I believe the subsoil organic carbon dataset developed in this study will have a significant impact on the relevant field. The paper is well organized and clearly presented. Below are my suggestions for the authors’ consideration.

1. Line 20: the word ‘we’ is missing.
2. Line 29: adding the range of soil depth would help strengthen this conclusion.
3. Line 33: the word “Global” should be in lower case; Grammer error in “which providing”.
4. Line 41: space between gas and (GHG) is missing.
5. Line 42: a period after the citation is missing.
6. Line 45: grammar error in “, which contributes”
7. Line 108: grammar error in " from”
8. The data collected from the literatures should be published as well for validation purposes and promote boarder application by other researchers.
9. While the authors have done a great job collecting literature data with a well global coverage. However, the density of study sites varies significantly across different regions. Please discuss the limitations of this data collection.
10. Section 2.2: as the logic flows from previous section to this one, it directs reader to believe that this section explains how the authors calculated SOC density and stock from the literature. It however seems to estimate gridded SOC stock via predicted soil β in the following section. If the latter is the main focus, consider relocating it to the right place (maybe after 2.5).
11. Section 2.3: Clarify whether the soil β values were directly obtained from the studies or calculated using Equations 3 and 4. Typically, soil β is calculated from these equations based on known SOC at different depths in the literatures, rather than the reverse. Clarification on this would be helpful.
12. Line 146: awkward wording.
13. Section 2.4: consider moving it after 2.5, creating a more logical sequence: extracting data from literatures -> building model to predict soil β -> preparing spatial data -> estimating SOC stock.
14. For 1221 soil profiles in 161 studies, the authors could make use of the variability of SOC in each study to estimate the uncertainty range of this global SOC dataset. Given the high heterogeneity of SOC, adding uncertainty estimates could enhance the value of this dataset. This is just a suggestion for the authors’ consideration.
15. The PF generally performs well across three ecosystems. However, it tends to over-estimate the lower β and under-estimate the higher β. The authors need to reset their model to improve it. If it cannot be resolved, an explanation and discussion of the potential impacts on predicted SOC, particularly regarding spatial distribution (e.g., even lower soil β in boreal grasslands as seen in Figure 3E), should be provided.
16. Figure 3: clarify that the numbers in panels d-f represent predictions to avoid confusion.
17. Line 304-306: consider moving this explanation to the discussion section.
18. Comparing the estimated SOC stocks with other studies across different ecosystems in terms of total numbers is valuable. Additionally, including comparisons with spatial maps would provide a more comprehensive validation of the dataset.
19. Line 414-415: awkward wording.

Wang et al. collected 1221 soil profiles to quantify the vertical distribution of soil organic carbon at global scale. The topic is important and interesting. However, I believe that there are several substantial concerns before publication.

In year of 2000, Jobbágy & Jackson (2000) quantified the vertical distribution of soil organic carbon with more than 2700 soil profiles up to 3 m. The current dataset just includes 1221 soil profiles, it is too small! The website had provided a lot of soil profiles at global scale, which may help to the current study (WoSIS Soil Profile Database | ISRIC).

Jobbágy, E. G., & Jackson, R. B. (2000). The vertical distribution of soil organic carbon and its relation to climate and vegetation. Ecological applications, 10(2), 423-436.

Jobbágy & Jackson (2000) had evidenced that the equation 3 had the worst performance in fitting the vertical distribution of soil organic carbon, the author should provide the rationality of the functions used.

The calculation of SOC density by Equation 1 had great limitations due to the faction of gravel content.

The calculation of global SOC storage is based on Equation 2 (𝑆𝑂𝐶 𝑑𝑒𝑛𝑠𝑖𝑡𝑦 ∗ area of cropland, grassland or forestland). This calculation of useless because of the small data set and greater uncertainty. I suggest that the author focuses on vertical distribution itself.

In 2.2 Global soil attributes calculation, why divide the soil profiles into 5 layers with 20 cm intervals? The vertical distribution of SOC can be quantified by correlation between SOC and depth directly.

The selection of the predictors for β needs clear motivation. For example, how did microbial biomass carbon and nitrogen influence the vertical distribution of β? The β is calculated by SOC, therefore, it is a bad choice including SOC as a predictor. The vertical distribution of SOC should be regulated by root or belowground net primary productivity (Xiao et al. 2023).

Xiao, Liujun, Guocheng Wang, Jinfeng Chang, Yaoyao Chen, Xiaowei Guo, Xiali Mao, Mingming Wang et al. "Global depth distribution of belowground net primary productivity and its drivers." Global Ecology and Biogeography 32, no. 8 (2023): 1435-1451.

Figure 3 showed greater bias of the current model in predicting the global pattern of β because the slope is not equal to one. In addition, observed (in the y-axis) vs. predicted (in the x-axis) regressions should be used (Guerschman & Paruelo, 2008).

Piñeiro, G., Perelman, S., Guerschman, J. P., & Paruelo, J. M. (2008). How to evaluate models: observed vs. predicted or predicted vs. observed?. Ecological modelling, 216(3-4), 316-322.

I believe that the writing of the current study needs to be improved.