The contribution of the research is having produced the first 1km global cropland dataset with long time span by employing the newly developed spatially explicit allocation algorithms. The theoretical framework of the method is convincing, and the results are well presented. This spatiotemporally continuous dataset will provide a new opportunity for better understanding the past global changes caused by ALCC.

Although the new global cropland dataset has been created to such a high resolution (especially for historical periods), it is necessary to recognize the uncertainty of this dataset that was unmentioned in this manuscript. It is suggested to make a brief discussion further to address the issue that both the cropland area per capita estimated by HYDE and the historical population datasets adopted by this study are with huge uncertainty.

Firstly, the reviewer acknowledged that there is no more reliable cropland area data at the global scale than HYDE up to now, but it should be noticed that the historical cropland area data in HYDE have great uncertainty that has already been proved in many countries. One of the main reasons is due to the unreliable estimation of per capita cropland area. The continuous improvement of cropland area data in versions of HYDE mainly relies on the modification of the historical cropland area per capita curve according to the quantitative regional reconstruction results or other related indexes. Although the authors gave a clear definition of cropland (line 118), which was also the same as the definition adopted by HYDE during its historical cropland area estimation. The definition without considering the unignorable amount of fallow land or crop rotation in history would cause obvious underestimation about the cropland areas in some countries. Especially in countries that are far less intensively cultivated (completely different from the traditional agricultural period in China), like some countries in Europe. Some studies have also pointed out that the area of cultivated land in Europe in HYDE is obviously underestimated. Thus, this would cause a smaller extent of historical cropland distribution and lower fractions in the gridded allocation datasets.

Second, the historical population dataset of HYDE is actually derived from the national or subnational statistical/estimated population amount by downscaling method, which is basically the same as the allocation algorithm of historical cropland. The huge uncertainty also existed in this dataset caused by its allocation principles. Since both studies have adopted the population factor in the allocation algorithm of historical cropland, the difference of gridded datasets between HYDE and this study is namely caused by the different usage of physiogeographic factors and their resolutions during the allocation. In future related researches, please cautiously use this unevaluated historical population dataset.

Additionally, the cropland results are displayed at a global scale, so the details cannot be seen clearly, and no administrative boundary was added on the regional maps (line 635-640, Fig4-6, it would be better to add some necessary labels and administrative boundaries on the map; it seems that the linear unit scale should not be added under the geographical coordinate system?).

This study targeted the construction of global cropland dataset starting from 10000 BCE and extending to the future in 2100 CE. Great effort was made to integrate, harmonize, and downscale multi-datasets to produce the final 1 km global dataset. The new dataset is expected to be very useful for a broad spectrum of studies and applications since it considers mapping the historical, as well as future, cropland distribution at a relatively optimal spatial resolution to such a large geographical scale and long period. However, going through the manuscript could raise several questions to the readers, which need to be considered by the authors. Some of those questions are listed as follow:

• Understandably, the date of 10000 BCE was the staring of farming, but integrated datasets (Table 1) were back only to 1950 CE, regardless of the population dataset sourced from HYDE? Hence, defining the suitability for cropland was totally dependent on the population data during the period before 1950, which makes the quality of the cropland mapping during the period totally dependent on the quality of the single layer of the population. Hence, the reader could question the added value and uncertainty when starting the production of the maps from 10000 BCE, with the lack of data covering this long period?
• Suitability map played a crucial role in mapping production. The authors clearly stated that the influence of the variables defining the land suitability for agriculture was not equal throughout the whole period. For instance, population was the key variable defining the land suitability in earlier dates due to the traditional farming practices and weak global links (as mentioned in line 169). However, another assumption in line 175 was made by authors that I believe it needs further explanation and justification. The assumption is that “the impact of population on cropland distribution is negligible for most regions in the future years (2010 CE-2100 CE)”, why?
• The comparison between the constructed cropland dataset and HYDE 3.2 dataset in Figure 7A showed a relatively close r value for recent years (1700-2000) but a high increase of RMSE value. A reader could expect that the error in quantifying cropland area would decrease when getting closer to the current time. But this is not the case in the constructed dataset, any justification from the authors about that?