Summary

Reliable mapping cropland area is essential for assessing and monitoring the sustainability of agriculture from regional to global scale. Here Tubiello et al., generated country-level crop area dataset based on six remote sensing-derived crop area dataset at the global scale. The generated dataset agreed well with FAO land use statistics; with the generated dataset, the authors conducted a comprehensive analysis regarding the crop area and its uncertainty. The manuscript is generally well organized, and the research is important. Here, I listed a few concerns regarding the manuscript.

Specific comments

1)    Line 47-49, since there have been cropland agreement maps in the previous study, is it suitable to refer to the new dataset in this study as “Cropland Agreement Mapping dataset”? It seems that the name did not reflect the difference between the dataset in this study and that of previous study. My understanding is that one major difference between the two datasets is that this study aggregated the previous dataset to the country level, correct?

2)    Line 50, 25% of what? I cannot understand this sentence.

3)    Line 86-87, actually, the R2 and MAE between the crop area of CAM and FAO can be compared with the performance between the crop area of a single crop data (one of the six crop area data) and FAO.

4)    Line 118-119, how did you aggregate the pixel-level information to national scale? What kind of coordinate system is used for the dataset? Did you consider the area differences across different grid cells or pixels? Please clarify it.

5)    Line 127, “Conversely” or “Similarly”?

6)    Line 131, “definitional bias would be higher for SAk classes” not clear. Here you mean “definitional bias would be higher for SAk classes with lower k values” correct? Please clarify it.

7)    Line 153, Pearson correlation coefficient is R instead of R2, here you mean R square?

8)    Line 171-174, it’s hard to follow this part. I assume many readers would have problems similar to me. What’s relative uncertainty, what does “communicate” here mean, is it a commonly used mathematical term? Here “dimension” corresponds to number of samples? Please clarify those terms or definitions and use plain language to explain what does this part exactly mean.

9)    Fig. 2-3, what does the red dashed line and black line represent, respectively? Please clarify it in the figure caption. Will the R2 between the crop area of CAM and FOASTAT be higher than that between the crop area of a single data (one of the six crop area data)? Will the NRME be lower by using multiple data ensemble mean?

10)    Fig.4, the label and unit of y-axis should be given

11)    Fig. 5, is this regression line statistically significant? Please show the p-value.

12)    Fig.7, please add necessary description in the figure caption to explain what does the size of the circle and color respectively mean.

13)    In addition to the results, brief discussion about how the findings could aid further development of land cover or crop area products could potentially make this study more influential.

         Tubiello et al., have presented an important contribution to the field through the development of a novel cropland area mapping dataset (CAM). This dataset, which aggregates information from six high-resolution remote sensing-based global cropland maps, includes an estimation of uncertainty at country and regional levels. The systematic analysis of agreement and disagreement among the six contributing maps has provided valuable insights into the underlying uncertainties associated not only with individual map accuracies but also with their definitional differences.

The manuscript is generally well-written and presents the data in a clear and effective manner. The identification of regions with large uncertainty could prove particularly useful for future endeavors in land cover and land use product development and data evaluation.

Nonetheless, there are a few points that could use further clarification:

Line 153: It is mentioned that R2 is the Pearson correlation coefficient, which seems to be a misunderstanding. R2 is the coefficient of determination.

Figure 1: I noticed a large area of low agreement in the Southeastern United States. It would be beneficial to readers if you could provide a more in-depth discussion about this. Specifically, what are the differences in how each of the six datasets defines 'cropland' in these areas? Are these differences primarily due to methodological variations or could there be other underlying factors contributing to the disparity?

Additionally, it would be helpful if you could discuss the limitations of your approach, including any potential biases and how they might be addressed in future research.

I hope these comments help to further strengthen the manuscript.