Line 174-175: “A spatially stratified sampling strategy based on quadrilateral grids was adopted to mitigate the effects of spatial autocorrelation.” Could the authors specify the spataially stratified sampling strategy?

Line 177: “more than 50,000 valid sample points were collected annually” How much of the sample points are from the field survey? Also, how did the authors distringuish winter wheat fields from other crop types based on visual inspection? Some examples can be provided.

Line 234-235: “all non-wheat pixels (Section 3.1) were classified into two types: non-wheat winter crops vs. non-winter crops and non-wheat spring crops vs. non-spring crops, according to their respective growth stages”: I think there are four types?

Line 238: What is the definition of the “spectral separability indices (SI)”?

Section 3.2: what would the accuracy be if you do not do the Selection of provincial feature set?

Line 256: It is not clear why (Yang et al. 2023) is cited here.

Line 264-269: it is not clear to me why does the model can map planeted winter wheat and harvested winter wheat by changing the time window of the feature set.

Line 269: “The final products include harvested area maps of spring and winter wheat for 15 provinces, as well as planted area maps of winter wheat for 10 provinces.” What cause the difference number of available provinces for harvested area maps and planted area maps of winter wheat?

The manuscript introduces CN_Wheat10, a 10 m resolution dataset of spring and winter wheat distribution in China for 2001–2023. It used a cross-regional training-sample generation method to address the lack of large-scale training data and a province-level feature-selection strategy to improve the regional adaptability of a random-forest classifier. Building on this, the authors generate separate spring- and winter-wheat maps and further derive planted- and harvested-area products, which I consider the key contribution of the work. However, it remains somewhat unclear whether the method is specifically tailored to address these mapping targets, even though the reported overall accuracies exceed 90% across years and regions. In addition, a more detailed description of the methodology and parameter settings, together with more rigorous validation, is needed to strengthen the study.

General comments:

1. One of the key contributions is mapping spring and winter wheat separately, but it is not clear how these two classes were distinguished from the outset. The authors mention using different time periods, yet it remains uncertain whether all regions were processed with distinct workflows or whether dominant spring/winter regions were predefined based on expert knowledge. The lack of overlap in the maps gives the impression that the latter approach might have been used, though this is only my inference. Since mixed-cropping areas do exist in China, clarification on this point would strengthen the manuscript.
2. For the planted vs. harvested area mapping, it is unclear whether the classification was guided by specific labels or simply distinguished by using growing-season vs. full-season time series. If it is the latter, both classifiers might depend on similar mid-season features (e.g., April in Figure 4). In that case, I would consider these products to represent in-season vs. end-season maps rather than true planted vs. harvested maps. It is also unclear whether the authors considered the logical constraint that the harvested area must be a subset of the planted area. Moreover, the evaluation of wheat reduction may mask systematic errors. For example, commission errors in mid-season maps and omission errors at end-season maps could lead to large deviations in area estimates. In such cases, the difference between planted and harvested areas cannot be treated as reliable evidence of wheat reduction (Figure 15). Most importantly, area estimation should be based on rigorous statistical approaches rather than simple pixel counting.
3. More elaborations are needed on sample generation, feature selection, and validation. The reason for choosing the CDL of Kansas and North Dakota should be clarified (e.g., NDVI curve comparison), along with the number of CDL-derived training samples and the generated spring/winter wheat pixels. It is also unclear whether the zone strategy followed provincial boundaries or the agro-ecological regions in Figure 1. If it was based on the province level, this may be problematic given phenological variance (Liu et al., 2024). Please also clarify whether the VH threshold was derived using samples independent from the validation data, and consider tuning the random forest parameters (Li et al., 2023).

Specific comments:

1. The manuscript provides harvested area maps for 15 provinces but planted area maps only for 10 provinces. What explains this discrepancy, and why were spring wheat planted area maps not produced given the similar workload and methodology?
2. Line 235: I found it difficult to understand the logic behind classifying non-wheat pixels into two types. This part could be explained more clearly.
3. Line 245： Why were SAR features not included in the feature selection process but instead added only after filtering the spectral features?
4. Figure 8: The y-axis could be adjusted to a more appropriate scale to make the accuracy values easier to interpret, for example by starting at a value higher than 0.6.
5. Figure 9: Does the y-axis represent average overall accuracy or another metric?
6. Since the temporal range of the current dataset is limited by Sentinel-2/1 data availability, is there potential to extend the mapping to longer time series using other sensors? This could be worth mentioning in the discussion. In particular, examining long-term dynamics of spring and winter wheat distribution is scientifically important for understanding cropping system shifts and their adaptation to climate change.
7. The current validation for spring wheat is not sufficient, and the WorldCereal dataset with spring cereals map could be considered to help compensate for this gap (Van Tricht et al., 2023).

Liu, Yifei, Xuehong Chen, Jin Chen, Yunze Zang, Jingyi Wang, Miao Lu, Liang Sun, Qi Dong, Bingwen Qiu, and Xiufang Zhu. 2024. “Long-Term (2013–2022) Mapping of Winter Wheat in the North China Plain Using Landsat Data: Classification with Optimal Zoning Strategy.” Big Earth Data 8 (3): 494–521. doi:10.1080/20964471.2024.2363552.

Li, H., Song, X.-P., Hansen, M.C., Becker-Reshef, I., Adusei, B., Pickering, J., Wang, Li, Wang, Lei, Lin, Z., Zalles, V., Potapov, P., Stehman, S.V., Justice, C., 2023. Development of a 10-m resolution maize and soybean map over China: Matching satellite-based crop classification with sample-based area estimation. Remote Sensing of Environment 294, 113623. https://doi.org/10.1016/j.rse.2023.113623

Van Tricht, K., Degerickx, J., Gilliams, S., Zanaga, D., Battude, M., Grosu, A., Brombacher, J., Lesiv, M., Bayas, J.C.L., Karanam, S., Fritz, S., Becker-Reshef, I., Franch, B., Mollà-Bononad, B., Boogaard, H., Pratihast, A.K., Koetz, B., Szantoi, Z., 2023. WorldCereal: A dynamic open-source system for global-scale, seasonal, and reproducible crop and irrigation mapping. Earth System Science Data 15, 5491–5515. https://doi.org/10.5194/essd-15-5491-2023