Data-Driven Artificial Intelligence Applications for Sustainable Precision Agriculture

Round 1

Reviewer 1 Report

The paper deals with an interesting topic related to the practical implementation of artificial intelligence methods in precision agriculture. I think the authors should consider whether this paper is actually a scientific paper or more of a review paper? maybe a pilot study?

The paper is well written, there is a clear and logical selection of content for each chapter and subsection of the paper.

Please answer the following questions. The answers will clarify some doubts of the reviewer.

Abstract:

Please write where the research was conducted.

Introduction

Please indicate how many minimum years should be included for AI-based models to produce reliable results.

Please add what minimum error values of AI models are acceptable in agricultural research.

Please state what potential limitations such models have.

Part: Data-driven reduction of inputs

Line 213: Is such data provided completely free of charge?

References:
I suggest adding DOI references to each article. 

All article:

Please comment on how farmers in Germany are supported when it comes to implementing AI methods in agriculture. Is training provided? Instructional sessions?

On what basis were the pilot farms selected for the different programmes? On the basis of the economic size of the farm?

Author Response

Point 1: Abstract. Please write where the research was conducted

Response 1: The paper provides a summary of the most recent research activities in the form of research projects implemented and validated by the authors in several European countries, with the objective of showing the already achieved results, the current investigations and the still open technical challenges.

Point 2: Introduction. Please indicate how many minimum years should be included for AI-based models to produce reliable results.

Response 2: The minimum number of years for each model AI-based model depends mainly on the type of crop and each phenotype. For example, models for the DEMETER project have been trained and tested with data collected during the monitoring of phenology in olive orchards in Tuscany (Italy) in 2008-2010. However, in the VINBOT project, the vines were manually assessed for canopy dimensions during the ripening period of the 2016 season.

Point 3: Introduction. Please add what minimum error values of AI models are acceptable in agricultural research.

Response 3: As for the Response 2, the minimum error values of AI models depend on the required precision of each variable of the model, which further depends on the type of crop and the application of the use case.

Point 4: Introduction. Please state what potential limitations such models have.

Response 4: As for the Response 2, the limitations of AI models depend on the type of crop and the application of the use case. For example, hidden bunches by vegetation in the VINBOT project led to underestimation of the actual yield.

Point 5. Part: Data-driven reduction of inputs. Line 213: Is such data provided completely free of charge?

Response 5: As it has been mentioned, the NaLamKI initiative has just started and the business model for its exploitation is still under development. This will include the way data will be provided (completely free, low-cost, paid) depending also on the type of application and user. What the initiative guarantees is that open interfaces will be offered for partners from agriculture, industry, and service suppliers from crop production.

Point 6. References: I suggest adding DOI references to each article.

Response 6: DOI references are provided for each paper.

Point 7. All article: Please comment on how farmers in Germany are supported when it comes to implementing AI methods in agriculture. Is training provided? Instructional sessions?

Response 7: Although this part is still to be defined once results are available, it is expected that training will be provided to all the involved stakeholders (partners from agriculture, industry, and service suppliers from crop production).

Point 8. All article: On what basis were the pilot farms selected for the different programmes? On the basis of the economic size of the farm?

Response 8: Most of the examples have been developed by large European consortia within R&D calls. Thus, there has been no further selection but the participation within the consortia.

Reviewer 2 Report

The authors presented a well-arranged review paper about the application of AI in agriculture from the side of data as the main pushing factor.

However, I believe the following comments are important to enhance the paper quality:

• The authors should state the possibilities, challenges, and possible solutions of applying PA, or AI in general in the agricultural sector in developing countries. For example, is the data-driven AI able to stand against the cheap labor in these countries?
• What are the possible problems arising from data privacy with cloud-based agriculture especially if the data is shared between farmers, businesses, and governmental agencies?
• I see that the authors brought examples about applying AI in agriculture in the EU, how about Japan, USA, and other developed countries? This should be also included as this paper should be a bit comprehensive.
• It would be important to create a table summary for the common algorithms, protocols, applications, or techniques drives from AI and used in various agricultural operations as we know this is already an established industry in several advanced countries.
• It would be good to enhance the resolution of figures 3, 4, 6, 8, 9, 10. Additionally, pointing to similar or common details would be explanatory in figure 2.           

Author Response

Response to Reviewer 2 Comments

Point 1: The authors should state the possibilities, challenges, and possible solutions of applying PA, or AI in general in the agricultural sector in developing countries. For example, is the data-driven AI able to stand against the cheap labor in these countries?

Response 1: The paper is based on European examples of the application of PA developed by the corresponding institutions of the authors. One of the main current concerns of the European problems is the undeniable raising of the average age of farmers and the forecasted problems of labour shortages in the agricultural sector due to lack of younger generations. Therefore, some of the statements that are made could be difficult to scale up to developing countries, especially the cheap labour issue remarked by the reviewer.

Point 2: What are the possible problems arising from data privacy with cloud-based agriculture especially if the data is shared between farmers, businesses, and governmental agencies?

Response 2: In order to tackle this issue, one objective of the NaLamKI initiative is the development of AI methods and GAIA-X compliant services, always ensuring data sovereignty along the entire value chain and interoperability between various central and decentralized cloud providers and users. In particular, the guarantee of data sovereignty, end-to-end networking and GAIA-X-compliant AI services increase acceptance by partners and farmers, thus harboring a high potential for innovation and making a significant contribution to environmental protection in connection with sustainable agriculture.

Point 3: I see that the authors brought examples about applying AI in agriculture in the EU, how about Japan, USA, and other developed countries? This should be also included as this paper should be a bit comprehensive.

Response 3: As it is mentioned in the papers, the examples are located in several European locations. Furthermore, the authors have added the corresponding locations to each of the examples following the suggestion of the reviewer. Due to the high variability of crops even within single fields, it may be unprecise to apply obtained conclusions to other countries.

Point 4: It would be important to create a table summary for the common algorithms, protocols, applications, or techniques drives from AI and used in various agricultural operations as we know this is already an established industry in several advanced countries.

Response 4: Many of the papers reviewed in the second section have already proposed detailed tables to summarize algorithms, protocols and applications of the use of the AI in agricultural sector. Our approach depicted in Figure 1 aims at providing an updated version of several European research projects that have been or are being run by the authors. Taken together, these examples represent the current abilities and future potential for AI applications in European agricultural research projects.

Point 5: It would be good to enhance the resolution of figures 3, 4, 6, 8, 9, 10. Additionally, pointing to similar or common details would be explanatory in figure 2.          

Response 5: Pictures have been replaced by HD ones.

Reviewer 3 Report

- Author submit the manuscript as Article, but the overall content of the manuscript looks like a review paper. Can add more detail on how the author conducts the experiment, material use, and more specific about the studied methodology.
- Author gives a conceptualization of the agriculture subdomains and AI-related technologies. However, it seems there is some important information is missing from the conceptualization  (figure 1), e.g., the connection between each other.
- This paper also discussed the use of Machine Learning (ML) prediction, but it does not show clearly what type of ML, how the ML works, what is the final result of ML.
- title use "Data-driven Application," but there are no sections explaining detail on material and method used in this manuscript.
- No discussion section clearly explains the author's results and concerns the previous study, hypothesis, and methodology.
-  It is necessary to rearrange the placement of sections in the manuscript to attract readers 
-  could not find the method, result, and conclusion in the abstract clearly.

Author Response

Point 1: Author submit the manuscript as Article, but the overall content of the manuscript looks like a review paper. Can add more detail on how the author conducts the experiment, material use, and more specific about the studied methodology.

Response 1: The objectives, methodology and material implemented are different for each of the research projects. When available, the reviewer is kindly referred to detailed descriptions in already published papers. For example, reference [23] describes the algorithms, materials and methods used to evaluate the evaluation trial.

Point 2: Author gives a conceptualization of the agriculture subdomains and AI-related technologies. However, it seems there is some important information is missing from the conceptualization  (figure 1), e.g., the connection between each other.

Response 2: In order to clarify the connection between agricultural operations and AI-based technologies, authors have modified Figure 1 in order to depict relevant agricultural operations as faced by the research projects, and the AI-based technological solutions for each of those projects.

Point 3: This paper also discussed the use of Machine Learning (ML) prediction, but it does not show clearly what type of ML, how the ML works, what is the final result of ML.

Response 3: The authors are not completely sure how to answer this question as detailed ML and DL algorithms have been mentioned in several of the sections. If further explanation is required, the reviewer is kindly referred to detailed descriptions in already published papers as mentioned in Response 1.

Point 4: title use "Data-driven Application," but there are no sections explaining detail on material and method used in this manuscript.

Response 4: As it has been mentioned in Response 1, materials and methods used in each of the pilots are different. When available, the reviewer is kindly referred to detailed descriptions in already published papers (listed above).

Point 5: No discussion section clearly explains the author's results and concerns the previous study, hypothesis, and methodology.

Response 5: In last years, the EU has actively undertaken R&D activities related to the application of AI for the digitization of agriculture. However, to the best of the authors’ knowledge, there are no recent surveys of re-search activities at European level related to the application of AI technologies for the agricultural sector. Following the approach from Bacco et al, this paper aims at providing an updated version of several European research projects that have been or are being run by the authors. Taken together, these examples represent the current abilities and future potential for AI applications in European agricultural research projects.

Point 6: It is necessary to rearrange the placement of sections in the manuscript to attract readers

Response 6: Some of the sections of the paper have been replaced and further extended in order to try to make the paper more attractive for potential readers.

Point 7: could not find the method, result, and conclusion in the abstract clearly.

Response 7: The content of the abstract has been changed following the suggestion of the reviewer.