The study emphasizes the importance of accurately assessing probable maximum precipitation (PMP) for high-risk water infrastructures, water resource management, and hydrological hazard mitigation. Using a fine spatiotemporal resolution precipitation dataset, the authors employ the improved Hershfield method to reveal the spatial distribution of 1-day PMP in China. They note a general decrease in national PMP from Southeast to Northwest. The research highlights the dominance of precipitation variability in North China and intensity in South China, with climate change expected to lead to a widespread increase in PMP (>20%) across the country under certain scenarios. Next, I offer comments indicating points where some clarifications could be helpful.

Section 2.2:

Note that it is not only the different products that can result in different estimates but also the resolution and methods used. Here, attempting to validate extrapolated results from the gridded HRLT resulted in bilinear interpolation adding extra uncertainty. The interpolation itself changes the values; known effects include smoothing out extremes (see e.g., 10.1175/JAMC-D-20-0259.1 for similar cases of regridding data). Additionally, the in-situ data are point measurements and can exhibit different behavior than precipitation at 1km x 1km. The authors could comment on potential effects relevant with their methodological choices.

Section 2.3:

There is a long literature regarding the evolution of the statistical method of PMP. I suggest the authors see the review of Salas et al. (2020; 10.1061/(ASCE)HE.1943-5584.0002003).

Section 2.5:

Maybe I'm missing something here, but are the CMIP6 projections used raw? Daily projections of precipitation can have very large biases and probably cannot be used for this purpose. If these are bias-corrected projections, then what type of bias correction method was used? Since bias correction methods can also affect the results (see e.g., Luo et al. (2021; 10.1002/joc.7294, or Dong & Dong (2021; 10.1007/s00382-021-05773-1)

Section 3.1:

"...which is related to the systematic overestimation of GPM IMERG product over China." You can also check the review by Tang et al. (2020; 10.1016/j.rse.2020.111697) on the topic.

I am not sure if the linear regression presented in Figure 4a, b, c has meaning. It is clear from the figures that the linear regression line is not a good model. The nonlinearity is apparent from the cloud of points. Thus, conclusions and discussions based on the fitted linear regression can be misleading.

Section 3.3:

Results shown here must be coupled with results from studies validating the performance of CMIP6 over China. As mentioned before, more details should be given on the projections used (raw, bias-corrected, what methods were used for BC, etc.). (see also Zhu et al. (2020; 10.1007/s00376-020-9289-1)

Section 4.1:

"Yang et al. (2018) discussed the potential of satellite data to estimate PMP in poorly gauged regions." Indeed, all these gridded products (and not only satellite-based) can offer an option when in-situ data are not available, but it must be clear that there are large differences in extreme precipitation values these products report (see Rajulapati et al. (2020; 10.1175/JHM-D-20-0040.1). The fact is that no single product can claim the best performance, and it seems the best option we have is to use many of them and quantify the uncertainty.

Section 4.3:

One additional point to mention could be the general limitations of PMP methods. What exactly do these methods offer compared to probabilistic methods? The literature on using probability methods in China is pretty rich (see, for example, a review by Gu et al. (2022; 10.1016/j.advwatres.2022.104144).

Overall, I believe this is a well-written and well-organized study that has its place in the literature, pending some minor amendments to clarify some points. I enjoyed reading the paper, and if PMP methods are still officially used in engineering practice in China, then this could be a useful study for practitioners. Most of my discussion points are optional, and the authors can skip them. The most important ones are related to clarifications about the CMIP6 outputs used and whether they have been biased corrected, etc.

Simon Michael Papalexiou

This study employs a high-resolution (1d, 1km) precipitation dataset to estimate spatially distributed PMP across China, and compares these estimates with two benchmark datasets. It focuses on analyzing the trend of PMP over time, rather than providing a static estimate, and extends its investigation into future changes in PMP based on climate model outputs. The discussion section is thorough, touching upon several interesting topics. Overall, the manuscript is well-written, offering insightful findings and valuable data pertinent to China. I have no significant concerns but would like to offer a few suggestions for consideration in the manuscript revision:

1. Line 15: The manuscript mentions that the dataset integrates observations with machine learning algorithms (Section 2.1). Including a brief explanation (e.g., a few words) here could offer valuable context to readers.
2. Paragraph Structure: The manuscript frequently utilizes long paragraphs. I suggest breaking these into shorter, more digestible sections to enhance readability.
3. Figure 3: To aid in comparison, I recommend employing a consistent color scale across all sub-figures.
4. Figure 6: The patchy patterns observed, particularly in panels (d) and (e), warrant further explanation. Clarifying these patterns could enhance the reader's understanding of the underlying data and analysis.
5. Figure 8: The substantial discrepancy between UserSMIP and LFMIP-pdLC raises questions about the reliability of the findings. Incorporating the other reviewer’s (Simon) comment for CMIP6 validation and bias correction might address these concerns and strengthen the manuscript.

Guoqiang Tang