Ground-based detection of Antarctic clouds: analysis of cycles and comparison with IASI products

Donat, Federico; Maestri, Tiziano; Fabbri, Elisa; Martinazzo, Michele; Bianchini, Giovanni; Del Guasta, Massimo; Di Natale, Gianluca; Palchetti, Luca; Masiello, Guido; Serio, Carmine; Liuzzi, Giuliano

doi:10.5194/egusphere-2025-2793

Preprints

https://doi.org/10.5194/egusphere-2025-2793

Preprints

03 Jul 2025

| 03 Jul 2025

Ground-based detection of Antarctic clouds: analysis of cycles and comparison with IASI products

Federico Donat, Tiziano Maestri, Elisa Fabbri, Michele Martinazzo, Giovanni Bianchini, Massimo Del Guasta, Gianluca Di Natale, Luca Palchetti, Guido Masiello, Carmine Serio, and Giuliano Liuzzi

Abstract. Over Antarctica, the identification of cloud layers from infrared satellite observations is extremely challenging due to the similarities in temperatures and radiative properties of the clouds and the underlying iced surfaces. Ground-based observations, collected by the Radiation Explorer in the Far InfraRed – Prototype for Applications and Development (REFIR-PAD) spectroradiometer operating at Concordia Station, Dome C, on the Antarctic Plateau are used to obtain scene classifications with the Cloud Identification and Classification (CIC) algorithm. The resulting cloud occurrence time series span the timeframe 2014–2020 showing cycles of 12 months (with maxima in December) and 6 months (with maxima in January and July), providing evidence of the semiannual oscillation of the Southern Hemisphere also in localized cloud occurrence. Similar harmonics are observed in the collocated surface temperature and pressure. Analysis of the cloud radiative effect shows that the far infrared downwelling radiance during the peaks of the semiannual oscillation is about twice as high as during its minima. Ground-based cloud classifications are compared to satellite-derived products of the Infrared Atmospheric Sounding Interferometer (IASI) flying on MetOp A, B, and C. Several IASI L2 cloud products (namely cloud tests, cloudiness summary and cloud phase) collocated with the Concordia Station geolocation are considered. The comparison regards more than 1,200 satellite observations from 2014 to 2020, and is conducted by means of a "one-to-one" correlation analysis and via the analysis of the observed cloud occurrences. The one-to-one analysis (conducted using temporally and spatially collocated measurements from IASI and REFIR-PAD) shows that, up to December 2019, the IASI products Artificial Neural Network (ANN) test and the Advanced Very-High-Resolution Radiometer (AVHRR) heterogeneity test are moderately correlated with ground classifications, while the Numerical Weather Prediction (NWP) test, AVHRR cloud fraction test, and flag cldnes are mostly anticorrelated. However, from December 2019, both the NWP test and the flag cldnes switch to positive correlation values. When the flag cloud phase is used as a scene classifier, a limited correlation is found up to December 2019 but significantly higher values are observed in 2020. Finally, it is shown that the IASI cloud phase classification (ice or mixed/liquid) is well correlated with the ground-based phase classification.

Received: 12 Jun 2025 – Discussion started: 03 Jul 2025

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Federico Donat, Tiziano Maestri, Elisa Fabbri, Michele Martinazzo, Giovanni Bianchini, Massimo Del Guasta, Gianluca Di Natale, Luca Palchetti, Guido Masiello, Carmine Serio, and Giuliano Liuzzi

Status: final response (author comments only)

RC1:
'Referee Comment on egusphere-2025-2793', Anonymous Referee #1, 16 Sep 2025
The manuscript documents an analysis based on a cloud mask product derived using an algorithm that ingests radiances from a ground-based spectroradiometer over Antarctica. The study describes annual variability and dominant frequencies of cloud coverage, as well as a validation of cloud mask products derived from IASI on the MetOP satellites.
Even though each topic (climatology and satellite validation) are, in principle interesting, they are not developed with the necessary depth to warrant publication in AMT. First, the findings on cloud annual cycle and the six-month cycle are well known, as the authors properly documented. The analysis of the far-infrared radiation is interesting, but it does not meaningfully add to what is known about cloud and climate variability of Antarctica. Moreover, these types of analyses fall outside the scope of AMT: “The main subject areas comprise the development, intercomparison, and validation of measurement instruments and techniques of data processing and information retrieval for gases, aerosols, and cloud.”

The validation of IASI products is better aligned with AMT’s subject areas. However, as a validation paper, the structure and analysis better match those for a conference proceeding than an article ready for peer review analysis. Generally speaking, the analysis is primarily the comparison of mean values and (Matthews) correlation coefficient, without an in-depth analysis of the causes for the discrepancy or how the analysis could inform algorithm developers or IASI users about best practices for using the satellite products in atmospheric research. Along this line, the repetition of findings in a summary section, and the exclusion of a discussion and conclusion sections, make this reviewer think that the authors rushed the submission of their manuscript.

Regarding the analysis, I argue that it is not possible to obtain statistically meaningful results by analyzing correlations or mean values at a monthly scale. From Fig. 6, especially for the spatiotemporal collocation, the number of samples per month are generally less than 14 samples. With just a few samples, it is very likely that the statistical values of MCC in Figs. 7 are negligible, and thus, no inferences should be made in terms of long-term relationships (times series) between REFIR-PAD and IASI.

As the authors, stated, the different spatial resolution of IASI products and the ground-based observation is a important impediment. The fundamental question here is whether REFIR-PAD retrievals are well suited for validating satellite products with a footprint of several km. And this is not a minor concern, as the FOV of REFIR-PAD is not even 5% the scale of the pixel resolution of IASI. Sometimes, the high sampling rate of ground-based instruments like lidars (ceilometer), can help translate temporal statistics to spatial ones, under the assumption of a constant wind. This is more challenging to do with REFIR-PAD, but it could be attempted. Regrettably, I am not convinced that this validation is methodologically sound. At least, the alternative hypothesis that the discrepancy between ground-based and satellite data is attributed to the dissimilar sampling characteristics cannot be falsified.

Why do some statistics improve after 2019? Again, this is a critical question that needs to be addressed.

If the goal is to evaluate sounder retrievals, then what is the purpose of including products derived from a broader band instrument such as AVHRR.

I am not providing a list of edits and suggestions at this time because I anticipate significant changes in a revised manuscript. That being said, the manuscript needs to be carefully proofread.
Citation: https://doi.org/10.5194/egusphere-2025-2793-RC1
- AC1: 'Reply on RC1', Tiziano Maestri, 03 Oct 2025
  
  See attached file
  
  Citation: https://doi.org/10.5194/egusphere-2025-2793-AC1
RC2: 'Comment on egusphere-2025-2793', Anonymous Referee #2, 23 May 2026

The article presents a data analysis from the ground-based Concordia Station, Dome C, on the Antarctic Plateau. The cloudiness from 2014 to 2021 is discussed and compared with measurements from the nadir viewing satellite IASIS.
The Atmospheric measurement technique challenge is clearly stated first sentence of the abstract. Authors carefully describe the methodology, with figure 2 for example, and expose the possible biases in their datasets with Figure 1 and 3 for example. The first part of the article describes in detail the different steps of this analysis. I liked figure 4 showing nice measurements made from 2014 to 2021.
However, I am not satisfied with the comparison between IASIS and the ground-based measurement. I would like the article to provide more explanations on these comparisons and especially on the reasons for the anticorrelations observed.
Intuitively, I would trust more the ground base measurements and the CIC analysis. Therefore, I agree to take the CIC/REFIR-PAD cloud products as a reference for the comparison, as was done Figure 7 to 14. I don’t understand why ANN, NWP, and cldnes >=3 are anticorrelated with CIC/REFIR-PAD and the authors don’t give any explanations. Maybe, a better description of the IASIS data product would help to understand why it is the case ?
It seems that cldnes is a combination of other indicators from what I understood (including ANN, NWP). Why do the authors focus on lower levels indicators like ANN and NWP ? What do we learn from it ?
In the opposite the “cloud phase” indicator seems independent from the other ones, because this one is well correlated with ground base measurements. Is it the case ? how “cloud phase” is measured on IASIS ?
As the 1^st reviewer said “the analysis is primarily the comparison of mean values and (Matthews) correlation coefficient, without an in-depth analysis of the causes for the discrepancy or how the analysis could inform algorithm developers or IASI users about best practices for using the satellite products in atmospheric research. Along this line, the repetition of findings in a summary section, and the exclusion of a discussion and conclusion sections …” I do also find that in this revised version, there is too much repetition in the summary section, and no clear conclusive statements are made. One such statement I would expect is ““cloud phase” is the best IASIS indicator for cloudiness” or “NWP in IASIS does not agree with ground base observation, which questions the simulated clear-sky radiance derived from numerical weather prediction in IASIS data product”, could we say that ? The “dramatic improvement in 2020” of IASIS data needs to be questioned also. I doubt that this was never discussed in any paper concerning IASIS data or IASIS new data products.

Small comments :
Several times, the reference style is not correct like L24 or L27 L34 L77 etc… where the date is missing for example : Liou 2022.
More description of the IASI instrument would be relevant : is a nadir-viewing satellite with wavenumber range X …What is the wavenumber range and the spectral resolution
L98 : It is only said that the classification is based on a machine learning technique, but I believe it is trained to detect slight differences in the radiance properties of the particles ? I would appreciate a word on the physics behind
MCC classification : it was not clear on how many scenes the MCC was calculated ?
L171 watch the formatting of the hypertext link
L250 : it might be interesting to mention that the 4 months harmonic period also occurs in temperature, even though it is slightly below the 99% confidence level. Especially to reenforce your novel observation of this 4-month oscillation.
L355 analysing or analyzing ?
Figure 4, 5, 7, 8 , 9,10 ,11 ,12,13,14, legend : replace words (“left upper panel” etc) with letters (a,b,c …), as labeled in the figure.
Appendix A : I did not understand the implications of this results, I don’t know what “Hit Rate” is and how it is helpful for the paper.

Citation: https://doi.org/10.5194/egusphere-2025-2793-RC2

Federico Donat, Tiziano Maestri, Elisa Fabbri, Michele Martinazzo, Giovanni Bianchini, Massimo Del Guasta, Gianluca Di Natale, Luca Palchetti, Guido Masiello, Carmine Serio, and Giuliano Liuzzi

Viewed

Total article views: 6,024 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
5,251	571	202	6,024	197	278

HTML: 5,251
PDF: 571
XML: 202
Total: 6,024
BibTeX: 197
EndNote: 278

Views and downloads (calculated since 03 Jul 2025)

Month	HTML	PDF	XML	Total
Jul 2025	630	94	35	759
Aug 2025	547	30	20	597
Sep 2025	2,660	57	43	2,760
Oct 2025	250	34	15	299
Nov 2025	180	40	20	240
Dec 2025	150	55	20	225
Jan 2026	317	85	15	417
Feb 2026	191	60	9	260
Mar 2026	259	85	17	361
Apr 2026	60	25	8	93
May 2026	7	6	0	13

Cumulative views and downloads (calculated since 03 Jul 2025)

Month	HTML	PDF	XML	Total
Jul 2025	630	94	35	759
Aug 2025	547	30	20	597
Sep 2025	2,660	57	43	2,760
Oct 2025	250	34	15	299
Nov 2025	180	40	20	240
Dec 2025	150	55	20	225
Jan 2026	317	85	15	417
Feb 2026	191	60	9	260
Mar 2026	259	85	17	361
Apr 2026	60	25	8	93
May 2026	7	6	0	13

Viewed (geographical distribution)

Total article views: 6,012 (including HTML, PDF, and XML) Thereof 6,012 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 25 May 2026

Short summary

The cloud occurrence over the Antarctic Plateau is characterized using ground-based interferometric data from 2012 to 2020. The results show a yearly pattern, and a six-month cycle linked to atmospheric oscillations. The cloud radiative forcing at far infrared doubles during cloud occurrence oscillation peaks. Infrared Atmospheric Sounding Interferometer (IASI) Level 2 products are compared to ground data, showing an improved agreement in cloud identification from year 2020.


Total:	0
HTML:	0
PDF:	0
XML:	0