetation state, and fuel moisture before predicting burned area fraction from meteorological, land-cover, fuel, and socioeconomic predictors. Historical reconstructions are driven by reanalysis and historical socioeconomic data, whereas future projections use CMIP6 climate simulations with Shared Socioeconomic Pathway-consistent land-use and socioeconomic inputs.
Comparison with the FireCCILT11 satellite-based burned area product shows agreement in spatial patterns, temporal variability, and multi-decadal trends (r = 0.87). By the late 21st century, global burned area is projected to decline by 6.7–21.9 % across the scenario ensemble, masking substantial regional divergence. Savanna burned area is projected to decline by 15.2–31.3 %, whereas temperate and boreal forests are projected to increase by 6.3–18.2 %, particularly under high-emission pathways. Tropical moist forests are projected to show weak declines under low-emission pathways (7.6–8.1 %) but increases under high-emission pathways (7.6–24.4 %). Extreme fire activity is projected to decline in tropical regions, but to increase in temperate and boreal biomes, with the strongest amplification generally occurring for the most extreme events and under higher-emission scenarios.
Sparky-Climate provides a consistent framework for analysing long-term changes in fuel conditions and fire regimes under alternative climate and socioeconomic pathways.
General comments
This report provides a clear, comprehensive description of the model structure and use, with the key results of model performance, future projections of burned area, and highlighted areas of projected extreme fire activity clearly evident from the abstract through to the discussion.
Specific comments
While I understand that ignition sources cannot represented in the model, the discussions throughout the paper of the effect of human ignitions are useful and necessary to understanding. Given that lightning ignitions are a major contributor to total burned area in some parts of the world, such as boreal forests, a discussion should also be included around why lightning is excluded from this model and the potential effect this could have on interpreting results. For example, lightning is commonly used in other fire models (e.g. Rabin et al., 2017).
Comments for specific sections are provided below:
Section 3.2:
The value of the correlation coefficients in the text throughout this section do not agree with the values presented on Figure 3. While most discrepancies are small, in aggregate this creates concern over which values are correct. Please amend either the text or figure to the correct values.
L303: r=0.86, whereas in figure 3, top right plot, r=0.87. A value of r=0.87 is also stated in the abstract.
L305: r=0.85, whereas in figure 3, the global burned area time series, r=0.79.
L319: TGS r=0.86, but in figure 3, r=0.76.
L319: FGS r=0.82, but in figure 3, r=0.68.
L321: TMBF r=0.54, but in figure 3, r= -0.20. If r=-0.20 is the correct value, I would suggest highlighting this low correlation in the text, as you have done for Deserts and Xeric Shrublands.
L322: TDBF r=0.52, but in figure 3, r=0.42.
Section 3.5:
All the probability ratios provided throughout the text in this section do not appear to align with the values provided in Figure 7 and so further clarification is needed. I have provided only 2 examples here.
The probability ratio is defined in section 2.4 as the probability of exceeding each reference threshold during the target period (Pt) divided by the corresponding probability during the reference period (Pr), i.e. Pt/Pr. Therefore, if I understand correctly, the percentage change in exceedance probability should be ((Pt – Pr)/Pr) * 100%. Which simplifies to (Pt/Pr – 1)*100.
Applying this formula, in line 422, the probability ratio (Pt/Pr) of exceeding the 99.9th percentile in TGS in the 1950s compared with the reference period is quoted as 1.05 however the corresponding percentage value in Figure 7 is 139%, instead of the expected 5%.
Similarly, in line 433-5: “In TGS, the probability of exceeding the 99.9th percentile declined to 0.08, 0.10 and 0.03 times the corresponding SSP-specific 2015-2024 baseline under SSP3-7.0, SSP4-3.4 and SSP5-8.5 respectively”. These values correspond to -34%, -36% and -52% respectively in Figure 7 so it is surprising that the percentages fall monotonically but the ratios do not.
Please clarify the calculation used to compute the percentage changes in exceedance probability used in Figure 7 and amend the values as necessary.
Section 4.4:
As above, please check the ratio values quoted throughout from section 3.5 as appropriate.
Technical Comments
Methods
L157: Please indicate which layer of soil moisture used as an input variable.
L253: 95th percentile values are not provided, but mean values are. Please amend as appropriate.
Section 3.3:
L352-3: I suggest rewording the phrase “stable but variable conditions”. Perhaps “[followed by a long period of comparatively stable] conditions, with continued interannual variability.”
L433: The phrase “99th and 99.9th percentile exceedances declined across all scenarios” does not apply to DXS according to figure 7, whereas there is a small increase in SSP1-1.9. Please amend this sentence.
L440: 95th percentile values are quoted, but not provided in Figure 7.
L445-6: “the probability of exceeding the 99.9th percentile threshold was 0.91”. This value is quoted in the discussion (section 4.4) as a probability ratio, please amend as appropriate.
Section 4.3:
L546-553: In this paragraph, I would suggest including discussion and reference to studies on the potential climatic drivers of reduced burned area in Africa. Such as Zubkova et al. (2019).
L607-9. Please provide a reference for this section. Such as Zheng et al. (2021) currently in your reference list.
Section 4.4:
L622-5: This sentence repeats results and could be dropped to improve readability.
L654: The 95th percentile values are not provided in results.
L662-668: This paragraph is a little repetitive of paragraph L635-642. I would suggest combining these paragraphs to improve readability. While I understand that ignitions cannot be represented in the model, lightning-ignited fires have been shown to be the dominant source of burned area in many high latitude regions, particularly boreal forests, and I would suggest including a discussion of the influence of lightning on burned area here, including the uncertainty in future lightning projections.
Figure 7:
The caption to references 95th percentile values that are not present, and does not reference mean values which are in the figure.
The mean values are not referred to in the results section, and 95th percentile values are referred to. Please update either the text or the figure and caption as appropriate.
References
Rabin, S. S., Melton, J. R., Lasslop, G., Bachelet, D., Forrest, M., Hantson, S., Kaplan, J. O., Li, F., Mangeon, S., Ward, D. S., Yue, C., Arora, V. K., Hickler, T., Kloster, S., Knorr, W., Nieradzik, L., Spessa, A., Folberth, G. A., Sheehan, T., Voulgarakis, A., Kelley, D. I., Prentice, I. C., Sitch, S., Harrison, S., and Arneth, A.: The Fire Modeling Intercomparison Project (FireMIP), phase 1: experimental and analytical protocols with detailed model descriptions, Geoscientific Model Development, 10, 1175–1197, https://doi.org/10.5194/gmd-10-1175-2017, 2017.
Zubkova, M., Boschetti, L., Abatzoglou, J. T., & Giglio, L: Changes in fire activity in Africa from 2002 to 2016 and their potential drivers. Geophysical Research Letters, 46, 7643–7653. https://doi.org/10.1029/2019GL083469 , 2019.