The manuscript submitted by Huo et al. provided a dataset of six soil trace elements (I think they tried to focus on micronutrients) across the Tibetan Plateau (TP). The element distribution was investigated, and the possible factors regulating their distribution were discussed. Meanwhile, they used AI models to predict and map these elements across the TP. In general, this dataset is important to understand micronutrient cycling in the third pole of the world. However, there are many big issues limiting its wide use.

>> The first issue is the analysis methods of the elements in the soil, which used the XRF to determine the element concentrations in the field. I think this method has a large uncertainty when using in the field, compared to traditional methods like ICP-OES and ICP-MS. Unfortunately, the authors did not give convincing quality control data to ensure the precision of the analysis.

>> Second, I do not why only six elements were considered in this study. There are many trace elements or even micronutrients in soil, and some others (e.g., B, V) are also important for plants or animals. These limited element types in this study are not helpful for users to apply further studies. Meanwhile, there is many key background information (specific elevation, vegetation type in each site, local climate) that was not provided in the manuscript. Such data are also the important part of the dataset.

>> Third, the data predicted, as the authors mentioned in the manuscript, also have many uncertainties. One of the reasons may be linked to selection of ecosystem types. In this study, many natural ecosystems were selected, but farmlands (a landscape associated with human disturbance) were ignored. This should result in the uncertainty of these element distributions on the TP.

>> Additionally, the section of Discussion was not well drafted, which is very superficial and lacks key evidence to support the discussion points.

Some other specific comments can be found below.

Specific comments:

Line 1: Normally, Fe cannot be termed as a trace element in soil (but in plants, it can be termed as micronutrient), and it has a high abundance in crust or soil like K, Ca, and Mg. Furthermore, in Lines 10-11, six elements are targeted in this study, but why only four of them is shown in the title? Even though there are six elements targeted, I think the dataset is still small. There are many kinds of trace elements in soil, such as toxic metals (e.g., Cd, Cr, Hg, Pb, Sb) and micronutrients (e.g., V, B). So, I strongly suggest the authors to adding more elements in the dataset. This will increase the application of the dataset and attract more attention.

Line 7: Micronutrients are totally different from trace elements (shown in the Title). Micronutrient is defined by plant demand, but trace element has a broader scope. As I mentioned above, some toxic elements cannot be termed as micronutrients, but they belong to trace elements. What does this dataset target to, micronutrients or trace elements? If you aimed to map micronutrients, the Title must be changed.

Line 10: How many samples were collected in the 526 sites? In other words, please provide the size of the dataset. Moreover, were all the soils from surface layer? How deep of the layer? At least, this basic information should be provided in the Abstract.

Lines 13-16: These results are too simple to summarize the characteristics of the elements, such as the concentration ranges, the reasons of the distribution, and/or potential application. Additionally, why did you only introduce the spatial patterns of Fe, Mn, and Zn, and how about the distribution of other elements? As a whole, the section of Abstract is too simple, and I cannot find more information of the dataset.

Line 20: I think this dataset only targeted to micronutrients, right? If so, the Title indeed needs to be changed to fit the contents or aims in the study.

Lines 20-22: Seriously, trace elements or micronutrients include more than those listed here. For example, BNF processes also need other trace and/or micro-elements like V, but in this dataset, many these kinds of trace elements were not considered. So, adding more elements is necessary for such dataset.

Line 34: Please make clear of “microelements” or “micronutrients”.

Lines 35-36: Add references here.

Lines 40: Still, “trace elements” or “micronutrients”?

Lines 44-46: According to the figure, the farmlands were not considered in the dataset. On the Tibetan Plateau, farmland is one of the most important land uses, and more strikingly, micronutrients in farmlands are particularly essential for crops and human health, as you mentioned in Lines 23-24. Unfortunately, this dataset ignored the data in such important landscape. So, it is necessary to add the data in the farmlands for meeting the aim in this dataset (see the Title in Line 1).

Lines 50-51: As I mentioned above, you ignored the agricultural ecosystem.

Lines 51-52: How did you realize “maintaining relative homogeneity in species composition, community structure, and habitat conditions”? I think this is not the necessity for the sampling in this study, because this dataset needs to represent the heterogeneity of the field on the TP. More importantly, if you had avoided these conditions, artifact disturbance must affect the analysis results of the element distribution.

Lines 53-54: I have a big concern for the sampling design in this study. Clearly, the soil development is totally different in the selected ecosystems. For example, in many forests, 0-10 cm soil may only cover organic layer with high organic matter or high concentrations in some elements like Cu, Zn, or Ni, but deficiency of some other micronutrients. However, in deserts or meadows, this soil may cover the A horizon or parent materials due to the weak pedogenesis. Such a disparity will result in totally different elements’ distribution in these ecosystems selected. So, the authors must provide the reasons for the sampling design in order to better direct the application of the dataset.

Line 54: Another concern for this sampling is that elevation and vegetation community are important factors for element distribution. However, this specific information was not provided.

Lines 56-62: I do not think this method can well analyze the element concentrations in soil like that of ICP-OES (or ICP-AES) and ICP-MS. The XRF method has a very large error, particularly used in the field. Nowadays, this instrument is normally used in the lab, after collecting the soil samples, because it is unstable for it when using in the field. So, the authors must provide serious and strict evidence for the determination of element concentrations by using this method, and some necessary comparisons must be done with other reported data in the soil from some similar sites on the TP. Then, the quality of the data must be strictly analyzed to make sure that the element concentrations are really accurate or reasonable. At least, now I do not see the quality control data in the manuscript, and I also do not think this method could obtain reliable concentrations for most of the elements analyzed.

Line 63: In Table 1, much more information should be complemented, such as more dominated plant species, elevation ranges, local climate. I suggest to establishing more columns to exhibit this information.

Lines 70-72: This TP method is wrong, but your method is to analyze bioavailable fraction of P. Still, you must provide necessary quality control data for the precision of the element concentrations. This is particularly important for the dataset.

Lines 73-76: Specify the method of CIA with necessary citation. You used XRF too much for the element analysis, but without necessary precision analysis. This is unacceptable.

Lines 79-80: Normally, when sampling in the field, slope, aspect, and elevation data can be recorded simultaneously. Why did you not get these data, but dependent on the online data? This will lead to more errors for them. The same case is also for the vegetation types (Lines 84-85).

Lines 88-91: Where are these data listed in your dataset, corresponding to your sites? Also, I strongly suggest the authors to providing an Excel file to exhibit all the data analyzed or compiled from online. This will help users easily obtain and cite the data.

Line 93: After your screening, how many data were left for the analysis below?

Lines 99-100: Some nutrients like P, S were not included in this analysis? Additionally, I do not think the anthropogenic disturbance can be totally represented by grazing intensity, because in some ecosystems like deserts or forests, very little grazing activity is there. Meanwhile, this dataset did not consider the data in farmlands, which subjectively removed the important human disturbance on the TP.

Lines 109-120: Re-organize the description of the results. If you tried to introduce the distribution of element concentrations (e.g., mean, standard error), specify all the values of each element. Do not make repeated description in two different paragraphs with different aims. Additionally, please do compare your data with other reports in the similar study areas. This can help to correct the data quality in your study.

Line 125: …vegetation… There are format errors in the manuscript.

Line 126: In Figure 3, was the statistical analysis conducted? If so, add the statistical results in the figure. The similar case is also for Figure 4.

Lines 125 & 147: In these two sections, elemental differences among vegetation types and lithology were analyzed. However, elevation and climate gradients are also very important for the element distribution. Why not exhibit the variations in each element concentration with them? This trend is different with the analysis in the section of 3.4 (Line 175).

Line 207: In this section, I have several concerns for the predicted results. First, because the dataset did not consider other landscapes (e.g., farmlands), the spatial patterns cannot be exactly representative on the TP. Second, even though the concentrations could be acceptable using the XRF method, I still suspect the reliability and reasonability of the data. Such a way to exhibit the concentrations of trace elements in the soil across TP may have limited reality for application. Third, despite the statistical analysis for the prediction in Lines 208-215, there will be many uncertainties (as the authors also mentioned in Lines 255-261) for the spatial distribution of the elements, even without considering the analysis precision of the XRF method.

Overall, I really warry about the future use of the dataset under the current results, which indeed ignored the precise analysis methods by high-precision equipment such as ICP-OES, ICP-MS. Meanwhile, the AI models also have many uncertainties for the data predicted, one being also closely related to the quality of the original element concentrations.

Lines 228-229: This comparison is not meaningful. As I mentioned above, you should make comparisons with other reports across the study areas, and then ensure the quality of the data. Then, you may make more comparisons with other reports worldwide.

Line 230: Because of your method for analysis of element concentrations, I cannot believe the conclusion of “deficient levels” here.

Lines 232-233: Seriously, what are the aims of this discussion or this conclusion? You did not analyze any specific fractions of elements in the soil or some other related research in the study area, and how can you conclude the increased degradation? This may mislead readers.

Lines 236-245: Where is the direct evidence of these discussion points? I don’t like discussion that lacks evidence from this study, but only based on points from cited references. As shown in your data, you have climate and weathering related data (e.g., MAP, CIA), and you should analyze these data and then make deep discussion. If the discussion is from your data, please show the relevant results in the form of figures.

Lines 246-254: Similar to those in Lines 236-245, this discussion is too superficial. These discussion points are very arbitrary and lack scientific basis and evidence.

General Assessment:

The study addresses a significant knowledge gap in high-altitude, cold-arid ecosystems. The application of machine learning for spatial prediction is appropriate and modern. The manuscript is generally well-structured, but several aspects require clarification, strengthening, and more in-depth discussion before it can be considered for publication.

Major Comments:

1. There is a critical ambiguity regarding the analytical method for the core six micronutrients (Fe, Mn, Cu, Zn, Ni, Mo). The text states a portable XRF was used in the field (Lines 55-56) but later describes lab-based wavelength-dispersive XRF on pressed pellets (Lines 59-61). The accuracy and validation of field-based XRF measurements for these elements, especially at low concentrations (e.g., Mo), must be explicitly detailed. The authors should clarify the protocol, report calibration metrics (R², RMSE) against certified standards, and specify if all element data came from the same method.
2. The use of relative importance metrics ('betasq') is a good start, but the analysis could be significantly strengthened. Consider using alternative methods (e.g., permutation importance from the Random Forest model itself) to cross-validate the reported driver rankings. Furthermore, the discussion of the U-shaped response to MAP (Line 191) is intriguing but remains qualitative. A more rigorous statistical exploration of these nonlinear relationships (e.g., using generalized additive models) would greatly bolster this key finding.
3. The poor performance of the models for Cu and Mo needs a more thorough discussion. Simply stating limited utility is insufficient. The authors should hypothesize why these elements are less predictable. Are the key drivers not captured in the predictor set? Is measurement error higher? Is their distribution more stochastic? This critical reflection is essential for a balanced interpretation of the results.
4. The results show significant lithological control for some elements (Fig. 4), yet climate is reported as the dominant driver in the importance analysis (Fig. 5). This apparent discrepancy needs reconciliation. The discussion should integrate these findings, explaining how regional climate patterns might override or interact with the inherent geochemical signal from the parent material across the vast plateau.
5. The high-resolution prediction maps (Fig. 7) are a key output. However, the manuscript does not provide associated uncertainty maps (e.g., prediction intervals). For users to properly utilize these datasets, an assessment and visualization of spatial uncertainty are crucial. Please add this or explicitly state it as a limitation.

Minor Comments:

1. The abstract mentions six microelements but the title and data availability specify only four (Fe, Mn, Zn, Ni). The title and abstract should be aligned. Either adjust the title to reflect the full study or refocus the abstract on the four well-predicted elements.
2. Figure 1b is described but not effectively explained in the caption. The relationship between the bars and dots (ecosystem area vs. sampling frequency) should be explicitly stated to justify the representativeness of the sampling strategy.
3. The terms "micronutrients," "microelements," and "trace elements" are used interchangeably. For consistency and precision, authors should choose the same word throughout the manuscript.
4. The data availability section provides a DOI, but this should also be formally cited in the main text (e.g., in the Methods or Results section) when the dataset is first mentioned.
5. Line 40 has a trailing comma after "Mo" ("...Ni, Mo,).").
6. The discussion on ecological implications (Lines 228-233) is good but could be slightly expanded. Briefly mention specific plateau processes that might be most sensitive to these micronutrient limitations