Climate Models Underestimate Pre-Industrial Fire Emissions
Abstract. An improved understanding of how fire regimes changed from the pre-industrial to the present-day is required to reduce uncertainty in anthropogenic aerosol radiative forcing. Direct observations of pre-industrial fires are scarce so here we evaluate assumptions about historical fire regime change using black carbon data from polar and alpine ice core records and global climate model simulations from the Coupled Model Intercomparison Project Phase 6 (CMIP6) and the Large Ensemble Community Project (LENS). Black carbon is a tracer of both natural and anthropogenic combustion, enabling comparison of fire and industrial combustion emission changes from the pre-industrial (1850 CE) to the present day (1980 CE). Polar ice core records indicate relatively modest increases in mean and median 1980 CE/1850 CE black carbon ratios of 1.27 and 1.15, compared with 2.89 and 2.64 in the CMIP6 ensemble, respectively. Alpine ice cores record larger increases (mean 2.31; median 1.93), compared with even larger increases simulated in CMIP6 in alpine regions (mean 3.59; median 2.50). We tested three possibilities to explain the difference between black carbon simulated by models and recorded in observations: (1) pre-industrial fire emissions are too low in models; (2) changes in aerosol transport over the Industrial Era using data from LENS; and (3) differences in modelled aerosol properties and deposition parameterizations using CESM2 data. Analysis of these different drivers revealed that the most effective method to reconcile the model-observation disparity is to increase pre-industrial fire emissions in the models and that transport or aerosol properties are of secondary consideration.
Competing interests: At least one of the (co-)authors is a member of the editorial board of Atmospheric Chemistry and Physics.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.- Preprint
(2788 KB) - Metadata XML
-
Supplement
(1074 KB) - BibTeX
- EndNote
Status: open (until 15 Sep 2026)
- RC1: 'Comment on egusphere-2026-2728', Anonymous Referee #1, 24 Aug 2026 reply
-
RC2: 'Comment on egusphere-2026-2728', Anonymous Referee #4, 24 Aug 2026
reply
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2026/egusphere-2026-2728/egusphere-2026-2728-RC2-supplement.pdf
-
RC3: 'Comment on egusphere-2026-2728', Anonymous Referee #3, 27 Aug 2026
reply
Overall comment:
In this study, the authors present a comprehensive comparison of ice-core records with historical black carbon simulations from global models, with the aim of understanding the large discrepancy between observed and modelled changes in BC from the pre-industrial period to the present day. The compilation of ice-core records and the attempt to evaluate emissions, aerosol processing/deposition, and transport within a common framework are valuable, particularly given the ongoing development of historical emission inventories for future CMIP experiments. The manuscript is generally well organized. However, I have concerns about whether the current analyses are sufficient to support the manuscript's central attribution that underestimated pre-industrial biomass-burning emissions are the primary explanation for the model–ice-core discrepancy. I therefore recommend major revision before the manuscript can be considered for publication in ACP.Major comments:
1) The manuscript presents the underestimation of pre-industrial biomass-burning BC emissions as its central conclusion, but I am not fully convinced that the analyses presented here uniquely support this attribution. The authors show a substantial discrepancy between ice-core and modelled PD/PI ratios. However, demonstrating that the models overestimate the historical change in BC deposition is not equivalent to demonstrating that underestimated PI fire emissions are the dominant cause. The manuscript itself subsequently discusses large uncertainties associated with BC aging, transport, wet/dry deposition, snow processes, and post-depositional modification, some of which could potentially produce systematic rather than random biases. I therefore think the causal language throughout the manuscript is stronger than warranted by the evidence. The authors should more rigorously distinguish what is directly demonstrated by the ice-core/model comparison from what remains an attribution hypothesis. Ideally, this could be strengthened through a quantitative perturbation or sensitivity analysis in which PI biomass-burning emissions are increased to determine whether, and by how much, the observed spatial distribution of PD/PI ratios can actually be reconciled with the models.
2) The major conclusion that pre-industrial fire emissions are underestimated in CMIP is not new. The potentially important advance here seems instead to be the systematic global comparison across 37 ice cores and the attempt to determine whether transport and deposition uncertainties can explain the discrepancy. The manuscript would be considerably stronger if it were reframed around this question. In particular, the authors should provide a much more explicit comparison with prior studies, which already used ice-core records and inverse modelling to reconstruct improved biomass-burning emissions. What does the present analysis demonstrate that could not be inferred from that study? Does the magnitude or spatial structure of the inferred CMIP bias agree with their reconstructed emissions?
3) I have concerns about treating the compiled ice-core dataset as a globally representative observational constraint without more careful consideration of its spatial representativeness and source footprints. The study combines 37 records distributed across the continents, but individual ice cores represent BC deposition influenced by potentially very different source regions, transport pathways, precipitation regimes, and local topography. For example, the Asian comparison contains only two ice cores, while South America appears to be represented by a single ratio. This makes statements concerning a “global” systematic bias potentially sensitive to the geographical distribution and weighting of the available records. I strongly recommend testing the robustness of the global mean and model–observation discrepancy to alternative weighting approaches and leave-one-region/site-out analyses.
4) The use of PD/PI ratios is understandable for harmonizing heterogeneous ice-core records, but it also introduces an important interpretive problem that deserves substantially more attention. The manuscript defines PI and PD using 25-year means corresponding approximately to 1850–1875 and 1965–1980, respectively, with the averaging period selected partly because it minimizes variability among the ice-core records. At the same time, the authors note that extremely small PI anthropogenic emissions produce very large PD/PI emission ratios and consequently introduce an additional “Anthropogenic+BiomassBurningPI” treatment. Ratio metrics are inherently sensitive to small denominators, and similar sensitivity could influence both spatial comparisons and the apparent magnitude of model bias. More importantly, agreement in a ratio does not establish agreement in absolute BC deposition: a model could reproduce PD/PI while substantially overestimating or underestimating both periods. I would therefore like to see the ratio-based analysis complemented, wherever observational units permit, by comparisons of absolute PI and PD BC levels or normalized anomalies. This would help establish whether the identified disagreement actually arises primarily from underestimated PI BC, overestimated PD BC, or some combination of the two.Minor comment:
1) The global map panels in Figure 4 should be enlarged to enhance the readability of the figure.Citation: https://doi.org/10.5194/egusphere-2026-2728-RC3 -
RC4: 'Comment on egusphere-2026-2728', Anonymous Referee #2, 31 Aug 2026
reply
This study uses a compilation of 36 ice core black carbon records to assess the relative changes in BC deposition from the pre-industrial to present day as simulated by CMIP6. The manuscript concludes that CMIP6 underestimates pre-industrial BC emissions, as higher pre-industrial BC emissions are needed to reconcile the observed ice core ratios with known increases in anthropogenic BC emissions.
Major Comments
As the title clearly states, the main conclusion is that CMIP6 climate models underestimate pre-industrial fire emissions. There is a bit more nuance to this conclusion in that it is truly the underlying fire emissions inventory used by CMIP6 (Van Marle et al., 2017) that is the primary cause of the underestimation, not necessarily any of the mechanics of the models themselves. The authors assess both components (the model itself and the underlying emissions inventory) in this manuscript, leading to my two main concerns.
First, the underestimation of pre-industrial biomass burning in Van Marle et al., 2017 (and thus CMIP6) is already well known, which the authors do acknowledge in the introduction. And there have been a number of other studies that have used ice core data—and for the most part, much of the same ice core data as used here—to reach this conclusion. These existing studies include Moseid et al., 2022, Liu et al., 2021, Eckhardt et al., 2023, Zhang et al., 2024, and at least two of these also go a step further to provide a solution to this problem in the form of modified biomass burning emissions inventories that are consistent with ice core data across the entire PI to PD. Thus, I am concerned about what is new here. In my view, it reaches the same conclusion as previous work without as robust of an assessment of the underlying data (e.g. it only examines the PI to PD ratio, while the previous work has considered how the emissions have changed continuously since the PI) and without providing much of a solution (e.g. a relatively broad conclusion that revised inventories and better modeling approaches are needed, while previous work has already provided modified emissions inventories that could help align CMIP models with ice core data).
Second, the authors rule out potential mechanics of the models themselves (atmospheric transport, deposition processes, etc.) as a primary cause of the underestimated pre-industrial emissions, but these analyses do not seem particularly in depth. For the section discussing the influence of transport (starting line 585), for example, the authors use the LENS ensemble that has small temperature perturbations to broadly assess atmospheric transport pathways. Are the small temp perturbations realistic of uncertainties inherent to the model and capable of generating meaningful differences in transport pathways? Is it possible to test any of the other multitude of parameters that could impact atmospheric transport of BC? What about atmospheric lifetime, scavenging, precip, wet vs dry deposition, etc? I would think that these factors could introduce more of a ‘biasing’ effect than temperature alone, which the authors show to introduce more variability but not a bias. For example, a shorter atmospheric lifetime for BC would presumably decrease overall concentrations at an ice core site with disproportionate impacts on natural and anthropogenic sources depending on their proximity to the site. The manuscript hints at this in section 4 on lines 685 but the discussion lacks detail.
All in all, the conclusion of this manuscript may be worth re-iterating, especially to prepare for CMIP7, but because 1) previous work using very similar datasets have reached the same conclusion while providing meaningful improvements to the erroneous emissions inventory underpinning CMIP6, and 2) the evaluation of the performance of the model seems to lack detail, I think significant revisions to clarify the novelty and strengthen the model evaluation would greatly strengthen the manuscript.
Other Comments
I do not see the point in including the levoglucosan record from Kilimanjaro. It is a completely different tracer than BC and I don’t think it is justified to treat it as a BC-equivalent record, especially since it is treated differently than the BC records with the LOESS filter. This also starts to get into the nuances of the BC measurements themselves. The majority of the cores (~26) were developed at the same lab with identical methods (Desert Research Institute with an SP2 integrated into a continuous flow analysis system) and therefore are presumably quite consistent. Do all the ice core records here represent a relatively consistent BC measurement across the 36-core compilation? Do they all use the same method?
The section discussing post-depositional modification of BC (starting on line 705) introduces uncertainties about the ice core records themselves. Are certain sites used in this study more prone to post-depositional impacts than others?
Fig 1- Is there any logic to the numbering of the sites? It would be nice if they corresponded to Table S3 to make it easy to identify the data source for each site.
Fig 2- How were the local regions determined? Are they all the same size?
Fig 3- The colorbar makes sense for the bottom left panel, but is pretty useless for the other three. Consider changing to make it clearer?
Would it be possible to underlay any of the maps with the CMIP6 PD/PI ratio for each grid cell? That may be very informative to visualize the spatial variability of the data this manuscript seeks to address.
Reference to Schmitt et al. 2019, TC Discussions- double check this exists, please.
Citation: https://doi.org/10.5194/egusphere-2026-2728-RC4
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 130 | 59 | 18 | 207 | 34 | 12 | 22 |
- HTML: 130
- PDF: 59
- XML: 18
- Total: 207
- Supplement: 34
- BibTeX: 12
- EndNote: 22
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Summary
Previous work has shown a mismatch between pre-industrial black carbon (BC) in climate model simulations and BC inferred from ice core records, suggesting that pre-industrial fire emissions in commonly used inventories may be underestimated. Here, the authors test that idea using 37 ice core datasets and three Earth system model ensembles, and conclude that underestimated pre-industrial fire emissions are indeed the main cause of the disagreement, with changes in transport and deposition playing secondary roles.
Overall, the manuscript fits well within ACP’s scope and addresses a clearly defined scientific question. The analyses appear rigorous, the presentation is clear, and the study is well grounded in the existing literature. I do have a few logistical questions below, but I am in general supportive of publication once these points are addressed.
Major comments:
Firstly, if I understand it correctly, the authors argue "underestimated PI fire emissions are the main explanation for the model–ice core mismatch" through the following chain of reasoning (lines 400 to 410): Inventories show a large PI to PD increase in anthropogenic BC, but little change in fire BC. Since ice core BC reflects the sum of sources, the ice core PD/PI ratio should increase strongly, yet it instead resembles the weak change implied by fire emissions, so the authors infer that PI fire emissions must be higher than inventories indicate.
However, the potential weakness is the implicit assumption that anthropogenic and biomass burning BC are transported and deposited to ice core sites with similar efficiency. Anthropogenic BC may be removed near sources or remain within the boundary layer, while biomass burning plumes can be lofted and delivered more efficiently, and may also align seasonally with snowfall and preservation. If models at ~100 km resolution do not capture these source dependent “accessibility” differences, then the ice core ratio resembling the fire ratio does not uniquely imply underestimated PI fire emissions.
Another concern is that the manuscript does not directly quantify the relative roles of emissions, transport, and deposition in driving the model-ice core mismatch but instead infers their importance indirectly.
In particular, I do not feel fully convinced by the transport discussion in Section 3.4. First, the LENS ensemble mainly samples internal meteorological variability within a single model framework, so the resulting spread is not a strong test of whether forced PI-to-PD changes in circulation and removal, or structural model biases in these processes, contribute to the mismatch. Second, even if “PI versus PD transport differences” are found to be small in LENS, this does not rule out other transport-related explanations, particularly source-dependent delivery. Anthropogenic and biomass-burning BC can differ in injection height, seasonality, and removal efficiency, and models may represent these differently, which could affect how strongly each source contributes to deposition at alpine and polar ice-core sites.