Review of “Satellite Observations Reveal Northern California Wildfire Aerosols Reduce Cloud Cover in California and Nevada Through Semi-Direct Effects” by Gomez et al.

In this study the authors use spatially collocated satellite products to investigate extreme fire events in the southwestern United States. The authors use several methods to isolate the response of the atmosphere to absorbing aerosol from fire events. Monthly-mean extinction profiles show enhanced extinction in months of high fire activity. Daily profiles of thermodynamic properties, subset for different large-scale meteorological conditions, show the troposphere over the region is warmer by several degrees K on days with high fire activity. Other variables are used to illustrate the coincident response of the water-vapor profile and cloud properties. TOA fluxes of shortwave radiation from CERES are used to link the changes in the atmosphere to a radiative effect. The theme of the manuscript is linking all these responses to an effect called the semi-direct effect. However, I do not believe the authors provide robust evidence of this. Due to this, and other issues raised below, I recommend major revisions in order for this manuscript to be considered for publication.  

Major comments:

The semi-direct effect. There needs to be a clear link between the presence of absorbing aerosol and changes to the cloud-field due to temperature. The vertical profiles of extinction show signs that smoke (and polluted dust) is enhanced in high fire-activity months. This dataset, as stated in the manuscript, is a monthly product on a very coarse grid. However, without coincident profiles of aerosol to directly compare against the increases in T, decrease in RH, and decrease in CF it is very difficult to make the SDE link robust. I don’t think this would be an issue if SDE was stated as a possible mechanism, but as it is the main conclusion (and title) of the manuscript I think this needs to be addressed. A few things that could be considered: The reader has no idea how important an increase of 0.005 km-1 extinction is (the value for the bulk of elevated smoke/polluted dust EC in Figure 2a). Is this an important magnitude? Could this amount of extinction result in the magnitude of T (2K throughout troposphere) seen in Figure 5? Also, where are the clouds? Do you see the reductions in cloud fraction in the same place as the enhanced T? Finally, CALIPSO provides profiles on much shorter timescale and spatial scales. It may be beyond the scope of the paper to do the whole analysis, but it would be good to use a single year or a few cases to really establish that the increase in T, decrease in CF etc are definitively collocated with absorbing aerosol.

Extreme fire events are associated with the release of BB aerosols but also heat. Most, if not all, of the findings in this study could be explained by the coincident release of heat that increase the temperature, reduce the RH, and reduce cloudiness. How do the authors know that this isn’t the case?

The semi-direct effect (the main theme of the paper) is not introduced until Section 4. This needs to be properly introduced.

Focus on high/low extreme surface pressure conditions. I think it is important to stratify the large-scale environment etc to isolate on the aerosol impact and the authors have done a good job attempting this. However, the bulk of the results persist with a focus on contrasting the extreme (10th/90th percentile) conditions. The conclusions are dependent on the high or low ps conditions, but we don’t know which conditions are more appropriate. Would it be useful to use the most common conditions so that it is more readily applicable to the real world? Or conditions that are more representative of future scenarios for the region of interest? Or at least include this element alongside the extreme conditions.

The aerosol-cloud interactions section (4.5) needs more attention. The conclusions are that ARI dominates over ACI and this is due to the semi-direct effect. The increase in IWP coincident with T is highlighted as a ‘fingerprint’ of a dominant radiative effect. Many studies have found that IWP may increase due to either ARI or ACI. This includes convective invigoration and other mechanisms. One of the primary mechanisms by which aerosols impact clouds via ACI is the suppression of the warm-rain process. What does the precip response show? Is this consistent? Does the cloud-field distribution change? Shallow cumulus may respond in a different manner than the deeper clouds.

The discussion section is missing any comparison with other studies and reads as an afterthought. As a reader I have no idea if this is complimentary to other studies or paradigm shifting. I am surprised this has been overlooked.

Minor comments:

L15: Frequency and severity of wildfires. Please include some values or statistics.

L31: “..altering the hydrological and radiative balance of the atmosphere.” Please include references.

L34. I don’t think a reference to methane fits within this manuscript.

L36: “..leading to increased SW forcing”. Which direction is the forcing?

L37: ”higher injection”. Higher in altitude or magnitude?

L37: Can you expand this to include other aerosol impacts on cloud microphysics? Such as LWP, the warm-rain process, convective invigoration?  

L43: What other parts of the world? Please give examples of other locations that may experience similar events.

L44: “aerosols primarily and secondarily emitted”. These terms need defining, especially ‘secondarily’.

L45: What are BB emissions dependent on? What are these models missing? What is a ‘proper’ parameterization of BB emission (L54)?

L48: Are there any studies that are relevant for the region you are looking at?  This section should provide us with a good base for establishing where our current level of understanding is.

L54: “dust aerosol from wildfire-cleared vegetation” needs to be properly introduced. Have other studies shown this process is an important source of uncertainty?

L67: What is fire dry matter emission? Is this expressed in GFED as fluxes of different aerosol species? Can you include this information in the manuscript and properly define DM.

L71: ‘burned area’ and ‘fire power’ need to be defined. Where does the fire power dataset come from? What is the difference between the two methods? If the burned area is from AQUA/TERRA does it not also suffer from cloud cover obstruction?

L81: Why is there a temporal uncertainty? Is this related to the method for calculating burned area?

L86: ”..GFED is from an older model”. Older model compared to what? What has changed? Is it likely to impact the conclusions of this study? Please state which version of GFED this study uses in Section 2.

L87: suggestion: replace accurate with robust.

L100: Are the retrievals instantaneous? If so what local time are the retrievals valid for?

L116: The SSF1Deg-Day product provides a daily mean value. Why was this used instead of the instantaneous SSF1Deg-Hr product? The latter is temporally collocated with MODIS and AIRS products on AQUA, whereas I believe the daily product will include interpolation of (AQUA) values across a synthetic diurnal cycle.

L137: Is it reasonable to assume all the variables are well-fit assuming a normal distribution?

L155: A lot of unnecessary detail in this section. Please consider reducing.

L196: Is there a more scientific word that can be used in place of ‘suspicion’?

L199: Would this be better placed in Section 3.2.

L215: Aha! Here is the fire-dust introduction. I suggest moving this to the introduction section.

L218: ‘Semi-direct effects’.  This term hasn’t been introduced. Given that this is a theme of the paper and is in the title I think it should be given a proper introduction.

L221: “dust and smokey aerosols”. Could you briefly explain how CALIPSO classifies this species?

L224: How important is an extinction of 0.01 to 0.02 km-1 of extinction? How does this compare to background extinction profiles?

L229: Negative anomalies. This requires more attention. I don’t think it is sufficient to point towards large error bars – especially given they are deemed statistically significant. If you suspect an abnormal month then it should be possible to check this.

L234: “fingerprints of SDE = warm cloud layer”. This example is one of many ways the SDE may manifest, and only holds true if the absorbing aerosol is located within the cloud layer. Please rewrite to make this clearer.

L236: CDF is already defined.

L243: “distribution of AOD is significantly different”. Are you referring to the spatial distribution? If so where do you show this is significant?

L267: MH20 is not defined in the main manuscript. Is this from AIRS?

L270/275: what is the mechanism driving low specific humidity in the high troposphere? Could you summarise the findings that are in section 2 of the supplement?

L282-283: This information is in the figure caption and doesn’t need to be re-introduced.

L284 (+288 +290): “Significantly leftward”. I don’t think a direction is a useful metric. Also, I’m not sure the figure shows CF liquid for ps90 under high DM is significantly suppressed (leftwards).

L286: “(at the 90% confidence interval)”. This has already been defined.

L287-288: The difference in CTH anomaly is significantly increased for ps90.

L293: “this creates conditions of cloud-free skies”. Do you observe largely cloud-free skies under ps90 DM10 conditions?

L321: Can you suggest a mechanism that explains the change in cirrus cloud fraction?

L329: “IWP scales positively with T, so this is a fingerprint of a dominate radiative effect”. How do you know this is not confounded by microphysical effects? Could IWP responses not be driven by ACI and temperature via ARI?

L334: There is a lot of observational evidence showing biomass burning aerosols can enhance cloud droplet number concentrations. Please include these studies as a contrast.

L338: “…reduce RH to the point where clouds are unable to form in the first place”. Can you see signs of this in the dataset you have? A simple histogram of LWP or CF would give you an idea of how the cloud field distribution is modified.

L346: Longwave effects. Is there a reason this analysis wasn’t extended to look at the LW or NET radiative fluxes at TOA available in the SSF1deg product?

L355: Results may be applicable to other Med climates. Please expand on this. Do they have similar atmospheric drivers?

L363: “few and far between / infancy”. There have been advances since 2016 to include interactive fire emissions. I suggest the authors include more recent studies.

All figures: Once something is defined (i.e., Northern California nCA, 10th percentile DM DM10) I don’t think you need to define it again in every subsequent figure. This would help clarify the figure captions. This is also true for the manuscript (i.e., pressure ps).

All figures: Not always easy to read the labels and axis labels. Scrolling through the pdf at 100% zoom will show you the plots that need amending.

All map plots: I suggest replotting the plots showing the spatial distributions so that the latitude and longitude cells are equivalent. In all plots the latitude is stretched. I also suggest making the ‘significance dot’ larger so it’s immediately clear.

Figure 5. Label for M_H2O includes a subscript (cl). What does this indicate?

Figure 7: Caption details for (a) and (b) are switched.

Figure 9: Colorbar labels disappear at bottom of range.

Figure 11. LWP and IWP wrong units shown.

In their paper, Gomez and coauthors stratify observations of meteorological parameters over northern California by surface pressure (as a proxy for fire weather) and fire emissions and find that for both pressure extremes analyzed, greater biomass burning emissions correspond to warmer temperatures, lower relative humidity, and lessened cloud fractions. They interpret this as evidence for a semi-direct effect that reduces cloud fraction from smoke absorption.

Unfortunately, no evidence is presented that clearly demonstrates this relationship is both causal and that it runs in the direction the authors assume. In particular, a reasonable hypothesis based on the same results would be that, for the same general synoptic conditions, random fluctuations that increase temperature (decrease RH) favor increased fire emissions.

The paper is not suitable for publication as written. One option to move forward would be to conduct some kind of analysis that would plausibly establish causality. For instance, are the temperature perturbations seen very large compared to variability within a given synoptic state? (And thus require an explanation beyond meteorological variability.) The other option, which is more feasible if the authors are to limit themselves to an observational analysis, would be to reframe the results as an interesting hypothesis versus a shown conclusion. The work would then argue for follow-up analysis using a model (perhaps a regional model given GCM limitations) that could assess the plausibility of effects of this size from a given smoke perturbation. This wouldn’t require much additional analysis, but would require a major rewrite of the title, abstract, and main text.

Specific comments:

Line 129: CALIPSO does have snapshots available; it’s fine if you don’t want to use them, but maybe specify that you’re just referring to gridded products here?.

Figure 5: The bars currently show one standard error, correct? They should be changed to 90% confidence if you want to use “significant” language.

Figure 7: Panels a/b are mixed up in the caption versus the figure.

Code availability: “Upon request” is not compliant with ACP standards, in my reading. I would advise creating a Github repository.