Please see the attached PDF

Review of “Regime-based Aerosol-Cloud Interactions from CALIPSO-MODIS and the Energy Exascale  2 Earth System Model version 2 (E3SMv2) over the Eastern North Atlantic” by Zheng et al.

In this work, the authors investigate aerosol-cloud interactions (ACI) in the Eastern North Atlantic (ENA) with satellite observations and nudged Energy Exascale Earth System Model version 2 (E3SMv2) simulations. In particular, the authors examine differences in liquid water path (LWP), droplet number concentration (N_d), and their covariance between observations and simulations. They find that, in general, there are systematic seasonal discrepancies between E3SMv2 and satellite observations of LWP and N_d that line up with prior studies. They also find the presence of the “inverted-V” in the models and observations, with a more pronounced V shape in the model output. To investigate the effect of ENA meteorology on these results, the authors employ a machine learning method to partition the data into 4 regimes based on synoptic conditions: pre-trough, post-trough, ridge, and trough. This more-targeted analysis reveals additional, interesting insights into the differences between the model and the observations and allow for more specific inferences on the importance of meteorology on ACI processes.

The paper is well-organized and detailed, showcasing valuable results that are interesting on their own merit and motivate exciting future research. The reanalysis-clustering method for regime analysis in particular I thought was a strong result with broad-reaching applications. ENA is strongly governed by transient synoptic weather systems, and I felt this methodology provided an interesting alternative to the usual methods of regime classification. While I think this is overall a strong paper, I do have some comments that I feel should be addressed before publication. These are detailed below. Good work!

General comments

1. One key problem I have with this paper is that I think it “buries the lede” in terms of its subject matter. In the abstract, the regime names seem to come out of nowhere, and the title gives no indication that synoptic regimes are a key piece of subject matter for this work. Synoptic systems as an important lens through which we should be viewing this data aren’t suggested until well into the paper (end of section 2) and aren’t truly discussed until section 4. While I understand the desire to maintain a clear narrative, I think highlighting this portion of the analysis more clearly in the abstract and motivating it more directly in the introduction would help strongly with readability.
2. In line with my previous comment, I feel the paper is a little lacking in terms of background on synoptic regime analysis with respect to ACI. While the authors do mention McCoy et al., 2020, which is important background for this work, it is done in the somewhat vague context of “atmospheric regimes” (particularly when “cloud regimes” and “meteorology regimes” are specified in the next few paragraphs). I think, generally, prior authors examining ACI in a synoptic meteorology context have taken a more cyclone-specific approach (as in, compositing around low-pressure centers), as opposed to the trough/ridge classification approach. Differentiating synoptic regimes from more general “atmospheric” regimes here would help with seeding this idea early on. Also, I think adding some additional analysis contrasting these results with the prior literature in this area (e.g., McCoy et al., 2020) would add some critical context to the results/conclusions of this paper. If the authors feel that the analysis presented is too novel to be usefully compared to prior analyses of ACI in synoptic scale contexts, then that needs to be defended more thoroughly in the manuscript.
3. Throughout the manuscript, there are many figures with errorbars. It is unclear to me outside of figures 4 and 7) what precisely the ranges on these errorbars represent. In a note on Table 3, the range values are described as propagated uncertainties of variables used in the covariance. Is that the case for the bounds of the errorbars in the remainder of the figures? Or are they representing other statistics of the distribution (e.g. standard deviation like Figs. 4, 7)? More thorough descriptions of these errorbars in their respective figure labels are necessary for reader understanding. Additionally, I think some more detail on the sources of uncertainty (as mentioned in Table 3) would be helpful for overall understanding. Specific uncertainty values are mentioned once in the paper on L139 for MODIS-retrieved LWP. It is clear, however, that other sources of uncertainty – from satellite observations and from simulations – are considered for calculating the bounds in Table 3 (and perhaps elsewhere). To the authors’ credit, the uncertainties inherent to the utilized satellite retrievals (particularly with respect to their analysis of the trough regime) are discussed in the paper. But to contextualize the results, more thoroughly and specifically describing the sources of uncertainty and how they’ve factored into the analysis is necessary.

Specific comments:

L185: What ERA5 outputs are used for nudging the model, specifically? What is the relaxation time? More information as to how the model has been nudged is necessary.

L189: You define σ_EXT already on L129

References: Mechem et al., 2018, is cited throughout the paper, but seems to be absent from your list of references.

Table S1: I found myself frequently referencing Table S1 during my review of this paper and feel that it would be a useful inclusion in the main manuscript.

This work uses a regime-based approach to investigate aerosol-cloud interactions, specifically LWP and Nd, in warm marine clouds. The data set incorporates 2006 to 2014 measurements from CALIPSO for aerosol properties with MODIS for cloud properties in comparison to E3SMv2 simulations with machine learning to cluster four synoptic regimes in the ENA. Overall, they find that model relationships match observations qualitatively, but have more sensitive relationships likely due to model representation of cloud processes. 

The paper reads very well with very few typos. The authors do a good job presenting necessary background and what's at stake in the introduction, starting with ACI and how it is measured then how it is modeled and how the two disagree. I understand how difficult it is to compress many of these complicated concepts such as precipitation suppression or enhanced entrainment-induced evaporation which introduce uncertainty, but this paper would benefit from more explanation of these processes. For the data and method and section, I think this paper would benefit from discussion of the satellite products chosen. The literature suggests that most satellites with same measurement types are in good agreement, but I'm curious about the omission of MODIS Terra as it would provide more temporal resolution. Figures 2 and 4 are great for showing that models qualitatively match Nd-LWP and aerosol extinction coefficient-Nd relationships in observations, but are more exaggerated. The comparisons between observations and model parameters is well summarized at the end of each section. 

While very thorough, I can't help but feel as though the section 3's focus on seasonal comparisons between satellite and model parameters and relationships is unnecessary to this body of work, especially with the end of that section mentioning its limitations and the need to separate by regime to disentangle meteorological variability which I think is the meat of this body of work. The comparisons between observations and the E3SMv2 models in each regime are thorough and any discrepancies have a one or more hypothesis supported by the literature. The paper concludes with many suggestions of next steps on how to improve model representation of observations, leaning on the difficulties of overcoming satellite measurement uncertainty. I think it would be useful to sort of rank future changes that are most feasible or important to implement. I'm excited to see future work in other regions, especially of the number of regimes found and how these relationships change, especially with more drastic differences in sources of aerosol. 

Specific comments:

L48 Nitpicky, but satellite remote sensing's main advantage over in-situ cloud measurements is the spatial extent. The temporal resolution is often worse, especially of polar orbiting satellites. 

L49 It would be good to explicitly cite more of the numerous papers advancing ACI using satellite data.

L67 It would help to introduce the inverted V-shaped relationship observed in satellite retrievals in the previous paragraph. It is currently explained in detail much later in L283. 

L159 It would help to include a typical range of relative error values in retrieval and r-values of comparisons if available to better quantify what is meant by significant and decent.

L242, I think figure 1 would benefit from a 5th column of annual statistics.

L288 It would be helpful to provide ranges of Nd of these regimes found in previous work for comparison.  

L294 Missing space after Nd. 

L332 The explanation of enhanced sea salt emissions with increased wind speed makes sense and should be observed globally, but is the seasonal dependency on the under and overestimation of the model aerosol extinction coefficient regionally dependent then if dust is a factor? Additionally, it would help to provide more context of where the transported dust in the ENA comes from. 

L365 It would be helpful to provide the sensitivity values from the cited studies for comparison. 

L513 It would be helpful to provide a range of the peak of the inverted V-shape across the regimes in text to highlight this point. 

L523 While I understand that ℒ0 is bulk, wouldn't separating by the peak of the inverted V-shape illustrate your points more of the rapid decrease? 

Figure 8 Nitpicky, but I think this figure would benefit from labels of the tickmarks between 10 and 50 as a significant part of the data is within that range.