Investigations into the effects of conservation agriculture, by means of straw return in this case, on soil organic carbon sequestration has increasingly been a subject of interest. This manuscript by Yan et al. has made attempts to include the amount of straw return as an explanatory variable for predictive modeling, mapping, and spatiotemporal analysis of SOM. From this perspective, this study certainly has the potential to contribute to a better understanding of how increasing organic inputs could help improve SOM content at regional scale.

However, I must say that, in my opinion, the approach adopted in this study is not of sufficient scientific soundness for it to be considered for publication in SOIL. My major concerns are summarized as:

General comments:

1. As outlined in the title, the main novelty of this manuscript would be the incorporation of straw return amount into the Digital Soil Mapping framework. But descriptions on how the straw return amount (the CRC factor) is quantified and mapped are lacking. The authors did give a reference (Huang et al., 2020; Liu et al. 2020 in Table 1 is not listed in the References) to the data source (I did not manage to get access to the paper), but a more detailed explanation should still be added into the paper. For example, what was the approach used for CRC mapping? What was the sampling design and sample size for CRC field quantification, at what time of year? Without this information, it is difficult to assess the data quality and thus the overall modeling approach.

2. As far as I understood, the authors only used the CRC data in 2018 to evaluate the effect of straw return on SOM content in the same year. I think this is problematic because the effect of organic inputs on SOM dynamics is a mid-to-long term process, so linking spatial variability of SOM to CRC at one timepoint seems a bit farfetched to me. I would suggest the authors to look into the cumulative effect of straw return on SOM over a longer time period.

3. For the RF model, the authors included the CRC factor for the year 2018, but the relative importance of CRC appears to be low. This again questions the validity of the approach and the CRC data. At the very least, the authors should compare the predictive performances of models with and without CRC, so as to demonstrate whether incorporating CRC improves the model performance.

4. The predictive performances (CC and RMSE) for the 2006 and 2018 RF models were not exceptional (Figure 2) – the accuracy was actually worse in 2018 after the addition of CRC. This means that the predicted SOM values are associated with large prediction errors and uncertainties, thus weakening the obtained results from direct comparisons for the purpose of SOM monitoring.

Specific comments:

1. In the Abstract, one should briefly mention the size and characteristics of the study area.

2. Line 35-40, what is the difference between conventional mapping and DSM? Doesn’t DSM also comprise the procedures you outlined in the first sentence of the paragraph?

3. Line 70-75, there is no mention of the sampling designs in 2006 and 2018. Also, what are the sample sizes?

4. Line 85, I suggest the authors to specify how NDVI and EVI were calculated? Annual mean or based on images from a specific month?

5. The authors should specify the statistical method used for significance tests for all the boxplots. Otherwise, it is difficult to evaluate the appropriateness of the comparisons on changes in SOM with varying straw return amount.

6. The entire Results and discussion section was more focused on the interpretation of the results. An in-depth discussion on the strengthens and weaknesses of the methodology is missing.

7. Overall, the writing of the manuscript should be improved.

The manuscript addresses an important topic: the effect of straw return on soil organic matter. However, the introduction is rather short and does not fully explain the spatio-temporal modelling of soil properties. Furthermore, the materials and methods lack precision and the protocols deviate from current practices. Why was the soil sieved at 0.25 mm, while the fine earth is generally defined as < 2 mm (line 78). There is also a confusion between SOM and SOC. The wet oxidation protocol should be explained more carefully, because these analyses determine the SOC and NOT the SOM content.  the effect of straw return (section 3.5) is difficult to evaluate. First of all, the term is not clearly defined. According to the materials and methods section it is the residue cover and not the percentage of the residue produced by the crop. Second, the statistical analysis of the effect (see Fig. 6) is poorly explained. If Fig. 6 displays the SOM content of the pixels in each class, these observations are not independent and therefore  cannot be pair wise compared using a statistical test (ANOVA or t test). This remark also holds for figures 7-10.

Lines 35-36 Would not it be better to express the functions of Jenny and SCORPAN with the dependent variable ‘soil property’ rather than ‘soil’. After all, ‘soil’ is a broad concept that cannot be quantified and you mention ‘soil properties ‘ in line 38.

 Lines 40-43 You have explained (not in great detail) the role of DSM for quantifying the spatial variation in soil properties. Here you also include the temporal component. This has to be explained in more detail.

 Lines 74 and 75 The sampling design and use of legacy data is not discussed, so it is difficult to interpret their effects on ‘prediction error’.

Line 82 Please provide the reference for the ‘Resource and environment data cloud platform’

Line 87 Although spectral indices such as NDVBI and EVI are well-known, this is much less the case for the NDTI and STI. Please specify these indices.

Line 89 If I understand correctly, you use the CRC of 2018 for all fields between 2007 and 2018? This is a strong assumption as it does not take differences in crop performance or crop type into account. Please describe more clearly that the CRC is not used as a co-variate, but you compare two datasets (with and without residue).

Section 2.4.2 The technique of geographical detector is not as widely known as e.g. random forest. The principles will have to explained in a couple of sentences.

Table 3 Please include a column with the number of samples.

Section 3.2 Please explain the abbreviations e.g. ‘(all-Y)’. As it stands the reader has to look them up in the figure caption. It is not clear either whether these statistics apply to the calibration or the validation data set.

 Line 180 and further on. Please define what you mean by ‘straw return content’. As far as I can see it is the straw cover and not necessarily the percentage of residues produced.

Figure 6 Please explain how the significance was calculated and what it means. Why did you not try to fit a regression and analyse the significance of the regression?