# RC1

General comments:

The paper describes a 10 - minute merged radar-gauge precipitation data set for South Korea with spatial resolution of 0.5 km and temporal coverage from 2016 to 2024. The manuscript is well written; the methods are comprehensible and the results regarding the data set quality are convincing. The data itself are very useful, well described and accessible.  I have only some minor comments for improvement (see below).

 Response) We sincerely thank the reviewer for the positive and constructive assessment of our manuscript and dataset. We have carefully addressed all the comments. In particular, we removed the unsupported statement in the abstract, revised Figure 2 for consistent notation and clearer data flow, simplified Figure 5, and clarified the reasons for the file organization and the division of the dataset into two repositories. Detailed responses are provided below.

Specific comments:

1. Line 24: “The dataset is evaluated through comparison with conventional geostatistical interpolation methods.” I could not find this comparison in the text. This would be an interesting and necessary exercise if this would be a research paper. Since this is a data set paper this comparison is not absolute necessary but this sentence should be removed.

Response) Thank you for pointing this out. We agree that the manuscript does not present a systematic comparison with conventional geostatistical interpolation methods. Therefore, we have removed the corresponding statement from the abstract (Lines 23-24).

1. Figure 2 can be improved:

In “radar rainfall field” and “gauge rainfall field” the symbols “R” and “G” should be introduced.

Sometimes variables are functions of (t) sometimes not. Be consistent.

Sometimes variables are in brackets some times not. Be consistent.

Once a dot is used for multiplication, once not. Be consistent.

The arrow from the left lower box should go out at the end and input the right box on top.

Response) Thank you for these helpful and detailed suggestions. We have carefully revised Figure 2 to improve the consistency and clarity of the notation and workflow. Specifically:

1. The symbols and have been explicitly introduced to denote the radar-derived rainfall field and the rainfall observed at the i-th gauge, respectively. The radar rainfall corresponding to the i-th gauge location is expressed (lines 130-132, p.5).
2. Time-dependent variables are now consistently expressed as functions of t (Figure2, in p. 6).
3. The use of parentheses in the mathematical expressions has been standardized throughout the Figure2.
4. The multiplication notation has been made consistent throughout the equations (in Figure2).
5. The arrow from the lower-left box has been repositioned so that it exits from the end of the box and enters the top of the “Residual computation” box (in Figure2).

1. Line 194: There are more than 470.000 single files. Experience has shown that handling (e.g. copying) thousands of single files is much more time consuming than handling bigger files. Is this really the best way to store these files for each 10-minute interval separately?

Response) We thank the reviewer for raising this important practical concern. The 10-minute file structure was adopted to remain consistent with the near-real-time production design of the RRBF framework, in which each precipitation field is generated and released as soon as it becomes available (approximately one minute of processing time per 10-minute analysis, as noted in Sect. 1). A similar practice is followed operationally by KMA, which distributes its composite radar reflectivity product (HSR) as individual files at 5-minute intervals. As one purpose of this dataset is to support future operational implementation, we intentionally preserved this per-interval file structure so that it directly reflects the output format of the underlying algorithm.

At the same time, we agree that distributing several hundred thousand individual files is impractical for end users. For this reason, the dataset is not distributed as loose 10-minute files; individual files are aggregated into monthly compressed archives (*.tar.gz) before being uploaded to Zenodo, which also imposes a limit on the number of files per upload. Users therefore download and extract data by month rather than handling individual 10-minute files. We have revised Sect. 4.3 to clarify this design rationale (in Lines 224-229), and the monthly archive structure is described in Sect. 6 (Data availability).

1. Figure 5: What is the additional information when using line-dots in addition to the bar plot. One of the two is sufficient.

Response) We thank the reviewer for this observation. We agree that displaying both the bar plot and the line-dot makers is redundant, as both convey the same annual values. We have revised Figure 5 (p. 13] to show the bar plot only, which is sufficient to represent the interannual variability described in the text.

1. Data availability: The data should not be in two separate publications. Why are they not stored in a single archive?

Response) We thank the reviewer for raising this point. The dataset was divided into two Zenodo repositories because of practical constrains imposed by the Zenodo flatform, namely a limit on the total number of files (<=100) and a limit on the total storage volume (<50 GB) permitted per DOI. Given that the full dataset (2016-2024) consists of approximately 473,000 individual 10-minute precipitation fields at 0.5km resolution, the total volume and file count exceeded the limits allowed for a single Zonodo upload. The dataset was therefore split into two parts (2016-2020 and 2021-2024) to comply with these platform-lavel constraints, rather than for scientific or organizational reasons. We have clarified this rationale in Sect. 6 (Data availability, in lines 420-421).   

General comment:

 This paper describes a tremendous amount of work. It is very well written – and therefore I have only few specific comments. Most of them can regarded as suggestions for future work.

Specific comments and corrections:

(1) When it comes to the interpretation of remote sensing data there is always a debate about the meaning of the term “resolution”. I understand it as “physical resolution”. You mention (line 72 …) a “spatial resolution of 500 m × 500 m” – which I would understand as the gridding of the data.

By comparing radar estimates with high-density station fields (~1.5 km distance), we found that radar estimates are smoother than the station-based estimated, which suggest that the “physical” resolution of the radar estimates is less than indicated.

I think it would be sufficient to mention that the radar data are available on a 500 m × 500 m grid – which does not necessarily represent the physical resolution of the data.

Response) We thank the reviewer for this clarification. We agree that the term "spatial resolution" as used in the manuscript refers to the native grid spacing of the HSR product rather than the physical resolution of the underlying radar measurement, which varies with range from each radar due to beam broadening. We have added a clarifying sentence in Sect. 2.1 to make this distinction explicit (in lines 72-73).

(2) We noticed quite some fundamental changes when comparing gauge data with estimates from single and dual polarized radar data. For the validation period in 2022 this is not important anymore (except Incheon), but in further work you could check there have been larger differences in the time before. It would be interesting to know, when this update happened within 2014(2016) - 2019.

Response) We thank the reviewer for these related and helpful suggestions. We have added a new table (Table B1, Appendix B, p. 20-21) summarizing, for each radar site in the KMA network, the operation start year, the year of dual-polarization upgrade, the antenna altitude, and the frequency band. As shown in the table, most radars were sequentially upgraded from single- to dual-polarization systems between 2014 and 2019, with Baengnyeongdo being the first site upgraded (2014) and Gangneung the last (2019) among the ten operational sites. The Incheon International Airport radar was separately upgraded to a dual-polarization system in August 2022. We have added a corresponding reference to Table B1 in Sect. 2.1, along with a brief description of this upgrade timeline and a note that potential impacts of these hardware and configuration changes on long-term reflectivity homogeneity cannot be fully excluded (lines: 86-90).

(3) Figure 1: Most radar stations are close to the coast, but some are further inland. Since there also altitude-dependent errors it would be interesting to know the altitudes of the station (e.g., in a table, showing also the replacement time (comment 2), or by mentioning the altitude range.

Response) Table B1 shows that the radar network spans a wide range of elevations, from low-lying coastal or airport sites (e.g., Gangneung, 99 m; Gosan, 103 m) to high-altitude inland mountain sites (e.g., Myeonbongsan, 1136 m; Gwangdeoksan, 1064 m), reflecting the mixed coastal and inland siting strategy visible in Fig. 1. We agree that altitude-dependent beam propagation and blockage effects are an important consideration, and we have noted this range explicitly so that readers can relate individual station characteristics to the spatial error patterns discussed in Sect. 5.2.3.

(4) Most gauge stations (AWS) just resolve 0. 5 mm increments, but you apply (e.g., line 142) a “minimum rainfall threshold of 0.1 mm h⁻¹”. I am not sure if this is fully consistent (but I would also not know how to make it better, without losing the light rain events).

Response) We thank the reviewer for this careful observation, which prompted us to clarify the basis for this threshold. The value of 0.1 mm h⁻¹ was applied uniformly across both the 10-minute and 60-minute analyses for consistency. This value corresponds exactly to the finest resolvable rainfall rate achievable with ASOS gauges (0.1 mm min⁻¹) when accumulated over a 60-minute window, and is therefore physically meaningful at that timescale. For the 10-minute accumulations, the minimum non-zero rainfall rate that can actually be reported by any gauge in the network is higher, owing to gauge resolution (0.6 mm h⁻¹ for ASOS and 3.0 mm h⁻¹ for AWS, once 10-minute accumulations are converted to hourly rates). As a result, the 0.1 mm h⁻¹ threshold functions equivalently to a simple non-zero criterion at the 10-minute timescale, while retaining direct physical meaning at the 60-minute timescale used in Table 2. We have revised Sect. 3.2 to clarify this distinction explicitly (Lines: 148-154).

(5) In Line 196 and Table 1 you use “Local Standard Time (LST)”, but in Line 74 you have already introduced the “Korea Standard Time (KST)” – which seems to be the same. I would suggest to use just one name throughout.

Response) We thank the reviewer for catching this inconsistency. We have replaced all instances of "Local Standard Time (LST, UTC+9)" with "Korea Standard Time (KST, UTC+9)" throughout the manuscript, including in Table 1 (p. 8-9) and Sect. 4.3, to ensure consistent terminology with the definition given in Sect. 2.1 (in line 207).

(6) Line 214: “(summer 2022” – there is just a missing “)”.

Response) We have corrected the section headings to “ …method (summer 2022)”. (line: 231)

(7) Line 222: “60 min accumulated rainfall” are these conventional hours, or can this be any 60 min interval? In the former case, you could consider just using “hours” ?

We thank the reviewer for pointing out this ambiguity. These are not conventional clock hours. The 60-minute accumulations are formed by summing six consecutive, non-overlapping 10-minute values from the underlying 10-minute time series, and the starting point of each 60-minute window is not necessarily aligned to the top of the clock hour (e.g., a window may span from 00:30 to 01:20). We have retained the term "60 min accumulated rainfall" throughout the manuscript and added a clarifying note at first use (Sect. 5.1.1) to make this explicit, rather than adopting the term "hours," which could incorrectly imply alignment to standard clock hours. (Lines 240, 333).

(8) Figure 6: The distribution in the wet year 2020 shows still considerable structure. Could this be related to typhoons?

Response) We thank the reviewer for this insightful comment. We agree that the pronounced spatial structure in the 2020 annual precipitation field may be partly related to typhoon-induced rainfall. The year 2020 was characterized by an exceptionally prolonged summer rainy season, followed by several typhoons that affected the Korean Peninsula in August and September. In particular, Typhoons Maysak and Haishen made landfall near Busan and Ulsan, respectively, and subsequently tracked through Gyeongsang and Gangwon Provinces, producing substantial rainfall over southern and eastern Korea, including the mountainous regions of Gangwon Province (KMA National Typhoon Center, 2021). These typhoon-related rainfall events likely contributed to some of the pronounced spatial features observed in the 2020 annual accumulation field.

However, we consider that the spatial distribution in 2020 cannot be attributed solely to typhoons. The annual accumulation integrates precipitation throughout the year, and the exceptionally wet conditions in 2020 were also strongly influenced by the prolonged summer rainy season and several other heavy-rainfall events. We have therefore added text in Lines 310-316 of the revised manuscript to clarify that the spatial structure in 2020 likely reflects a combination of seasonal precipitation variability, orographic and coastal effects, and typhoon-related rainfall.

(9) Line 357: You mention the bias on Jeju Island. Isn’t there the highest mountain in entire South Korea? I might be wort mentioning this.

Response) We thank the reviewer for this helpful suggestion. We have added a brief note in Lines 384-388 of the revised manuscript that this bias likely reflects beam blockage near Hallasan (1,947 m), the highest mountain in South Korea, located near the center of Jeju Island, which requires the HSR algorithm to substitute higher-elevation scans and can lead to underestimation in the Z–R estimates. The RRBF estimates are less affected, since gauge observations are directly incorporated into the bias correction.

(10) Table 2 shows results for different threshold values, but I could not find a discussion in the text. It would be nice to summarize these results.

Response) We thank the reviewer for pointing out this omission. We have added a summary of the threshold-dependent results in Lines 346–351 of the revised manuscript. Specifically, we note that as the rainfall threshold increases, the negative bias of the Z–R estimates becomes progressively larger (NBE from −13.13% to −30.73%) and their correlation with gauge observations decreases (COR from 0.7703 to 0.6852), reflecting greater underestimation by radar-only estimates during higher-intensity rainfall. In contrast, the RRBF estimates maintain a bias close to zero across all thresholds (NBE within ±3.3%) and show only a modest decline in correlation (COR from 0.9251 to 0.8699), demonstrating that the benefit of gauge-based bias correction persists even at higher rainfall intensities.

(11) Line 388: I assume that the partition into two data sets is due to the high data volume?

Response) Yes, the reviewer's assumption is correct. As clarified in the revised Sect. 6 (line 420-421), the dataset was split into two Zenodo repositories due to platform-imposed limits on the number of files and total storage volume per DOI.

(12) Do you plan to update the dataset in the future? If yes then it would be good to indicate this.

Response) We thank the reviewer for raising this point. At present, we do not have specific plans to extend the dataset beyond the 2016–2024 period covered in this study.

All the best for your future work!

Response) We sincerely thank the reviewer for their thoughtful and constructive comments, as well as their kind wishes. We greatly appreciate the time and care taken in reviewing our manuscript.