Initial Submission:

Recommendation: Minor Revisions.

Comments to Author(s):

Manuscript Number: egusphere-2025-3549

Manuscript Title: Understanding the Spring Cloud Onset over the Arctic sea-ice

Authors: Jean Lac et al.

Overview and general recommendation:

This manuscript uses 13 years of satellite lidar observations and in-situ measurements from the MOSAiC campaign to examine the drivers of the Arctic spring cloud onset. One main goal of the manuscript is to understand the drivers of the spring cloud onset. Specifically, the role of moisture advection vs. warming due to seasonal increases in solar insolation and temperature advection. Supporting previous literature, the authors find a relatively small role for moisture advection. They additionally show that the timing of the spring cloud onset is associated with warming from ice-dominated cloud regimes T < -13C to liquid-dominated cloud regimes T > -13 C. The authors consequently infer that atmospheric warming is responsible for the spring cloud onset.

I find this manuscript well-written and motivated. The application of long-term satellite records to the spring cloud onset is valuable and interesting. The addition of in-situ ground-based lidar and radiosonde observations complements this top-of-atmosphere perspective nicely and care is taken to make fair comparisons between these data sources. The authors demonstrate that moisture advection explains some variability but not the overall spring cloud onset and show that cloud phase is a strong function of temperature. However, there is relatively little focus on isolating the drivers of this temperature increase and it is ascribed to either solar insolation or remote temperature advection without much analysis. Specifically, both solar insolation and temperature advection are listed as potential drivers of the spring cloud onset, but manuscript’s discussion of these processes is inconsistent and often lacking. I think the discussion of these processes should be clarified in the introduction and discussed with an eye towards future change in the discussion/conclusion. With these changes and other suggestions detailed below, I believe that the manuscript can be accepted with minor revisions. Beyond those edits, however, I think that the impact of this work would greatly benefit from the inclusion of some additional analysis to separate between remote and local drivers of the spring cloud onset. These suggestions along with other more minor comments are included below. While not necessary, I leave decisions regarding these suggestions to the editor and authors.

Comments are formatted as:

Line number: “Text”

Specific Comment

Line 116: "ice cloud layers"

Lines 12 – 14: “Overall…April”

The authors do not demonstrate that solar heating is responsible for the spring cloud onset. The contributions from local heating and remote temperature advection are not explored here.

Lines 50 – 51: “As spring…persist”

The role of solar insolation vs. remote transport is not examined here. Is there previous literature you can point to that identifies solar insolation as the dominant driver of low-level atmospheric warming in the spring?

Lines 51 – 56: I think it is valuable here to describe how the temperature dependence of cloud phase is mediated by dynamics and aerosols in addition to the WBF process. For example, Shaw et al. (2022) and Gjelvsik et al. (2025) both studied how model aerosol schemes impact the temperature-dependence of Arctic cloud phase while using CALIOP observations as ground-truth.

Line 58: remove comma

Line 81: CALIPSO sampling has high spatial and temporal resolution, but the spatial sampling is limited due to the small footprint size. How often does CALIPSO obtain a complete observation of the study area and how do the authors handle incomplete data at daily resolution?

Line 102 (equation 2): Is additional nomenclature needed to indicate the threshold versus the attenuated total backscatter in the cross-polarized direction?

Lines 107 - 109: I had to read this sentence multiple times before understanding and recommend reorganization. e.g. When a layer with SR > 30 is located between 720m and 3200m only 17% of the underlying profiles are fully attenuated, leading to unclassified layers near the surface.

Line 116: "ice cloud layers"

Line 116: suggest adding "at each isotherm"

Line 131: i.e.

Lines 134-135: “but does not…SR > 30”

This additional clause is a bit confusing and I recommend deleting it or moving it to a separate sentence.

Line 145: replace comma with "and" between temperature and relative humidity.

Line 151: Some technical understanding is assumed here and should be described. far-range channel is not previously defined/described. Complete overlap is also not defined.

Line 172: saturate  saturated

Line 176: extra period "section. 2.1"

Line 195: Unclear what "missing saturation w.r.t. liquid 85% of the time" means. Is this a classification error and if so can you define it more clearly?

Figure 1 caption:

Does each pixel have daily data? If not, what frequency of data do most pixels have and how are missing data accounted for?

panel instead of pannel

Line 213: until the 7 May   until 7 May

Lines 219 - 222: I think that this information should be included as supplemental content if it is discussed.

Figure 2: Ice clouds also appear to be least frequent in the atmospheric temperature inversion. Can you discuss the role of atmospheric dynamics here as opposed to the focus on atmospheric temperature?

Lines 236 - 237: wording is confusing, perhaps a comma is missing between 6% and below?

Lines 238 - 240: Does this mean that high ice clouds must be increasing since cloud ice clouds stay constant (line 236)? If this is the case, I would partition the probable thin ice clouds into categories below and above 3.2km.

Lines 240 - 241: Can the authors explain the mechanism why?

Lines 241 - 244: “In addition…720m”

Can the authors explain why this matters/why it supports their conclusions?

Lines 245 – 248: The structure and intent of this paragraph is confusing. I suggest rewriting this and the previous paragraph for cohesiveness and clearly explaining the author's hypothesis on the importance of atmospheric ice particles.

Figure 3: I recommend setting the x-axis maximum to May 30 as in Figure 1, splitting ice clouds into categories below and above 3.2km, and adding a second y-axis to show the seasonal evolution of solar insolation at 70 and 82 degrees latitude.

Figure 4: Panel a: "time serie”  “time series"

Panel c: Annotated events should be labeled more clearly.

Label: "capture --> captures"

Lines 254 - 265: This paragraph only considers advection as a moisture source, but what about local sources like the melting surface? If this contribution is known to be small please state that with an appropriate reference.

Lines 270-272: If these data are described then they should also be shown in a referenced figure.

Lines 265 – 280:

This paragraph is really interesting.

So anomalous moisture advection superimposes variability onto the existing seasonal cycle but does not modify it?

Could the existing seasonal cycle be driven by the advection of moist static energy? If temperature is the driving factor, the energy needed to raise temperatures can either be sourced locally (e.g. solar warming) or remotely (advection from lower latitudes). This is an important distinction since these two sources may respond differently to global warming.

I recommend repeating the moisture advection analysis (Figure 4) with low-level moist static energy. This should quantify the role of all advective processes to the spring cloud onset. Additionally, increases in atmospheric heat content due to the absorption of solar radiation by the atmosphere can be calculated from CERES observations (or ERA5 fields) to quantify the role of the seasonal solar cycle. Here, radiation absorbed by the surface can be assumed to go into phase changes and ignored. This analysis should allow the authors to disentangle these processes as drivers of the spring cloud onset.

Figure 5: caption: review color labels. The “white” color in panel a. appears tan to me.

Line 284: Citation should be in parentheses

Line 303: Why do all cloud fraction values decline steeply at 0C?

Lines 320 – 322: After previously focusing on solar insolation, temperature advection is now discussed. This is not consistent with the narrative described in the abstract and introduction, which focuses entirely on solar insolation. Please review all discussion of these climate processes and ensure they are consistent throughout the manuscript.

Lines 336 - 339: This seems to contradict your conclusion in lines 262-263. Can you clarify/reconcile these statements?

Line 355: expanse – expense

Lines 368 – 374: “What does this conclusion say about the future of the spring cloud onset and sea ice melt onset? Does it imply that the contribution of the cloud onset will not push melt onset earlier?” I think an understanding of solar vs. advection driven warming would be especially valuable here since these processes will change very differently under global warming.

Figure C1: Caption should describe what the bolded and other lines represent

Figure D1: Caption should describe what the filled region represents and why the dashed box is present.

Review of manuscript number 2025-3549

"Understanding the spring cloud onset over the Arctic sea ice"

by Jean Lac et al.

This study is based on the analysis of satellite observations of clouds and sea ice concentration in order to explain the cloud onset occurring from March to May in the Arctic.

The authors used the CALIOP-GOCCP lidar observations from 2008 to 2020 to determine cloud layer phases (liquid, ice and unclassified) and their occurrences. Sea ice concentration is determined from NSIDC observations, and temperature and humidity data are obtained from ERA5 reanalyses. In addition, analyses of ground based data (lidar observations and radiosoundings from the MOSAIC campaign) are included.

The results highlight that, in addition to the supply of moisture contributing to the increase of cloud occurrence during Arctic spring, the rise in air temperature (due to more solar radiation from March to May) plays a major role in the increase of the occurrence of  liquid containing cloud. The authors conclude that the increase of temperature alters the balance of cloud phase formation, favoring the liquid phase formation at the expense of ice phase.

General comments:

The paper is well written and well structured. The observations, as well as the methodology, are well described. In particular, the authors take care to apply the same method for both space and ground based observations. They demonstrate that water vapor  transport is not a limiting factor for the spring cloud onset. They suggest that ice production processes lead to a depletion of moisture in early spring. They assume that, later during spring, the increase in air temperature is responsible for the transition from ice dominant clouds to liquid dominant clouds.

In my opinion, the goal of this study is clear and of interest to the scientific and Arctic communities. The data and methodology used are very well appropriate. Nevertheless, the main conclusion highlighting that liquid cloud amount increase with temperature on a global average is rather well established, and not particularly very new.

The study could be made more robust and substantially improved by accounting for the interannual variability of cloud occurrences and by investigating deeper the link between cloud onset and additional environmental parameters (for example: warm air intrusions, dynamics and stability conditions, surface coupling…). Below are some ways to enhance the study before publication.

First, the results are based on averages over the entire Arctic sea ice region and a long period of 13 years. The authors should study more in detail the interannual variability of cloudiness, sea ice concentration and thermodynamical parameters (temperatures, humidity). The extensive database (13 years, almost the entire Arctic region) makes such an analysis possible. For example: is there a link between the interannual variability of sea ice coverage and cloud occurrences (and phase), as well as temperature, humidity amount and their transport?

A second point is that the authors should investigate (in link with the interannual variability of cloud occurrences mentioned above) additional parameters influencing cloud occurrence variability. For example, since the study focuses on low clouds, it is important to examine their coupling with the surface. In addition, the link of cloud occurrence with warm air intrusions should be analyzed, including back trajectories to assess air mass transport.

Finally, the present study considers the entire sea ice covered surface only. This area is very large, and figure 1 suggests that the increase in cloud occurrence is more pronounced in certain regions (for example near open oceans: Barents, Kara and Chukchi seas). Therefore, an analysis of the spatial variability of cloud occurrence in specific representative regions with different environmental conditions would be valuable. This could also include areas over open ocean or land, which present highly contrasting conditions in terms of temperature and humidity.

Specific comments:

L 30 : “depending on definition of melt onset”: please give a brief summary of the commonly used definitions.

L 62 : Please explain how cloud properties could be influenced by boundary-layer processes.

L 98: Please indicate the units of ATB.

L 106-107: Multilayers clouds are also present. How are you sure that these unclassified are composed of liquid and not ice?

L 116: Why figure 6 is mentioned before fig 3, 4 and 5 ?

L 118-122: As the authors explain that unclassified layers are clouds, why these layers are not taking into account in the cloud phase ratio?

L 118: You should change “cloud phase ratio” into “Ice phase ratio” or something else referring to “ice phase”, according to its definition. It will be more convenient to interpret.

L 132: The limit for low-level thick clouds (or ice cloud ratio) in terms or SR is not very pronounced in the figure 3 ( or in figure D1). Could you explain why you used the value of SR = 30 as a threshold? Would using another value significantly change the results?

L 127-136: Please, explain with more arguments the interest of the histogram and the two categories you highlight?

L162-167: There was some confusion when reading this section. I understand that CALIPSO-GOCCP may miss some thin ice clouds when SR<5 and you use the ground based observations for ensure that ice clouds could be detected when SR > 3. But, how can you be sure that the satellite and ground based lidars have the same cloud sensitivity? Could you for example provide the frequency distributions of the SR for the two systems (co-located along the MOSAIC track)?

Figure C1: I see more clouds for SR between 1 and 3 on the figure ? You wrote the opposite in the text? This figure need more explanation, and you could add it in the manuscript.

L 183: “To understand low cloud formation………humidity and temperature.” Other parameters may impact the cloud formation, such as dynamics, surface coupling, stability, local sources or long range transport of aerosols… (see my major comment above) 

Section 2.1: I suggest to include open seas and MIZ areas. (see the major comment)

L 260-262: “Moreover….relative to March”. This sentence is unclear. Please, rewrite.

L 265 – 280: For the blocking event, could you provide maps similar to figure F1 to identify the moisture decrease? In addition, do you have hypothesis for this cloud occurrence increase since it is not due to moisture advection ? A more detailed meteorological analysis of the situation would be helpful. Back trajectories could provide useful insights insight into air mass transport, and give an indication of aerosol transport. Moreover, this section well illustrates that cloud occurrences and moisture transport can vary substantially on a daily scale, and that this variability has to be considered, in addition to the annual averages presented in the paper.

Figure 4: Blue light curves not really visible. Please improve.

L 284: “from” missing between “observations” and “Andreas et al”

L 319:  How does figure 7b change if you use only the cloud layer to determine the temperature distributions?

The manuscript investigates the Spring cloud onset over the Arctic sea ice and examines different mechanisms (two of which were proposed in past studies) that could drive that increase in cloud cover. For this, the authors use 13 years of CALIOP lidar observations to identify seasonal trends in cloud cover over the Arctic sea ice. Reanalysis data is used for temperature and humidity calculations, while ground-based lidar measurements and radiosoundings during MOSAiC are used to compare the CALIOP climatology to a case study in Spring 2020. Several lines of argumentation point to the fact that the Arctic Spring cloud onset is connected to an increase in temperature and, consequently, a decrease in the efficiency of ice cloud processes, which drives a transition towards low-level, liquid-containing clouds.

Seen in the context of climate change, a Spring cloud onset that is mainly connected to a temperature increase might get triggered earlier in the coming decades, with potentially large implications for the radiative balance in the Arctic. Leveraging a long time series of spaceborne lidar observations, this paper marks an important and timely contribution to this research field. Thus, the manuscript fits very well within the scope of the journal.

The manuscript is structured in a very clear way, and the figures support the line of argumentation. I structured my review in the following way: I will start with a couple of general comments that I think should be addressed by the authors, followed by more specific comments below. Finally, I will give some hints for technical corrections.

General comments:

• Due to the polar orbit of CALIPSO, there is a well-documented sampling bias at higher latitudes that can influence statistics when profiles are aggregated over a larger domain covering different latitude bands. The much larger number of profiles close to 82°N will dominate the overall sample domain statistics, if not addressed e.g. by utilizing random sampling with replacement. This might be especially important as these areas are farthest away from the moisture source of the ocean. I suggest that the authors consider this sampling bias and address it accordingly or, if already implemented, specify the how in the Data section (Sect. 2).
• Connected to the first comment, I feel like uncertainties related to the datasets used could be discussed more comprehensively. The sensitivity study and SR thresholds as well as the radiosounding intercomparison between MOSAiC and ERA5 go in that direction, but statements regarding the uncertainty in cloud cover estimates are missing. In addition, how do e.g. multi-layer cloud situations complicate the analysis or is the study limited to single-layer clouds? I am not so familiar with CALIPSO-GOCCP, but maybe that is already handled? I suggest adding a bit of information here.
• One the main arguments regarding the driver of the Spring cloud onset in this study is the availability of enough moisture to ensure saturation w.r.t. ice already way before the Spring cloud onset in March. However, for these low clouds below 3.2 km, based on Fig. 2, it looks like the clouds are still well within the mixed-phase temperature range above -38°C. It is my understanding that most of the ice clouds that form above -38°C also require saturation w.r.t. water (unless the ice crystals have fallen from above and originally formed via pore condensation and freezing as the nucleation pathway). However, this process requires ice-nucleating particles, which are quite rare in the Arctic. Having that in mind, I wonder whether the saturation-argument w.r.t. ice is truly sufficient to explain the consistent ice cloud cover?
• I am wondering about the moisture intrusion analysis. It looks like this is done using a vertically integrated quantity, which is dominated by the boundary layer moisture coming in from over the ocean, pushing the cold dome farther north. Would it not be more accurate to look at the moisture at the cloud levels investigated here? Otherwise, one might not see a cloud response as the moisture intrusion investigated here happens mainly below the cloud layer.
• Regarding the latitudinal dependence of the timing of the Spring onset: I agree that this would in principle be an argument against some of the hypotheses discussed in the study. However, and related to my comment above, I wonder about the definition of the contour for the calculation of the moisture flux and the validity of relating it to a domain-averaged low cloud cover (Figure 4). I suggest using only intrusions from the ocean to the sea ice (and exclude the Canadian Arctic for example). Otherwise, the signal might get averaged out. Fig. 1 (bottom center panel) suggests that the Spring cloud onset indeed starts in those regions (Atlantic sector and Pacific sector).

Specific comments:

Line62: I think it would be helpful to mention a specific example of how boundary-layer processes can influence low-level cloud properties here.

L102: Please provide a bit more detail here, such as that cloudy layers with cross-polarized ATB above this threshold are considered ice clouds, and layers with values below the threshold are considered liquid clouds.

L122: Please add how the statistics differ from previous studies, as I believe this to be important information for the reader.

L133: Category identification. Could you please specify why you chose altitude < 1.0 km as an additional criterion for the low-level liquid clouds? From the inset of Fig. 3 it looks like an extension to 1.5 km could be feasible to increase statistics, but maybe the logarithmic colorbar for the frequency is tricking me a bit? It would be interesting to hear the reasoning behind. In addition, I like the two ways of looking at it, altitude vs. SR (Fig. 3) and phase ratio vs. SR (Fig. D1) histograms. However, it would be more illuminating in my eyes if Figure D1 would also be color-coded with the frequency, or maybe even better: have the same altitude vs. SR plot from the inset in Fig. 3, but instead of color-coding by frequency, color-code it by phase ratio. That way it should become immediately clear that the low-clouds with SR>30 are liquid-containing clouds.

L133: Could you please include what percentage of cloud fraction is covered by the two categories introduced here, to get a feeling for how many cloudy layers are not included in one of the two categories?

L165 and Figure C1: I do like the sensitivity test of ice cloud fraction to the SR threshold, but it is unclear to me what the different lines represent in Figure C1? Please specify in the text and also the figure caption.

L193: For the computation of the biases between ERA5 and MOSAiC radiosondes, why did you choose altitudes of 2km instead of the 3.2km, which is used for the computation of the cloud statistics? I think it would be best to be consistent here.

Section 2.3: The fact that ERA5 assimilated the MOSAiC radiosondes makes me wonder whether this section is a bit misleading. If you compare the temperature and humidities for one location that is fine, but the author’s themselves mention that ERA5 likely performs worse at other locations and times. I understand why the section is included, and why it is hard to evaluate reanalysis in the Arctic. However, I feel stating the absolute biases like this is a bit misleading.

Section 3.2: I suggest including Fig. 3 as a sub panel in Fig. 2, as they are directly connected and having them side-by-side will foster comprehension by the reader. At the same time, I suggest removing the inset figure and make it a separate figure in the Data section (Section 2), potentially with a second histogram color-coded by phase ratio instead of frequency (see my comment to L133 above). That way methodology and results are a bit better separated in my eyes. It would also help as the SR frequency histograms are already discussed in Section 2.1 and are used for the definition of the two categories.

Data availability: LWCRELIDAR data is only mentioned in the data availability section, but it is unclear where this data has been used in the study?

Technical corrections:

Figure A1: Please increase the font size of axis and colorbar labels, caption: ‘based on the hourly solar angle only’ (remove additional ‘only’)

L85: Sections 4 and 5 explore […]

L98: 2.5 x10-3

Equation 2: Please double-check the equation, because if I compare with Cesana and Chepfer (2013), there a several missing minus signs in the exponents of 10, also it should be made a bit clearer that this is x10-3 or x10-4 etc.

Figure 2: Colorbar label liquid-containing cloud fraction closing bracket for unit ‘]’ missing

L119: 1°C temperature bins

Figure 3: inset maximum scattering ratio label reads 100,000, whereas caption mentions range from 3 to 10,000. Please correct accordingly and check the main text for consistency.

L134: optically thinner

Figure D1: Please specify that this is the ratio between ice and all clouds.

Figure B1: legend entries ‘resol vertical’

L169, L191: radiosonde

L176: Section 2.1 (remove additional dot)

L190, L196: radiosondes

Figure 2 caption: Colorbars are from […] for the liquid-containing […]

Figure 3: the inset is covering parts of the low-level thick liquid line.

Figure 4: caption ‘and captures most of the Arctic sea-ice extent in spring’

L254/255/257/258: x10^13. Please also correct all other instances.

Figure 6: Caption and legend label: liquid-containing clouds instead of liquid clouds

L308: we further observe

L323: from generally ice-cloud dominated in March to liquid-containing cloud dominated in May

L351: over liquid-containing clouds

L355: at the expense

L370: before the end of the century