Spatially-explicit monitoring of vegetation and soil fractions is critical for understanding terrestrial ecological processes. There are thus many global vegetation fractional cover products including ENVISAT, CYCLOPES, GEOV, MuSyQ, GLASS, and CGLS. The authors provided a unified monthly fractional vegetation-soil nexuses product during 2001-2022 with MESMA. As a user, we would like to rely on more precise satellite-derived fraction data. Therefore, I pay more attention to the credibility of the new fraction products presented in this paper. This paper merely presented the evaluation of the estimates of vegetation and soil fractions datasets in Section 3.1 and Section 4.1, respectively. However, this validation is inadequate. It’s difficult to persuade me that the presented product is superior to existing fraction products. Thus, further validations are required.

MESMA needs a large library of endmembers representative of each ground component. In Section 2.2.1, the paper depicted the selection of endmembers. I suggest that supplement some RGB images of these selected pure pixels.

Please clarify the reason that the mean error and mean absolute error were used as accuracy metrics.

Please provide additional information on the source of GLCVRD, such as GLCVRD (Bruce Pengra et al., 2015).

Could you please provide an explanation for the red dot appearing at the top of Figure 3?

Could you please provide a detailed procedure for fractional vegetation-soil estimates compared to GLCVRD?

Line 332 “estimates vegetation and soil fractions”: estimates of vegetation and soil fractions

Monthly vegetation and soil fractions products are important to understand the global landcover change and evaluate the impacts of climate change and human activities on the terrestrial ecosystem. The topic is interesting and the MESMA algorithm may be a practical method for decoupling the mixed pixels. I acknowledge the potential of the algorithm; however, I have some major concerns regarding the dataset itself, including its significance, validation methods, as well as the overall organization of the article. Given that the ESSD journal focuses more on the dataset, I am unable to provide a positive evaluation.

1. For significance and usefulness, the product is derived using the MODIS data with a resolution of 500m. Actually, there are a lot of available global or regional higher-resolution land cover products and FVC products. As a user, why do I have to select your product for the analysis? I think using higher-resolution products can acquire more reasonable results. There is not enough evidence in the article that convince me of the indispensability of your dataset at this stage. More content indicating the importance should be included in the Introduction.

2. The current classification system comprises only five types. While I understand that the algorithm proposed by the authors can effectively decompose mixed pixels, I am not sure if it is enough for the practical analysis only based on these categories.

3. The superiority of your dataset is only proved by the comparison with the NDVI and MOD44B product and I think is not adequate for the validation and drawing a reasonable conclusion that your model is more accurate. 

4. I think the accuracy of the algorithm shown in Fig.3e-h is not promising. For the BS/DA/IS type, when the reference values are low, why does the algorithm show a pronounced overestimation? I think the authors should explain it in the Discussion. Comparison with other methods or products is necessary here to prove the superiority of your dataset. Additionally, high R-squared values for the IS type are likely attributable to the data being excessively scattered along the X-axis and more samples should be included for this type. 

5. The current analysis in the paper is predominantly focused on the Chinese region and its surroundings. Moreover, much of the work is centered around validating existing conclusions. Given that the authors have generated long-time-series global products, it is suggested to incorporate more globally relevant and newly discovered findings.

6. The organizational structure of the paper needs revision. Currently, the Method, Results, and Discussion sections are intermingled with contents from other sections. For example, Lines 150-153, should be moved to the Introduction; Lines 312-313, Line 318-320 explain the further reasons, moving them to the Discussion should be better; Section 4.1 just describes the intercomparison of different products or methods, instead of discuss the potential reasons, I think it should be described in the Results section. 

Some technical problems:

Lines 44-45: The full name of PV and NPV should be presented for the first time.

Line 72-73: indexes-->indices; leaf area index (LAI) should be a structural variable for the vegetation instead of vegetation index; The abbreviation for Normalized Difference Vegetation Index (NDVI) should be mentioned here. 

Line 80: human -->human activity?

Lines 138-140: the introduction of the spectral mixture analysis model should be moved to the Introduction section.

Lines 153-155: You only mentioned the applicability of this classification system to tropical rainforests and drylands. Its suitability for global application is unclear.

Figure 2: The wavelength of B1-B7 should be specified. 

Line 204: MESMA has already been declared in the Introduction, so it is not necessary to use its full name here.

Line 137 and Line 231: RMSE is defined twice. Different subscripts should be used for distinction.

Lines 236-245: The introduction of the Mann-Kendall test should not be presented here, because the methods section should emphasize your own work. It is suggested to move it to the supporting information.

Line 262: “reasonably satisfactory” is not an object expression. Only presenting the numbers is enough here.

Lines 264-266 should be moved to the Discussion.

Line 274-277: Where is Figure 3 i-k? The full name of LAI should be mentioned here.

Line 294: ×106 km2 -->×10⁶ km²

Line 312: result from -->results from

Line 378: RESE -->RMSE

Sun et al. generated global monthly vegetation and soil fraction maps during 2001-2022 using a spectral mixture analysis method. Application of this method in global scale is interesting and deserves a scientific publication. However, the proposed product at current stage is not enough to publish on ESSD and the accuracy of this product is not very well demonstrated, from the perspective of a dataset and users. There are some main concerns:

1. The selection of endmember spectra is critical in this method. The authors establish a library of endmember spectral using a nested framework, combined with MODIS derived endmember spectra used in previous studies. However, if the proposed method is valid enough, why is it necessary to use those from previous studies? What are the differences between spectra decided in this study compared to existing ones? How about the proportion of endmember from your method and previous studies in the final library? Furthermore, the authors used a hierarchical clustering method to generate sub-groups of endmembers. This clustering method is quite dependent on the input parameters. Among a lot of clustering algorithm, why did you select this one? How about the input parameters, as well as the performance of its accuracy?
2. Only five endmembers are select to represent the surface. More evidences are necessary to demonstrate the selection of five endmembers, e.g., group of current land classifications.
3. The validation step is very important to convince the users to use your datasets, instead of other existing products. However, the validation is not enough. For example, the authors used GLCVRD reference dataset which was produced in 2010 to validate the new products during the period during 2001-2022. How did you consider the inconsistency between time period and land changes during this long period? I would suggest the authors to improve the validation section, either by cross-validation with other existing products or by independent datasets.
4. The structure of this manuscript should be re-organized, especially the results and discussion. Put all figures in results section. In discussion section, I expect more deep analysis and discussions on the intercomparison with existing products, the advantages and disadvantages of the new products, and uncertainties.

Some minor comments:

• Line 85-86: this is not the key scientific question of this study. More emphasis should be started from issues of existing global vegetation/soil fraction products, instead of application of spectral mixture analysis method in global scale.
• Line 135: Spectral mixture analysis, ‘s’ should be capitalized.
• Line 166: what is the resolution of this land cover product? There are multiple land cover products, why do you select this one? How did you overlap the climate classification zones and the land cover products when the resolution is quite different?
• Line 180-181: While ***, especially in vegetation growing seasons. This sentence seems not complete.
• Line 192: 35 GV spectra, should be PV spectra?
• Lien 212-213: how long it took to generate one global map on GEE?
• Line 260: It is strange that Sahara Desert and polar regions had higher RMSE. These regions were well-known for unique land cover type.
• Figure 3.e-h are not enough to support your conclusion.
• Section 3.1: add more analysis for pixels that were estimated inaccurately.
• Lie 358: indicate the resolution of MODIS pixel.

This paper aims to provide a globally comprehensive record of monthly vegetation and soil fractions during the period 2001–2022 using the spatiotemporally adaptive MESMA methods at the powerful Google Earth Engine (GEE) platform. However, in my view, some issues should be resolved. Please find my detailed comments below.

The overall structure of this article appears to be somewhat confusing. I recommend considering adjustments to enhance the clarity and coherence of the presentation.

In your paper, you have utilized Köppen-Geiger climate classification maps Version 1, while Version 2 is also available at https://www.gloh2o.org/koppen/. Could you please provide some insights into the rationale behind selecting Version 1 over Version 2? Additionally, considering the abundance of global land cover products, what motivated the choice of MCD12Q1? 

In this paper, authors selected the typical sites employed for standardized endmembers selection based on 2020 MCD12Q1. Could you elaborate on the decision-making process behind employing this particular year for standardized endmembers selection?

When comparing your estimated vegetation and soil fractions dataset, along with NDVI, fractional PV, and NPV, against the MOD44B vegetation continuous fields product, I noticed that you specifically chose to focus on the tree and non-tree vegetation components. Could you elaborate on the reasoning behind this selective comparison? 

Could you please provide more detailed reasons or methods explaining why you decided to verify PV and NPV as a single category in your study? Additionally, I noticed that you exclusively chose GLCVRD reference data from 2010 for validation purposes. Could you please elaborate on the decision-making process behind selecting this specific year for validation? 

Given that your product has a spatial resolution of 500m and a monthly temporal resolution, while the MOD448 product has a spatial resolution of 250m and a yearly temporal resolution, could you please provide a detailed description of the comparison process and the methods employed between them?

Please revise the abstract of this paper, some sentences are confusing. For example, line 32-34: “Sustainably managing terrestrial ecosystems requires an increased understanding of these structurally and functionally heterogeneous multi-component information and their changes, but we remain lack of such records of fractional vegetation and soil information at global scale.”.

Abbreviations are preferably not used in the abstract of a paper, and if they are used in the abstract, the full name of each abbreviation should be presented the first time it appears. For example, Lines 44-45: “PV and NPV”.

Line 124: Please correct “croplands and mosaics of croplands and natural vegetation;” to “Cropland/Natural Vegetation Mosaics”.

Line 137 and 231: Please add the serial number to the formula in this paper. For example, RMSE was defined twice.

Line 171: Why did the authors choose the threshold 'D > 0.7' in this study, and are there any references supporting this decision?

Line 268-277: Could you please provide the location or clarification for Figure 3 i-k? I would like to ensure that all relevant chart information is accurate and accessible.

Line 331 and 348: The subtitle '4.1 Compared with other datasets and traditional SMA model' seems inconsistent with 'Figure 7: Comparisons with other datasets and LSMA models.' Could you please clarify whether the LSMA model is considered one of the traditional SMA models, and ensure consistency in the expression used?

Line 378: “RESE” --> “RMSE”.