Please see attached file

The topic has significant interest for both weather and climate applications. The paper is well written and organized. One suggestion is some more detailed discussions on the ground truth data (DARDAR) could be added for limitations and further improvements. A few minor revisions below, mostly for more clarification, will also help to improve the manuscript for publication.

Line 5 Abstract: Add (PRObabilistic cloud top Phase retrieval for Seviri) to ProPS

Line 30-34: Add full name of GOES: Instead of GOES-R/S, GOES-R series or GOES-16/17/18: Add sensors: ABI and AHI with references, like SEVIRI. GOES -> GOES-R

Line 40:” the Lidar-Radar cloud product DARDAR…”  Add a brief summary for readers who are not familiar with this data, space-borne data derived from CloudSat-CALIPSO, even though the details are followed in the next section but it appears first here.

Line 46-47: For “DARDAR as ground truth” - If the data is temperature only-based, still limitations especially for supercooled and mixed? If any, it would be better to include some discussions on the of the ground truth in the data or conclusion sections.

Line 76-77 Add more details. Did the authors use the cloud top phase info in DARDAR from CloudSat-CALIPSO as is? Assumed there was no further consideration on the lower layer  phase from the active-sensors for passive radiometers like SEVIRI, correct? What exactly mixed phase is defined?

Line 54 and nighttime eval.in sect 8: Have you ever considered 3.9 um or a channel difference including this, particularly for nighttime retrievals?

Line 65:  “DARDAR (liDAR/raDAR, Delanoë and Hogan, 2010)” repeated in Intro and here.

Line 66: Probably need references for these satellites, CloudSat and CALIPSO, although they are well know.

Line 72: MET-9 -> maybe better to write the fullname, Meteosat-9?

Line 79: “…observed (see Mayer et al. (2023) for details).” - more info about the phase data used as a ground truth in this study will be desirable here, briefly from Mayer et al. (2023) if needed.

Line 98: “combinations, at probability distributions are used” – more clarification?

Figure 2: Hope the figure quality with larger fonts can be added in the final version.

Line 135: M2 means the other channel measurement?

Line 157: “most likely”: Any minimum threshold which may give 'uncertain'?

Eq 7: “season” - How was it quantified in the computation?

Line 201: “parameters A introduced above in Sect. 7 “ -> need to be corrected?

Line 203-205: Not very clear. Could you explain more details? It is explaining Fig. 3, right?

Line 207: (BT: - What about putting this acronym to the place where it appears first, and using BT for consistency?

Figure 3: Higher resolution one with bigger fonts, please. The figure caption doesn't seem very straightforward, please more clarify it for better understanding?

Line 2014: “the brightness temperature” -> BT10.8

Line 217: BT -> BT10.8

Line 276: Bayesian retrieval methods  -> specify which cloud property retrievals

Line 376: Fig. 6  -> Figure 6

Line 379: “…retrieval detects (most) clouds which…” Can we think the retrieval method is for cloud detection in the first place, not just cloud phase discrimination?

Line 379: Add the full name for ITCZ, even though most of us know what it is.

Line 523-524: Again, addition to using 3.9 um info, have you ever considered additional environment parameters from ERA-5 such as low level humidity and SST/low level atmos temperature combinations?

Appendix B: I would think this discussion can be part of the main section 8.

References: Cover 1999: Some info still missed.