The authors have made a thorough, comprehensive yet clearly explained evaluation of OLCI CHROMA cloud parameter retrievals.

I recommend publication with only minor remarks.

Introduction. The authors have used ARM ground-based data but not ACTRIS Cloudnet ground-based data (https://cloudnet.fmi.fi/, Illingworth et al. (2007)), although this network delivers very similar data, among which CTH and cloud phase data, and is also used for satellite cloud product validation (e.g., Compernolle et al., 2021, Vinjamuri et al. (2023)). Please provide some rationale why ACTRIS Cloudnet was not considered. Please also include a paragraph with references on how ARM and Cloudnet was previously used in satellite cloud product validation.

Page 6, line 148. Please provide some numbers how the CTP 60 mb uncertainty goal translates to CTH uncertainty. This is currently provided on page 13, line 273, but it should (also) go here.

Page 6, line 148. Please provide reference for standard atmospheric profile.

Page 6, line 149. Where do these numbers for COT uncertainty goals 25%, 35% come from? Also from Werdell et al 2019? Please indicate.

Page 7, line 181 and page 8, line 1. 'the retrieved CTH'. Retrieved by the satellite, right? Please make more clear by stating 'where h is the CTH retrieved by the CTH'

As indicated by the authors, the CTH retrieved by the satellite can have an error, in which case the parallax correction will also have an error. Have the authors considered to combine the CTH of ARM (which is the ground-truth) with the viewing zenith angle of the satellite to estimate the correction? Is there a reason why this was not considered?

Table 1 and 2. This table has 3 main categories: 'COT<3', 'single-layer COT>=3', 'multi-layer COT>=3'. However, from fig 3 it is clear that there is also a strong dependence on CTH.

I would therefore propose to have within each category, below each row with 'All', a row for low clouds (e.g., CTH<3 km or CTH<4 km) and a row for high clouds (CTH >3 km or > 4 km).

Page 12, line 255-256.  "while the ARM data may be sensitive to lower cloud droplet concentrations (and retrieve a higher top) than satellite remote sensing, this would not account for a multi-km error"

In Sneep et al. (2008), figure 1, large differences are obtained between the cloud pressure from MODIS based on thermal infrared, and those based on retrieval using the O2 A absorption, using O2-O2 absorption and rotational Raman scattering. 

Given these large differences, is it therefore not imaginable that the CTH negative bias of OLCI CHROMA is because it is based on O2 A absorption? Note also that in the S5P TROPOMI Cloud routine validation, large discrepancies are obtained between the cloud top height from OCRA\ROCINN CAL (based on O2A absorption) and that of Cloudnet (Lambert et al., 2025, table 2).  

Page 20, line 379. The authors motivate why they use CTP instead of CTH. Still it would be good to have results for CTH as well, as it gives a point of intercomparison with the OLCI vs ARM comparisons. In particular, it would be interesting to have a figure similar to Fig 3 (CTH difference vs CTH) for OLCI CHROMA vs MODIS and vs SLSTR, to check if discrepancy increases with CTH.

References

Illingworth et al. (2007). Illingworth, A. J.; Hogan, R. J.; OtextquotesingleConnor, E. J.; Bouniol, D.; Delanoë, J.; Pelon, J.; Protat, A.; Brooks, M. E.; Gaussiat, N.; Wilson, D. R.; Donovan, D. P.; Baltink, H. K.; van Zadelhoff, G.-J.; Eastment, J. D.; Goddard, J. W. F.; Wrench, C. L.; Haeffelin, M.; Krasnov, O. A.; Russchenberg, H. W. J.; Piriou, J.-M.; Vinit, F.; Seifert, A.; Tompkins, A. M. & Willén, U.Cloudnet. Bulletin of the American Meteorological Society, 88(6), 883-898. https://doi.org/10.1175/BAMS-88-6-883

Vinjamuri et al. (2023). Vinjamuri, K. S.; Vountas, M.; Lelli, L.; Stengel, M.; Shupe, M. D.; Ebell, K. & Burrows, J. P. Validation of the Cloud_CCI (Cloud Climate Change Initiative) cloud products in the Arctic. Atmos. Meas. Tech., 16(11), 2903-2918. https://doi.org/10.5194/amt-16-2903-2023

Compernolle et al. (2021). Compernolle, S.; Argyrouli, A.; Lutz, R.; Sneep, M.; Lambert, J.-C.; Fjæraa, A. M.; Hubert, D.; Keppens, A.; Loyola, D.; O'Connor, E.; Romahn, F.; Stammes, P.; Verhoelst, T. & Wang, P. 

Validation of the Sentinel-5 Precursor TROPOMI cloud data with Cloudnet, Aura OMI emO_2--emO_2, MODIS, and Suomi-NPP VIIRS Atmos. Meas. Tech., 2021, 14, 2451-2476

https://doi.org/10.5194/amt-14-2451-2021

Sneep, M.; de Haan, J. F.; Stammes, P.; Wang, P.; Vanbauce, C.; Joiner, J.; Vasilkov, A. P. & Levelt, P. F. Three-way comparison between OMI and PARASOL cloud pressure products J. Geophys. Res., Wiley-Blackwell, 2008, 113, D15S23

https://doi.org/10.1029/2007jd008694

Lambert et al. (2025), Quarterly Validation Report of the Copernicus Sentinel-5 Precursor Operational 

Data Products #27: April 2018 –  May 2025. 

Lambert, J.-C., A. Keppens, S. Compernolle, K.-U. Eichmann, M. de Graaf,  

D. Hubert, B. Langerock, M.K. Sha, E. van der Plas, T. Verhoelst, T. Wagner,  

C. Ahn, A. Argyrouli, D. Balis, K.L. Chan, M. Coldewey-Egbers, I. De Smedt,  

H. Eskes, A.M. Fjæraa, K. Garane, J.F. Gleason, J. Granville, P. Hedelt,  

K.-P. Heue, G. Jaross, ML. Koukouli, E. Loots, R. Lutz, M.C Martinez Velarte,  

K. Michailidis, A. Pseftogkas, S. Nanda, S. Niemeijer, A. Pazmiño, G. Pinardi,  

A. Richter, N. Rozemeijer, M. Sneep, D. Stein Zweers, N. Theys, G. Tilstra, O. 

Torres, P. Valks, J. van Geffen, C. Vigouroux, P. Wang, and M. Weber. 

S5P MPC Routine Operations Consolidated Validation Report series, Issue #27, 

Version 27.01.00, 227 pp., 15 June 2025.

Review of the paper "Evaluation of cloud height, optical thickness, and phase retrievals from the CHROMA algorithm applied to Sentinel-3 OLCI data" by Sayer et al.



This paper introduces the CHROMA algorithm which was developed to retrieve cloud information such as cloud height and optical thickness. As the authors explain, this algorithm was intended to be used for the OCI instrument on the new PACE mission. The algorithm was applied to OLCI data instead. The results presented in this paper are therefore also interesting to the OLCI community.



The topic of this paper is certainly fitting for AMT. The paper is interesting to read and well written. I think this paper deserves publication.



I came across a few minor typos:



- page 6, line 125: "although is" -- this should probably be something like "although it/this is"?

- page 7, line 172: "0.1^oresolution" --> "0.1^o resolution" (space missing)

This paper presents a comprehensive and timely evaluation of the CHROMA algorithm’s cloud products (CTH, COT, and phase) derived from Sentinel-3 OLCI data. The study effectively uses ground-based ARM data and intercomparisons with other satellite products to assess the algorithm’s performance, providing guidance for the future application of CHROMA to PACE OCI data. The methodology is sound, the analysis is thorough, and the paper is well-written and clearly structured. This work will be of great interest to the community and is well-suited for publication in AMT. I recommend its publication with only a few typos I identified.

Minor Corrections

• Line 39: ‘select’ -> ‘selected’
• Line 116: ‘KAZRASRCL’ is a typo and should be ‘KAZRARSCL’
• Line 217: The phrase "implying larger outliers" is slightly awkward. Suggest rephrasing to "which implies the presence of larger outliers..."
• Line 289: ‘counfound’ is a typo for ‘confound’.
• Figure 9 Caption: ‘arithmeric mean’ should be corrected to ‘arithmetic mean’.

This paper is well-written and covers an important topic.  The authors

validate cloud property retrievals from OLCI, including a validation

of their uncertainty.  I was not very familiar with oxygen-based cloud

top height retrievals myself, and found the paper interesting to read.

I have a few clarification questions, but otherwise it can be published

with minor revisions.

§2.1, page 4, line 97: by "spectrally correlated uncertainty", do the authors mean an error covariance?

§3.1, page 8, line 195: the phrasing in this sentence is a bit confusing.  You mean you use

medians when analysing the difference?  And you don't only use medians, your tables also

contain RMSE.  I think this sentence would fit better in §3.2.

§3.2, page 8, line 211: I suggest to replace "disproportionately" with "strongly", as I

don't know in proportion (or disproportion) to what NSA would be overrepresented, or

"disproportionately overrepresented" is a pleonasm here.

§3.2, table 1, page 9: The score for multi-layer GUC for F_60 and F_ED is very low at 0.00.

§3.2, page 12, line 262: "it is not clear how differences from Voigt would be integrated",

could this be estimated?

§3.2, page 13, line 268: fixed although FGD is not robustly retrievable, couldn't you do

better than a fixed value?  For example, a fixed value per cloud type?

§3.2, page 14, figure 5a: I find the results in this figure surprising.  At COT between

10 and ~60, an increase in CTH is associated with an increase in FGD.  But FGD is defined

as the vertical cloud fraction between cloud top and surface.  Shouldn't that mean that FGD

should be expected to be close to 1 for very low clouds?

§3.2, page 14, line 289: "not be globally representative", the authors could consider

if A-Train or EarthCare data can be used.  You comment on this in the conclusions, but

it could be mentioned in the beginning why this was not possible for this study.

§3.3, page 15, figure 6: I understand why uncertainty estimates are poor for multi-layer

clouds, and this is commented on in the text.  But why are they so poor for COT<3,

apparently even worse than for multi-layer clouds?

§3.3, page 15, lin 313: uncertainty estimates are not skillful.  Could this be improved?

§3.4, page 18: those F_ED values are very low, in particular for MFRSR matchups.  This

is true even for single-layer clouds.  Are the uncertainy estimates overconfident?  But

this does not show from Figure 8b?  The table and figure seem to show conflicting messages

about the reliability of the uncertainty estimates.

§3.5, page 19, line 365: As the authors note, the phase comparisons may not be generalisable.

This makes them hard to use.  I'm not sure if they should be included in the paper.

§3.5, page 20, line 370: I am surprised that satellite data sets are more likely to label

mixed-phase clouds as liquid than as ice, even though they are sensitive near the top where

ice should be relatively dominant.  Why does this happen?

§4, page 20, line 379: Why is COT averaged geometrically but CTP arithmetically?  I would

understand arithmetic for CTH, but CTP drops exponentially with CTH, so I would think a

geometric mean to be more appropriate.

§5, page 24, line 480: OLCI CTH uncertainty estimates only have skill for COT > 3

single layer clouds.  For others they don't have skill.  This line promises a bit too much.

§5, page 25, line 509: Note that OLCI also has H₂O absorption channels at 0.91 µm.  Could

they be used if the H₂O profile is known?

Editorial:

line 289: counfound → confound

line 345: remove "to"

line 431: missing ) after µm and µ should not be italic

line 435: differed → differ