This paper presents a framework for classification of properties of marine stratocumulus clouds, by simply placing the clouds in space defined by three corners of optical thickness classes. The method is demonstrated by application to one year of satellite data, and large-eddy simulation, and thereby gives insight into cloud processes and reveals how cloud susceptibility to aerosol perturbations depends on cloud morphology. Several quite remarkable conclusions can be drawn regarding magnitudes and even sign of radiative effects of aerosol-cloud interaction.

The paper is well written, the methodology overall clearly and reliably described and the results very relevant. It presents a useful perspective for estimating susceptibility, for classifying and determining representativity of means, and for refining classifications based on often used discrete classifications from ambient quantities.

I would recommend for publication in ACP after only minor revisions, according to the following comments.

General comments

1. Scope and focus

The title can be read as having a focus on marine cloud brightening as a climate intervention method, which is actually mentioned only at the very end of the paper as one application of relevance to the results. One way to go would be to discuss the results in relation to MCB literature more but I would rather recommend emphasizing some of the general findings more. Perhaps even remove or replace the “suggests weak cloud brightening potential” in the title to make it more general, to reflect the broad application of the method and results, not limiting it to MCB.

Also, arguably, MCB would be a candidate for the strong local aerosol perturbations that are explicitly said not to be considered (line 95-96), and if the authors want to turn the focus more to MCB that should be discussed too.

There are many noteworthy results and conclusions, such as how the scene mean is not representative of the morphology (line 157-158), that S_LWP is negative everywhere (line 212), that S_LWP dominates S_twomey (line 253), that S_net near zero is the most common response (line 253). I think these results should be emphasized, and also could be discussed more in relation to previous literature. For instance for the sign of S_LWP, what might be the explanation for many earlier estimates of positive S_LWP in precipitating cases, and V-shaped sensitivity? There are some hints e.g. in line 225 and 288-289 to sampling biases, but it would be interesting to dig deeper, if possible.

2. Data and methods

a) The method is demonstrated with one year of satellite data (line 52), and this is what the results are based on. Why is not more data used? Do you think it matters? Is there any sensitivity in the results to the year chosen? Does the screening and selection (lines 54-59) affect the results?

b) Is it not possible to reach a certain state (or morphology) in different ways? The LES simulations present one plausible path, but discussing this equifinality could be interesting. Could it matter to the susceptibility-properties ascribed to the categories?

c) S_net is defined as the in-cloud susceptibility (eg line 25), which seems to mean that cloud fraction adjustment is not taken into account, only LWP adjustment. This is also indicated by equation 1, that doesn’t have a term dlnA_c/dlnCF*dlnCF/dlnN_d. But equation 8 in Bellouin et al (2019), that the nomenclature follows, does on the other hand include a term for cloud fraction adjustment to N_d. It is not clear how the satellite derived S can distinguish the in-cloud and total contributions to albedo-changes and exclude S_CF, and hence why the residual between theoretical S_Ac and satellite observed S_net is attributable to S_LWP (as stated in line 107, and then used to estimate S_LWP). Might this approach to estimating S_LWP as a residual affect the results?

Specific and technical comments

Line 17: Why not introduce r_e annotation for cloud droplet size, it appears fist on line 57

Line 21, 24: How certain is S_Ac, compared to S_LWP described as "uncertain" (cf. Bellouin et al 2019)?

Line 42 says S_LWP greater in small MCCs, with the explanation that the adjustment is active in cloud cores rather than peripheries. But wouldn't we expect smaller horizontal-extent clouds to have more periphery relative to cell core? The Zhou and Feingold (2023) reference given rather states that small MCC cores have higher N_d, and more evenly distributed LWP. Maybe the explanation could be expanded a bit, to clarify the mechanism.

Line 68: Please specify the A_c used, if and how it is based on radiative fluxes for clear sky and all sky, or on cloud properties, or other

Line 66: It wouldn’t hurt to include Fig A1 int the main manuscript

Line 78: Fractions of cloudy pixels, should rather be percentage

Line 125: S_CF notation has not been defined before

Line 136-137 repeats line 43-45, and again in line 158-159

Fig 1: Where is this/these scene/s (lat, lon)?

Line 166: “represent -> represents”

Line 179: It is not clear from Fig 2 c/d that nighttime thickening follows the same trajectory as the thickening during the previous night, the two trajectories marked with arrows are rather separated. Please clarify this statement.

Line 188: The phrasing is a bit unclear here. I think you mean that the information from a satellite snapshot placed in the ternary diagram can give information about the state of the cloud scene, as different parts of the diagram represent different processes. Directly from the satellite snapshot, without the help of the ternary this arguably *can’t* be inferred. Perhaps rephrase to clarify. The same formulation appears in line 282.

Line 209: Would anything else be possible, than clouds forming in a thin state, and dissipating via a thin state?

Figure 2: Caption could say what contours (in b) represent. That information is currently only in later figures. Also, Figure A2 and Fig A3 refer to figure 2a (for what the contours represent, presumably), whereas Fig 3 and Fig 4 refer to Figure 1a. Please correct and/or clarify in the figure captions.

Fig A3: What is REF here? This could be made consistent with text (line 240) referring to A_c.

Fig 4b: Diagram says S_twomey but text says S_Ac, this would be good to make consistent

Line 270: “fig 2c and 2c” -> “fig 2c and 2d”

Line 271: “are found “-> “is found”

Line 296: Is it possible to give an error estimate on that number?

This paper seeks to better understand cloud adjustments to aerosol perturbations, with a focus on the response of the liquid water path. Noting that the likely adjustment in different cloud types or morphologies will vary, they propose a novel method for constraining for this factor, using the distribution of cloud optical depth. They find that in many cases, the intrinsic cloud responses (Twomey and LWP adjustment) offset each other, leaving the overall cloud field response to be driven by cloud fraction changes, which they find can be negative under some circumstances.

This is a novel and clever decomposition that aims to solve a long-standing problem. It is clearly in scope for Atmospheric Chemistry and Physics. I have a few points, primarily about the framing of some of the results, but following these I think it would be suitable for publication.

Main points

The method for calculating S_LWP is supported by theoretical considerations, but I am a little unclear as to how well this decomposition works in practice. Almost all of the terms are determined by the average bin optical depth, other than S_net (which is determined from observations).  To what extent is S_net also being determined at a constant optical depth? Given the authors define their cloud type based on the properties that are expected to vary under an aerosol perturbation, is this restricting the magnitude of the response? I appreciate the authors have already made considerable effort to demonstrate this, but perhaps it is possible to demonstrate (potentially using some synthetic data), that this method is able to extract the required susceptibilities?

To what extent is the albedo-Nd relationship due to meteorological covariations? This has been some uncertainty about it in the past. Would we expect it to be free of the same issues that drive the negative Nd-LWP relationship (Arola et al., 2022)? Would the use of a microwave LWP product to determine the Nd-LWP relationship address some of the issues raised in this work (as it doesn't depend directly on the LWP)?

The negative cloud fraction response to aerosol is interesting, being different from almost all previous studies. I would note that the Meskhidze et al (2009) result is likely due to a regression to the mean effect. Analyses that controls for the initial cloud state shows a (very weak) increase in cloud amount during the day (Gryspeerdt et al., 2014a; Pugsley et al, 2025). While a don't think the results presented here are affected by the same bias, it does make them almost unique (and so perhaps worth some additional scrutiny).

Minor comments

L14 - Is this not a bit more broadly relevant? If I understand it correctly, it would suggest that almost all of the net aerosol forcing would have to come from the cloud fraction response too?

L28 - There is some earlier work in this area from Landsat studies, e.g. Davis et al, JAS, 1997 (10.1175/1520-0469(1997)054<0241:TLSBIS>2.0.CO;2).

L37 - To what extent is this different? Presumably cloud morphology controlled by meteorological factors? To me, a strong aspect here is that morphology is an observed parameter, unlike the large scale meteorology (which will have biases and uncertainties). I would note that there have been some studies that investigated susceptibilities as a function of observed cloud properties (which is related to the method here; e.g. Gryspeerdt et al., GRL, 2012; Langton et al., GRL, 2021)

L55 - SPI filters have been shown to help improve retrieval biases (Grosvenor et al, Rev. Geophys., 2018). Are these not used due to the morphology decomposition? How does this affect the distribution of biases in the results?

Fig. 1e - These scenes are relatively important as they have a reader decide what cloud types each part of the diagram refers to. How were they chosen?

L162 - To what extent is this a CF-optical depth relationship, rather than the more complex morphology-dependant relationship?

L183 - I am not quite clear what the LES diagrams are demonstrating. There is a small loop which looks like it is related to the diurnal cycle (daytime breakup), but I am not clear what the transition to/from the bottom left of the diagram represents? Is this the spinup of the simulation? A night time breakup is more unusual.

L192 - I agree that precipitation must play an important role in cloud breakup, but this might not be the whole story, as there is relatively little precipitation during the day (Burleyson et al, JAS, 2013), which might explain the lack of an aerosol impact on cloud development during the daytime breakup period (Pugsley et al, PNAS, 2025).

Fig. 2 - Assuming the authors are using python, the use of a mesh plotting command (e.g. pcolormesh) might help make these plots more triangular

L214 - Some studies have cast doubt on the causality of the Nd-LWP relationship (e.g Gryspeerdt et al., ACP, 2019; Arola et al, 2022). Does this method avoid the retrieval biases that have been hypothesised to drive this?

L219 - Does this relationship depend on cloud top RH? This would add confidence that the negative relationship is due to entrainment, rather than a retrieval bias/effect.

L237/Fig. 4b - S_Ac is called S_Twomey in the figure

Fig. 4b - What are the units for S_Twomey?

L239 - I think S_Ac is defined in Eq. 2 such that it would be at a maximum for the smallest Ac

L242/Fig. 4c - The S_net pattern looks very similar to S_LWP. this appears to be due to the S_Twomey/S_Ac values having very little variability.

L295 - Given the small overall relationship in the most common cases, why is this not consistent with other methods (or is it)?

Fig A2 - This should be in same colorscheme as Fig 4a to make comparison simpler.