The authors are in a unique position to have access to 1-min rainfall data for a large number of stations (>60,000) in China.  The authors used the 1-min data to compute the peak 30-min intensity and storm energy in accordance with recommendations from the RUSLE/RUSLE2 manual. Based on what is presented, the data are of good quality and data analysis were rigorously undertaken.  Compared to previously published R-factor maps, the one produced by the authors has the potential to be regarded as the definite map of the R-factor for the period considered (2014-2022), because 1-min data were used to compute EI30 directly for so many stations.

As a data product publication, the quality of data product, particularly the margin of error and associated uncertainty are of great importance.  The authors compared their R-factor map to that from Panagos et al (2017) and Yue et al. (2022), and concluded that the previous R-factor values for this region (China) were overestimated by 31%-65%.  While Panagos et al. (2017) used hourly rainfall data and bias correction was crude, Yue et al (2020) and Yue et al. (2022) addressed the effect of data resolution thoroughly and  Yue et al (2020) used in fact 1-min data for 62 sites in China to bias-correct estimated R-factor values.

The main reason for me to recommend Major Reason is that authors need to explore/explain why there are systematic differences between the erosivity map submitted and that published in Yue et al. (2022), and the implications of using 1-min data for only 10 years for the uncertainty associated with computed R-factor values.

It is imperative to compare EI30/R-factor values for the 62 sites used in Yue et al. (2020) that underpins the map in Yue et al. (2022).  The record length for many of the 62 stations was much longer than 10 years.  Could the record length have contributed the discrepancy between these maps? Could rainfall erosivity have decreased over the past 10 years?

For R-factor calculations, procedures, definition, and equations, recommended for RUSLE2 were closely followed, which is great.  However, the record length for the calculated R-factor (10-year) was short. The recommended record length is minimum 20 years to have a reliable estimate of the R-factor.  I would actually argue for 30+ years just like the mean annual precipitation to define 'climate' of a region.  Too short a period of 10 years is particularly relevant and problematic for areas of low precipitation and fewer erosive rainfall events in western China for instance.

Authors either wait for 10+ years to produce a more reliable R-factor map or compare and explain the discrepancy with a view to improve the uncertainty associated with the latest attempt at erosivity mapping for China in spite of the fact that 1-min rainfall data have just become more widely available for the past 10 years.

Minor and editorial comments:

I have attached an annotated pdf with minor comments and highlighted where attention to English expression, and grammar is required.

References:

Yue, T. et a. (2020) Effect of time resolution of rainfall measurements on the erosivity factor in the USLE in China. International Soil and Water Conservation Research. 8: 373-382. 

The manuscript “Rainfall erosivity mapping in mainland China using 1-minute precipitation data from densely distributed weather stations” presents rainfall erosivity maps over China where 1-minute precipitation data was used. While the overall content is suited for this journal, the manuscript requires quite some work before further consideration for publishing. Remarks:

• The manuscript needs to be checked for grammatical errors. The introduction, for example, needs further work.
• The authors only present their erosivity product (in countable figures, in fact) without discussing the implication their new maps would have compared to the already available ones, e.g. on soil/land conservation practices, etc. The whole manuscript therefore just reads like a short report presenting a rainfall erosivity map (and some brief inter-comparisons) over China without really providing the reader with any new information.
• The authors only present a qualitative assessment of their grid maps and do not provide any quantitative measures in their performance evaluations. They need to quantify (in numbers; say in terms of correlation coefficient, rmse, bias, …) how their erosivity product compares to the others.

Specific comments:

L20:  “… overestimate China’s mean annual rainfall erosivity by 31%—65%,  …” - So here you assume your product is the reference?

L37: “… challenge due to the unrealistic for a dense  …” – Grammar.

L38: “… simply the calculation” – do you mean ‘simplify the calculation’?

L39-40: “Various E-I models have been developed, employing linear, polynomial, exponential, logarithmic, and power-law functions…” 

Various EI models? Where is the literature on these developed models? You need some citations here.

L41-43: “Studies have indicated that E values derived from 1-hourly in-situ precipitation data tend to underestimate those obtained from 1- minute data by approximately 10% (Agnese et al., 2006; Yin et al., 2007) …”

Maybe expound on this. According to the cited studies, why does 30 min precipitation underestimate the kinetic energy compared to 1 minute rainfall data?

L44-46: “…Compared to the radar remote sensing-based E values, the multi-year averaged annual rainfall kinetic energy calculating using E-I method was smaller with biases ranging from -6.17% to -12.5% across distinct regions worldwide.”

So the authors of the cited literature assumed RS-based E values were correct? It is known that remote sensing (RS) products can have shortcomings, as they rely on non-exact methods that are developed to translate the remotely-sensed signals to a variable of interest. Why did they use data derived using RS methods as the reference?

L48: “…When the in-situ data is used, I30 tend to be increasingly underestimated with increasing time intervals of precipitation data”   - Citation/justification needed

L49: “It has reported  …” >> it has *been* reported

L50-51: “… Consequently, the in-situ precipitation data with 1-minute temporal resolution are the best suitable data for deriving I30 of an rain event.  …”

an >> a

The authors just state this without providing any reasons/explanations; why is precipitation at 1-min resolution suitable?

L50-56: “Recent years,  there occurs some gridded precipitation  datasets with high temporal resolution. However, it should be caution  when the gridded data are directly used to calculate I30, because large underestimation in I30 has been widely reported . “ – *in recent years ; this whole section needs to be revised/rephrased (and checked for grammatical mistakes).

L58-59: “Based on the analysis presented, the following conclusion can be drawn: The  bias in estimating the I30 of individual rainfall events is significantly larger than that for estimating E under the latest available datasets. The estimation error of I30 is the most crucial source of inaccuracies in determining rainfall erosivity.  ” – based on the analysis presented where? This section needs to be rephrased/restructured as it is difficult to follow.

L71: “…different precipitation events” – such as?

L71-… : “Thus, this study aims to develop …”

Too short. The authors need to introduce the study with a bit more detail.

L79: “…integrity level exceeding 90% …”

What is this integrity level exactly? How is it defined?

L90: Fig 1a: “number of staton” >> number of stat*i*ons

L98: “…interpolated spatially into 0.5° grids by using the Thin Plate Spline method …”

Any justification why this interpolation method was used in the CMA grid data? … the authors use Kriging instead when spatially interpolating their grid erosivity products; have they considered using a Spline-based method similar to CMA? What is the implication of using one over the other especially in relation to erosivity spatial interpolation?

Also, how did the authors reconcile the 0.25deg resolution of your grids (see section 2.1.1) with the 0.5deg CMA grid in your comparative analyses?

L130-133: “…cold season (January to March, November to December). Subsequently, the median and standard deviation of event rainfall erosivity are computed for the warm and cold seasons…”

The authors define 2 cold season periods. Are the 2 cold seasons separated/differentiated when performing these std dev. and median analyses?

L153-156: “Kriging … the impacts of the spatial interpolation method on the accuracy of the R factor map in these regions …”

Other than kriging, have the authors tested other interpolation methods on their ‘quality-checked’ precipitation datasets? E.g. the spline-based method used in the CMA grids.

L169-172: “…(ERA5) precipitation data … This study used this released gridded dataset to calculate the mean annual rainfall erosivity from 2014 to 2020 for the Dawang-Chayu area …”

This ERA5 dataset should be described in the earlier data description section. Also, why didn’t the authors use ERA5-Land, which is at a relatively higher spatial resolution than ERA5?

L180-…: “The newly generated R factor map over mainland China is compared with the existing maps… When compared with the map developed by Yue et al. (2022), the correlation is good overall, but our calculated values are significantly lower.”

This is too short. Correlation coefficient (r) and the coefficient of determination (r^2) only show the 1:1 agreement between data and does not really provide the reader with any new information. Needs further discussion on why the high correlation with both products (Panagos et al., and Yue et al.’s product), and why the underestimation relative to Yue et al.’s product.

Also, the authors just mention the high correlation without quantifying the correlation (r) value. There needs more comparisons and discussions where other performance metrics (such as RMSE, Bias, etc.) are also included.

L195-…: “…watersheds, there is no consistent pattern (Figure 5b ).” - Again, too short. Why the inconsistency between the products? The authors should discuss why their results are different from earlier works/products.

L199-203: Figure 5a is not described in the figure caption. Also, this figure (Fig5a) should part of the site[s] description section. It is not one of the results from this study.

The authors also define the interquartile range, but do not use this anywhere in the text. Some details in the caption are also unnecessary – e.g., the IQR description and “the plot and line box are the average and median values, respectively “ --- this is indeed how box plots are designed/constructed so no need to mention the obvious.

L205: https://doi.org/10.n888/Terre.tpdc.301206 - This doi is not accessible