Cropland Productivity Evaluation: A 100 m Resolution Country Assessment Combining Earth Observation and Direct Measurements

Round 1

Reviewer 1 Report

 

General comments: This paper which mainly presenting cropland productivity evaluation combining earth observation and direct measurements. The data validation is done with the total of 80000 cultivated fields of Hungary for 5 years collected during the field survey and MODIS GPP dataset, which made this work relatively interesting and powerful. There is always a lot of contribution of such work for world-wide cropland productivity evaluation. However, there are still some points that need to be well rephrased.

 

1．Please, rephrase and merger the information in the introduction section, although the authors have presented focused literature review.

 

2. Did the authors used the specific metrics that considers the imbalance between the test dataset and training dataset? Why selected 90% data as the training data and 10% as the test data? How to make this selection randomly？(LN210-213).

 

3. Please, improve the quality of figures 4、7, and make it more proper read and clear understanding.

 

4. Please, pay attention to superscript and subscript, such as CaCO3 in LN308、318.

 

5. Can you give more information for why the power of sunflower productivity estimation was such low?

 

Author Response

Dear Reviewer 1,

 

Thank you very much for evaluating our manuscript, and providing valuable comments to improve our manuscript. We very much appreciate your positive and helpful comments.

We revised the manuscript by accepting all the suggested changes. We copied your comments below and marked our responses in red.

 

General comments: This paper which mainly presenting cropland productivity evaluation combining earth observation and direct measurements. The data validation is done with the total of 80000 cultivated fields of Hungary for 5 years collected during the field survey and MODIS GPP dataset, which made this work relatively interesting and powerful. There is always a lot of contribution of such work for world-wide cropland productivity evaluation. However, there are still some points that need to be well rephrased.

1．Please, rephrase and merger the information in the introduction section, although the authors have presented focused literature review.

Following your suggestions, we have completely revised the introduction section.

2. Did the authors used the specific metrics that considers the imbalance between the test dataset and training dataset? Why selected 90% data as the training data and 10% as the test data? How to make this selection randomly？(LN210-213).

We provided more information on the validation process in the revised manuscript (L218-220). Our approach is based on the consideration of the large sample size, which allows testing based on the 90-10 subdivision of the dataset. Indeed, in case of smaller sample size a biased model, or a model with higher uncertainty may have been resulted and the applied validation process might not have been the most suitable. However, in our case the internal variability of both the modelling and the validation datasets covered the variability of cropland in Hungary and proved to be sufficient for the model building.

3. Please, improve the quality of figures 4、7, and make it more proper read and clear understanding.

Thank you for the note, we improved the quality and readability of these figures.

4. Please, pay attention to superscript and subscript, such as CaCO3 in LN308,318.

We checked the whole text to fixed the superscripts and subscripts.

5. Can you give more information for why the power of sunflower productivity estimation was such low?

Yes indeed, thanks for suggesting. We provided additional information in discussion section of the manuscript.

Reviewer 2 Report

Based on multiple remote sensing and yield data, a 100 m resolution cropland productivity evaluation was conducted by random forest method. The authors found that significant difference in the accuracy between the productivity prediction of different crops and developed a feasible method for cropland productivity assessment at the country scale.

There is seldom research has focus on such topic, so the current study is on a topic of relevance and general interest to the readers of the journal. However, based on my personal evaluation, there are some flaws in the method and Analysis. Thus, the authors should clarify some of them and have a very major revision for the method. After that, the results could be further considered. Therefore, I recommend that a major revision is warranted.

 

Before publication, this manuscript could be further improved by taking the following proposals into consideration.

1) Offering more information for the whole work process in this study, especially about the AIIR field database, corresponding data preparation and processing.

2) It’s necessary to discuss the result in this study. In my personal opinion, the discussing section should be revised because the research highlights and results of this paper are not well explained. Some necessary literatures should be added for this section.

3) Better to modify the writing and tense of the full text. Moreover, the paragraph formatting should be checked carefully, such as Line 291 and 299.

4) Results need to have more specific data descriptions to support the conclusion of this study.

5) It should be noted that the structure and description of introduction need a major modification. In this part, the authors should highlight the necessity of this study. For example, compared to traditional cropland productivity evaluating method and model, the reason and advantage for using random forest method.

6) As shown in Fig3 (flowchart) and corresponding description of method, the remotely sensed GPP dataset at 500 m was used for correcting yield data and result validation. For this part, the authors used downscaled method to generate 100 m resolution GPP based on 500 m MODIS GPP data. In particularly, the topographic and spatial scaling effect could significant affect the GPP estimation, the authors need to pay attention to this and add some sentences to address it in the discussion. Some references are recommended in following section, which may provide some reference to help the authors and readers getting a better understand for this problem.

 

Some references:

https://www.sciencedirect.com/science/article/pii/S0304380010005211

https://www.sciencedirect.com/science/article/abs/pii/S0022169422014032

 

L11: “with 100 m spatial resolution (100 m)”, the 100 m in parentheses is redundant.

L88: “a.s.l.,”, is it redundant or what is its mean ?

L111: “Plant phenology indices based on long term (15 years) … … to produce GPP dataset for the whole country”. How to obtain GPP by PPI ? Moreover, is it mean that the MODIS GPP product (MOD17) was used？The data source and methodology needs to be supplemented here.

L121: The temperature and precipitation data cover range from 1951-2013, while the AIIR was 1985-1989. I am doubt about the study period of this study, the authors should to clarity it. Moreover, the period of these data used in this work should add some words to address it.

L95: More information should be added to describe the AIIR data, such as the format of it (point or shape).

Line 157: Please adding some references.

Line 177: Please describe the method for GPP and DEM downscaling.

Line 219: How to establish the general productivity model?

Tab 1 and 2: Please modify the table format. In my opinion, table of three-dash lines is better.

Fig 2: The figure is illegible, especially for the legend of subfigure 1 and 4.

Author Response

Dear Reviewer ,

 

Thank you very much for evaluating our manuscript, and providing valuable comments to improve it. We very much appreciate your positive and helpful comments.

We revised the manuscript by accepting all the suggested changes. We provide your comments below and marked our responses in red below them.

 

Based on multiple remote sensing and yield data, a 100 m resolution cropland productivity evaluation was conducted by random forest method. The authors found that significant difference in the accuracy between the productivity prediction of different crops and developed a feasible method for cropland productivity assessment at the country scale.

There is seldom research has focus on such topic, so the current study is on a topic of relevance and general interest to the readers of the journal. However, based on my personal evaluation, there are some flaws in the method and Analysis. Thus, the authors should clarify some of them and have a very major revision for the method. After that, the results could be further considered. Therefore, I recommend that a major revision is warranted.

Before publication, this manuscript could be further improved by taking the following proposals into consideration.

1) Offering more information for the whole work process in this study, especially about the AIIR field database, corresponding data preparation and processing.

We have improved the methods section, including the items you pointed at.

2) It’s necessary to discuss the result in this study. In my personal opinion, the discussing section should be revised because the research highlights and results of this paper are not well explained. Some necessary literatures should be added for this section.

We have improved the discussion with additional references and comparison of previous results of others with ours. We also tried to improve the research highlights.

3) Better to modify the writing and tense of the full text. Moreover, the paragraph formatting should be checked carefully, such as Line 291 and 299.

We have carefully reviewed the whole text and corrected the language errors and typos. We have also tried to improve the flow of the text.

4) Results need to have more specific data descriptions to support the conclusion of this study.

We have rethought and expanded the results section to make it easier to understand and presented data in a way that hopefully supports the conclusion in a straightforward way.

5) It should be noted that the structure and description of introduction need a major modification. In this part, the authors should highlight the necessity of this study. For example, compared to traditional cropland productivity evaluating method and model, the reason and advantage for using random forest method.

Following your suggestions, we have completely revised the introduction section.

6) As shown in Fig3 (flowchart) and corresponding description of method, the remotely sensed GPP dataset at 500 m was used for correcting yield data and result validation. For this part, the authors used downscaled method to generate 100 m resolution GPP based on 500 m MODIS GPP data. In particularly, the topographic and spatial scaling effect could significant affect the GPP estimation, the authors need to pay attention to this and add some sentences to address it in the discussion. Some references are recommended in following section, which may provide some reference to help the authors and readers getting a better understand for this problem.

Some references:

https://www.sciencedirect.com/science/article/pii/S0304380010005211

https://www.sciencedirect.com/science/article/abs/pii/S0022169422014032

Following your suggestions, we have added more information on the downscaling of GPP data in the Methods section and additional sentences on the issue in the Discussion section.

L11: “with 100 m spatial resolution (100 m)”, the 100 m in parentheses is redundant.

Thank you for pointing at it. We fixed it.

L88: “a.s.l.,”, is it redundant or what is its mean?

We have changed the short version of ‘altitude above sea level’, to a more common phrase “above see level”.

L111: “Plant phenology indices based on long term (15 years) … … to produce GPP dataset for the whole country”. How to obtain GPP by PPI ? Moreover, is it mean that the MODIS GPP product (MOD17) was used？The data source and methodology needs to be supplemented here.

Yes, we used MOD17 product to conduct the GPP values across the country. We made it clear in the text now.

L121: The temperature and precipitation data cover range from 1951-2013, while the AIIR was 1985-1989. I am doubt about the study period of this study, the authors should to clarity it. Moreover, the period of these data used in this work should add some words to address it.

We did not want to use the mean precipitation and temperature data characteristic of the 1985-1989 period, but to consider a wider time window that would provide more general information about climatic characteristics, both for training the model and for applying it to mapping. Therefore, we used the monthly mean data of the period 1951 and 2013.

L95: More information should be added to describe the AIIR data, such as the format of it (point or shape).

We improved our text with more information about the AIIR dataset. “The sampling for the soil tests was carried out in such a way that the parcels were divided into 12 ha sections and then, along the diagonals of the selected sections soil samples were taken from at least 20-20 locations using the so-called parallel sampling method. The subsamples were taken homogenised, so that an average sample was taken from the subplots of each agricultural field. For areas with a slope greater than 12%, average samples were taken separately for each (upper, middle, lower) section of the slope, taking into account erosion and different soil nutrient supply. The database was digitized in 2000 and in 2014 was upgraded to a modern geo-spatial database (point data with coordinates). We have selected the points that still fall on arable land at the time of our study.

Line 157: Please adding some references.

We added the references and improved the clarity by deleting these maps and writing the online reference of the digital maps.

Line 177: Please describe the method for GPP and DEM downscaling.

We added more information about that: “The GPP data, which was originally produced on a 500 m resolution was downscaled to 100m resolution and normalised to values between 1 and 100. The downscaling was performed by nearest neighbor resampling method. The SRTM data, which was originally produced on a 30 m resolution was generalised to 100m resolution with bilinear interpolation technique.”

Tab 1 and 2: Please modify the table format. In my opinion, table of three-dash lines is better.

We modified our tables.

Fig 2: The figure is illegible, especially for the legend of subfigure 1 and 4.

We deleted these maps and wrote the online reference of the digital maps.

Reviewer 3 Report

Review of the manuscript: remotesensing-2113631

Cropland productivity evaluation: a 100m resolution country assessment combining earth observation and direct measurements (authors: Csikós, N.; Szabó, B.; Her-mann, T.; Laborczi, A.; Matus, J.S.; Pásztor, L.; Szatmári, G.; Takács, K.; Tóth, G.)

The manuscript presents a methodology for a quantitative assessment of land suitability for the agricultural area of Hungary, focused on high-input agriculture. Crop-specific - in relation to the three main crops in the country: wheat, maize and sunflower - and general productivity indices were estimated applying a Random Forest technique, at a 100 m resolution. In addition, the relative weight of the different factors considered (related to soil, orography, climate and management) was established using the R bclust package. Both crop-specific and general weights were computed, and results allowed to identify the most important factors for each crop and at a more general level.

General Comments

The manuscript is interesting and fully falls within the scope of the Journal. The topic is contextualized with a sufficient literature review and study objectives are well defined. The abstract is sufficiently informative, and most of the keywords are appropriate (see Specific comments). The description of the method is well structured, but some important details about the data are missing. In particular, it is not clear which phenological indices were used and which period the MODIS time series cover, while the method adopted to combine GPP and yield data it is not specified. As for weather data, instead of using the 1951-2013 series, I suggest considering a more recent period of almost 30 years (e.g. the 1981-2010 climate normal), that can be more representative of the period of analysis (which seems to start from 1985). In addition, it is not clear how the 10-days data of precipitation and temperature were used. Another question concerns the choice of bioclimatic indices (why not to test some more crop-specific indices?). Finally, the choice to refer to a rather old dataset for information on cropland management is not justified enough: how sure can we be that agricultural management has not changed since the date of this dataset? The section on Results and discussion is well organized, even though graphs about model validation should be better explained (See specific comments). Discussion and conclusion are clear and well written, I suggest to also highlight the limitations of the study, in particular an accuracy of 40% does not ensure a fully operational use for monitoring aims. Other minor changes needed are related to English typos, some unclear sentences to be better explained and the citations, some of which do not follow the journal standards.

Specific comments:

Page 2, Line 26: I suggest to add the keyword “land evaluation” and/or “land suitability” instead of “evaluation”

Page 2, Line 71: I suggest deleting “While”

Page 4, Line 111: “plant phenology indices” is too generic, it is important to specify which indices have been used and the time period covered by the remote sensing data.

Page 7, Line 177-178: downscaling is often challenging, some details on the procedure followed may be useful.

Page 9, Line 259: replace “explain” with “explains”.

Page 9, Line 260: I am not sure that an accuracy of 40% is enough for a country scale assessment, I think discussing the results obtained by other indices such as MAPE can help to evaluate this point.

Page 10, Figure 4. The graphs are not clear, in particular the title of Y axis is confusing: are you sure the regression is between observed data and observed data, or the Y axis refers to some predicted data?

Page 11, line 292: I suggest replacing “reflects” with “reflecting”

 

Page 11, line 300: I suggest replacing “reflects” with “reflecting”

 

Author Response

Dear Reviewer,

 

Thank you very much for evaluating our manuscript, and providing valuable comments to improve our manuscript. We very much appreciate your positive and helpful comments.

We revised the manuscript by accepting all the suggested changes. We copied your comments below and marked our responses in red.

 

Cropland productivity evaluation: a 100m resolution country assessment combining earth observation and direct measurements (authors: Csikós, N.; Szabó, B.; Her-mann, T.; Laborczi, A.; Matus, J.S.; Pásztor, L.; Szatmári, G.; Takács, K.; Tóth, G.)

The manuscript presents a methodology for a quantitative assessment of land suitability for the agricultural area of Hungary, focused on high-input agriculture. Crop-specific - in relation to the three main crops in the country: wheat, maize and sunflower - and general productivity indices were estimated applying a Random Forest technique, at a 100 m resolution. In addition, the relative weight of the different factors considered (related to soil, orography, climate and management) was established using the R bclust package. Both crop-specific and general weights were computed, and results allowed to identify the most important factors for each crop and at a more general level.

General Comments

The manuscript is interesting and fully falls within the scope of the Journal. The topic is contextualized with a sufficient literature review and study objectives are well defined. The abstract is sufficiently informative, and most of the keywords are appropriate (see Specific comments).

The description of the method is well structured, but some important details about the data are missing. In particular, it is not clear which phenological indices were used and which period the MODIS time series cover, while the method adopted to combine GPP and yield data it is not specified.

We have improved our Methods/Datasets section to clarify this important point.

As for weather data, instead of using the 1951-2013 series, I suggest considering a more recent period of almost 30 years (e.g. the 1981-2010 climate normal), that can be more representative of the period of analysis (which seems to start from 1985).

We did not want to use the mean precipitation and temperature data characteristic of the 1985-1989 period, but to consider a wider time window that would provide more general information about climatic characteristics, both for training the model and for applying it to mapping. Therefore, we used the monthly mean data of the period 1951 and 2013.

 

In addition, it is not clear how the 10-days data of precipitation and temperature were used. Another question concerns the choice of bioclimatic indices (why not to test some more crop-specific indices?). Finally, the choice to refer to a rather old dataset for information on cropland management is not justified enough: how sure can we be that agricultural management has not changed since the date of this dataset?

We fixed the unclear part of the meteorological data. “The Central-European FORESEE meteorological database [36], that covers the whole area of the country in 0.1x0.1 degree grid, was used to derive mean temperature and precipitation totals at monthly intervals (between 1951 and 2013).”

10-day data was a mistake, we only used monthly data.

We have added a few sentences to the methods (AIIR dataset) to make this clear.

The section on Results and discussion is well organized, even though graphs about model validation should be better explained (See specific comments).

Discussion and conclusion are clear and well written, I suggest to also highlight the limitations of the study, in particular an accuracy of 40% does not ensure a fully operational use for monitoring aims.

Thank you for these suggestions. We improved these in discussion in L371-380.

Other minor changes needed are related to English typos, some unclear sentences to be better explained and the citations, some of which do not follow the journal standards.

Specific comments:

Page 2, Line 26: I suggest to add the keyword “land evaluation” and/or “land suitability” instead of “evaluation”

According to you suggestion we changed the keywords.

Page 2, Line 71: I suggest deleting “While”

Thank you for calling our attention to it. It is now fixed.

Page 4, Line 111: “plant phenology indices” is too generic, it is important to specify which indices have been used and the time period covered by the remote sensing data.

We improved this section.

Page 7, Line 177-178: downscaling is often challenging, some details on the procedure followed may be useful.

We added more information about that: “The GPP data, which was originally produced on a 500 m resolution was downscaled to 100m resolution and normalised to values between 1 and 100. The downscaling was performed by nearest neighbor resampling method. The SRTM data, which was originally produced on a 30 m resolution was generalised to 100m resolution with bilinear interpolation technique.”

Page 9, Line 259: replace “explain” with “explains”.

Thank you for calling our attention to it. It is now fixed

Page 9, Line 260: I am not sure that an accuracy of 40% is enough for a country scale assessment, I think discussing the results obtained by other indices such as MAPE can help to evaluate this point.

We wrote more about this in discussion in L371-380.

Page 10, Figure 4. The graphs are not clear, in particular the title of Y axis is confusing: are you sure the regression is between observed data and observed data, or the Y axis refers to some predicted data?

We have changed the titles of the axes to make them clearer. We have added a newer version of this figure.

Page 11, line 292: I suggest replacing “reflects” with “reflecting”

Thank you for calling our attention to it. It is now fixed

Page 11, line 300: I suggest replacing “reflects” with “reflecting”

Thank you for calling our attention to it. It is now fixed

Reviewer 4 Report

Dear authors,

Thank you for your research of quantitative assessment of soil biomass productivity with 100 m spatial resolution on Hungary coverage. This investigation is very relevant to the sustainable development of agriculture. 

You can find some remarks and comments below:

Lines 11-12 - You can use spatial resolution data (100m) only once.

L. 17 - Did you use only the RF method for this study?

L. 26 - Please, replace the keywords "cropland", "productivity", and "evaluation" because the title contains those words. Keywords have to contain different words than the title. You can use "Hungary" as the keyword. 

L. 35 - You should put references on national and international programs.

L. 66 - Please, add below additional data about significant crop types, for example, structure croplands in Hungary.

L. 97 - Do you mean around 80 thousand? Fields? 

L. 151 - Figure 2 has too tiny legends. Did you improve maps from dosoremi.hu?

L. 233 – Please, improve the reference.

L. 240 – Please, improve the reference.

L. 266-277 – You can improve the title of figure 4. 

L. 295 – The legend of the map has to be bigger. 

L. 334 – You should add comparing your results with other studies. 

L. 378-391 – Conclusions must be improved. 

Author Response

Dear Reviewer,

 

Thank you very much for evaluating our manuscript, and providing valuable comments to improve our manuscript. We very much appreciate your positive and helpful comments.

We revised the manuscript by accepting all the suggested changes. We copied your comments below and marked our responses in red.

 

Thank you for your research of quantitative assessment of soil biomass productivity with 100 m spatial resolution on Hungary coverage. This investigation is very relevant to the sustainable development of agriculture.

You can find some remarks and comments below:

Lines 11-12 - You can use spatial resolution data (100m) only once.

Thank you for your note. Done as suggested.

1. 17 - Did you use only the RF method for this study?

Yes, just random forest has been used. Clarified this more in the revised text.

1. 26 - Please, replace the keywords "cropland", "productivity", and "evaluation" because the title contains those words. Keywords have to contain different words than the title. You can use "Hungary" as the keyword.

We deleted them and used others, including Hungary and Gross Primary Productivity as keywords.

1. 35 - You should put references on national and international programs.

Thank you we named some national and international program in the revised manuscript.

1. 66 - Please, add below additional data about significant crop types, for example, structure croplands in Hungary.

We have added some information about this in the Introduction and Methods sections.

1. 97 - Do you mean around 80 thousand? Fields?

Yes, the AIIR database covers the full agricultural area of the country and contains information about 80,000 parcels in Hungary. (In the 80’s it was and exceptional good system, established for planning nutrient management and agricultural production.)

1. 151 - Figure 2 has too tiny legends. Did you improve maps from dosoremi.hu?

We deleted the Figure 2, because these maps are available in better quality on the website.

1. 233 – Please, improve the reference.

Thank you for calling our attention to it. It is now fixed.

1. 240 – Please, improve the reference.

Thank you for calling our attention to it. It is now fixed.

4. 266-277 – You can improve the title of figure 4.

We improved the Figure 4. title: Model validation results for estimating total cropland area, wheat, maize and sunflower productivity. Results were significant on a 0.01 level (GPP means Gross Primary Productivity).

1. 295 – The legend of the map has to be bigger.

We increased the font size of the text.

1. 334 – You should add comparing your results with other studies.

Following your suggestion, we have added some references to the discussion and compared our result with them.

1. 378-391 – Conclusions must be improved.

We wrote additional text in the conclusion to improve it.

Round 2

Reviewer 1 Report

This manuscript has made proper revisions.

Author Response

Thank you for your earlier positive and helpful review.

Reviewer 2 Report

Thank you for the author's efforts to revise, the quality of the revised paper has been greatly improved, before publication, this manuscript could be further improved by taking the following proposals into consideration.

1、         L65-67 can be move to study area.

2、         The basis for the selection of independent variables in the random forest method and the importance of each independent variable should be clarified more clearly. The dynamic change of the independent variable limits the universality of the method

3、         The estimation accuracy of each model is only 40%, which is too low, the spatial resolution of meteorological data is too coarse, the resolution of TerraClimate is 4638.3m，the worldCLIM. Sentinel-2 land cover has 10m resolution. Therefore, more high spatial resolution data can improve the accuracy of prediction.

 

4、         The standardization of the charts in the article needs to be further improved, how to make the charts more beautiful, please refer to ‘Early-season mapping of winter wheat and garlic in Huaihe basin using Sentinel-1/2 and Landsat-7/8 imagery’, the paper use RF to construct two model to classify the winter wheat and garlic, and have high classified accuracy.

5、         The font and color of the axis titles of each figure should be unified, such as Figure 3c and d

6、         The legend in Figure 3 can be placed in the frame

7、         A\B\C\D fonts in Figure 4 are too large

8、         The size of the compass, legend, and scale bar of each map should be coordinated to increase the standardization of drawing

Author Response

Response to Reviewer Comments

 

Dear Reviewer!

Once again, thank you very much for your valuable comments to improve our manuscript.

We have copied your comments below and marked our responses in red.

 

1. L65-67 can be move to study area.

We have moved lines 65-67 to the study area section.

2. The basis for the selection of independent variables in the random forest method and the importance of each independent variable should be clarified more clearly. The dynamic change of the independent variable limits the universality of the method

You are correct that a clearer explanation of the basis for the selection of independent variables and the importance of each independent variable would improve the understanding of our methodology. Our standpoint was that only those variables can be tested, which are available for the whole country, namely those soil variables which are mapped and those climatic and terrain variables, which are also available for the whole area of Hungary. However, we agree that if the assessment was performed for a smaller area with more (or different) variables, we could set up a model that fits better for this small area.

Regarding the dynamic change of independent variables, we agree that this can limit the universality of the method. Our proposal is to perform the analysis time to time, so the changes can be implemented in future productivity models, hopefully using richer databases with higher spatial, temporal and thematic detail.

We included additional text to the revised manuscript to reflect your points. 

3. The estimation accuracy of each model is only 40%, which is too low, the spatial resolution of meteorological data is too coarse, the resolution of TerraClimate is 4638.3m，the worldCLIM. Sentinel-2 land cover has 10m resolution. Therefore, more high spatial resolution data can improve the accuracy of prediction.

Regarding the accuracy of the models, we understand that a 40% estimation accuracy may not meet some expectations. However, it is important to note that our study was focused on exploring the potential use and importance of these independent variables for a country scale prediction. The heterogeneity we had to deal with is partly reflected in our Table 1. However, there is a great degree of heterogeneity in soil units shown in the table, eg. in texture, pH, horizons etc, plus, in one third of the area, of which soil units are not represented in Table 1, a wide range of specific soil types are included, eg. a variety of saline and sodic soils, drained organic soils, shallow rendzina soils, arenosols etc. (for an impression please see: https://esdac.jrc.ec.europa.eu/Awareness/Documents/EU_Presidency/poster1_en.pdf) Furthermore we could use a limited number (even if we believe that those are the most important ones) of soil properties in our assessment, as mentioned above. 

Regarding the spatial resolution of the meteorological data, we acknowledge that higher resolution data can potentially improve the accuracy of predictions. However, the resolution of 1 km data (WorldClim) was chosen based on the available data for the full study period. We agree that new sentinel data can improve the predictions and we hope that our planned repeated testing in the future will prove your suggestion and justify our intention to do so.

4. The standardization of the charts in the article needs to be further improved, how to make the charts more beautiful, please refer to ‘Early-season mapping of winter wheat and garlic in Huaihe basin using Sentinel-1/2 and Landsat-7/8 imagery’, the paper use RF to construct two model to classify the winter wheat and garlic, and have high classified accuracy.

As suggested in the 4 points below, we have rearranged our figures, font type, size, colour have been harmonized to make them consistent.

We also referred to the article in the introduction and discussion.

5. The font and color of the axis titles of each figure should be unified, such as Figure 3c and d

Thank you for your suggestion! The font and color of the text is indeed different, we have corrected it.

6. The legend in Figure 3 can be placed in the frame

We have moved the legend inside the frames.

7. A\B\C\D fonts in Figure 4 are too large

As requested, we have reduced the font size in Figure 4.

8. The size of the compass, legend, and scale bar of each map should be coordinated to increase the standardization of drawing

We made the necessary revisions to ensure that the size of the compass, legend, and scale bar are coordinated in each map in the article.

Reviewer 3 Report

 

The updated manuscript has been considerably improved: the writing is clearer and more effective in all sections.

Most suggestions have been incorporated. The revised sections of methods and results are clearer and more comprehensive. In particular, the new version of the flowchart in figure 2 (ex figure 3) clarifies some important details on methods. However, further information should be still added to improve the understanding (see specific comments). Other minor changes needed are related to English typos, some unclear sentences to be better explained

Specific comments

Page 2, line 58: “as well” should be deleted

Page 2, line 60: please, add a comma between “conditions” and “need”

Page 2, lines 65-68, from “Winter…”  to “…assessment.”: this sentence is not a general consideration, but it concerns the analysis presented in the manuscript; I suggest moving it at the end of the introduction, where the study objectives are reported. (page 3, lines 95-102)

Page 3, line 98: I suggest changing “mai” with “main” and adding “to” before “produce…”

Page 3, line 101: please, add “to” before “propose…”

Page 4, line 137: some references and further explanations on the parallel sampling method could improve readability.

Page 5, lines 160-163, from ”We used…”: this sentence should be rephrased, I suggest replacing “because” with a colon.

Page 6, line 166-167: I suggest replacing “in 0.1x0.1…” with “with a 0.1x0.1…” and “precipitation totals at monthly intervals” with “total precipitation at monthly scale”

Page 6, line 169-170: how was this downscaling (from 0.1 degrees to 100 m) carried out?

Page 8, line 253: I suggest 1) adding “was computed” between “GPP data” and “by taking” 2) replacing “the two datasets” with “the two normalized datasets”

Page 8, line 266-267: this sentence is incomplete

Page 9, lines 301-307: I suggest to call this 10% of data as “validation data” to avoid confusion with the other 10% of test data used to calibrate the model

Page 13, line 335: I suggest replacing  “on a 0.01 level” with “at the 0.01 level”

Page 13, line 342: I suggest to write “followed by” instead of “than follow”

Page 19, Figure 6: In the y axis, the variables Cor.EOV.x and Cor.EOV.y are not very clear: I suggest calling them “East coordinate” and “North coordinate”

Page 20, line 415: write “seem” instead of “seems”

Page 20, lines 430-433, from “While the R2 value….” to “other studies”: it should be specified that this comment concerns only wheat.

Page 21, line 465: I suggest replacing “is surpassing” with “outweighs”

Page 21, line 500-501: repetitions (“also”) should be avoided

Page 22, line 512: I suggest writing “at 100 m resolution” instead of “on 100 m resolution”

Page 22, line 519: this line should be rephrased

 

Author Response

Response to Reviewer Comments

 

Dear Reviewer!

Once again, thank you very much for your valuable comments to improve our manuscript, we appreciate your positive and helpful comments.

We have copied your comments below and marked our responses in red.

 

Most suggestions have been incorporated. The revised sections of methods and results are clearer and more comprehensive. In particular, the new version of the flowchart in figure 2 (ex figure 3) clarifies some important details on methods. However, further information should be still added to improve the understanding (see specific comments). Other minor changes needed are related to English typos, some unclear sentences to be better explained

Specific comments:

Page 2, line 58: “as well” should be deleted

Fixed

Page 2, line 60: please, add a comma between “conditions” and “need”

Fixed

Page 2, lines 65-68, from “Winter…”  to “…assessment.”: this sentence is not a general consideration, but it concerns the analysis presented in the manuscript; I suggest moving it at the end of the introduction, where the study objectives are reported. (page 3, lines 95-102)

Thank you for your suggestion, we mention that also at the end of the introduction and we moved this sentence at the end of the study area.

Page 3, line 98: I suggest changing “mai” with “main” and adding “to” before “produce…”

Fixed

Page 3, line 101: please, add “to” before “propose…”

Fixed

Page 5, lines 160-163, from ”We used…”: this sentence should be rephrased, I suggest replacing “because” with a colon.

We have fixed it according to your suggestion.

Page 6, line 166-167: I suggest replacing “in 0.1x0.1…” with “with a 0.1x0.1…” and “precipitation totals at monthly intervals” with “total precipitation at monthly scale”

We corrected this sentence.

Page 6, line 169-170: how was this downscaling (from 0.1 degrees to 100 m) carried out?

The downscaling was performed by bilinear resampling method. We added this sentence where you indicated.

Page 8, line 253: I suggest 1) adding “was computed” between “GPP data” and “by taking” 2) replacing “the two datasets” with “the two normalized datasets”

Thank you, we have improved it based on your suggestions.

Page 8, line 266-267: this sentence is incomplete

We connected this sentence to the next one.

Page 9, lines 301-307: I suggest to call this 10% of data as “validation data” to avoid confusion with the other 10% of test data used to calibrate the model

We fixed that.

Page 13, line 335: I suggest replacing  “on a 0.01 level” with “at the 0.01 level”

Fixed.

Page 13, line 342: I suggest to write “followed by” instead of “than follow”

Fixed.

Page 19, Figure 6: In the y axis, the variables Cor.EOV.x and Cor.EOV.y are not very clear: I suggest calling them “East coordinate” and “North coordinate”

According to your suggestion we changed the text of the figure 6.

Page 20, line 415: write “seem” instead of “seems”

Fixed

Page 20, lines 430-433, from “While the R2 value….” to “other studies”: it should be specified that this comment concerns only wheat.

Thank you for your note, we fixed that.

Page 21, line 465: I suggest replacing “is surpassing” with “outweighs”

We fixed that according to your suggestion.

Page 21, line 500-501: repetitions (“also”) should be avoided

Second also has deleted.

Page 22, line 512: I suggest writing “at 100 m resolution” instead of “on 100 m resolution”

Fixed.

Page 22, line 519: this line should be rephrased

We rephrased this sentence.

Reviewer 4 Report

Dear authors, thank you so much for improving the article.

It will be more apparent if you change the legend's text in figure 4. All parts of figure 4 have the similar text "Mean weighted land evaluation value ..." and A, B, C, and D mean specific data.

The text of figure 6 looks quite small for reading.

Author Response

Dear Reviewer!

Once again, thank you very much for your valuable comments to improve our manuscript.

We changed the Figure 4 and 6 according to your suggestions.