| 08 Sep 2026
Combustion controls and plume structure of wildfire aerosol optical properties observed by TROPOMI
Abstract. Wildfire smoke influences air quality and climate through the scattering and absorption of solar radiation by aerosol particles, yet the factors controlling aerosol optical properties and their spatial variability within smoke plumes remain uncertain. Here, we use satellite observations of aerosol optical properties and trace gases from the TROPOspheric Monitoring Instrument (TROPOMI) to investigate the influence of combustion conditions on aerosol optical properties and their spatial variability within wildfire smoke plumes during the Australian 2018–2019 and 2019–2020 bushfire seasons.
Substantial differences in aerosol optical depth (AOD) and single-scattering albedo (SSA) are observed among major land cover types. Wildfires characterized by stronger flaming combustion, inferred from nitrogen dioxide (NO2) and carbon monoxide (CO) column density ratios (ΔNO2/ΔCO), are associated with more absorbing plume-averaged aerosol populations, whereas smoldering-dominated fires are associated with more scattering smoke and larger AOD. These relationships indicate that combustion conditions exert a first-order control on aerosol absorptivity, while aerosol loading is more strongly influenced by total emissions. Within individual plumes, aerosol loading and trace gas abundance remain concentrated in NO2-defined plume cores, whereas SSA exhibits much weaker spatial gradients. Independent box-size analyses further show a systematic tendency toward higher SSA away from plume cores and source regions while ΔNO2/ΔCO decreases, indicating that these spatial changes occur concurrently with atmospheric processing. Although chemical processing, dilution, and mixing cannot be separated, these observations provide constraints for evaluating model representations of wildfire emissions, plume evolution, and smoke radiative effects.
Overview
The manuscript by Dix et al., “Combustion controls and plume structure of wildfire aerosol optical properties observed by TROPOMI,” examines TROPOMI NO₂ and CO, as well as TROPOMI aerosol optical depth (AOD) and single-scattering albedo (SSA), in Australian wildfire smoke plumes and investigates differences among four vegetation classes. The manuscript is well written and, for the most part, scientifically sound. However, I have some concerns regarding the quality and uncertainties of the TROPOMI aerosol products, and their suitability for some of the applications and interpretations presented in the manuscript. I believe the manuscript fits well within the scope of Atmospheric Chemistry and Physics (ACP) and should be of interest to the journal's readership. I recommend publication after addressing the comments and concerns outlined below.
General comments
This manuscript addresses interesting and important topics: the relationship between trace gases, burning stage, aerosol properties, as well as the transformation within the fire plumes. The use of the high spatial resolution of TROPOMI to investigate these relationships is quite valuable and interesting. However, I have several concerns regarding the aerosol products used in the analysis and, in particular, the extent to which cloud contamination and the spatial definition of the plume may affect the results and their interpretation.
I have some concerns regarding the quality of the TROPOMI aerosol products used in this study, particularly AOD and SSA. Based on previous experience with TROPOMI aerosol retrievals, the TROPOMI AOD has a very high correlation with cloud fraction (and also Torres et al, 2020 recommend rigorous cloud screening. I therefore recommend that the authors provide additional evidence demonstrating that the AOD enhancements identified in the analysis are genuinely associated with the fire plume and smoke aerosols rather than clouds.
At a minimum, the cloud fraction should be shown alongside AOD in Fig. 2. In fact looking at Fig. 2 I think that the enhancement of AOD east of the fire might be due to clouds rather than smoke aerosols. A coincident RGB image of the study region (for example, from NASA Worldview) would also be useful for interpreting the spatial patterns shown in Fig. 2 and determining whether clouds are present in the region of enhanced AOD.
The authors should check if the filters used for TROPOMI AOD are enough for sufficient cloud screening or if additional filters need to be included.
If possible, I would also recommend comparing the TROPOMI AOD with another aerosol product, such as MODIS or VIIRS, ideally for all fires but at least for the example shown in Fig.2 . This would provide some confidence that the observed spatial patterns are robust and not an artifact of the TROPOMI retrieval. This would provide some quantitative validation and would substantially strengthen the analysis.
I have even greater concerns regarding the interpretation of the SSA results. The range of SSA values reported in the manuscript (approximately 0.84–0.90) is relatively small, and it is not clear whether this range is meaningful given the uncertainty of the retrieval.
The authors should provide information on the expected uncertainty of the SSA product and discuss whether the differences observed in the analysis are larger than the retrieval uncertainty. For example, Torres et al. (2020) compared TROPOMI aerosol SSA with AERONET observations and found very limited correlation – it looks like a cloud of points where the retrieved values appear to fall within a broadly reasonable range.
I would therefore encourage the authors to be cautious about drawing conclusions from the SSA without further validation of this product. The analysis could still be shown but at the very least the conclusions and discussions need to reflect this and need to address that TROPOMI SSA showed no correlation with coincident AERONET SSA observations (Torres et al., 2020).
I would also be cautious about replacing TROPOMI SSA with VIIRS SSA without further evaluation. Although VIIRS provides a useful independent aerosol product for comparison, published evaluations have also identified substantial limitations in VIIRS SSA retrievals. Wang et al., 2023 (10.1016/j.scitotenv.2023.161667) report poor performance of the VIIRS SSA products. Thus, an independent comparison with VIIRS could be useful to strengthen the analysis but again conclusions need to be drawn cautiously.
The inclusion of vapor pressure deficit (VPD) is an interesting quality to add to the analysis. However, VPD primarily describes the atmospheric evaporative demand and therefore provides an indication of atmospheric dryness, but it does not directly represent the moisture content of fuels (VPD should not be described as a direct indicator of "fuel-drying conditions” l. 115).
Instead of VPD , or in addition it would be useful to consider established fire-weather indicators such as the Fire Weather Index (FWI) and/or the Fine Fuel Moisture Code (FFMC), which are more directly related to fuel moisture and fire danger. These could provide a more physically meaningful assessment of the relationship between fuel conditions and aerosol/fire emissions.
These variables could potentially be calculated using ERA5 meteorological inputs, or can even simply be downloaded from the Copernicus Data Store (DOI: 10.24381/cds.0e89c522; https://ewds.climate.copernicus.eu/datasets/cems-fire-historical-v1?tab=download) .
I have some concerns about the spatial approach used to characterize the plume and its evolution (as presented in Fig. 6, l. 402ff). In particular, the analysis appears to average quantities over relatively large boxes that may contain both plume-affected and non-plume-affected observations. This makes the physical interpretation of the resulting relationships somewhat difficult.
For example, the finding that a smaller fraction of a larger analysis area is affected by the plume is largely expected simply because the analysis area increases. The analysis boxes appear to include areas both upwind and downwind of the fire (with the fire at its centre). Consequently, the aerosol observations in the analysis box may include substantial contributions from clouds or background aerosol/trace gases that are unrelated to the fire plume; or stemming from another fire in the area.
If the objective is to investigate dilution, aerosol transformation, or changes in aerosol properties with distance from the source, I think the analysis would be more physically meaningful if it were performed specifically along the downwind plume. Ideally, the observations could be characterized as a function of distance from the source or, even better, approximate plume age. Plume age can be approximated with wind speed and distance to the source. This would allow to distinguish changes associated with plume evolution from spatial variability in background aerosol or cloud contamination. Averaging over an entire box that only partially intersects the plume might not be very meaningful. But perhaps I did not understand how this analysis was done.
Specific comments
Line 100: Why were the TROPOMI data gridded to 0.05x0.05? Please include a rationale for this spatial resolution here. Also, how are multiple TROPOMI observations/overpasses handled when constructing the gridded dataset? Is the analysis based on daily observations, and was any spatial or temporal gap filling applied?
Line 103: Does the "best" quality flag remove cloud-contaminated observations, or is a separate cloud filter required? Please clarify and verify this with respect to the TROPOMI aerosol product algorithm. (see general comments)
Line 115: The term "fuel-drying conditions" is not correct here. VPD is primarily a measure of atmospheric dryness/evaporative demand rather than fuel moisture.
Line 175: How is the local background concentration/value defined? Please provide a more detailed description of the methodology used to determine the background.
Line 182: The manuscript refers to the "long-term average NO₂ column." Please clarify the spatial extent of the box used for this calculation (the "analysis box") and the time period over which the long-term average is calculated.
Section 2.5 – Retrieval limitations and uncertainties: The discussion should also consider the effect of aerosols on the TROPOMI NO2 air-mass factor (AMF). The TROPOMI NO2 retrieval assumes an aerosol-free atmosphere when calculating the AMF, while aerosols can substantially affect the radiative transfer and therefore the retrieved NO2 column. In the TROPOMI NO2 algorithm, aerosols are indirectly accounted for as clouds. This issue is particularly relevant for biomass-burning plumes, where aerosol loading can be substantial. Griffin et al. (2021) (https://doi.org/10.5194/amt-14-7929-2021) – discuss the impact of accounting for aerosols on the NO2 AMF (and consequently VCD) in fire plumes.
The uncertainties from the AOD and SSA validation from Torres et al., 2020 could also be explicitly stated in this section, as well as the uncertainties for TROPOMI CO and NO2 in fire plumes.
Figure 5: It is somewhat confusing that the x-axes in panels (a) and (d), and in panels (b) and (e), are not the same. It may be useful to plot AOD against both FRP and NO₂, and similarly examine SSA as a function of FRP and NO₂. If this is too much for the main manuscript, it could also be included in the Supplement.
Figure 6: The meaning of the y-axis label "Fraction of plume value > reference box" is not clear.