The authors present a high-spectral-resolution clear-sky ultraviolet radiation dataset for China, which features unprecedented 0.5 nm spectral sampling across the 280-400 nm range. The integration of reanalysis data with physical radiative transfer model (SMARTS) represents a robust approach. Its credibility is further strengthened by comprehensive validation against ground-based CERN stations and CERES satellite products. The dataset appears to be a valuable resource, particularly given its spectral resolution and spatial coverage over China. The manuscript merits publication after addressing the following minor reversion.

Abstract: Please correct "Earth’s Radiant Energy System" to the full name "Clouds and the Earth's Radiant Energy System".

Section 2.2: The sun-earth distance correction factor have two different letters, “S” and “s”; standardise the format.

Section 2.3: Phrase “concentration and distribution” is redundant, as the distribution characteristics of errors already encompass the concept of central tendency.

Figure 3: Part of the site name is obscured in Figure 3.

Table 1: How are the variables presented in Table 1 incorporated into the SMARTS model?

Figure 4: The unites of solar radiation in Figure 4 are inconsistently represented, with both “Wm^-2” and “W m^-2” being used.

Figure 5: The same issue as in Figure 4.

Line 263: “The normally strong aggrement ...” is an inaccurate statement.

Section 4.2: The conclusion could be enhanced by providing additional analytical insights regarding the distinctive features and relative performance of the two products, which would help users make informed selections.

Line 315: “-50.06% to -40.90% per AOD unit” change in UV irradiance require clarification.

Section 4.4.1: Replace the imprecise term “near 400 nm” with the scientifically accurate designation of either “UV-A range (315-400 nm)” or “long-wave interval (380-400 nm)”.

Section 4.4.2: The comparative analysis of the three regions would be strengthened by establishing a unified reference baseline, such as expressing their UV radiation levels as deviations from the national average value.

Figure 12: Please briefly quantify the radiation differences between transitional seasons (Spring/autumn).

Line 389: “considerable altitude” is not appropriate.

Line 431: “coinciding with” inappropriate wording.

This study presents a 10 km high-resolution hourly surface solar ultraviolet (UV) radiation dataset under clear-sky conditions across mainland China from 1981 to 2023. The dataset is reconstructed using the SMARTS spectral model based on ERA5 and MERRA2 reanalysis data. To validate its performance, the authors conduct comprehensive statistical comparisons with the CERES UV product, demonstrating clear advantages in terms of spatial-temporal resolution and data quality. Using this dataset, the study further analyzes the spatial distribution characteristics of clear-sky UV radiation intensity over China at multiple temporal scales, including hourly, daily, and annual means. It also discusses possible reasons for poor fitting at specific stations, as well as the seasonal and interdecadal variations in UV radiation. The results fill a data gap by providing a high-resolution, long-term, and hourly UV radiation dataset for China, which holds value for future studies on the effects of UV radiation on human health and the natural environment. The manuscript is generally well-organized, and the methods are rigorous and sound. My comments are as follows:

Major Comments:

(1) The validation of this dataset relies on 37 stations, which are unevenly distributed and heavily biased toward inland areas. In contrast, coastal regions—where ultraviolet (UV) radiation interacts with ocean-atmosphere processes (e.g., sea salt aerosols, high humidity, and differences in surface albedo between land and ocean)—are severely underrepresented in the validation network. This introduces uncertainties regarding the dataset’s accuracy in coastal regions, which are crucial for both terrestrial and marine-related applications. If available, the validation dataset should be expanded to incorporate UV observation data from coastal meteorological stations or marine research platforms. Additionally, a sensitivity analysis should be conducted targeting coastal-specific parameters.

(2) The manuscript identifies aerosol optical depth (AOD), total column ozone (TCO3), and forecast albedo (FAL) as key drivers of UV radiation (Section 4.3), but not account for the modulation of these drivers by ocean-atmosphere feedback processes in coastal regions. In Sections 2.2 and 3.2, it is necessary to clarify whether the MERRA-2 AOD data include sea salt aerosol components, or if terrestrial AOD data are used for coastal grids. If sea salt aerosols are not explicitly modeled, the limitation should be discussed, and adjustments for future research should be proposed. Finally, it is recommended to cite relevant literature on ocean-atmosphere interactions and UV radiation to strengthen the scientific basis.

Minor comments:

(1) The term "Methodologys" should be corrected to "Methodology"?

(2) Line 209: The phrase “compared with 196,170 observed data points from 37 CERN…” is unclear. A clearer expression could be: “Out of the 196 original observed data points from 37 CERN during 2005–2013, 170 were selected after applying …”

(3) Line 245 and elsewhere: Ensure that all subplot borders are clearly visible and consistent in line width throughout the figures.

(4) Line 325: The color differences among the lines in Figure 8 are not sufficiently distinguishable; please enhance the contrast or adjust the color scheme for clarity.

(5) Line 362: In Figure 10(a), only the top layer is clearly visible, while the lower layers are largely obscured. As a result, this subplot may not be very informative. The similar Figure A1 in the appendix presents the information more effectively.

(6) Ensure that all energy flux units are uniformly expressed as "W m⁻²".

(7) Add the definition of FAL (Forecast Albedo) and explicitly state whether it is land-focused.

(8) Provide a table to summarize the characteristics of the validation stations, including latitude, longitude, elevation, and climatic conditions.

Comments on “First High-Resolution Surface Spectral Clear-Sky Ultraviolet Radiation Dataset …” by Qi et al.

The development of a reliable Solar Ultraviolet (UV) radiation dataset is undeniably crucial, underpinning a wide range of applications from biophysical modeling to public health assessments. The dataset presented by Qi et al., covering mainland China from 1982 to 2023 and generated using the SMARTS (Simple Model of the Atmospheric Radiative Transfer of Sunshine) spectral model, represents a significant and valuable contribution to the community. The validation results against ground observations, which show a strong performance, are commendable.

However, two major points regarding the methodology and presentation of results require clarification and revision, as detailed below.

1. Precision and significance of reported digital numbers

The study frequently employs an excessive number of digital numbers (e.g., reporting the correlation coefficient R as 0.919). While precision is generally desirable, the number of significant digits must be meaningful and justified by the data quality and the inherent uncertainty of the model and observations.

Please review the entire manuscript and uniformly apply a statistically meaningful number of significant figures to all reported metrics (e.g., R, RMSE, bias, and model parameters). For instance, given typical measurement and model uncertainties, reporting R to three decimal places may imply a false level of precision where R=0.919 is not meaningfully distinct from R=0.923. The chosen precision should reflect the uncertainty of the estimated value.

1. Rationale for linear regression methodology

The linear regression analysis, as displayed in Figures 4(a), 5(a), 6(a)(b), and 7(a)(b), appears to be conducted directly on the entire scatter plot of data points. When dealing with large sample sizes, this approach can lead to a regression bias that is unduly influenced by the density distribution of the samples, particularly at the extremes.

I recommend considering an alternative or supplementary approach: first compute the conditional mean (i.e., bin the observed data along the x-axis and calculate the mean of the modeled data for each bin) and then perform the linear fit on these conditional mean values. This technique can effectively suppress the sampling bias and provide a more robust characterization of the central tendency relationship between the modeled and observed values, especially when the sample size is large. A discussion on the impact of this methodological choice should be included.

Other minor comments

Line 145: The physical significance and unit of the parameter z should be explicitly stated in the text where it is first introduced or used.

Figure 13: The interpretation of trends, particularly in long-term, non-stationary time series, is highly sensitive to the chosen data range and the trend methodology. The authors should acknowledge this sensitivity and briefly discuss the potential impact of their chosen methods, perhaps referencing relevant literature. (e.g., Zhao et al., On the trend, detrending, and variability of nonlinear and nonstationary time series, PNAS, 2007, 104 (38) 14889-14894).

Code Sharing: To maximize the utility and reproducibility of this valuable work, the authors are strongly encouraged to convert the SMARTS model implementation used for this study into a Python-based package and share the code with the community. Python facilitates easy integration with modern data science and AI/Machine Learning models, significantly enhancing the model's accessibility and impact.