This study constructed a gridded dataset for crop water consumption in China. The advantages of this dataset are the separate estimation of blue and green evaporation (E), transpiration (T), the use of the local information on irrigation type per crop, and the long time series. Even though comparisons with other studies on ET look good, I have many concerns about methods that are not clear in the paper. Without this information, it is hard for me to conclude on the dataset quality and reliability. I suggest a major revision and reconsideration for publication unless the authors can address the following concerns/questions.

Major comments:

1. The aquaCrop model setting is absent.

1-A) There are different versions of AquaCrop and available on different platforms, e.g. Windows interface for FAO AquaCrop v7, v6 or before, and Python and MATLAB open source versions. Which AquaCrop version was used for this study?

1-B) As I understand, publicly available AquaCrop cannot simulate perennial crops. How did the authors deal with perennial crops using AquaCrop, e.g. tea, and apple?

1-C)  Using AquaCrop needs crop characteristics as input, e.g. the maximum canopy cover, canopy cover decline coefficient, canopy growth coefficient, and many others. Where did these inputs come from and what are the inputs?

1-D) What do the authors mean by calibrated AquaCrop? How did the authors calibrate AquaCrop? Section 2.3.1 does not explain the calibration on AquaCrop.

2. Reliability in separating E and T.

2-A) Distinguishing T from ET is not traditionally in water footprint studies. One challenge is that the ratio of T to ET depends not only on crop growth, and irrigation type but also on field management, e.g. weeds, soil fertility, mulching. As for large-scale assessment, there is always lacking data for field management, thus, it is difficult to estimate T and E separately. However, total ET is more robust to field management, so lacking such data will not be a big issue for total ET. Here the authors try to explicitly distinguish E and T from ET but still have very limited ability to describe the field management in the model. In this sense, how do the authors evaluate the reliability of separating E and T?

2-B) Did the authors further distinguish the color of E and T? and how?

3. Irrigation and soil moisture assumption. In Table 5, comparing the previous studies, this study has a higher water footprint. According to my experience using these datasets, they generally overestimate water consumption compared to hydrological models and recent crop models for some crops because of two reasons (maybe more): first, they assume irrigation once there is a water deficit, even though the water deficit would be tiny; second, when setting the initial soil moisture, they assumed the field capacity of the soil moisture at the beginning of each year. Both are unrealistic and will lead to overestimation of water consumption. How do the authors set irrigation rules and soil moisture in the model and did the authors have a spin-up for the model?

4. Usability and quality of the benchmark

4-A) Figure 6 shows the benchmark at the grid level. However, if we have a close look at the graph, we find usually, the resolution is the provincial level, e.g. the whole province is efficient or not. This is because the calibration and scaling were on the provincial level. If it is the case, scaling factors seem to play a critical role here other than irrigation, climate, and so on. What do the authors think of the reliability of the benchmark in this case?

4-B) The whole of China was divided into two climate zones for the benchmark without considering soil type. This seems oversimplified to me. Furthermore, this dataset provides the benchmark for each year, but the idea for a benchmark is to provide a kind of efficient “reference”. So, why not derive using the whole time series? Imagine one wants to use the benchmark for future analysis, which year should be selected? How do the authors suggest using their benchmark?

Minor comments:

1. In the title, “consumptive water footprints”. By definition, water footprint is consumptive water use. Is consumptive redundant here? Or “consumptive water use”?

2. Line 95-102. The authors basically only rely on the global datasets for land use from the year 2000 and then scale to each year. Are there other better datasets that have better quality/time coverage in China?

3. What is CC in Equation 11? Canopy cover?

4. I didn’t see the causal relationship between the number in lines 222-228 and why it is important to distinguish irrigation type. Lines 263-269 do explain some.

5. In the paper, e.g. line 240, the authors discussed the monthly blue and green water consumption but is not available in the dataset. Some items are available monthly and some annually. What are the criteria to decide which data is open to the public or not? And the dataset lacks projected coordinate system information.

6. Line 245-247, why sprinkler irrigation has the highest value? Because of crop type?

7. Line 298-300, can the authors explain why the benchmarks of some crops are higher in arid zones and others are higher in humid zones?

8. in GAEZ+2015 and MapSPAM2010, they also have other crops, did the authors include them in the comparison? And GAEZ+2015 is developed for the year 2015 and MapSPAM2010 is developed for the year 2010, why not use the corresponding years other than the average between 2001 and 2008?

Technical corrections

1. Table 3 doesn’t seem consistent with the text. I only checked for rice. In Table 3, the total water consumption of rice is 81847+58979+4629+5540=150995 M m3 and in the text, it is 143 G m3.

This paper calculated the WFCP from physical model and then analyzed the water footprints of major crops in Chine between 2001 and 2018, it is a useful topic and the analysis workflow is fine. But my major concern is about the quality of the data, I think there are data with better quality can be used as input and generate more reliable dataset for analysis. Therefore, my opinion is that the paper needs a major revision and a new version of the dataset should be generated before the analysis as the current version make no sense in scientific community.

1. Some concerns about the input dataset:

(1) Irrigated and rainfed crop areas between 2001 and 2018 is generated from the combination of MIRCA 2000 with 5 arcmin resolution and the year book data, but currently, there are actually a lot of new dataset about irrigated cropland from satellite data, for example, Zhang et.al (https://www.sciencedirect.com/science/article/abs/pii/S0022169422009830) 2022, and Zhang et.al 2022 (https://www.nature.com/articles/s41597-022-01522-z). Maybe there are still other data products, I just searched in Google and found the above two datasets. All these irrigation data could be verified and then resampled to 5 arcmin, and used as input data. Basically, the spatial distribution of irrigation land generated from satellite are more reliable than a 15-year-old coarse resolution data and yearbook. As the irrigation cropland is the basement of all the following data generation, please keep the pace with new datasets in the geo-agricultural community.

(2) The crop phenology for the major crop need to be further specific. For example, the winter wheat, the planting date of winter wheat is Oct 15, and all winter wheat are assumed to have the same phenology characters. But this is not the situation in practice, and the phenology of winter have significant difference even in North China Plain with different sowing, emergence and dormancy date. It is essential to use more detailed phenology data as input.

1. Data comparison

Figure 9, Please include RMSE in the Figure, when analyzing quantitative results, RMSE is more commonly used to evaluate overestimation and underestimation. Although it seems the value of R square is good here, I found the number of high ET value is small, this indicates the high R square value in this Figure do not make sense either, Please include more high ET-value pixels in the analysis.

1. Separation of E and T

The separation of E and T can not be verified because there is no validation data. Please address this in discussion section, pay attention to include the uncertainty analysis of E and T separation.

Line 183, Can you show the map of humid (AI > 0.5) and arid (AI < 0.5) zones?

Line 194, What remote sensing data have you used for comparison?

Line 316~319, I think the two factors addressed here do not make sense as the data used in the analysis have significant shortage, please use your new version to compare.

The article " A gridded dataset of consumptive water footprints, evaporation, transpiration, and associated benchmarks related to crop production in China during 2000–2018" established a gridded dataset of monthly green and blue water footprint of crop production (WFCP), evaporation and transpiration et al. based on AquaCrop model. On the whole, the paper has important research significance and much work was done, but there are some shortcomings in the paper, and it is suggested to modify.

1. In Lines 186-187, “The statistical yearbook only has crop production statistics on the provincial level. Therefore, we calibrated crop production at the provincial scale.”. The statement is not accurate. Statistics data on the city level can be found in the provincial statistical yearbook.
2. In your research, AquaCrop model was used to simulate crop production more than 21 crops. Wheat, corn and rice are the main food crops in China. In crop production, irrigation and fertilization are both important management measure to improve crop yield. However, only irrigation was considered. In my opinion, it is necessary to consider the effects of irrigation and fertilization on crop production in different regions when simulating crop production process using AquaCrop model.
3. As we all know, AquaCrop is a farm-scale model. In your research, the crop production at the provincial scale was calibrated and validated by statistical yearbook. I don’t think this is a good idea. It is recommended that some representative sites were selected to verify the AquaCrop parameters.
4. Whether the parameters are consistent in AquaCrop model under different irrigation methods?
5. Lines 391-392. The crop parameters do not vary with simulation time and space. It is inaccurate for the whole Country to use only one set of crop parameters to simulate crop yield. In my opinion, different crop parameters should be used for each province.
6. The discussion needs to be rewritten. The results of this study can be compared with those of others, and the reasons for the similarities and differences can be analyzed. For example, in section 4.1, it is inadequate just compared with the evapotranspiration result with SEBAL model, which is a single-source model. In your research, evaporation and transpiration has been estimated. It will be more meaningful to verify evaporation and transpiration based on the two-source model.

Minor questions:

1．In Line 40, what’s the mean of “CWASI”.

2．In Lines 130, 135 et al., no spaces before “where …”.