Dear Authors,

here are some comments I had to post to finalize to review process which took than due. Really sorry for this.

General comments

As clearly outlined in the Abstract, the objective of the study is threefold: to evaluate hydrometeor partition ratios retrievals, to exploit the combination of Dual Frequency (DF) satellite radar and Dual Polarization ground radar (GR) observations for estimating HPR based on satellite DF observations and to improve ground-based radar estimates of HPR.

NEXRAD S-band polarimetric radar observations are matched with satellite radar (SR) dual-frequency measurements of the Global Precipitation Measurement (GPM) satellite mission collected from 2014 to 2023.

The data, regardless of the platform (ground- or satellite-), undergoes a complex processing procedure to compensate for multiple error sources before being resampled into a common spatial domain.

The work is articulated in the following main steps:

1. Dual polarization ground-based radar data (NEXRAD) and dual-frequency Satellite-based radar data (GPM-DPR) are both resampled in so-called superobbed volumes to make the data resolution comparable. This aggregated volume contains hundreds of high-resolution radar gates.
2. The high-resolution radar gates contained in the superobbed volumes are classified with the Hydrometeor Classification algorithm proposed by Park et al. (2009) which assigns a dominant hydrometeor type (e.g. light rain, snow, hail, etc.).
3. For each aggregated volume S_i, it’s determined the number of pixels N(HM_k) in which each hydrometeor class HM_k is dominant
4. For each class, the quasi-HPRs (qHPRs)are calculated as N(HM_k) normalized by the total number of valid pixels in the superobbed volume.
5. These qHPRs are subsequently used both to train and to validate the HPR models on superobbed volumes from SR and GR observations. Consequently, the qHPRs represent a fundamental component of the study.

Major comments

This methodological choice raises a concern regarding potential circularity:

• The HPR estimates from GR are evaluated against qHPRs that are themselves derived from the same GR data (at higher resolution), which may lead to an overestimation of model performance.
• There is a risk that the HMCᴾ model merely reproduces the classifications already present in the GR dataset, rather than demonstrating genuine improvement or generalization beyond the initial HMC algorithm.
• As a result, the validation lacks independence, which undermines the robustness of the conclusions regarding the model’s effectiveness in the GR context.

To mitigate this issue, it would be advisable to use qHPRs solely for training, and to validate the GR-based HPR estimates using independent data sources (e.g., in situ observations, radiosonde profiles, or alternative/independent radar systems).

Minor comments

1. Abstract. In my opinion, there are too many acronyms in the abstract (about 40 in 15 lines), which makes it difficult to read. The abstract is the first thing people read. Please try to simplify it while ensuring it remains effective.
2. Introduction, pag. 1, line 18. I question the use of the adjective “fundamental”. I'm not sure that HMC can really be considered fundamental to QPE at least at the range gate level. Do you mean from an operational or scientific point of view? Who actually applies HMC before estimating QPE? In stratiform precipitation, the application of VPR correction is much more important than any HRM classification. In convective precipitation, what is the overall impact of excluding certain hail spot from the rain field?
3. To make the reading easier I suggest to focus the manuscript on the proposed methodology, moving the description of GR and SR data pre-processing (Sections n. 2.1, 2.2) to the Appendix.