Identifying Peach Trees in Cultivated Land Using U-Net Algorithm

Round 1

Reviewer 1 Report

The paper is interesting and well-written.

If possible I would recommend to provide a wider, international background to this study.

How is that problem managed in other countries? How could the solutions presented in this study be applicable and beneficial to other regions in the world?

Author Response

Revision details and the response to the review comments on the manuscript entitled “Identifying Non-Grain Production in Cultivated Land using U-net Algorithm” with manuscript ID: land-1769021

 

We are grateful for the opportunity to revise our manuscript based on the comments made by the editors and reviewers. We have taken care to address all the comments point-by-point below and have incorporated the suggestions in the manuscript.

Reviewer #1:

Point 1:

The paper is interesting and well-written. If possible, I would recommend to provide a wider, international background to this study. How is that problem managed in other countries? How could the solutions presented in this study be applicable and beneficial to other regions in the world?

Response 1: Thank you for your constructive and valuable suggestions. We have made additions to the Introduction section to provide a wider scope with an international background for the study. The revised portions (lines 24-26), (lines 27-29), (lines 40-44), and (lines 47-49) are as follows:

1. “Since the 1960s, the amount of cultivated land per capita worldwide has decreased to 0.21 hectares [2].”
2. “Some scholars have found that there is not much room for global grain production to in-crease [3]; consequently, the protection of cultivated land has become a major issue both in China and worldwide.”
3. “Many cultivated land protection policies have been widely used in European and American countries and some developing countries, such as the "Green Belt Policy" in the UK, the "Physical Planning Act" in Germany, and the approach of land-use control in the US [12], the "Revitalizing Rainfed Agriculture network" in India [13] and the Low Carbon Emission Plan in Agriculture in Brazil [14].”
4. “However, this issue remains prominent in China and other developing countries with similar agricultural planting structures.”

In addition, our study demonstrated the feasibility of using a combination of UAVs and the U-net algorithm to economic forests in cultivated land. Combined with the U-net algorithm, the NGP index can intuitively utilize spatial information and express the NGP problem quantitatively. Therefore, the NGP index can provide a new quantitative method for this issue. The method used in this study can be alleviate the NGP problem of cultivated land in China and other developing countries with similar agricultural planting structures.

 

Author Response File: Author Response.pdf

Reviewer 2 Report

Please see the attached document

Comments for author File: Comments.pdf

Author Response

Revision details and the response to the review comments on the manuscript entitled “Identifying Non-Grain Production in Cultivated Land using U-net Algorithm” with manuscript ID: land-1769021

 

We are grateful for the opportunity to revise our manuscript based on the comments made by the editors and reviewers. We have taken care to address all the comments point-by-point below and have incorporated the suggestions in the manuscript.

Reviewer #2:

Point 1:

The title seems to be short. The authors proposed a new method to identify non-grain production. Some indication of that may be included in the title.

Response 1: Thank you for your suggestion. As you have pointed out, our research proposes a new method to identify NGP. Therefore, we have revised the title of the article to “Identifying Non-Grain Production in Cultivated Land using U-net Algorithm."

 

Point 2:

Keyword: I suggest removing the word “China” from the keyword. The country name should not be the keyword. Please add another keyword.

Response 2: Thank you for your comment. We have removed the mentioned word and added another keyword, "land policy."

 

Point 3:

Introduction:

• In this section, the author should present the relevant studies conducted in the recent past to identify the research gaps. Then they should mention the contribution they are going to add in this paper. I suggest adding one paragraph on this issue.
• Line 27: Conversion to what?
• Line 36-37: Does China’s policy incentivize NGP? You should explain.

 

Response 3: Thank you for your comment and suggestions.

• A deficiency in the introduction section was that the research gap was not clearly stated. To fix this problem, we have added literature (lines 82-84 and 87-90), tried reorganizing the writing of the introduction, and attempted to show the research gap and how to bridge this gap. The changes to the introduction have been marked in red.

“The rapid development of NGP necessitates higher requirements for precise quantitative monitoring. To address the issue of NGP, the agricultural management department must understand the NGP within its jurisdiction. Currently, NGP is primarily investigated using a statistical analysis of farm household data from questionnaire surveys [17, 18]; remote sensing data is used to examine its spatial distribution. Many time-consuming field surveys are required to quantify the conversion of traditionally cultivated crops to NGP. To improve efficiency, four types of remote sensing data are used to examine the spatial distribution of NGP: 1) Data from the 3rd National Land Resource Survey, derived from visual interpretation of satellite images from Gaofen-2, have been used to obtain NGP information [16]. 2) Operational Land Imager images from the Landsat 8 satellite with a spatial resolution of 30 meters, combined with UAV images, were used to separate the grassland and trees in the cultivated land [19]. 3) Aerial photos with a spatial resolution of less than 1 meter have been used to visually interpret the NGP in cultivated land and to analyze the type and spatial distribution of NGP, combined with land use planning, annual land use survey data, and soil maps [20]. 4) As a type of NGP, agricultural plastic greenhouses can be accurately visually interpreted using high-resolution Google Earth imagery, and the extraction accuracy was 97.20% when combined with deep learning algorithms [21]. Currently, Greenhouses and ponds, which significantly differ from cultivated land, are relatively easy to identify; however, identifying economic forests remains challenging. Economic forests and non-grain crops can easily be confused with cultivated land during visual interpretation (Figure 1). Thus, the agricultural management department urgently needs a method to identify NGP across large areas of cultivated land, especially for economic forest,in near real-time, at a low cost, and with high precision [22]. Therefore, a new monitoring method is required to distinguish NGP from cultivated land.”

• We have elaborated the text here to explain the ‘conversion’ (lines31-33) as follows: “However, due to accelerated industrialization and urbanization, approximately 2.94×10⁵ ha of cultivated land in China is converted into construction land yearly [5].”
• Thank you for this question. The central government has always discouraged the NGP of cultivated land, but some local governments will formulate special agricultural and industrial policies to encourage local rural economic development and increase the income of local rural laborers. Like the "One Village One Product" strategy (OVOP) in the Guanzhong Plain of Shaanxi Province, the OVOP strategy not only highlights the regional characteristics, but also improves the local per capita income level, and promotes the local per capita income level the prosperity of the agricultural and rural market economy. However, under the influence of the OVOP strategy, the area of cultivated land for grain crops has been continuously reduced, and the sustainable development of the land has been destroyed. Therefore, the agricultural and industrial policy launched by the local government is an essential factor that directly affects the farmers' choice to grow non-grain crops on cultivated land. There will be some conflicts between the development of the local economy and the farmland protection policy of the central government. Therefore, it is necessary to adjust and regulate the behavior of local governments through technical means.

To provide better understanding, we have added (lines 67-72) in the introduction section as follows:

“thus, the OVOP strategy is a significant cause of the NGP expansion [6], although it can improve the income level of residents. This phenomenon contradicts the central government's policy of discouraging NGP. To prevent this situation from becoming more complicated, fast and convenient assessment of the NGP data is crucial for the central government to regulate the behavior of local governments.”

 

Point 4:

Materials and methods:

• Figure 2: Please add the north sign for maps a and b.
• Section 2.3: Here, the author should highlight their specific contribution (e.g. improving the U-net algorithm)

 

Response 4: Thank you for these suggestions.

• We agree with your comment. We have included the north sign for maps a and b.
• The suggestion that Section 2.3 should highlight our specific contribution is quite appropriate. We have reorganized the language in Section 2.3 (lines 175-178). We took full advantage of the pixel-based features of the U-net algorithm and calculated the area of different types of ground objects according to different pixel This pixel-based identification method plays a crucial role in designing the NGP index (Section 2.4) to determine the severity of the NGP problem and further solves the problem of how to assess the severity of NGP problems. Therefore, this combination of the U-net algorithm and NGP index can provide a new method for solving the current issues in agricultural management.

Lines 175-178: “The advantage of the U-net algorithm is that, when labeling the NGP, the output should include different types of boundaries, and each pixel should be given a category label. This pixel-based technique allows the network to learn and provide predictions for each pixel.”

 

Author Response File: Author Response.pdf

Reviewer 3 Report

The manuscript is well prepared and presents an effective methodology and experimental design to assess the effects of data availability on phenology-based rice mapping in China via deep learning. For me, it is an excellent initiative and is on the way to being published. However, some points need to be clarified to improve the relevance and the potential of citation. Therefore, I would like to present some adjustments that can help to improve the study.

Introduction, Material and methods, and Results are well defined. However, the current Discussion is small. As the Authors have diverse interesting results, all of them need to be well discussed. Also, all of them need to be well discussed also with the state-of-the-art. The authors referenced a few papers in this section. In the current form, this section overshadows the ability of the method. Improving this section, highlighting differences regarding other methods, will valorize the study and substantially expand the citation potential. Therefore, this section should be improved. A recommendation is to emphasize the novelty of the current work, making a clear distinction of the assumptions to highlight originality. Examples of topics to be explored in the discussion section:

Lines 319-325: Please, what characteristics made our method more suitable for this purpose than others, such as those cited? Here, is a great chance to discuss and highlight why your method is better, highlighting what factors may have conditioned improvements in our results in comparison with others.

Emphasize the potential and capability of the method to be applied to other study areas and for evaluating other LULC classes.

Limitations are presented only in Conclusion. It can be more suitable in the Discussion, highlighting trade-offs and potential strategies to overcome the presented adversities.

- What features or intrinsic characteristics of the method benefitted the identification of NGP in cultivated land using the U-Net algorithm and the estimation of NGP of cultivated land in projecting plots?

Also, may be relevant to emphasize topics that are reinforced in the Conclusions:

- The fact that NGP in cultivated land has expanded quickly owing to economic and policy incentives and should be treated as an emerging potential threat to grain security in China. NGP is influencing the local society, economy, and environment? What vulnerabilities did NGP introduce to local agriculture and food security?

- Insights for conducting agricultural management assessments. How is it possible to consume this method (combination of identification method and NGP index) as a tool for agricultural management to perform routine inspections for NGP in cultivated land, thereby replacing field measurements to achieve significant labor savings?

- Insights for agricultural management/interventions to mitigate NGP influence;

Moreover, it would be relevant to answer:

- For territorial planning purposes, there are intraseasonal variations of land management practices on a seasonal scale? If yes, what are the driving factors? Also, are agricultural management or edaphoclimatic conditions significantly varying between the seasons or within the region?

Author Response

Revision details and the response to the review comments on the manuscript entitled “Identifying Non-Grain Production in Cultivated Land using U-net Algorithm” with manuscript ID: land-1769021

 

We are grateful for the opportunity to revise our manuscript based on the comments made by the editors and reviewers. We have taken care to address all the comments point-by-point below and have incorporated the suggestions in the manuscript.

Reviewer #3:

Point 1:

Introduction, Material and methods, and Results are well defined. However, the current Discussion is small. As the Authors have diverse interesting results, all of them need to be well discussed. Also, all of them need to be well discussed also with the state-of-the-art. The authors referenced a few papers in this section. In the current form, this section overshadows the ability of the method. Improving this section, highlighting differences regarding other methods, will valorize the study and substantially expand the citation potential. Therefore, this section should be improved. A recommendation is to emphasize the novelty of the current work, making a clear distinction of the assumptions to highlight originality. Examples of topics to be explored in the discussion section:

• Lines 319-325: Please, what characteristics made our method more suitable for this purpose than others, such as those cited? Here, is a great chance to discuss and highlight why your method is better, highlighting what factors may have conditioned improvements in our results in comparison with others.

 

• Emphasize the potential and capability of the method to be applied to other study areas and for evaluating other LULC classes.

 

• Limitations are presented only in Conclusion. It can be more suitable in the Discussion, highlighting trade-offs and potential strategies to overcome the presented adversities.

 

• What features or intrinsic characteristics of the method benefitted the identification of NGP in cultivated land using the U-Net algorithm and the estimation of NGP of cultivated land in projecting plots?

 

Response 1: Thank you for your comments and suggestions.

• We have added the characteristics that made our method more suited for this purpose compared to others in the discussion section (lines 338-340) and (lines 343-349) as follows:

92. a) lines 338-340: “this method of distinguishing NGP types based on pixels was different from extracting the same land-over type using different spectra, but the result is comparable with the overall accuracy of 92.81% achieved by Xu et al. [34].”
93. b) lines 343-349: “Such a label with the stem of the tree provides a reference for the selection of the characteristics of the forest trees. On this basis, this method combines the characteristics of the phenological period of economic forests. It was found that this new method can effectively solve the problem of difficulty in visual interpretation of economic forests and food crops. This provides an effective method for distinguishing economic forests and food crops in the process of monitoring NGP and determining the expansion area of economic forests.”

• We have added the potential and capability of this method to be applied to other study areas and for evaluating other LULC classes in the discussion section (lines 375-378) and (lines 380-384) as follows:

1. a) lines 375-378: “U-net is among the most popular algorithms in land use/land cover (LULC) research, which combines several benchmark data sets to achieve state-of-the-art performance based on spectral and spatial information with limited training data [46, 47].”
2. b) lines 380-384: “Moreover, the application of the pixel-based U-net to different densities of grassland, shrubs, and sparse forests, which are more confusing land-use types, is still limited to an extent. The algorithm combined with phenology in our study may provide a new way of thinking, but the appropriate method needs further exploration.”

• We agree with this comment. Limitations have been moved to the discussion section, highlighting trade-offs and potential strategies to overcome the presented adversities.
• We have added features of the method that benefitted the identification of NGP in cultivated land using the U-Net algorithm and the estimation of NGP of cultivated land in projecting plots in the discussion section (lines 358-360) and (lines 369-372) as follows:

1) lines 358-360: “Moreover, peach trees have typical phenological characteristics, which meet the application requirements of this method for NGP monitoring.”

2) lines 369-372: “The phenological period of economic forests is an important feature that helps distinguish them from windbreaks, pond fish farms, and vegetable greenhouses. The canopy, flowers, and shadows of the phenological period in this study were taken as the main labels."

 

Point 2:

Also, may be relevant to emphasize topics that are reinforced in the Conclusions:

• The fact that NGP in cultivated land has expanded quickly owing to economic and policy incentives and should be treated as an emerging potential threat to grain security in China. NGP is influencing the local society, economy, and environment? What vulnerabilities did NGP introduce to local agriculture and food security?

 

• Insights for conducting agricultural management assessments. How is it possible to consume this method (combination of identification method and NGP index) as a tool for agricultural management to perform routine inspections for NGP in cultivated land, thereby replacing field measurements to achieve significant labor savings?

 

• Insights for agricultural management/interventions to mitigate NGP influence;

Moreover, it would be relevant to answer:

For territorial planning purposes, there are intraseasonal variations of land management practices on a seasonal scale? If yes, what are the driving factors? Also, are agricultural management or edaphoclimatic conditions significantly varying between the seasons or within the region?

 

 

Response 2: Thank you for your constructive and good suggestions.

• NGP is influencing the local society, economy, and environment. We have added vulnerabilities of NGP introduced to local agriculture and food security in the Conclusion section (lines 401-402) as follows:

“NGP results in the reduction of cultivated land area and total grain production as well as the destruction of cultivated land quality.”

 

• We have added to conduct an agricultural management assessment in the Conclusion section (lines 407-410) as follows:

“When the NGP area is more than 30%, it means that the problem of NGP already exists. Under long-term monitoring, the agricultural management department needs to formulate a plan for management in time according to the actual situation.”

 

• The NGP influence can be mitigated through agricultural management or interventions, such as growing cash crops in cultivated land (soybeans, sesame, peanuts, cotton, etc.). These cash crops are easily affected by seasonal climatic conditions, and the planting structure can be adjusted during a season. Therefore, the impact of planting cash crops on cultivated land is easy to recover. However, our research object is economic forest (apple tree, peach tree, pear tree, etc.), which accounts for the largest proportion of NGP area. Generally, it needs to be planted for many years and harvested in a specific season every year. Therefore, the impact of this type of NGP on cultivated land is also difficult to recover.

 

Author Response File: Author Response.pdf

Reviewer 4 Report

The manuscript described an application of U-net in pixel classification for peach trees.

Although the manuscript was concisely written and communicates to readers well, I don’t see genuine efforts being made towards conducting a high-quality study. NGP includes various agricultural products, while only peach trees were involved in the study. It is inappropriate to have the title, abstract and introduction focus on the general term NGP. A comprehensive review of existing studies classifying NGP using aerial images is missing, which is necessary for the justification of the study. Image data were collected within a single day at one location, which is insufficient to prove the generalizability and reliability of U-net for peach trees, let alone for NGP. The method of the study lacks sophistication, as only two classes were involved for the classification task. Certain aspects of the method were not described in detail. For example, how the image labels were prepared was unspecified.

For the reasons above, I do not recommend the manuscript for publication.

Author Response

Revision details and the response to the review comments on the manuscript entitled “Identifying Non-Grain Production in Cultivated Land using U-net Algorithm” with manuscript ID: land-1769021

 

We are grateful for the opportunity to revise our manuscript based on the comments made by the editors and reviewers. We have taken care to address all the comments point-by-point below and have incorporated the suggestions in the manuscript.

 

Reviewer #4:

Point 1:

The manuscript described an application of U-net in pixel classification for peach trees. Although the manuscript was concisely written and communicates to readers well, I don’t see genuine efforts being made towards conducting a high-quality study. NGP includes various agricultural products, while only peach trees were involved in the study. It is inappropriate to have the title, abstract and introduction focus on the general term NGP. Image data were collected within a single day at one location, which is insufficient to prove the generalizability and reliability of U-net for peach trees, let alone for NGP. The method of the study lacks sophistication, as only two classes were involved for the classification task. Certain aspects of the method were not described in detail. For example, how the image labels were prepared was unspecified.

 

Response 1:

Thank you for your comment. We are very grateful for the opinions of the experts. We are also very happy to elaborate on the theme of our article further. NGP includes many agricultural products, but there are three main types of NGPs: economic forest, pond fish farming, and greenhouse vegetable production. The economic forest occupied large-scale cultivated land, mainly based on fruit products with high economic values, which began to replace traditional grain crops. Pond fish farming and greenhouse vegetable production will decline the quality of cultivated land to a certain extent. As a type of NGP, agricultural plastic greenhouses can be accurately visually interpreted using high-resolution Google Earth imagery, and combined with deep learning algorithms; the extraction accuracy was 97.20%. Greenhouses and ponds, which significantly differ from cultivated land, are relatively easy to identify. Actually, our team has visual interpreted all the plastic greenhouse and ponds to identify the distribution of NGP in Pinggu district. However, identifying economic forests remains challenging. Visual interpretation can easily confuse economic forests and non-grain crops with cultivated land. Thus, the agricultural management department urgently needs a method to identify NGP across large areas of cultivated land, in near real-time, at a low cost, and with high precision.

Our starting point is to solve the more prominent problems in NGP. The research plan is to conduct investigations in areas where NGP problems are common. This area needs to meet the conditions for the encroachment of farmland by economic forests to become an NGP problem. Therefore, we chose the Pinggu District of Beijing as the main research area. During the actual investigation, we found that the economic forests planted in the NGP area of Pinggu District are mainly peach trees.

In addition, our study demonstrated the feasibility of using a combination of UAVs and the U-net algorithm to economic forests in cultivated land. Combined with the U-net algorithm, the NGP index can intuitively utilize spatial information and express the NGP problem quantitatively. Therefore, the NGP index can provide a new quantitative method for this issue. However, as you have pointed out, this approach still has some limitation in dealing with general NGP problems. In the future, we would be interested in exploring this suggestion, researching economic forests, and other types of identification.

 

Point 2:

A comprehensive review of existing studies classifying NGP using aerial images is missing, which is necessary for the justification of the study.

Response 2:

A deficiency in the introduction section was that the research gap was not clearly stated. To fix this problem, we have added literature (lines 82-84 and 87-90) and attempted to show the research gap. We found that most of NGP, such as plastic greenhouses and fish pond, are relatively easy to identify by the methods mentioned in these literature, and the gap is that economic forests be confused with cultivated land. The changes to the introduction have been marked in red.

“The rapid development of NGP necessitates higher requirements for precise quantitative monitoring. To address the issue of NGP, the agricultural management department must understand the NGP within its jurisdiction. Currently, NGP is primarily investigated using a statistical analysis of farm household data from questionnaire surveys [17, 18]; remote sensing data is used to examine its spatial distribution. Many time-consuming field surveys are required to quantify the conversion of traditionally cultivated crops to NGP. To improve efficiency, four types of remote sensing data are used to examine the spatial distribution of NGP: 1) Data from the 3rd National Land Resource Survey, derived from visual interpretation of satellite images from Gaofen-2, have been used to obtain NGP information [16]. 2) Operational Land Imager images from the Landsat 8 satellite with a spatial resolution of 30 meters, combined with UAV images, were used to separate the grassland and trees in the cultivated land [19]. 3) Aerial photos with a spatial resolution of less than 1 meter have been used to visually interpret the NGP in cultivated land and to analyze the type and spatial distribution of NGP, combined with land use planning, annual land use survey data, and soil maps [20]. 4) As a type of NGP, agricultural plastic greenhouses can be accurately visually interpreted using high-resolution Google Earth imagery, and the extraction accuracy was 97.20% when combined with deep learning algorithms [21]. Currently, Greenhouses and ponds, which significantly differ from cultivated land, are relatively easy to identify; however, identifying economic forests remains challenging. Economic forests and non-grain crops can easily be confused with cultivated land during visual interpretation (Figure 1). Thus, the agricultural management department urgently needs a method to identify NGP across large areas of cultivated land, especially for economic forest,in near real-time, at a low cost, and with high precision [22]. Therefore, a new monitoring method is required to distinguish NGP from cultivated land.”

 

We have attempted to improve the manuscript and made changes in accordance with the reviewers’ suggestions. These changes will not influence the content and framework of the paper. We marked the changes with red in the revised manuscript. We appreciate the efforts of the editor and the reviewers made and hope that the corrections are satisfactory. Once again, thank you very much for your comments and suggestions.

 

Author Response File: Author Response.pdf

Round 2

Reviewer 4 Report

In my previous comments, I mentioned that “image data were collected within a single day at one location”, which the authors did not address in their response. It is irresponsible to draw any general conclusions regarding the feasibility of a model when the model is developed using such as a biased dataset. Any experienced modeler should know that such models would likely perform poorly when being tested on external data. Additional data collected at different times and locations are necessary for model validation or test, if not also for model training, unless the authors can provide evidence to support the robustness of their model.

The authors also did not address my previous comment on how the image labels were prepared.

Little efforts were put into the revision of the manuscript. I suggested a comprehensive review in introduction, which is necessary to help readers identify the knowledge gaps in current literature and justify the current study, while the authors only added two more references.

My opinion on using the term NGP in the article remains the same. Regardless of what the authors’ ultimate goal is, what was actually being classified in the study are peach trees. It is inappropriate and misleading to use the general term NGP. “NGP” should be replaced with “peach trees” throughout the manuscript.

In my opinion, from a technical perspective, the manuscript cannot be revised into a publishable form without significant additional data collection and model development work.

Author Response

We are grateful for the opportunity to revise our manuscript based on the comments made by the editors and reviewers. We have addressed all the comments point-by-point below and incorporated the suggestions in the manuscript. All the revised text is in a red-colored font in the revised manuscript. The responses to all comments have been prepared and attached herewith.

Reviewer #4:

Point 1:

 

In my previous comments, I mentioned that “image data were collected within a single day at one location”, which the authors did not address in their response. It is irresponsible to draw any general conclusions regarding the feasibility of a model when the model is developed using such as a biased dataset. Any experienced modeler should know that such models would likely perform poorly when being tested on external data. Additional data collected at different times and locations are necessary for model validation or test, if not also for model training, unless the authors can provide evidence to support the robustness of their model.

 

Response 1: Thank you for your constructive and valuable suggestions. In previous comments, you suggested that a multi-point, the multi-temporal dataset should be used as the initial data for the model, which is a very reasonable suggestion and can indeed increase the robustness of the model. We consider that the more the feature to be identified in the model is different from other features, the better the identification may be, so we did not choose other phenological periods when economic forests grow naturally but chose the flowering period. Compared with the cultivated land and windbreak features, the variability of peach tree images during the flowering period is more remarkable, so we only obtained images of economic forests during the one phenological period of flowering. Chen et al. also showed that economic forests and grain crops were easily confused. Therefore, we chose peach trees in flowering as the primary basis for differentiation and highlighted the flowering phenological periods in rows 148-150, lines 380-382, and 393-396, respectively, so the data were obtained at specific times. The identification results showed a more desirable solution to the problem of differentiating between the two. Based on your suggestions, we have added to the description of limitations in the discussion section, and we will continue to implement it in future work with additional multi-location, multi-temporal data to improve the robustness of the model further.

1. Chen, S.; Liu, Q.; Chen, L.; Li, J.; Li, Q. Review of research advances in remote sensing monitoring of grain crop area. Transactions of the Chinese Society of Agricultural Engineering. 2005, 06, 166-171.
2. Yin, P.; Fang, X. Assessment on Vulnerable Regions of Food Security in China. Acta Geographica Sinica. 2008, 10, 1064-1072.

 

2)lines 148-150: “Therefore, the economic forest in the study area was mainly peach trees, which were chosen as the main object of study, and the phenological characteristics of peach trees were analyzed.”

3)lines 380-382: “Moreover, peach trees have typical phenological characteristics, which meet the application requirements of this method for non-grain production monitoring.”

4) lines 393-396: “The flowering phenological period of economic forests is an important feature that helps to distinguish them from windbreaks, pond fish farms, and vegetable greenhouses. In this study, the canopy, flowers, and shadows of the flowering period were taken as the image labels."

5)lines 417-421 in discussion: “To improve the robustness of the model, we will further obtain data from more different locations within the flowering period and then build a dataset to explore the improvement of the model with data from more locations within the same phenological period. Furthermore, adding data from other phenological periods makes it suitable for the model.”

Point 2:

The authors also did not address my previous comment on how the image labels were prepared.

 

Response 2: Thank you for your suggestion. We added a paragraph explaining how the image labels were prepared in section 2.2. The revised portions are as follows:

“The problem of detecting and quantifying peach trees can be viewed as a semantic segmentation task. There are two categories to be distinguished based on the required tasks: peach and non-peach trees. The segmentation performance relies on the quality of the labeled data and segmentation algorithms, mainly by observing orthomosaic images of RGB images collected by the UAV. Then, the peach tree areas in the three experimental study areas were manually marked with digitizing polygon outlines, including branches, peach blossoms, and tree shadows. Therefore, the peach trees in the three typical phenomenon experimental areas in Pinggu District can be used for training and verification, respectively.”

Point 3:

Little efforts were put into the revision of the manuscript. I suggested a comprehensive review in introduction, which is necessary to help readers identify the knowledge gaps in current literature and justify the current study, while the authors only added two more references.

 

Response 3: Thank you for your comments. Based on your suggestion, we have moved the opening paragraph of Section 2.3 to the introduction Section, which is an introduction to machine learning applications and deep learning in arable research. Moreover, we have made some adjustments to the previous literature to better help readers identify the knowledge gaps in current literature and justify the current study. The revised portions are as follows:

“In recent years, machine learning methods have been preliminarily applied to the cultivated land problem. Based on the hierarchical structure of the model, machine learning techniques can be divided into surface learning (involving methods such as support vector machines or SVMs), logistic regression, and multi-structure deep learning (such as convolutional neural networks). Cai et al. used remote sensing images and an SVM to identify cultivated, and non-cultivated land in arid areas in China [17]; this method of recognizing cultivated land provides a good reference for obtaining information on crops and economic forests. However, traditional SVM learning methods have limitations in extracting different spectral changes within the same land cover type, such as vegetated and non-vegetated cultivated land [18]. Additionally, Ramezan et al. found that, for SVM algorithms for land cover classification, the accuracy decreases when the sample size increases [19]. Thus, Xu et al. proposed an end-to-end high-resolution U-net method to extract cultivated land from Landsat Thematic Mapper images [20]. This method can improve the extraction accuracy for cultivated land but lacks basic information on economic forests and non-grain production. The advantage of the U-net algorithm is that, when labeling, the output should include different types of boundaries, and each pixel should be given a category label. This pixel-based technique allows the network to learn and provide predictions for each pixel. Therefore, it is necessary to select typical ground features as category labels to take advantage of the U-net algorithm.

The rapid development of non-grain production necessitates higher requirements for precise quantitative monitoring. To address the issue of non-grain production, the agricultural management department must understand the non-grain production within its jurisdiction. Currently, non-grain production is primarily investigated using a statistical analysis of farm household data from questionnaire surveys [21, 22]; remote sensing data is used to examine its spatial distribution. Many time-consuming field surveys are required to quantify the conversion of traditionally cultivated crops to non-grain production. To improve the efficiency of the investigation, four types of remote sensing data are used to examine the spatial distribution of non-grain production: 1) Data from the 3rd National Land Resource Survey, derived from visual interpretation of satellite images from Gaofen-2, have been used to obtain non-grain production information [16]. 2) Operational Land Imager data from the Landsat 8 satellite with a spatial resolution of 30 meters, combined with UAV images, were used to separate the grassland and trees in the cultivated land [23]. 3) Aerial photos with a spatial resolution of less than 1 meter have been used to visually interpret the non-grain production in cultivated land and to analyze the type and spatial distribution of non-grain production, combined with land use planning, annual land use survey data, and soil maps [24]. 4) As a type of non-grain production, agricultural plastic greenhouses can be accurately visually interpreted using high-resolution Google Earth imagery, and the extraction accuracy was 97.20% when combined with deep learning algorithms [25]. Currently, Greenhouses and ponds, which significantly differ from cultivated land, are relatively easy to identify; however, identifying economic forests re-mains challenging [26, 27]. Economic forests and non-grain crops can easily be confused with cultivated land during visual interpretation (Figure 1). Thus, the agricultural management department urgently needs a method to identify non-grain production across large areas of cultivated land, especially for the economic forest, in near real-time, at a low cost, and with high precision [28]. Therefore, a new monitoring method is required to distinguish non-grain production from cultivated land."

26. Chen, S.; Liu, Q.; Chen, L.; Li, J.; Li, Q. Review of research advances in remote sensing monitoring of grain crop area. Transactions of the Chinese Society of Agricultural Engineering. 2005, 06, 166-171.
27. Yin, P.; Fang, X. Assessment on Vulnerable Regions of Food Security in China. Acta Geographica Sinica. 2008, 10, 1064-1072.

 

Point 4:

My opinion on using the term NGP in the article remains the same. Regardless of what the authors’ ultimate goal is, what was actually being classified in the study are peach trees. It is inappropriate and misleading to use the general term NGP. “NGP” should be replaced with “peach trees” throughout the manuscript.

 

Response 4: Thank you for your comment and suggestions. It was indeed the peach trees when identified and classified in our research, so following your advice, we have changed the “NGP” to “peach trees” in the title, abstract, and main body of the article.

 

 

Author Response File: Author Response.pdf