the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 08 May 2026
Distribution of cloud geometrical properties and types over the Mediterranean: insights from a decade of CloudSat and 1 year of EarthCARE measurements and ERA5 comparison
Abstract. Space-based radar observations from the CloudSat mission are used to characterize cloud properties over the Mediterranean during 2007–2017. CloudSat’s Cloud Profiling Radar (CPR) provides vertically resolved cloud measurements, enabling analysis of cloud geometry and type by altitude and month. Low clouds (<1 km a.s.l.) occur up to four times more frequently over sea (32–39 %) than land (8–12 %). Over land, clouds are most common between 1–4 km, dominated by Stratocumulus (36 %), with Altocumulus and Altostratus also present. Over sea, clouds peak at 0–2 km, with Stratocumulus most prevalent, especially in the West Mediterranean. High-level clouds peak at 8–11 km, reaching frequencies of ~30 % across the region. Seasonally, Stratocumulus peaks in winter and autumn, while Cumulus maximizes in July–August, particularly in the East Mediterranean. Altostratus and Nimbostratus are least frequent in summer, whereas Altocumulus peaks over East Mediterranean in July–August. High clouds show maxima in spring–early summer and autumn, with minima in midsummer. CloudSat column water and ice are compared with ERA5 reanalysis. Both datasets show similar spatial and seasonal patterns, with higher values over the western and central Mediterranean. Liquid water differences (−0.004 to −0.008 kg/m²) indicate good agreement, while larger ice discrepancies (up to −0.024 ± 0.017 kg/m²) suggest underdetection of thin clouds in ERA5. Preliminary results from the EarthCARE mission are presented but are limited by shorter records. Overall, this study supports improved Climate Data Records and continuity between past, current and future satellite observations.
- Preprint
(1752 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 05 Aug 2026)
- RC1: 'Comment on egusphere-2026-2384', Anonymous Referee #1, 25 Jun 2026 reply
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 336 | 97 | 14 | 447 | 21 | 21 |
- HTML: 336
- PDF: 97
- XML: 14
- Total: 447
- BibTeX: 21
- EndNote: 21
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Overview
This review examines the preprint of the article titled “Distribution of cloud geometrical properties and types over the Mediterranean: insights from a decade of CloudSat and 1 year of EarthCARE measurements and ERA5 comparison”, by Voudouri et al.
The study exploits ten years of CloudSat observations (2007-2017), to generate and compare cloud statistics on three regions of the Mediterranean basin (western, central, and eastern). Then, it uses the same dataset to check the ERA5 water and ice content products, quantifying a possible bias in the reanalysis. Finally, it presents cloud statistics after one year of EarthCARE data (10/2024 - 10/2025) to compare with the previous results from CloudSat observations.
The article’s primary contribution lies in proposing an approach that bridges the use of atmospheric observations with the improvement of models and reanalyses. It also contributes a broad overview on how clouds distribute within the Mediterranean basin by region and time of the year. The article’s approach and subject matter can be well-suited for publication in the ACP journal after some major revisions.
In the following sections, I propose several actions that could help strengthen the conclusions of the manuscript, with the aim of enhancing its scientific value and maximizing its impact on the community. I have focused on suggestions that can be addressed using the methods and tools already developed in this study, requiring little to no additional data or external information.
Major comments
On the cloud statistical aspects, I only have some minor comments stated in the next section of this review. My main concerns are on how comparisons with ERA5 are handled, and how EarthCARE data is exploited.
I think that in this form, there is a missed opportunity to highlight how active remote sensing instruments (especially radar, lidar) are different from the observations ERA5 uses to constrain the model outputs. Section 2.3 should indicate which observations ERA5 uses, at least to constrain water and ice content, since the differences in the observation techniques are important to investigate where the biases come from (especially for ice retrievals). In this same line, I advice also looking at the ensemble spread or uncertainty estimate from ERA5, especially to analyze the results form Figure 7(a). Could the dispersion come from data with more or less weighting of observations with respect to the model?
Continuing with the ice, you mention the possible existence of thin ice clouds (cirrus) as a possible reason for lower ERA5 ice estimation, yet you do not provide a quantitative estimate on how large this effect can be, or any proof. You could better support this statement by using CloudSat itself to estimate how much ice is present in such thin clouds, and see if the values compare with the difference you observe with ERA5. I think this is important because ice difference could also come from the ERA5 reanalysis methodology and the observations used, which include passive microwave normally much more sensitive to water than to ice, and possible model biases. This is also why I insist that it is important to highlight the observational aspects of ERA5.
A quantification of the error attributable to the under-detection of cirrus clouds would substantially strengthen the value of your statistical approach (especially since you already identify cloud types). It would also provide insight into the extent to which active remote-sensing instruments can contribute to improving reanalyses, and it would help guide future reanalysis developments by providing clues on the sources of these errors.
About section 3.6, I agree with your observation in lines 311 and 312. Cloud statistics for just one year of observations are of relatively minor interest since you would not be able to get monthly patterns. I think this section should be reformulated, and in my opinion it could be of much more value if it focused on checking if you get the same offsets you found in Figures 7 and 8 with CloudSat. This could also shed another light on the possible source for the ice bias. Is it the ice retrieval algorithm on the satellite that could cause the error? Checking with two different instruments using two different retrieval algorithms can help to reinforce or discard this conclusion.
Regarding other aspects of the article presentation: the English usage is correct, the language is appropriate for the subject, and the text is clear and easy to understand. The main reading problems arise when observing the figures.
In general Figure text is very small, and in some cases very difficult to read even on the computer screen with zooming tools. This work needs a rethinking on how figure legends are handled (maybe use a single caption per figure with bigger letters). For example Figures 4 and 5 do not need to repeat the same legend on every graph. It could be shown just once with a bigger caption size. The same idea applies to the colormaps on Figures 2 and 6. In cases where the colormap range matches there is no need to repeat the colorbar taking up space and reducing figure size. All axes titles and numbers should also be enlarged until they are at least comparable to the font-size of the main text.
Minor comments
- Lines 80-84: the objectives could be better highlighted. Perhaps with an enumeration.
- Line 136: by ice cloud you mean cirrus only? Or also other ice clouds? Due to the relatively low reflectivity range also mentioned.
- Line 140: you mention little differences in the central Mediterranean. More precision is needed.
Figure 2: What are the noisy blue dots on the top heights? Especially for the western Mediterranean? Also why did you keep the noise pixels under 30 dBZ? If there is no reason I advise to indicate somewhere clearly that it is noise, or to remove them using the mask you mention in line 149, and modify text accordingly.
Line 197: How do you handle the ~500m blind zone of CloudSat above the surface? If there is no solution, it could be better to indicate that you can only detect clouds in the 500m-2 km layer, and indicate how this could impact the statistics (for example sea-fog may be under–detected).
Lines 236-244: You don’t mention exactly what values make a cell “empty”, only that the threshold is considerably lower. Please be more precise. I also think it would be nice to have a map similar to Fig 6 (g) showing the count number or highlighting the “empty” regions, to provide an idea of how CloudSat observations spatially cover the defined Mediterranean regions.
Lines 361-362: with the results presented so far, it is not possible to conclude that the error comes from missing thin clouds.