Feature Comparison and Optimization for 30-M Winter Wheat Mapping Based on Landsat-8 and Sentinel-2 Data Using Random Forest Algorithm

Round 1

Reviewer 1 Report

Feature Comparison and Optimization for 30-M Winter Wheat Mapping Based on Landsat-8 and Sentinel-2 Data Using Random Forest Algorithm

Reviewer comments.

Dear Editors

Based on your review request, I have gone through the manuscript. Authors mapped winter wheat using Landsat8 and Sentinal-2 data using Random Forest algorithm. Authors applied their method on a small study area. Which is week part in this paper otherwise paper is well written. However Authors need to revise manuscript. I recommend major revision.

Dear Authors.

Manuscript on “Feature Comparison and Optimization for 30-M Winter Wheat Mapping Based on Landsat-8 and Sentinel-2 Data Using Random Forest Algorithm” is well written. However there are several comments need to be addressed for improvement of this manuscript.

Ln 15-16

Authors stated “Mapping 30-m winter wheat cropland can provide beneficial reference information necessary for understanding food security.”  But applied for a small study area.

Ln 33-35

Algorithm applied on a small study area. How can we generalize this for a large regions?

Ln 60-65

Authors provided web link for the GFSAD30 cropland products.

These products have publications to cite. Better to cite articles instead of website.

Ln-76-77

Spectral features[4 ],  Spatial features[5,6], temporal features[7,8]  and auxiliary features[9]

What are these features, give an example for each category.  It is easy to understand to readers.

Ln 79-80

“Furthermore, this can help retrieve higher-value, deeper information at larger scales[11] “

Authors cite only one reference here. There are several recent publications on large scale studies using Random forest approach applied on Landsat data. Please cite up-to-date literature (recent publications).

Ln-87-88

The random forest algorithm has been used extensively in land-cover classification applications [16-21].

Authors didn't cite any recent articles on RF classification for example Xiong et al., 2017, Teluguntla et al., 2018

Ln 101

Recent literature not cited. So needs to be updated. 

Ln 124: Table 1

 Row 4: USGS (GFSAD30 & Landsat

No references cited for these products. Recently I have seen some papers are published on these products. Better to cite them.

Ln 168

Authors did used any ground reference data for training. Without ground data, how you rely on secondary data for training.  

Ln: 349 (Table 3)

Authors shown only OA, PA and UA, but not shown number of samples used for accuracy assessment. Which is very important. Please provide full details of validation samples used.

Ln 429 (Figure11)

What is on Y axis? only mentioned TOA is it NDVI or EVI? please revise the y axis label.

Line 59:

Replace “Sier” with “Syr Darya”

Author Response

Point 1: Ln 15-16

Authors stated “Mapping 30-m winter wheat cropland can provide beneficial reference information necessary for understanding food security.”  But applied for a small study area.


Response 1: Thank you for pointing this out. We have supplement this point in Section 4.5, which is related to the discussion of advantages and limitations of approach in this article. (Annotation 60) (Page 21, line 1)

This paper proves that crop recognition and extraction with high accuracy can be achieved at regional scale by using limited and reasonable combination of feature spaces. To a certain extent, this paper only achieved 30-m winter wheat mapping in a small study area. Based on the experimental results of this paper, there may be some limitations in large-scale crop extraction research. However, the idea of simplifying and optimizing feature space can also provide a reference for crop extraction on a large scale to a certain extent.

Firstly, through the analysis and discussion of feature space in this paper, we can choose efficient feature variables as input variables of classifier from the two aspects of spectral and temporal features in future large-scale regional mapping.  (As we discussed in 4.2) Secondly, for crop information extraction in large regions, considering operational efficiency and accuracy, it is usually necessary to divide large areas into certain surface units, then calculating information for each surface unit separately and finally integrating the results of partitioning into the results of the whole large area. In this paper, the influence factors of the extraction accuracy of three different regions discussed in 4.3 can provide some ideas for the future research of information extraction by block. For example, when large regions cover farming areas, urban mixed areas and hilly areas, we can separate them according to their different geomorphological and environmental characteristics. For separated farming areas, it is helpful to focus on building features based on band information. As for urban mixed areas, attentions can be paid on features that based on spectral index. Through this targeted feature building process, it can help to achieve rapid and accurate information extraction.

Of course, information extraction at large regional scale is also facing a variety of other problems and challenges. Possible factors are still under consideration in the process of implementation to further improve the theory and research methods.

Point 2: Ln 33-35

Algorithm applied on a small study area. How can we generalize this for a large regions?

Response 2: Thank you for pointing this out. According to you suggestion, we have made related discussion about how we can generalize this for a large regions in the future studies in Section 4.5. (Annotation 61) (Page 21, line 1)

Point 3: Ln 60-65

Authors provided web link for the GFSAD30 cropland products.

These products have publications to cite. Better to cite articles instead of website.

Response 3: We agree to this point and many thanks for pointing this out. We have added related references in this point. (Annotation 4) (Page 2, line 23)

Point 4: Ln-76-77

Spectral features[4],  Spatial features[5,6], temporal features[7,8]  and auxiliary features[9]

What are these features, give an example for each category.  It is easy to understand to readers.

Response 4: We agree to this point and many thanks for pointing this out. We have added related references in this point. (Annotation 8) (Page 2, line 34-37)

At the same time, this part also had a further summary in Section 4.2, which is easy for readers to understand.

 Section 4.2:

“(1) a basic original broad band, such as the visible green, red, and near-red bands; (2) a narrow band carrying key information for vegetation, such as the red edge band (band selection mainly depends on the spectral resolution of the satellite sensor); (3) principal component analysis used to produce a band consisting of the new dimension obtained by the spatial transformation of the original band (moreover, it can reduce the dimensionality of the original wave band); and (4) a vegetation index obtained using band calculations in the key position of the spectral curve for different ground objects.”

Point 5: Ln 79-80

“Furthermore, this can help retrieve higher-value, deeper information at larger scales[11] “

Authors cite only one reference here. There are several recent publications on large scale studies using Random forest approach applied on Landsat data. Please cite up-to-date literature (recent publications).

Response 5: We agree to this point and many thanks for pointing this out. We have added related references in this point. (Annotation 10) (Page 2, line 39)

Point 6: Ln-87-88

The random forest algorithm has been used extensively in land-cover classification applications [16-21].

Authors didn't cite any recent articles on RF classification for example Xiong et al., 2017, Teluguntla et al., 2018

Response 6: We agree to this point and many thanks for pointing this out. We have added related references in this point. (Annotation 13) (Page 2, line 48)

Point 7: Ln 101

Recent literature not cited. So needs to be updated. 

Response 7: Thanks for pointing this out. We have added related references in this point. (Annotation 15) (Page 3, line 8)

Point 8: Ln 124: Table 1

 Row 4: USGS (GFSAD30 & Landsat

No references cited for these products. Recently I have seen some papers are published on these products. Better to cite them.

Response 8: Thanks for pointing this out. We have added related references for the related products. (Annotation 21) (Page 3, line 34)

Point 9: Ln 168

Authors did used any ground reference data for training. Without ground data, how you rely on secondary data for training.

Response 9: Thank you for pointing this out. Training data is exactly important for classification model. In order to ensure the reliability of the sample, we control it in two aspects.

Firstly, in Section 4.3, we pointed out that “The corresponding types of samples were marked in the same coordinate position as the satellite images based on the crop types in the reference data.” The reference data used is the data of cultivated land utilization in the whole year of 2016 and the boundary of land blocks is stable in one year, so as long as the crop types on the classification map are compared, the corresponding types of land blocks can be judged on the remote sensing images of the corresponding locations. In this way, the quality of most training samples can be guaranteed.

Secondly, we considered four cases occurred when random sampling points were scattered on remote sensing images in Section 4.2. These occurred when: 1) wheat pixels covered the sampling point, 2) non-wheat pixels covered the sampling point, 3) sampling points were near wheat pixels, and 4) when sampling points were near non-wheat pixels. For the first two cases, the pixel blocks in which the sample points fell were directly marked as training samples for the corresponding categories. For the latter two cases, the pure wheat and non-wheat pixel blocks that were closest to the sampling point were manually selected. In this interactive and positive action, we can control the quality and purity of sample data for training.

Point 10: Ln: 349 (Table 3)

Authors shown only OA, PA and UA, but not shown number of samples used for accuracy assessment. Which is very important. Please provide full details of validation samples used.

Response 10:We agree with this comment. We have added related information about our validation samples. (Annotation 44) (Page 10, line 27-28 ~ Page 11, line1-2)

Point 11: Ln 429 (Figure11)

What is on Y axis? only mentioned TOA is it NDVI or EVI? please revise the y axis label.

Thank you for pointing this. The Y axis mentioned is NDVI and is revised to right version after carefully check.

Response 11: Thank you for pointing this out. We now rectified the figure into right version and the Y axis mentioned is NDVI. (Annotation 54) (Page 17, line 23)

Point 12: Line 59:

Replace “Sier” with “Syr Darya”

Response 12: Thank you for pointing this out. We now rectified the figure into right version under your suggestion. (Annotation 3) (Page 2, line 17)

We appreciate for your warm work earnestly, and hope that the correction will meet with approval. Thank you very much for your comments and suggestions.

Author Response File: Author Response.docx

Reviewer 2 Report

Dear authors

are some comments about your paper, mostly editorial.

→ means "replace with". For instance

16: performed → conducted

means : replace "performed" by "conducted" on line 16

Best regards

R.

52 (corn, wheat, and rice) → (corn i.e. maize, wheat, and rice)

Comment: “Corn” is an Americanism, just like “Guinea corn” which is used for “sorghum” in anglophone West Africa. The name of the plant is “maize”. Please use “maize” or explain that “corn” is actually maize, as in my suggestion above. Since later parts of your paper actually refer to maize, I suggest to drop “corn” completely and to use the proper name “maize”.

67

Comment: “productivity” is an ambiguous term and is best avoided. Do you mean “yield” or “production”. Please use either “yield” or “production” or define what you mean by “productivity”.

69: concentrated at low resolutions

Comment: Not clear. Do you mean “confined to low resolutions” or “applicable only at low resolutions”?

73-74: gradually led to the multi-type →  gradually led to the adoption of the multi-type

78-79:  This results in the production of high-value agricultural remote sensing data products.

Comment: This has already been said several times. There’s no need to repeat it again.

84: technique based → technique frequently based

Comment: Random forest does not always use bagging

99: absorption characteristics → interference

110: and ensuring food security → help improve food security

Comment: no satellite has ever nor will ever ensure food security. If we are lucky, it can help, i.e. contribute to food security

114: especially in economically developing countries → also in economically developing countries

116-7: due to the spectral similarity between wheat crops and the heterogeneity of crop planting regions

Comment: not clear. I assume you don’t mean the “similarity between wheat and the heterogeneity”, which are different things. So you probably mean the similarity between several types of what crops. What is this? Winter wheat and summer wheat? You need to reword this.

128: high latitude

Comment: drop this. Latitude is not a relevant factor.

129: less → low

143: north to south → south to north

Comment: the river may meander a bit, but if flows from Czechia to the North Sea, which is S → N

143-4 cultivated areas representing a typical winter wheat crop field in Germany. → cultivated, typically with winter wheat.

153-4

Indicate the year too, not only the day

158-9: The table in (Figure 2) displays the phenology period of the main crops → Figure 2 displays the phenology of the main crops

Comment: winter wheat and rape phenology is not correct. The crops are planted in autumn, not in January. Also: what is “early_Age” and “Mid_Age”????

174-6: It also provided validation data for manually delineated ground truth polygons on a high-resolution map downloaded from Google Earth.

Comment: this not clear. Did you delineate the polygons or did you find the polygons prepared by someone else? If this is the case, you must indicate the details of the source.

185: n trees were carried out as n training sets → n trees were used as training sets

Comment: one cannot “carry out” trees.

195

Comment: this reminds me of 116-7 above. What do you mean by “type of wheat”? Maybe barley? Rye? Different wheat varieties? Spring/winter wheat?

283: enough → sufficient

291: higher than other seasons → higher than in other seasons

318: It is revealed in (Figure 7) that the classification results → Figure 7 shows the classification results

326: Results revealed that the lowest overall accuracy → The lowest overall accuracy

Comment: this paragraph has too many “revelations” (“reveal” is used 4 times, I think). The word is too strong. “Revealing” is making known something which was hidden or secret. Use “show”, “display”, “confirm”, “indicate” etc.

330: can meet → is adequate for

Comment: the wording above is my interpretation of “meet”

331: Results revealed that the lowest → The lowest

353: precision results among → precision among

373: Fig. 10 → Figure 10

410: Vegetation is one of Earth’s features that have an observable growth cycle → Vegetation is one of Earth’s features that have an observable cycle

Comment: this is not quite true. It applies to annual crops only. The Amazon forest, the high mountains or the deserts have very little of a “growth” cycle. Drop “growth”.

411: and surface climatic zone → and on climate.

411-2: Crop is one of special vegetation types that are → Crop are a special vegetation type that is

472: can be better → is best

484: were observed to exhibit distinct → exhibit distinct

486: mountainous → hilly

Comment: there are no mountains in Sachsen -Anhalt

p { margin-bottom: 0.1in; line-height: 120%; background: transparent none repeat scroll 0% 0%; }

Author Response

Point 1: 16: performed → conducted

means : replace "performed" by "conducted" on line 16

Response 1: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 11) (Page 2, line 42)

Point 2:52 (corn, wheat, and rice) → (corn i.e. maize, wheat, and rice)

Comment: “Corn” is an Americanism, just like “Guinea corn” which is used for “sorghum” in anglophone West Africa. The name of the plant is “maize”. Please use “maize” or explain that “corn” is actually maize, as in my suggestion above. Since later parts of your paper actually refer to maize, I suggest to drop “corn” completely and to use the proper name “maize”.

Response 2: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 2) (Page 2, line 10)

Point 3:67

Comment: “productivity” is an ambiguous term and is best avoided. Do you mean “yield” or “production”. Please use either “yield” or “production” or define what you mean by “productivity”.

Response 3: Thank you for pointing this out. We mean the “yield” and have rectified this into right version under your suggestion. (Annotation 5) (Page 2, line 25)

Point 4:69: concentrated at low resolutions

Comment: Not clear. Do you mean “confined to low resolutions” or “applicable only at low resolutions”?

Response 4: Thank you for pointing this out. We mean the “confined to low resolutions” and now rectified this into right version under your suggestion. (Annotation 6) (Page 2, line 28)

Point 5:73-74: gradually led to the multi-type →  gradually led to the adoption of the multi-type

Response 5: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 7) (Page 2, line 31,32)

Point 6:78-79:  This result in the production of high-value agricultural remote sensing data products.

Comment: This has already been said several times. There’s no need to repeat it again.

Response 6: Thank you for pointing this out. We have deleted this sentence under your suggestion. (Annotation 9) (Page 2, line 38)

Point 7:84: technique based → technique frequently based

Comment: Random forest does not always use bagging

Response 7: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 12) (Page 2, line 44)

Point 8:99: absorption characteristics → interference

Response 8: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 14) (Page 3, line 8)

Point 9:110: and ensuring food security → help improve food security

Comment: no satellite has ever nor will ever ensure food security. If we are lucky, it can help, i.e. contribute to food security

Response 9: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 17) (Page 3, line 19)

Point 10:114: especially in economically developing countries → also in economically developing countries

Response 10: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 19) (Page 3, line 27)

Point 11:116-7: due to the spectral similarity between wheat crops and the heterogeneity of crop planting regions

Comment: not clear. I assume you don’t mean the “similarity between wheat and the heterogeneity”, which are different things. So you probably mean the similarity between several types of what crops. What is this? Winter wheat and summer wheat? You need to reword this.

Response 11: Thank you for pointing this out. We now rectified this into the version of “…due to the spectral similarity between different wheat varieties” to meet your suggestion. (Annotation 20) (Page 3, line 29)

Point 12:128: high latitude

Comment: drop this. Latitude is not a relevant factor.

Response 12: Thank you for pointing this out. We now have dropped this to meet your suggestion. (Annotation 24) (Page 4, line 8)

Point 13:129: less → low

Response 13: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 25) (Page 4, line 8)

Point 14:143: north to south → south to north

Comment: the river may meander a bit, but if flows from Czechia to the North Sea, which is S → N

Response 14: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 22) (Page 4, line 5)

Point 15:143-4 cultivated areas representing a typical winter wheat crop field in Germany. → cultivated, typically with winter wheat.

Response 15: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 23) (Page 4, line 5,6)

Point 16:153-4

Indicate the year too, not only the day

Response 16: We agree to this point. We now have indicated the year to meet your suggestion. (Annotation 28) (Page 5, line 1-3)

Point 17:158-9:

The table in (Figure 2) displays the phenology period of the main crops → Figure 2 displays the phenology of the main crops

Response 17: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 29) (Page 5, line 4)

Comment: winter wheat and rape phenology is not correct. The crops are planted in autumn, not in January. Also: what is “early_Age” and “Mid_Age”????

Response: Thank you for pointing this out. The mean crop growth period in the area was investigated from the Growth stages of mono-and dicotyledonous plants released by Federal Biological Research Centre for Agriculture and Forestry in 2001. This is a credible document about the extended BBCH-scale, which is explained as ‘The extended BBCH-scale is a system for a uniform coding of phenologically similar growth stages of all mono- and dicotyledonous plant species. It results from teamwork between the German Federal Biological Research Centre for Agriculture and Forestry (BBA), the German Federal Office of Plant Varieties (BSA), the German Agrochemical Association (IVA) and the Institute for Vegetables and Ornamentals in Grossbeeren/Erfurt, Germany (IGZ).’

According to the principal growth stage of wheat displayed in this paper, as well as related paper of this area--- ‘A new method for crop classification combining time series of radar images and crop phenology information’ authored by Damian Bargiel. These two papers were cited in our paper of reference number 38 in section 2.2 and number 8 in Introduction section.

We also supplement definition about “Early_Age” and “Mid_Age”in section 2.2. (Annotation 30) (Page 5, line 6-11)

Point 18:174-6: It also provided validation data for manually delineated ground truth polygons on a high-resolution map downloaded from Google Earth.

Comment: this not clear. Did you delineate the polygons or did you find the polygons prepared by someone else? If this is the case, you must indicate the details of the source.

Response 18: Thank you for pointing this out. We have added the details of the source in this part to meet your suggestion. (Annotation 32) (Page 5, line 24-27)

Point 19:185: n trees were carried out as n training sets → n trees were used as training sets

Comment: one cannot “carry out” trees.

Response 19: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 34) (Page 6, line 14)

Point 20:195

Comment: this reminds me of 116-7 above. What do you mean by “type of wheat”? Maybe barley? Rye? Different wheat varieties? Spring/winter wheat?

Response 20: Thank you for pointing this out. We now rectified this into the version of “…winter wheat or the others… inside the respective pixel boundary” to meet your suggestion. (Annotation 35) (Page 6, line 24)

Point 21:283: enough → sufficient

Response 21: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 38) (Page 9, line 16)

Point 22:291: higher than other seasons → higher than in other seasons

Response 22: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 40) (Page 9, line 24)

Point 23:318: It is revealed in (Figure 7) that the classification results → Figure 7 shows the classification results

Response 23: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 42) (Page 11, line 8)

Point 24:326: Results revealed that the lowest overall accuracy → The lowest overall accuracy

Comment: this paragraph has too many “revelations” (“reveal” is used 4 times, I think). The word is too strong. “Revealing” is making known something which was hidden or secret. Use “show”, “display”, “confirm”, “indicate” etc.

Response 24: We agree with that comments. We now rectified this into different words under your suggestion. (Annotation 45) (Page 11, line 18)

Point 25:330: can meet → is adequate for

Comment: the wording above is my interpretation of “meet”

Response 25: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 46) (Page 11, line 22)

Point 26:331: Results revealed that the lowest → The lowest

Response 26: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 47) (Page 11, line 23)

Point 27:353: precision results among → precision among

Response 27: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 49) (Page 15, line 2)

Point 28:373: Fig. 10 → Figure 10

Response 28: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 50) (Page 16, line 6)

Point 29:410: Vegetation is one of Earth’s features that have an observable growth cycle → Vegetation is one of Earth’s features that have an observable cycle

Comment: this is not quite true. It applies to annual crops only. The Amazon forest, the high mountains or the deserts have very little of a “growth” cycle. Drop “growth”.

Response 29: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 51) (Page 18, line 3)

Point 30:411: and surface climatic zone → and on climate.

Response 30: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 52) (Page 18, line 4)

Point 31:411-2: Crop is one of special vegetation types that are → Crop are a special vegetation type that is

Response 31: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 53) (Page 18, line 4,5)

Point 32:472: can be better → is best

Response 32: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 56) (Page 19, line 24)

Point 33:484: were observed to exhibit distinct → exhibit distinct

Response 33: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 57) (Page 19, line 37)

Point 34:486: mountainous → hilly

Comment: there are no mountains in Sachsen –Anhalt

Response 34: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 58) (Page 19, line 39)

We appreciate for your warm work earnestly, and hope that the correction will meet with approval. Thank you very much for your comments and suggestions.

Author Response File: Author Response.docx

Reviewer 3 Report

Whereas crop specific remote sensing is an important topic, it is not clear how this study contributes towards advancement in this field.  Authors do not describe why they chose the specific study area.  Table 3 has a list of features with various subscripts that are not well explained or why they were chosen.  Authors also look at classification as a function of zone, but does not explain why they decided to look at these specific categories.  Results are not discussed in the context of prior work in this area. The conclusion section does not point out novel aspects of this research or how it compares to prior studies

Author Response

Point 1: Whereas crop specific remote sensing is an important topic, it is not clear how this study contributes towards advancement in this field.

Response 1:

Thank you for pointing this out. As for crop specific remote sensing is an important topic,  this paper conducted experiment using the prevalent random forest algorithm for its robust ability in cropland information extraction to achieve fine mapping precision at 30-m resolution, This study can contribute to advancement in this field for the below reasons:

In view of the heavy feature construction in machine learning classification, this paper explores the reasonable and efficient feature combination method for crop information extraction under multi-features (MFs), multi-feature setups (MFSs), multi-patterns (MPs) input in multi-zones. This paper also provided an idea that the MFSs combined with the three to four feature types, which derived from spectral reflectance and crop growth period characteristics, will enable high-precision extraction of 30-m winter wheat plots. For future crop remote sensing information recognition, it can provide an efficient way to construct feature space.

Secondly, this

paper explores the potential factors affecting the extraction results in three different regional environments. Specifically, for farming areas the recognition accuracy can be improved by improving the spectral resolution of images; for urban mixed areas where many kinds of land and objects are mixed, the accuracy can be improved by   constructing effective spectral index; for hilly areas, temporal information between different vegetation may be the main factor affecting classification accuracy due to complex vegetation landforms. Through this discussion, in the future crop information extraction work, we can construct feature space according to the local regional environmental characteristics, so as to ensure accuracy and improve work efficiency.

Based on these two points, this study will be helpful to provide positive research ideas for future regional scale crop extraction.

Point 2: Authors do not describe why they chose the specific study area.

Response 2:

Thank you for pointing this out. The study area chosen in this paper is the central part of Germany. The reasons for choosing this area are as follows:

Firstly, this region is a typical crop growing area in Germany, and it is a suitable choice for studying crop information extraction.

Secondly, the fragmentation of crop cultivation plots in Germany is very high. When Landsat-8 and entinel-2 are used as data sources, this feature can highlight the advantages of multi-spectral data and random forest method in retaining detailed information. The applicability and reliability of the experimental method will be improved when it is moved to larger blocks when the experimental accuracy of extraction in fractured areas is high.

At the same time, for the fragmented land in Germany, the mixed pixel problem will give no guarantee to the accuracy of crop extraction with the use of moderate resolution data such as MODIS.

From these two aspects, the German research area and remote sensing data sources are complementary to each other in a sense, so this paper chooses the research area in Germany.

Point 3: Table 3 has a list of features with various subscripts that are not well explained or why they were chosen.

Response 3: Thank you for pointing this out. We have added more information about this part in Section 2.4.2. (Annotation 37) (Page 7, line 33-42 ~ page 8, line 1-5)

Point 4: Authors also look at classification as a function of zone, but does not explain why they decided to look at these specific categories.

Response 4:

Thank you for pointing this out. These 3 zones are farming area, urban mixed area and forest area, which have their own unique characteristics.

The background environment of farming area is relatively single, with continuous crop distribution. There are vegetation, water, buildings, roads and other surface elements in urban mixed areas, and the background environment is complex. The characteristics of hilly areas are mainly manifested in the diversity of vegetation and the diversity of vegetation growth cycle. At the same time, the cultivated land is dispersed and fragmented. These three zones are also widespread in other areas and are three typical growth environments for the object of crops.

Therefore, the accuracy of crop extraction in these three small regions is discussed in this paper, so that we can construct specific feature space for different typical crop growth environment in the future research.

Point 5: Results are not discussed in the context of prior work in this area.

Response 5: Thank you for pointing this out. We have added the related work in this area from prior time as well as cited the corresponding references to the work in order to meet your kindly comments. (Annotation 26) (Page 4, line 16, 17)

Point 6: The conclusion section does not point out novel aspects of this research or how it compares to prior studies

Response 6:

Thank you for pointing this out. We have supplemented the conclusions in this respect to illustrate the novel aspects of our research in Section 5. (Annotation 62) (Page 21, line 12-15)

Firstly, this paper discusses the construction of feature space and feature combination method, and concludes that it does not need too many kinds of features to participate in classification to achieve high accuracy. It is pointed out that ‘three- to four-feature type was an important factor in feature combinations, derived from spectral reflectance and growth periodicity.’ This provides a positive way for us to simplify feature construction and improve classification efficiency in the future.

Secondly, the potential factors of image crop classification accuracy are discussed in three typical crop growing areas. In the future research, we can construct and extract features according to the characteristics of local crop growth environment, so as to improve the classification speed and ensure high classification accuracy.

Finally, this paper combines the red-edge band features of Sentinel-2 data in the construction of feature space. Experiments show that using red-edge information as crop extraction feature is effective, which has a certain scientific value in the future to play the role of red-edge band in crop extraction research.

Generally speaking, this paper discusses the optimization of crop extraction features and puts forward some views, which has positive reference significance for crop information extraction research.

We appreciate for your warm work earnestly, and hope that the correction will meet with approval. Thank you very much for your comments and suggestions.

Author Response File: Author Response.docx

Reviewer 4 Report

Comments:

Lines 43. Please avoid using acronyms used in the text only once. Please delete SDGs.

Line 114. This sentence is out of place if we consider that the study area is in Germany. Please delete.

Lines 141-144. Probably this paragraph can be inserted at the beginning, then explanations of Germany agriculture explanations make more sense.

Lines 174-177. Please add references regarding the horizontal accuracy of high-resolution Google Earth images. It's not obvious when you use these images (only these) to derive ground truth points for testing (lines 215-217).

Line 180. Please add in the text (wherever authors want) an explanation regarding the software or platform used, like ArcGis in line 197. It would be interesting to know if a commercial software is used, and if the same operations/procedures can be performed properly with free software.

Lines 220-225. I suggest the authors move this phrase to the beginning part or in the paragraph 2.2. It seems unusual depicts firstly operations such as TOA (line158), training and testing, and algorithm used, and only at the end illustrate the workflow.

Line 289. Please check this reference.

Line 300. Figure 5 looks too big; authors should consider reducing the size, for example by inserting the letters ((a); (b); etc.) near the picture. Moreover, the order of the letters is incorrect, ((a) – (d) – (b)).

Line 318. Please delete brackets when referring directly to the Figure in the text. The same afterwards (etc. Table 3).

Line 349. Table 3 has text elements that are not distinguishable (etc. UA (, or PA(. Please correct.

Line 453. This is interesting for readers, please insert references.

Discussions. Discussions should be more controversial and interpreted in comparisons of previous studies and of the working hypotheses; implications and findings should be discussed also considering future research directions.

Author Response

Point 1: Lines 43. Please avoid using acronyms used in the text only once. Please delete SDGs.

Response 1: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 1) (Page 2, line 1)

Point 2: Line 114. This sentence is out of place if we consider that the study area is in Germany. Please delete.

Response 2: Thank you for pointing this out. We have deleted this sentence to meet your suggestion. (Annotation 9) (Page 2, line 38)

Point 3: Lines 141-144. Probably this paragraph can be inserted at the beginning, then explanations of Germany agriculture explanations make more sense.

Response 3: We agree with your comment. We have moved this paragraph to the beginning of Section 2.1 to meet your suggestion. (Annotation 27) (Page 4, line 22)

Point 4: Lines 174-177. Please add references regarding the horizontal accuracy of high-resolution Google Earth images. It's not obvious when you use these images (only these) to derive ground truth points for testing (lines 215-217).

Response 4: We agree with your comment. We have added detailed information about the high-resolution Google Earth images used in the article to meet your suggestion. (Annotation 36) (Page 7, line 13-15)

Point 5: Line 180. Please add in the text (wherever authors want) an explanation regarding the software or platform used, like ArcGis in line 197. It would be interesting to know if a commercial software is used, and if the same operations/procedures can be performed properly with free software.

Response 5: We agree with your comment. We have added detailed information about the software and its version used in our research to meet your suggestion. (Annotation 31) (Page 5, line 25)

Point 6: Lines 220-225. I suggest the authors move this phrase to the beginning part or in the paragraph 2.2. It seems unusual depicts firstly operations such as TOA (line158), training and testing, and algorithm used, and only at the end illustrate the workflow.

Response 6: Thank you for pointing this out. In order to keep the consistency of whole paper, we decided to delete the part of TOA in section 2.2, by the reason of this part is more relative to the data pre-processing. Meanwhile, we moved the flowchart and its illustration word to the beginning of Section 2.4, which is the detailed information of the whole method. (Annotation 33) (Page 6, line 7)

Point 7: Line 289. Please check this reference.

Response 7: Thank you for pointing this out. We have corrected the format mistakes into the right version of reference format. (Annotation 39) (Page 9, line22)

Point 8: Line 300. Figure 5 looks too big; authors should consider reducing the size, for example by inserting the letters ((a); (b); etc.) near the picture. Moreover, the order of the letters is incorrect, ((a) – (d) – (b)).

Response 8: Thank you for pointing this out. We now have adjusted the size of Figure 5 to look more suitable to the page. In terms of the order of the letters, there is no omission here and the letter (d) following (a) was just designed for a neat layout because of the smaller size of (a) and (d). (Annotation 41) (Page 9, line 27 ~ page 10, line 1-3)

Point 9: Line 318. Please delete brackets when referring directly to the Figure in the text. The same afterwards  (etc. Table 3).

Response 9: We agree with that comment. We have deleted the redundant brackets in this paragraph to meet your kindly suggestion. (Annotation 43) (Page 11, line 11)

Point 10: Line 349. Table 3 has text elements that are not distinguishable (etc. UA (, or PA(. Please correct.

Response 10: Thank you for pointing this out. We have corrected the format mistakes into the right version of reference format. (Annotation 48) (Page 14, line 7,8)

Point 11: Line 453. This is interesting for readers, please insert references.

Response 11: We agree with that comment. In this paper, we mentioned that NDVI and EVI sequences can be used as characteristic variables in remote sensing classification technology. Among them, literature 8, 9, 23, 28 and so on are all related studies on remote sensing information classification using vegetation index series such as NDVI. At the same time, we supplement the literature on this point, which is 63 64. (Annotation 55) (Page 19, line 5)

Point 12: Discussions. Discussions should be more controversial and interpreted in comparisons of previous studies and of the working hypotheses; implications and findings should be discussed also considering future research directions.

Response 12: Thank you for pointing this out. In the discussion section, we have added Section4.4 to explore the opportunity of object-based method, another method relative to pixel-based classification. At the same time, the advantages and limitations of this method as well as the opportunities and challenges are discussed in Section 4.5. (Page 20, line 21-41 ~ page 21, line 1-3)

We appreciate for your warm work earnestly, and hope that the correction will meet with approval. Thank you very much for your comments and suggestions.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Dear Authors

A small correction required for Table 1 of row 4. Authors didn't not provide appropriate citations for USGS GFSAD30. 

Citations are available in the following links.Please update citations.

https://croplands.org/gfsadce30info

https://lpdaac.usgs.gov/about/news_archive/release_gfsad_30_meter_cropland_extent_products

Author Response

Point 1: A small correction required for Table 1 of row 4. Authors didn't provide appropriate citations for USGS GFSAD30.

Citations are available in the following links. Please update citations.

https://croplands.org/gfsadce30info

https://lpdaac.usgs.gov/about/news_archive/release_gfsad_30_meter_cropland_extent_products 


Response 1: Thank you for pointing this out. We have updated citations in Table 1 to meet your comments. (Annotation 2) (Page 3, line 36)

As Professor Aqiang Yang has given us great assistance in this round of revision, I wonder if it is appropriate or possible to add Mr. Yang to the list of authors. We’ll be very appreciated if this replenishment can be allowed and we’ll also be thankful if not.

We appreciate for your warm work earnestly, and hope that the correction will meet with approval. Thank you very much for your comments and suggestions.

Reviewer 4 Report

Comments:

Line 7, page 6. Winter wheat.

Lines 13-15, page 7. How did you assess this horizontal accuracy? You don't need to do the validation, it is enough to cite paper that have already done horizontal accuracy of GE images. Please consider that having very high accuracy is not important for training/testing on open fields of wheat, maize, etc. Please refer on adequate references.

Line 17, page 11. Zone1, correct to Zone 1.

Line 30, page 20.   (As we discussed in 4.2)???

Author Response

Point 1: Line 7, page 6. Winter wheat.

Response 1: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 3) (Page 6, line 7)

Point 2: Lines 13-15, page 7. How did you assess this horizontal accuracy? You don't need to do the validation, it is enough to cite paper that have already done horizontal accuracy of GE images. Please consider that having very high accuracy is not important for training/testing on open fields of wheat, maize, etc. Please refer on adequate references.

Response 2: Thank you for pointing this out. In order to support the reliability of the Google Earth images used in validation process, we have added references that are related to previous horizontal accuracy of GE images research to meet your suggestion. (Annotation 4) (Page 7, line 13-16)

Point 3: Line 17, page 11. Zone1, correct to Zone 1

Response 3: Thank you for pointing this out. We now rectified this into right version under your suggestion. (Annotation 5) (Page 11, line 17)

Point 4: Line 30, page 20.   (As we discussed in 4.2)???

Response 4: Thank you for pointing this out. It is a writing mistake and we have deleted this point to meet your suggestion. (Annotation 6) (Page 20, line 30)

As Professor Aqiang Yang has given us great assistance in this round of revision, I wonder if it is appropriate or possible to add Mr. Yang to the list of authors. We’ll be very appreciated if this replenishment can be allowed and we’ll also be thankful if not.

We appreciate for your warm work earnestly, and hope that the correction will meet with approval. Thank you very much for your comments and suggestions.

Author Response File: Author Response.docx