Integration of Remote Sensing and Machine Learning for Precision Agriculture: A Comprehensive Perspective on Applications

Response to Reviewer 1 Comments

1. Summary

Thank you very much for taking the time to review this manuscript. We gave a detailed reply below and highlighted the content.

2. Point-by-point response to Comments and Suggestions for Authors

Comments 1: Title: Change "remote" to "Remote".

Response 1: [Integration of Remote Sensing and Machine Learning for Precision Agriculture: A Comprehensive Perspective on Applications] Thank you for pointing this out. We agree with this comment. Therefore, We have revised the title. [The details can be found in lines 2-4 on the first page of the revised manuscript.]

Comments 2: L3: The phrase "A New Perspective of Comprehensive Application" seems incomplete. Could the authors revise the title to make it more connected to the first part?

Response 2: [Integration of Remote Sensing and Machine Learning for Precision Agriculture: A Comprehensive Perspective on Applications] Thank you for pointing this out. We agree with this comment. Therefore, We have revised the title. [Change "with the Integration" to "Integration of" to make the title more concise and smooth. Delete "A New Perspective of Comprehensive Application" because this part of the expression is not clear enough, we replace the more direct expression "A Comprehensive Perspective on Applications", which better reflects the content and focus of the article.]

Comments 3: L10: The connection between "resource shortage" and "climate change," and "the contradiction between human and land" is unclear. Please consider revising this.

Response 3: [With the global population growth, resource shortage and climate change, the contradiction between human and land has become increasingly prominent, these factors constitute a major challenge to the traditional agricultural model. There is an urgent need for a new mode of agricultural production that can achieve efficient, environmental protection and sustainable development.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [Add "these factors together constitute a major challenge to traditional agricultural models" to clarify the links between "global population growth", "resource shortages" and "climate change" and how they affect traditional agricultural models. The "urgent need" for a new mode of agricultural production to meet these challenges was emphasized.]

Comments 4: L15-17: This statement might be incorrect as many papers have been published on this topic. Please consider revising it. A suggestion is to highlight a specific aspect that other papers have not addressed.

Response 4: [Although there are many researches on the application of remote sensing technology and machine learning in precision agriculture, there are relatively few comprehensive and systematic studies on the integrated application of these two technologies. In view of this, this study conducted a systematic literature search of keywords from Web of Science, Scopus, Google Scholar and PubMed databases.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [We avoid the absolute claim that there is "almost no research", but instead point out that there are relatively few studies, with particular reference to the lack of a comprehensive and systematic review of the integrated application of the two technologies. In addition, it is pointed out that a specific aspect of this study is the in-depth discussion of the latest progress and potential application of remote sensing technology and machine learning in precision agriculture.]

Comments 5: L17-18: The statement “this study adopts the method of literature retrieval and case analysis” is unclear. Please revise for clarity.

Response 5: [In view of this, this study makes a systematic literature search of keywords from Web of Science, Scopus, Google Scholar and PubMed databases, and analyzes in detail the integration and application of remote sensing and machine learning algorithms in different fields of precision agriculture. In addition, the challenges and development prospects of the combination of remote sensing and machine learning in precision agriculture are discussed.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [We correct the incorrect expressions such as "using the method of literature retrieval and case analysis".]

Comments 6: Please numerically label the findings of this work in the abstract, e.g., (1), (2), etc.

Response 6: [The results show that: (1) because of their characteristics, different types of remote sensing data are significantly different in meeting the needs of precision agriculture, in which hyperspectral remote sensing is the most widely used, accounting for more than 30%. The application of UAV remote sensing has the most potential, accounting for about 24%, and showing an upward trend. (2) Machine learning algorithm has obvious advantages in promoting the development of precision agriculture, in which support vector machine algorithm is the most widely used, accounting for more than 20%, followed by random forest algorithm, accounting for about 18%. (3) the integration of remote sensing technology and machine learning can significantly improve the accuracy and efficiency of agricultural monitoring and identification, disease and pest stress detection, land / soil management and crop yield prediction.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [We digitally mark the main findings of the study in order to clearly distinguish each discovery, maintain the content of each discovery, and ensure its relevance to the theme of the entire paragraph.]

Comments 7: L32-34: The sentence “Under the background of the rapid change of global climate, the increasingly tense geopolitical situation and the rising risk of local war, agricultural practice is facing unprecedented uncertainties and challenges” lacks references. Please be cautious when mentioning “local war” and check this statement. Additionally, the current form lacks necessary references. Specifically, briefly mention the uncertainties and challenges in agriculture (suggest 10.1016/j.scitotenv.2024.174289) and provide references on global climate change (suggest 10.3389/fenvs.2023.1304845,10.1016/j.jenvman.2024.121375,and10.1109/jstars.2024.3380514).

Response 7: [In the context of rapid global climate change, agricultural practice is facing unprecedented uncertainties and challenges, such as climate warming, sea-level rise, drought and flood and other extreme hydroclimatic frequent occurrence [1-4].] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [We removed the description of "local war" to avoid unnecessary controversy and provided references to the uncertainties and challenges facing global climate change and agriculture.]

Comments 8: L40: Provide the full name of RS as Remote Sensing.

Response 8: [This progress not only changes the traditional mode of agricultural production, but also aims to optimize agricultural inputs (seeds, water resources, chemicals) through the application of advanced technologies (such as Remote Sensing, Machine Learning algorithms, Agricultural Robots, etc.), effectively manage crop variability, maintain or even increase yields, and cleverly avoid potential losses, thereby improving the efficiency and profitability of agricultural systems [4-6].] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [The full name of RS (Remote Sensing) is provided at line L40.]

Comments 9: L44: Specify the year for Yomo et al.

Response 9: [For example, Yomo et al. found in their 2023 research that the maximum likelihood algorithm based on Landsat images is used to classify land use and land cover. By using the multilayer perceptron neural network-Markov chain modeling method, the results show that the overall accuracy (Kappa coefficient) is as high as 92% [7].] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [The year is specified for Yomo et al at line L44.]

Comments 10: L55: Provide the full name of ML before using the abbreviation.

Response 10: [Therefore, for most countries in the world, it is very necessary to actively promote the coordinated development of precision agriculture, Remote Sensing(RS) and Machine Learning (ML), ensure agricultural production safety, and strictly abide by the food safety red line.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [We provided its full name "Machine Learning" before using the ML abbreviation.]

Comments 11: L58: Please include references.

Response 11: [In recent years, people have made multi-dimensional and deep exploration and efforts in the development of precision agriculture, which is mainly reflected in the research and application of new technologies, covering many key links such as personnel training, policy support and so on [12,13].] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [We are in "in recent years, people in the development of precision agriculture." Added references, 10.1016/j.atech.2024.100416. And 10.1016/j.jafr.2024.101048.]

Comments 12: While the paragraph (L57-71) introduces the advantages of using RS and ML, the following paragraph (L72-87) expands on the idea. The paragraph (L57-71) is lengthy; please consider shortening it and, if possible, combining it with the paragraph (L72-87).

Response 12: [In recent years, people have made multi-dimensional and deep exploration and efforts in the development of precision agriculture, which is mainly reflected in the research and application of new technologies, covering many key links such as personnel training, policy support and so on [12,13]. The aim is to overcome the shortcomings of traditional agriculture, such as time-consuming and laborious, improper use of resources, unstable crop yield and environmental pollution [14,15]. In this context, it is actually a complex and critical challenge to monitor crop growth and conditions in different locations and environments in real time, accurately and at multiple scales, and to use data with different time resolutions to meet a variety of purposes. In fact, it is a complex and critical challenge to respond quickly to extreme events according to changing climate conditions [16,17]. Fortunately, remote sensing technology has developed rapidly in agriculture, forestry, hydrology, environmental protection and other fields because of its unique advantages (such as synchronization, timeliness, spatio-temporal continuity and large-scale observation capability) [18-22]. It can continuously identify, measure and analyze the characteristics of target objects located on, above or even below the earth's surface, without the need for sensors to directly contact observed targets or events [23]. This not only greatly improves the efficiency of agricultural information acquisition, but also provides strong support for dealing with agriculture under extreme climate, so that crop managers can take timely measures to reduce the impact of disasters and ensure the safety and stability of agricultural production[24,25]. In addition, with the maturity of remote sensing inversion algorithms (such as linear regression, PROSAIL physical model, neural network), inversion data sets based on remote sensing images have also appeared, such as inversion product data sets based on MODIS images, Landsat images and Sentinel-2 images: water quality and water environment elements inversion products, vegetation parameter inversion products, land surface temperature inversion products and soil parameter inversion products. It provides reliable and rich data sources for agricultural remote sensing related research [26-28].] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [We combine two paragraphs into one paragraph to improve coherence and fluency, simplify some sentences and make paragraphs more compact.]

Comments 13: L72-75: Provide references for the use of remote sensing-based advancements in agriculture and forestry, hydrology (suggest 10.1016/j.atmosres.2023.106923 and 10.3390/rs15041030), and environmental protection (suggest 10.1007/978-981-19-1600-7_77 and 10.3390/rs15194762).

Response 13: [Fortunately, remote sensing technology has developed rapidly in agriculture, forestry, hydrology, environmental protection and other fields because of its unique advantages (such as synchronization, timeliness, spatio-temporal continuity and large-scale observation ability) [18-22].] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [We are in "in recent years, people in the development of precision agriculture." Added references, 10.1016/j.atech.2024.100416. And 10.1016/j.jafr.2024.101048.]

Comments 14: L98: Introduce the full name of UAV before using the abbreviation.

Response 14: [It is worth mentioning that the emergence of Unmanned Aerial Vehicle (UAV) marks a new era of remote sensing.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [Provided its full name "Unmanned Aerial Vehicle" before using the UAV abbreviation]

Comments 15: L103: Specify the year for Marques et al., Bah et al. (L106), and Yang et al. (L107).

Response 15: [For example, Marques et al have overcome the limitation of limited spectral coverage based on UAV in 2024, especially in low light, fog or smoke conditions to achieve real-time, efficient and distributed accurate monitoring [38]. Bah et al used drone images to detect weeds in the field in 2017 with an accuracy of more than 90% [39]. Yang et al used UAV image information to identify rice lodging based on decision tree algorithm in 2017, and the overall accuracy was 96.17% [40].] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [Specify the year for Marques et al., Bah et al., and Yang et al.]

Comments 16: L110: The term ML should be introduced earlier in the text.

Response 16: [As the core means of dealing with agricultural remote sensing information, ML model has been widely used and deeply studied in recent years. It is a data analysis method that allows computer systems to automatically learn patterns and rules from data without explicit programming.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [ML is introduced at the beginning of the text.]

Comments 17: L130: Consider choosing a more formal word than “pity”.

Response 17: [We deleted the expression “it is a pity that there is little comprehensive and systematic review in the integration ap-plication of the hot research field of precision agriculture”] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.

Comments 18: Figure 1 caption: Specify what (a) and (b) represent. Additionally, Figure 2 is currently of very low quality; consider replacing it.

Response 18: [(a) changes over time of 1200 peer-reviewed high-quality documents retrieved based on keywords; (b) changes over time of more than 300 peer-reviewed high-quality documents confirmed as references after screening] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [We have updated the meaning of figures 1 (a) and (b). In addition, replace figure 2]

Fig2. The geographical distribution of precision agriculture research based on the integration of remote sensing technology and Machine Learning the color depth directly reflects the number of research.

Comments 19: L166: Provide references.

Response 19: [There is no doubt that the application of RS in agriculture has greatly promoted agricultural change. This technology enables us to collect global data on the earth's surface regularly and remotely, providing unprecedented convenience for agricultural production and management.[52-54]] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.[We have added literature, including 10.1016/j.geomorph.2020.107055, 10.1016/j.asr.2015.10.038 and 10.1016/j.compag.2018.08.032]

Comments 20: L195: Refer to the comment for L103.

Response 20: [In their research in 2024, Ren et al used the characteristics of UAV to obtain crop growth status quickly and accurately in small and medium-sized areas, combined with Kalman filtering algorithm to assimilate WOFOST model, significantly improved the accuracy of yield simulation of different treatment schemes, and provided more accurate and reliable yield prediction information for agricultural producers [67].] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [The year 2024 is specified for Ren et al., and modified according to the 104th line format. Make the narrative clearer and more complete]

Comments 21: L204: The term UAV should be introduced earlier in the text.

Response 21: [It is worth mentioning that the emergence of unmanned aerial vehicle (UAV) marks a new era of remote sensing. UAV is a kind of unmanned small aircraft, which is often used to carry remote sensing equipment for aerial data acquisition.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.[ It is worth mentioning that the emergence of unmanned aerial vehicles (UAV) marks a new era of remote sensing. UAV is a kind of unmanned small aircraft, which is often used to carry remote sensing equipment for aerial data acquisition. "]

Comments 22: Figure 4 is of poor quality; it is difficult to see anything, especially in plot (a).

Response 22: [We have revised the quality of figure 4, especially the details in figure 4 (a).] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.

Figure 4. Remote sensing data commonly used in precision agriculture. (a) the distribution characteristics of different types of remote sensing data with time, and marked with different types of colors; (b) based on the literature retrieved in this paper, the statistical analysis of the proportion of all kinds of remote sensing data sources.

Comments 23: L223: Refer to the comment for L103.

Response 23: [The concept of ML can usually be traced back to Alan Turing's classic 1950 research paper on the possibility that machines can exhibit behavior similar to human intelligence [69]. This concept continued to develop in the following decades and gradually became a vital branch of computer science [70].] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [The content is modified according to line 103, so that the coherence of the sentence is guaranteed.]

Comments 24: Figures 5 and 6 are of poor quality and difficult to discern. Consider improving their quality.

Response 24: [We improved the quality of figure 5 to ensure that they were clearly discernible and removed the reference to figure 6] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.

Figure 5. The distribution of the most commonly used machine learning algorithms is obtained based on the keywords "machine learning" and "precision agriculture".

3. Response to Comments on the Quality of English Language

[ We have revised the English of the full text to ensure that the expression is correct and reasonable and conforms to the norms of academic papers.]

Response to Reviewer Comments

1. Summary

Thank you very much for taking the time to review this manuscript. We gave a detailed reply below and highlighted the content.

2. Point-by-point response to Comments and Suggestions for Authors

Comments 1: The article does not present an adequate methodology.

Response 1: [We revised the detailed description of the databases used to retrieve relevant literature (from Web of Science, Scopus, Google Scholar and PubMed, etc.), keywords ("precision agriculture", "machine learning", "remote sensing", "agricultural monitoring", "pest detection", "land use and management", "yield prediction" and "agricultural sustainable development", etc.), and updated the literature (in the past decade). In addition, we used typical cases. It can fully demonstrate the advantages of remote sensing combined with machine learning in precision agriculture.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.

Comments 2: At no point do the authors make clear the steps that were followed to define the articles that make up the review. The key words, selection criteria and other issues relating to the methodology were not described. Nor does the introduction contextualize the importance of the work.

Response 2: [The updated summary and keywords as well as the introduction are for your review: Abstract: With the global population growth, resource shortage and climate change, the contradiction between human and land has become increasingly prominent, these factors constitute a major challenge to the traditional agricultural model. There is an urgent need for a new mode of agricultural production that can achieve efficient, environmental protection and sustainable development. Precision agriculture (PA) is an agricultural practice that uses modern information technology to realize accurate management and decision support of agricultural production process. The practice of PA, which combines remote sensing technology with machine learning algorithm, provides a new possibility to solve these challenges. Although there are many researches on the application of remote sensing technology and machine learning in PA, there are relatively few comprehensive and systematic studies on the integrated application of these two technologies. In view of this, this study makes a systematic literature search of keywords from Web of Science, Scopus, Google Scholar and PubMed databases, and analyzes in detail the integration and application of remote sensing and machine learning algorithms in different fields of PA. In addition, the challenges and development prospects of the combination of remote sensing and machine learning in PA are discussed. The results show that: (1) because of their characteristics, different types of remote sensing data are significantly different in meeting the needs of PA, in which hyperspectral remote sensing is the most widely used, accounting for more than 30%. The application of UAV remote sensing has the most potential, accounting for about 24%, and showing an upward trend. Machine learning algorithm has obvious advantages in promoting the development of PA, in which support vector machine algorithm is the most widely used, accounting for more than 20%, followed by random forest algorithm, accounting for about 18%. (2) in the PA application of remote sensing fusion machine learning, the lack of acquisition and processing of high-quality remote sensing data, poor model interpretation and generalization ability are the main challenges at present; The possible future development is mainly focused on promoting agricultural intelligence and automation, strengthening international cooperation and sharing and sustainable transformation of achievements. To sum up, this study can provide new ideas and reference for remote sensing combined with machine learning in promoting the development of PA.

Keywords: Agricultural monitoring; disease and pest detection, land use and management; yield prediction; agricultural sustainable development

1. Introduction

In the context of rapid global climate change, agricultural practice is facing unprecedented uncertainties and challenges, such as climate warming, sea-level rise, drought and flood and other extreme hydroclimatic frequent occurrence [1-4]. At the same time, the global population is expected to reach 8.7 billion by 2030 and climb to 9.7 billion by 2050, which undoubtedly puts tremendous pressure on global food production [5]. However, it is gratifying that in recent years, with the increase of investment in science and technology and agricultural research, the development of PA has achieved certain results [6,7]. This progress not only changes the traditional mode of agricultural production, but also aims to optimize agricultural inputs (seeds, water resources, chemicals) through the application of advanced technologies (such as remote sensing, machine learning algorithms, agricultural robots, etc.), effectively manage crop variability, maintain or even increase yields, and cleverly avoid potential losses, thereby improving the efficiency and profitability of agricultural systems [8]. For example, Yomo et al found in 2023 that the maximum likelihood algorithm based on Landsat-8 remote sensing images is used to classify land use and land cover, and by using multi-layer perceptron-Markov chain modeling method, the results show that the overall accuracy (Kappa coefficient) is as high as 92% [9]. This study shows that the accuracy of agricultural monitoring and recognition can be significantly improved by integrating advanced remote sensing technology and machine learning algorithm. In addition, another study shows that the integrated learning random forest classifier is used to study the progressive lodging sensitive characteristics of rice types based on multi-spectral (444-842 nm) fusion UAV remote sensing technology, with an overall accuracy of 96.1% [10]. These examples prove the effective application of advanced technologies and algorithms in PA. Although PA has many advantages, limitations such as information accuracy, large amount of data, operational complexity and high initial cost cannot be ignored [11]. Therefore, for most countries in the world, it is very necessary to actively promote the coordinated development of PA, remote sensing (RS) and machine learning (ML), ensure agricultural production safety, and strictly abide by the food safety red line.

In recent years, people have made multi-dimensional and deep exploration and efforts in the development of PA, which is mainly reflected in the research and application of new technologies, covering many key links such as personnel training, policy support and so on [12,13]. The aim is to overcome the shortcomings of traditional agriculture, such as time-consuming and laborious, improper use of resources, unstable crop yield and environmental pollution [14,15]. In this context, it is actually a complex and critical challenge to monitor crop growth and conditions in different locations and environments in real time, accurately and at multiple scales, and to use data with different time resolutions to meet a variety of purposes. In fact, it is a complex and critical challenge to respond quickly to extreme events according to changing climate conditions [16,17]. Fortunately, RS technology has developed rapidly in agriculture, forestry, hydrology, environmental protection and other fields because of its unique advantages (such as synchronization, timeliness, spatio-temporal continuity and large-scale observation ability) [18-22]. RS is a technology that can obtain the information of the earth's surface without physical contact. It uses sensors to capture and record electromagnetic radiation signals reflected, emitted or scattered from the earth's surface from a long distance, and then continuously identify, measure and analyze the characteristics of target objects located on, above or even below the earth's surface by analyzing these signals [23]. This not only greatly improves the efficiency of agricultural information acquisition, but also provides strong support for dealing with agriculture in extreme weather, so that crop managers can take timely measures to reduce the impact of disasters and ensure the safety and stability of agricultural production [24,25]. In addition, with the maturity of RS inversion algorithms (such as linear regression, PROSAIL physical model, neural network), inversion data sets based on RS images have also appeared, such as inversion product data sets based on MODIS images, Landsat images and Sentinel-2 images: water quality and water environment elements inversion products, vegetation parameter inversion products, land surface temperature inversion products and soil parameter inversion products. It provides reliable and rich data sources for agricultural RS related research [26-28].

It is well known that most agricultural RS data are information provided by visible light and near infrared radiation reflected (or transmitted) by plants, which are measured according to wavelength, that is, spectral reflectance [29]. According to the change of vegetation, the spectral data commonly used in PA include visible light (400 nm), near infrared (700 nm) and short-wave infrared (1300 nm) [30,31]. In addition, multi-spectral remote sensing and hyperspectral remote sensing to meet different needs have also been proved to be effective means of plant phenotypic analysis, crop index acquisition and stress monitoring [32]. For example, European Sentinel-2, ENVISAT MERIS, French SPOT satellite and NOAA AVHRR satellite data, India's Hyperion, China's GF series data and HJ remote sensing data have been widely used [33-35]. It is worth mentioning that the emergence of unmanned aerial vehicle (UAV) marks a new era of RS. UAV is a kind of unmanned small aircraft, which is often used to carry remote sensing equipment for aerial data acquisition. It can provide more abundant and comprehensive spectral, spatial and temporal resolution data, vegetation height data and multi-angle observation, and has the characteristics of high efficiency, convenience, low cost and strong adaptability [36]. There have been many successful cases in crop classification, weed detection and vegetation monitoring, which prove the feasibility of UAV in PA [37]. For example, Marques et al have overcome the limitation of limited spectral coverage based on UAV in 2024, especially in low light, fog or smoke conditions to achieve real-time, efficient and distributed accurate monitoring [38]. Bah et al used drone images to detect weeds in the field in 2017 with an accuracy of more than 90% [39]. Yang et al used UAV image information to identify rice lodging based on decision tree algorithm in 2017, with an overall accuracy of 96.17% [40].

As the core means of dealing with agricultural remote sensing information, ML model has been widely used and deeply studied in recent years. ML is a data analysis method that allows computer systems to automatically learn patterns and rules from data without explicit programming. Researchers tend to use ML as an integrated framework for feature collection and classification, prediction, or decision support [42]. With the improvement of big data's computing power, many classical algorithms have been optimized and improved, and new models and methods continue to emerge [43]. Common ML methods include decision tree (DT), support vector machine (SVM) and logical regression (LR). The core of these methods is to find optimized statistical information ways, so as to automatically and efficiently solve practical problems such as classification and regression [44]. In addition, the convolution neural network (CNN) method based on ML has unique advantages in the field of image processing. It can automatically extract deep features from images and achieve accurate classification or recognition tasks [45]. Because of its unique data expression ability, these technologies can learn and extract valuable information automatically, thus effectively avoiding the complexity and subjectivity brought by traditional methods, and greatly improve the efficiency and generalization of processing multi-platform RS data [46]. It is with these advantages that ML has attracted more attention from agricultural researchers and experts, and listed it as the engine factor for the development of PA [47-51].

In view of this, this study conducted a preliminary literature search based on the keywords "RS", "PA" and "ML", and screened out more than 1200 related research articles. Through quantitative analysis of these documents, we have observed an overall upward trend in the number of related publications over the past two decades (as shown in figure 1 (a)). In order to ensure the comprehensiveness and depth of the research, we further conducted a more accurate literature search with "agricultural monitoring", "pest detection", "land use and management", "yield prediction" and "agricultural sustainable development" as key words. After rigorous screening, more than 330 high-quality peer-reviewed journal articles published in the past 10 years have been identified (as shown in figure 1 (b)). Although the number of research papers in this field is on the rise, there is still a lack of systematic review on the integration of RS technology and ML in PA. From the perspective of international cooperation, researchers in China, the United States, Brazil and other countries have made significant contributions to the application of RS technology and ML in the field of PA. In addition, there is an obvious imbalance in the spatial distribution of these studies, and there are great differences in the volume of research activities among different regions (as shown in figure 2). Therefore, through the in-depth analysis and summary of the existing research results, it is very important to systematically summarize the application status of RS and ML in PA, and discuss the current challenges and possible development directions in this field.

Figure 1. Retrieve the distribution of published high-quality research papers over time. (a) changes over time of 1200 peer-reviewed high-quality documents retrieved based on keywords; (b) changes over time of more than 330 peer-reviewed high-quality documents confirmed as references after screening.

Figure 2. The geographical distribution of precision agriculture research based on the integration of remote sensing technology and ML the color depth directly reflects the number of research.

] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.[We revised the structure and presentation of the full text, added detailed literature retrieval methods, and updated the literature as a peer-reviewed high-quality paper in the field of remote sensing combined with machine learning in precision agriculture.]

Comments 3: The sentences need to be revised to make the statements and the text as a whole clearer.

Response 3: [The full text is as follows for your review: 1. Introduction

In the context of rapid global climate change, agricultural practice is facing unprecedented uncertainties and challenges, such as climate warming, sea-level rise, drought and flood and other extreme hydroclimatic frequent occurrence [1-4]. At the same time, the global population is expected to reach 8.7 billion by 2030 and climb to 9.7 billion by 2050, which undoubtedly puts tremendous pressure on global food production [5]. However, it is gratifying that in recent years, with the increase of investment in science and technology and agricultural research, the development of PA has achieved certain results [6,7]. This progress not only changes the traditional mode of agricultural production, but also aims to optimize agricultural inputs (seeds, water resources, chemicals) through the application of advanced technologies (such as remote sensing, machine learning algorithms, agricultural robots, etc.), effectively manage crop variability, maintain or even increase yields, and cleverly avoid potential losses, thereby improving the efficiency and profitability of agricultural systems [8]. For example, Yomo et al found in 2023 that the maximum likelihood algorithm based on Landsat-8 remote sensing images is used to classify land use and land cover, and by using multi-layer perceptron-Markov chain modeling method, the results show that the overall accuracy (Kappa coefficient) is as high as 92% [9]. This study shows that the accuracy of agricultural monitoring and recognition can be significantly improved by integrating advanced remote sensing technology and machine learning algorithm. In addition, another study shows that the integrated learning random forest classifier is used to study the progressive lodging sensitive characteristics of rice types based on multi-spectral (444-842 nm) fusion UAV remote sensing technology, with an overall accuracy of 96.1% [10]. These examples prove the effective application of advanced technologies and algorithms in PA. Although PA has many advantages, limitations such as information accuracy, large amount of data, operational complexity and high initial cost cannot be ignored [11]. Therefore, for most countries in the world, it is very necessary to actively promote the coordinated development of PA, remote sensing (RS) and machine learning (ML), ensure agricultural production safety, and strictly abide by the food safety red line.

In recent years, people have made multi-dimensional and deep exploration and efforts in the development of PA, which is mainly reflected in the research and application of new technologies, covering many key links such as personnel training, policy support and so on [12,13]. The aim is to overcome the shortcomings of traditional agriculture, such as time-consuming and laborious, improper use of resources, unstable crop yield and environmental pollution [14,15]. In this context, it is actually a complex and critical challenge to monitor crop growth and conditions in different locations and environments in real time, accurately and at multiple scales, and to use data with different time resolutions to meet a variety of purposes. In fact, it is a complex and critical challenge to respond quickly to extreme events according to changing climate conditions [16,17]. Fortunately, RS technology has developed rapidly in agriculture, forestry, hydrology, environmental protection and other fields because of its unique advantages (such as synchronization, timeliness, spatio-temporal continuity and large-scale observation ability) [18-22]. RS is a technology that can obtain the information of the earth's surface without physical contact. It uses sensors to capture and record electromagnetic radiation signals reflected, emitted or scattered from the earth's surface from a long distance, and then continuously identify, measure and analyze the characteristics of target objects located on, above or even below the earth's surface by analyzing these signals [23]. This not only greatly improves the efficiency of agricultural information acquisition, but also provides strong support for dealing with agriculture in extreme weather, so that crop managers can take timely measures to reduce the impact of disasters and ensure the safety and stability of agricultural production [24,25]. In addition, with the maturity of RS inversion algorithms (such as linear regression, PROSAIL physical model, neural network), inversion data sets based on RS images have also appeared, such as inversion product data sets based on MODIS images, Landsat images and Sentinel-2 images: water quality and water environment elements inversion products, vegetation parameter inversion products, land surface temperature inversion products and soil parameter inversion products. It provides reliable and rich data sources for agricultural RS related research [26-28].

It is well known that most agricultural RS data are information provided by visible light and near infrared radiation reflected (or transmitted) by plants, which are measured according to wavelength, that is, spectral reflectance [29]. According to the change of vegetation, the spectral data commonly used in PA include visible light (400 nm), near infrared (700 nm) and short-wave infrared (1300 nm) [30,31]. In addition, multi-spectral remote sensing and hyperspectral remote sensing to meet different needs have also been proved to be effective means of plant phenotypic analysis, crop index acquisition and stress monitoring [32]. For example, European Sentinel-2, ENVISAT MERIS, French SPOT satellite and NOAA AVHRR satellite data, India's Hyperion, China's GF series data and HJ remote sensing data have been widely used [33-35]. It is worth mentioning that the emergence of unmanned aerial vehicle (UAV) marks a new era of RS. UAV is a kind of unmanned small aircraft, which is often used to carry remote sensing equipment for aerial data acquisition. It can provide more abundant and comprehensive spectral, spatial and temporal resolution data, vegetation height data and multi-angle observation, and has the characteristics of high efficiency, convenience, low cost and strong adaptability [36]. There have been many successful cases in crop classification, weed detection and vegetation monitoring, which prove the feasibility of UAV in PA [37]. For example, Marques et al have overcome the limitation of limited spectral coverage based on UAV in 2024, especially in low light, fog or smoke conditions to achieve real-time, efficient and distributed accurate monitoring [38]. Bah et al used drone images to detect weeds in the field in 2017 with an accuracy of more than 90% [39]. Yang et al used UAV image information to identify rice lodging based on decision tree algorithm in 2017, with an overall accuracy of 96.17% [40].

As the core means of dealing with agricultural remote sensing information, ML model has been widely used and deeply studied in recent years. ML is a data analysis method that allows computer systems to automatically learn patterns and rules from data without explicit programming. Researchers tend to use ML as an integrated framework for feature collection and classification, prediction, or decision support [42]. With the improvement of big data's computing power, many classical algorithms have been optimized and improved, and new models and methods continue to emerge [43]. Common ML methods include decision tree (DT), support vector machine (SVM) and logical regression (LR). The core of these methods is to find optimized statistical information ways, so as to automatically and efficiently solve practical problems such as classification and regression [44]. In addition, the convolution neural network (CNN) method based on ML has unique advantages in the field of image processing. It can automatically extract deep features from images and achieve accurate classification or recognition tasks [45]. Because of its unique data expression ability, these technologies can learn and extract valuable information automatically, thus effectively avoiding the complexity and subjectivity brought by traditional methods, and greatly improve the efficiency and generalization of processing multi-platform RS data [46]. It is with these advantages that ML has attracted more attention from agricultural researchers and experts, and listed it as the engine factor for the development of PA [47-51].

In view of this, this study conducted a preliminary literature search based on the keywords "RS", "PA" and "ML", and screened out more than 1200 related research articles. Through quantitative analysis of these documents, we have observed an overall upward trend in the number of related publications over the past two decades (as shown in figure 1 (a)). In order to ensure the comprehensiveness and depth of the research, we further conducted a more accurate literature search with "agricultural monitoring", "pest detection", "land use and management", "yield prediction" and "agricultural sustainable development" as key words. After rigorous screening, more than 330 high-quality peer-reviewed journal articles published in the past 10 years have been identified (as shown in figure 1 (b)). Although the number of research papers in this field is on the rise, there is still a lack of systematic review on the integration of RS technology and ML in PA. From the perspective of international cooperation, researchers in China, the United States, Brazil and other countries have made significant contributions to the application of RS technology and ML in the field of PA. In addition, there is an obvious imbalance in the spatial distribution of these studies, and there are great differences in the volume of research activities among different regions (as shown in figure 2). Therefore, through the in-depth analysis and summary of the existing research results, it is very important to systematically summarize the application status of RS and ML in PA, and discuss the current challenges and possible development directions in this field.

Figure 1. Retrieve the distribution of published high-quality research papers over time. (a) changes over time of 1200 peer-reviewed high-quality documents retrieved based on keywords; (b) changes over time of more than 330 peer-reviewed high-quality documents confirmed as references after screening.

Figure 2. The geographical distribution of precision agriculture research based on the integration of remote sensing technology and ML the color depth directly reflects the number of research.

2. Materials and Methods

2.1 Remote sensing data in precision agriculture

There is no doubt that the application of RS technology in agriculture has greatly promoted agricultural reform. This technology enables us to collect global data on the earth's surface remotely on a regular basis, providing unprecedented convenience for agricultural production and management [52-54]. Through a variety of sensors, we can directly or indirectly obtain almost all the key elements of agricultural practice, from crop growth to soil moisture monitoring, pest and pest early warning to yield prediction. At the same time, the wide geographical coverage and diversified resolution of RS technology also provide valuable data support for agricultural production and management [55]. As shown in figure 3, remote sensing satellites with different resolutions play different key roles in different PA practices, and rely on different characteristics and advantages to comprehensively serve the specific needs of PA from many angles [56,57]. With the continuous updating and upgrading of remote sensing sensors, agricultural managers and practitioners will continue to benefit from the in-depth application of RS technology, for example, remote sensing data show high practicability and effectiveness in evaluating and monitoring agricultural practice [58,59].

In general, when obtaining RS data, the value of RS images with appropriate resolution, band, reliable quality and cost-effectiveness can be maximized if they are selected according to specific agricultural problems [60,61]. For example, in the health assessment of urban trees using WorldView-3 satellite RS data, although the overall growth dynamics of trees can be monitored, there are some limitations in capturing the specific details of the dynamics of growth time in different parts of trees [62]. However, in another study, the use of Landsa-8 RS images with a spatial resolution of 10 to 30 meters provided a promising solution for disease detection in mixed forests in southern China [63]. In other studies, for the detection of plant diseases and pests infecting vegetation, the detection accuracy does not seem to be satisfactory based on visible light (780nm) data [64]. In a 2023 study by Zhu et al., although the use of drone RS can confirm the importance of red-light bands and adjacent bands, it has not achieved the desired results in the investigation of plant diseases and pests invading vegetation [65]. However, it is gratifying that multi-spectral RS data with rich bands and a wide range of wavelengths can capture subtle changes in infected plants affected by diseases and insect pests, thus showing excellent ability in early pest detection [66]. In their research in 2024, Ren et al used the characteristics of UAV to obtain crop growth status quickly and accurately in small and medium-sized areas. By assimilating RS data with WOFOST model effectively by Kalman filter algorithm, the accuracy of yield simulation of different processing schemes is significantly improved, and more accurate and reliable yield prediction information is provided for agricultural producers [67].

In the practical application of PA, according to different requirements and application scenarios, commonly used RS data sources include hyperspectral, multispectral, thermal infrared remote sensing, LiDAR remote sensing, SAR remote sensing and UAV remote sensing and so on. As shown in figure 4, the application of various RS data sources in PA is shown in detail, including the time distribution and proportion of RS data sources in PA. These informations undoubtedly provide valuable reference for agricultural managers and practitioners, not only help them have a more comprehensive and in-depth understanding of the characteristics and applicability of various RS technologies, but also provide strong support for them to make scientific and reasonable decisions in practical work.

Figure 3. Based on the comprehensive application framework of different remote sensing satellites in precision agriculture.

Figure 4. Remote sensing data commonly used in precision agriculture. (a) the distribution characteristics of different types of remote sensing data with time, and marked with different types of colors; (b) based on the literature retrieved in this paper, the statistical analysis of the proportion of all kinds of remote sensing data sources.

2.2 Overview of ML algorithms in Precision Agriculture

The concept of machine learning (ML) can usually be traced back to Alan Turing's classic research article published in 1950, that is, the possibility that machines can exhibit behaviors similar to human intelligence [68,69]. This concept continued to develop in the following decades and gradually became a vital branch of computer science. The core principle of ML is to automatically learn and sum up the rules in the input data, and realize the accurate prediction or classification of unknown data by extracting key features and constructing mapping functions [70]. In addition, as the core component of artificial intelligence, ML gives computer systems the ability to perform a variety of tasks efficiently, and continues to promote the innovation and development of intelligent technology [71]. Generally speaking, ML mainly contains three elements, namely: model, objective function and optimization algorithm. The model explains the correlation between input and output and the meaning and range of the parameters, the objective function measures the difference between the model prediction and the actual results, and the optimization algorithm minimizes or maximizes the objective function by iteratively adjusting the parameters. as a result, the best model parameters are obtained [72,73]. According to different types of learning, ML can be divided into four main categories: supervised learning, unsupervised learning, semi-supervised learning and reinforcement learning [74,75]. As shown in Table 1, the applications of the four main categories of algorithms in PA and their scope of application are listed and described in detail.

Table 1. Common Machine Learning algorithms and references in the Application Field of Precision Agriculture.

Figure 5. The distribution of the most commonly used machine learning algorithms is obtained based on the keywords "machine learning" and "precision agriculture".

Recent studies have shown that with the improvement of computing performance and the enhancement of massive data sets, ML has shown strong application capabilities in many fields, especially in the field of PA [120,121]. In particular, a series of emerging algorithms and technologies such as deep learning (DL), intelligent optimization, neural networks, computer vision and data enhancement continue to emerge. It has not only injected strong impetus into the field of ML, but also provided rich opportunities for all stages of agriculture. They enable agricultural practitioners to respond more effectively to challenges and achieve set goals [122]. As shown in figure 5, the frequency distribution of the algorithm is obtained by searching the keywords "ML" and "PA". From the chart, we can see that ML is widely used in the field of PA, and support vector machine algorithm has the highest frequency, accounting for more than 20%, followed by random forest algorithm, accounting for about 18%.

In addition, in the specific application of PA, different algorithms have given full play to their own advantages, and achieved a series of encouraging results. For example, Sladojevic et al proposed a new plant leaf disease detection and classification model based on deep convolution neural network. The model can accurately identify 13 different plant diseases and effectively distinguish plant leaves from the surrounding environment, which provides a powerful tool for plant health monitoring [123]. Li et al have made remarkable progress in the field of vegetable disease detection. They propose a lightweight network improvement algorithm based on YOLOv5s. The algorithm effectively eliminates external interference and significantly enhances the ability of multi-scale feature extraction, thus improving the scope and performance of disease detection [124]. Ashwinkuma et al. developed a convolution neural network based on the optimal mobile network, which is used to automatically detect and classify plant leaf diseases. The experimental results show that the CNN model performs well, the maximum accuracy is 0.985, the recall rate is 0.9892, the accuracy is 0.987 and the Kappa coefficient is 0.985 [125]. Yu et al use DL target detection technology to extract image feature information through complex network structure to achieve non-destructive recognition of crop diseases. Compared with the traditional method, this technique has higher recognition accuracy, faster detection speed and good stability in the visible light range [126]. Ang et al. creatively used Landsat-8 time series satellite images, combined with ML and normalized difference vegetation index (NDVI), successfully developed a new method, which effectively proved its value in yield prediction [127]. Aydin et al tested gradient lifting methods such as XGBoost, LightGBM and CatBoost for soil sample classification, and achieved high classification accuracy of up to 90%. Compared with previous studies, the prediction accuracy has been significantly improved [128].

3. Results

3.1 Agricultural monitoring and identification

Recent scientific research shows that the integration of RS technology and ML methods has brought remarkable progress to the application of agricultural monitoring and identification. RS technology can efficiently obtain crop planting area, growth status and other important information, while ML technology can accurately detect targets and extract features from these rich RS data, so as to achieve fine identification and classification of crops. With the continuous integration and development of these two technologies, the related agricultural problems, such as horticulture line detection [129], crop detection and classification [130,131], vegetation distribution [132,133], will usher in a new turning point. For example, Zhao et al improved the standardized precipitation evapotranspiration index (SPEI) by integrating RS data, and the results show that the new SPEI can greatly enhance the capacity of agricultural drought monitoring [134]. Lyu et al used EO-1 Hyperion images, combined with multi-terminal spectral mixing analysis (MESMA) and fully constrained least square pixel mixing (FCLS) techniques, to successfully identify typical vegetation species and improve the accuracy of grassland degradation monitoring [135]. Xiao et al fused Sentinel-2 and MODIS RS images using the enhanced spatio-temporal adaptive reflection fusion model (ESTARFM), and then accurately obtained the spatial distribution of irrigated rice fields by random forest (RF) algorithm. Based on the Penman-Monteith model and making full use of the daily observation data of the meteorological station, the dynamic monitoring of water resources in the critical irrigation period has been realized, and remarkable results have been achieved [136]. This application increases the feasibility of spatio-temporal fusion of multi-source RS data and makes it possible to continuously monitor the irrigation dynamics of paddy fields on a large scale.

In addition, it is very important to grasp the planting and distribution of crops on a large scale in a timely and efficient manner. Although scholars have done a lot of research on the basis of low and medium resolution RS, due to the widespread existence of mixed pixels and the lack of red edge bands, these techniques are difficult to effectively identify small plots of farmland, resulting in unsatisfactory recognition accuracy [137,138]. However, the research of Guo et al in this field has brought new breakthroughs. Using GF-6 WFV images, they constructed several decision tree models, which not only efficiently obtained the information of crop planting area and its spatial distribution, but also significantly improved the accuracy of image recognition [139]. In their latest research, Zhang et al used GF-1 RS images, combined with advanced multi-scale segmentation algorithms, to improve the accuracy of identifying forest types in Engebe ecological demonstration areas, and used nearest neighbor classification and random forest (RF) classification respectively, and compared the recognition results. The results showed that the effect of random forest classification was better, and the Kappa coefficients obtained in two consecutive years were 0.92 and 0.90 [99] respectively. Through a lot of research, people have reached a consensus: ML algorithms, including random forest (RF), support vector machine (SVM), artificial neural network (ANN), decision tree (DT) and so on, have greater potential in agricultural monitoring and recognition, and can significantly improve the efficiency and accuracy of monitoring and recognition [140-147].

3.2 Stress detection of diseases and insect pests

Crop diseases and insect pests are not only one of the core factors affecting plant yield and quality, but also one of the main causes of crop damage. As emphasized by the United Nations Food and Agriculture Organization, the damage caused by diseases and insect pests to agro-ecosystems should not be underestimated. Unfortunately, in early agricultural production, the problem of diseases and insect pests is often marginalized and not given enough attention, resulting in huge economic losses [148]. Therefore, the implementation of efficient and accurate detection of plant diseases and insect pests not only plays a vital role in ensuring the health of agro-ecosystem, improving crop yield and quality, and reducing economic losses, but also plays a key role in the development of PA. in recent years, it has attracted the attention of many scholars [149]. The premise to achieve this goal is to accurately detect and classify diseases and insect pests, and to distinguish different types and estimated quantities, in order to implement accurate pest prevention and control strategies. Traditional pest monitoring mainly depends on the manual identification of insect experts or technicians, this method is not only subjective and labor-intensive, but also not practical in large-scale applications [150]. However, with the popularity of sensors and embedded devices on the Internet, the combination of RS and ML has opened up a new way for the detection of diseases and insect pests in modern agriculture [151]. Mahanta et al obtained rich spectral features of vegetation based on a variety of sensor devices, and used ML models to identify spectral patterns related to specific diseases. Finally, the evaluation of the health degree of insects invading the forest was realized, and the detection efficiency was greatly improved [152].

Figure 6. Based on IDS map, MODIS map, Landsat map and hyperspectral response technology, the evaluation process of detecting insect infestation forest health was realized.

Most diseases and insect pests have the characteristics of concealment, latent, infectivity and uncertainty, which undoubtedly increase the difficulty and cost of control and bring great challenges to agricultural production [153,154]. However, it is worth noting that RS data from satellite sensors show that plants affected by the disease can be distinguished in a relatively short period of time by their spectral characteristics different from healthy plants [155]. The figure 6 shows the process of pest monitoring based on different types of sensors. More importantly, through further use of ML for analysis, we can not only determine the degree of damage, but also accurately identify the type of disease [156,157]. In the early detection of diseases and insect pests, many researchers tend to use traditional methods, that is, to establish empirical statistical models between diseases and insect pests and related factors (such as environment, climate, soil and vegetation index) [158,159]. In order to achieve effective monitoring of diseases and insect pests. These methods include multiple linear regression, partial least square regression, support vector regression and random forest regression. For example, Ebrahimi et al use support vector regression to detect parasites in crop canopy images, which greatly improves the detection accuracy [160].

In addition, through the comprehensive use of advanced image processing techniques such as image segmentation [161], feature extraction [162], target detection [163-166] and classification, researchers can solve complex problems in plant disease detection more accurately and efficiently [167]. Image segmentation technology is used to distinguish normal and abnormal leaves in RS images, while feature extraction is to extract meaningful information from segmented regions, such as color, texture, shape, etc. [168,169]. They provide a more detailed and accurate analysis method for disease detection [170,171]. For example, studies by Zhang et al have shown that TinySegformer model can provide a robust and practical solution for large-scale agricultural pest detection because of its high efficiency, accuracy and lightweight [172]. The interactive segmentation method based on GrabCut proposed by Lu et al., which is applied to field RS, can quickly extract locust images from various parts [173]. Barbedo et al. proposed an automatic detection algorithm for wheat scab based on hyperspectral technology. the algorithm has more than 91% classification accuracy and shows excellent robustness under a variety of complex factors such as shape, direction and shadow [174]. Mumtaz et al. combined with optical RS, image processing and depth learning methods to accurately detect and grade wheat rust [175]. Bao et al. proposed a RS method of UAV based on DDMA-YOLO, which can not only reduce the workload and time consumption of pest detection, but also effectively improve the detection efficiency [176].

Many studies have shown that the image fusion method can greatly enhance the accuracy of vegetation disease detection by fusing image information from different sensors or multi-stage processing [177-181]. For example, some scholars apply DL technology to the fusion of RGB images and segmented images, and develop a multi-head DenseNet architecture. After the strict verification of the public data set and the application of 50% discount cross-validation technology, the method shows excellent performance, and all the evaluation indicators have reached a very high level. for example, the average accuracy, recall, accuracy and F1 score reached 98.17%, 98.17%, 98.16% and 98.12% respectively [182]. Based on the multi-source fusion images of UAV and visible light, Ma et al successfully constructed a variety of ML models, which significantly improved the accuracy of cotton Vickers wilt detection [183].

It is worth mentioning that some researchers have adopted the improved DL algorithm framework for plant disease detection, and achieved remarkable results [184-187]. Dong et al. creatively proposed an effective scale-aware network architecture (ESA-Net) based on low-cost RS images [188]. After strict verification, ESA-Net showed excellent performance in plant disease detection, and achieved strong competitive results. Amarathunga et al proposed a new architecture based on visual converter (ViT), which integrates the attention mechanism driven by domain knowledge and effectively improves the accuracy of micro-pest detection and recognition at the species level [189]. Ye et al designed an end-to-end automatic disease detection framework based on multi-scale MA-UNet model and single-phase image based on UAV aerial photography data and Landsat 8 satellite RS markers, which greatly improved the efficiency and accuracy of disease monitoring [190].

3.3 Management and analysis of soil and land

As the cornerstone of human survival and development, land not only carries the key mission of agricultural production, provides us with food to maintain our livelihood, but also is an indispensable key prerequisite for ensuring human well-being [60,191]. Therefore, the management and analysis of soil and land resources is particularly urgent and important. In the field of soil monitoring and management, traditionally, we rely on field survey methods to obtain the spatial distribution data of soil groups [191]. However, these methods have many shortcomings, such as long monitoring period, high cost, complex operation procedures, many subjective judgment factors, and relatively limited accuracy [192]. Therefore, using traditional methods for soil monitoring is not only time-consuming and labor-consuming, but also may be difficult to meet the needs of modern soil management for accuracy and efficiency [16]. With its more accurate, richer and more professional characteristics, RS has brought revolutionary changes to soil monitoring and management activities. It provides multi-temporal images, enabling us to fully capture dynamic changes in land and soil characteristics [193-195]. In addition, RS has a wide range of data sources with large amounts of information and high accuracy, providing unprecedented possibilities for accurate assessment of soil conditions [196]. The use of advanced ML technology can achieve efficient and accurate processing and analysis of RS data, so as to realize the automation of data processing and feature extraction. It is very important to improve the efficiency and accuracy of soil and land management [197-199].

A survey found that the application of various types of RS data provides convenience and opportunities for soil management [200]. At the same time, in different RS soil applications, multi-spectral RS is the most widely used in soil [201]. Duan et al used the mean value of reflectivity and entropy texture parameters extracted from Landsat- 8 image, combined with MLC, SVM, ANN and RF ML, to identify soil groups in depth, and achieved good results. Zhou et al proposed a general ML method based on spatio-temporal constraints by using Sentinel-1 and Sentinel-2 data in 2024 [202]. Through verification, its accuracy and practicability have been fully affirmed [203]. Musasa et al made a detailed review of soil problems in arid environments in 2023, clearly pointing out that the Landsat satellite mission plays an indispensable role in promoting soil assessment and monitoring [204]. In addition, in view of the significant challenges such as insufficient information acquisition and limited measurement accuracy in the early soil moisture monitoring technology [205], the introduction of ML technology is a revolutionary change, which greatly makes up for these deficiencies [206,207]. In addition, high-resolution data show significant applicability in soil applications, especially in soil resource estimation and mapping [208-211]. Moreover, UAV shows great potential in soil analysis and evaluation, and many studies have fully proved its effectiveness in practical application. For example, Bertalan et al., through the mapping of soil moisture based on drones, deeply revealed the spatial heterogeneity of soil moisture and provided strong support for PA [212]. Menzies Pluer et al used UAV to draw the spatial distribution model of farmland soil characteristics and nutrient concentration, which provided a novel and low-cost method and new idea for soil management [213]. In addition, scholars also pointed out that the combination of UAV data fusion and ML is very important for accurate field estimation of soil texture [214-216]. At the same time, in many studies on the integrated application of RS and ML in soil management, we found that the discussion of soil organic carbon and salinity is also an eye-catching direction [217-221].

As one of the key factors of global ecological change, land use / land cover (LULC) has a far-reaching impact on the balance of ecosystem and the sustainable development of human society [222]. It represents the different ways in which human beings maximize the use of land resources and deal with related resources, and is very important for land management and analysis [223]. Therefore, in the research field of land management and analysis, we pay special attention to the temporal and spatial distribution of LULC and its applications. It goes without saying that the application of ML to RS data is of great significance for efficient and accurate land management and analysis [224]. On the one hand, traditional land management and analysis methods are often time-consuming and costly, and it is difficult to provide up-to-date information on various land use / land cover changes [225]. On the other hand, with its strong data acquisition and processing ability, RS can extract high-resolution multispectral information covering large areas that are difficult to access in real time, making land management and classification more cost-effective and time-saving [226,227]. Figure 7 shows the whole process of determining farming patterns using Landsat-8 and MODIS RS data, greatly improving the efficiency of agricultural practices [228]. In recent years, with the continuous development of ML technology, it is becoming more and more popular in mapping, analysis and land spatio-temporal analysis of LULC changes using RS data [229,230]. Examples of land management and analysis based on different ML methods include: random forest [100,101,231], support vector machine [102,103,232], decision tree [90,91,233], maximum likelihood classification [234,235], artificial neural network [97,99,236], convolution neural network [237,242], hybrid multiple model [243,246].

Figure 7. A general process for determining tillage patterns and cultivated / non-cultivated land areas based on multi-source remote sensing data.

3.4 Prediction and decision-making of crop yield

Determining crop yield information plays an indispensable role in crop field management, and crop yield prediction is one of the important cornerstones to ensure food security [247,248]. Traditional crop yield prediction methods usually involve destructive sampling, which not only wastes a lot of human and material resources in practical application, but also is inefficient and cannot meet the needs of the development of modern PA [249]. In order to overcome this bottleneck, we conducted an in-depth and systematic review of the literature, which covered many aspects, such as RS data sources, biological and abiotic factors, physical and chemical parameters, modeling methods and so on. the aim is to provide a more accurate and efficient crop yield prediction scheme and provide strong support for the sustainable development of yield prediction and decision-making.

We have learned that there are differences in the applicability and accuracy of operational assessment of crop status and yield based on different ML algorithms and RS data from different sources. Multi-spectral and medium-resolution RS represented by MODIS data are widely used in early crop yield prediction, and show potential uses [250,251]. Hyperspectral data have unique advantages in prediction, especially data from Landsat-8 satellites and hyperspectral imagers. Related studies have shown that they show great potential in yield prediction of crops such as citrus, wheat, corn, sugarcane and so on [252-256]. In addition, airborne LiDAR and high spatial and temporal resolution images are more suitable for crop yield prediction in fine abundance models [257-260]. A number of studies have shown that UAV data provide accurate and efficient support for PA prediction, especially in crop yield estimation accuracy and phenotypic analysis [261-267]. As shown in figure 8, Yang et al predicted maize yield based on UAV multispectral images combined with ML technology, revealing the great potential of UAV in yield prediction [268].

Figure 8. Prediction Framework of Maize yield based on UAV Image and Machine Learning.

In addition, as an integral part of PA practice, yield forecasting usually does not exist in isolation, but is the result of the interweaving and interaction of climate, soil, water, diseases and insect pests, management and other factors. For example, in an in-depth study, Anwar et al revealed that Australian wheat yields are extremely sensitive to climatic factors [269]. Bai et al. made it clear that assessing the impact of extreme weather on crop production is a key prerequisite for exploring agronomic measures to address climate change, and that fluctuations in climate variables closely related to crop production can have a profound impact on regional and global food production [270]. The importance of soil as a key factor affecting crop yield cannot be ignored. By combining RS data with ML, we can evaluate soil properties more accurately and, taking into account cost-effectiveness and time-benefit, achieve accurate prediction of crop yield [271,272]. Fry et al discussed the spatial variability between field soil properties and soybean yield and found that there was a significant correlation between different soil properties and changes in soybean yield, mainly affected by soil texture and organic carbon content in topsoil (the first 20cm) rather than surface topography [273]. In order to explore the actual effect of water on yield, Zain et al failed to consider the adaptability of the model, which led to adverse results [274]. In another study, Wang et al. developed an accurate polynomial function model, which can effectively adapt to the characteristics of irrigation and application in different areas, provides a scientific guidance strategy for water and fertilizer management, and realizes the accurate prediction of crop yield combined with advanced ML [275]. In addition, the impact of diseases and insect pests and management on yield estimates is also of concern [276-278].

In recent years, the combination of RS technology and ML for crop yield estimation and decision-making has become a research direction with great potential and prospect. The integration of this field not only improves the accuracy and efficiency of crop yield estimation, but also provides strong technical support for the fine management of agricultural production [265]. In this process, the selection of crop physical and chemical parameters is particularly important, which is directly related to the accuracy and reliability of the yield prediction model. Commonly used physiochemical parameters include vegetation coverage (FVC) [279], photosynthetically active radiation absorption (FPAR) [280-282], evapotranspiration (ET) [283-285], leaf area index (LAI) [251,286,287], chlorophyll content [288-290], and various vegetation indices (VIs). Such as normalized difference vegetation index (NDVI) [291-293] and enhanced vegetation index (EVI) [294]. These physical and chemical parameters and indexes are not only widely used in actual agricultural production, but also closely related to yield estimation.

In addition, the crop yield prediction model is also constantly adapting to a variety of new situation changes. For example, although the early traditional ground survey methods and sampling statistics methods based on empirical knowledge have experienced a lot of research and practice, they cannot meet the needs of improving the accuracy of production prediction and reducing costs [295]. With the application of crop growth model and data assimilation model, the yield prediction accuracy will be greatly improved [296]. For example, Zhang et al and Kheir et al have made yield predictions based on APSIM crop model, and achieved remarkable results [297,298]. In addition, the WOFOST model also performs well in crop yield prediction, and a number of studies have revealed its potential use in forecasting [299,300]. The SAFY model provides a new perspective and idea for the estimation of crop yield in a large area [301]. However, crop models are not perfect. They may be limited in large-scale applications, errors are easy to accumulate, and there are some problems such as over-fitting [302]. Similarly, ML models may encounter fitting problems in the training process, especially in the case of small data sets or improper feature selection. Fortunately, ML and data assimilation methods provide new solutions to the problems in crop models and ML [303]. By combining RS data and crop model, and with the help of ML optimization, we can not only make up for the shortcomings of the model in some aspects, but also significantly improve the prediction accuracy and enhance the applicability. This innovative method is gradually being widely concerned and favored by researchers [304-306].

4. Discussion

4.1 Current challenges

Acquisition and processing of multi-source RS data: Although the current RS data sources are rich and diverse, including free and open-source medium and low-resolution MODIS data, and the access is relatively convenient, high-quality, high-resolution RS data is still scarce [307]. This is mainly due to the fact that RS data are easily affected by meteorological factors, equipment factors, terrain factors and other factors in the process of RS data collection, which makes it difficult to guarantee the quality of RS data [308]. At the same time, the acquisition of RS data is also faced with many challenges, such as data accessibility, timeliness, integrity, data privacy and imperfect RS industry chain, which greatly restrict the development of agricultural RS [309]. In addition, the processing process of RS data is a highly specialized technical task, and each link needs profound technical details and professional operators to carry out accurately. For example, if data preprocessing, multi-source data fusion, data interpretation and application are not handled properly, it may not only damage the accuracy and reliability of data, but also lead to unnecessary cost and time investment [310]. The establishment of RS database can not only provide rich and high-quality data resources, meet the urgent needs of ML for a large number of data, but also achieve data standardization, rapid retrieval and scientific research sharing. However, the existing agricultural RS database is still in the stage of development, facing some problems, such as low degree of data standardization, limited scale of data set, uneven data quality and so on. However, with the continuous progress and innovation of technology, these problems may be solved step by step. For example, the application of multi-source RS fusion data and higher resolution sensors [312,313].

Interpretability and generalization of the model: The interpretability and generalization ability of ML models are still two major challenges that restrict its development. Generally speaking, the interpretability of ML models is easier to explain than DL models, which is determined by the complexity of the model structure [314]. Although some newer models may lead to improvements in accuracy, the process of understanding and accepting the model will also face challenges for agricultural practitioners. For this reason, in the application of PA, we can choose models with intuitive decision-making process, such as decision trees or rule-based models, which are often easier to be understood and accepted by agricultural practitioners because of their relatively simple structure. it can not only promote the transparency of model decision-making, but also accelerate the transfer of knowledge from technical experts to first-line producers, which is very important to improve the science and efficiency of agricultural practice [315]. In addition, providing information about the features, variables and algorithms that affect the results of the model is also a way to enhance the interpretability of the model, such as sensitivity analysis of model parameters to obtain the calibration of key parameters, or parameter weights based on prior knowledge [316]. Under the action of many factors, such as crop growth environment, varieties, soil conditions and climatic conditions, the model shows insufficient extraction of fine features when facing new data [317]. Usually through data enhancement, model integration and the introduction of regularization technology to optimize the generalization ability of the model, it can effectively prevent the model from fitting and maintain stable performance in the face of new data [318].

4.2 prospects for the future

Trend of intelligence and automation: With the application of intelligence and automation technology in PA, many problems existing in traditional agricultural models, such as insufficient data sets, inaccurate analysis and untimely decision-making, have been solved. This is due to the use of high-precision multi-source RS data, further enhanced data preprocessing to improve data quality, the expansion of data sample diversity and the integration of expert knowledge, all of which improve the accuracy of intelligent decision-making [310,320]. In addition, the development of smart agricultural equipment and the training and advocacy of farmers have lowered the technical threshold and contributed to the widespread dissemination of these technologies [321]. The intelligent fusion of multi-source RS data effectively solves many agricultural data problems, such as ensuring the consistency of data formats, optimizing processing speed, improving the stability of algorithms, and enhancing the generalization ability and interpretability of the model. Thus reducing the uncertainty in the whole process of agricultural production [322]. RS and ML are expected to achieve cross-border integration with advanced technologies such as the Internet of things (IoT), human-computer interaction visualization, data assimilation and blockchain [323,324]. This will further promote the development of comprehensive monitoring, yield prediction and disease monitoring in PA, and provide more accurate, efficient and sustainable solutions for agricultural production.

Data sharing and multidisciplinary interaction: In the context of global connectivity, international cooperation and data sharing mechanisms are strengthening day by day, and PA applications integrating RS and ML should go beyond geographical restrictions [325]. Different countries and regions should work together to share more accurate RS data and smarter ML algorithms to address global agricultural issues such as climate change, food safety and other challenges [319,326]. In addition, ML algorithms are closely combined with geoscience RS technology, automatic machinery and intelligent robots are used to realize the intelligence and precision of field management, and experts in many fields such as agricultural economics, ecology and physical science are integrated at the same time to form a set of comprehensive agricultural management system [321,327,328]. They can not only improve crop yield and quality, but also effectively reduce the use of chemical fertilizers and pesticides, speed up the transformation of agricultural achievements, protect the ecological environment, and promote the green transformation of agriculture. It is worth noting that it is a very important step to ensure that the research results of PA can really benefit farmers and realize the transformation from theoretical knowledge to scientific and technological practice [328,329]. We can form a bottom-up driving force by establishing cooperative relations with agricultural research institutions and cooperating with agricultural enterprises, using their market channels and technical support capabilities to popularize new technologies, encourage farmers to adopt new technologies through policy guidance and support, and encourage farmers' groups to participate in the application of new technologies [330]. In addition, combining this science and technology with the United Nations Sustainable Development goals (SDGs) can not only enhance the scope of science, but also promote the development of practice from a broader perspective, for example, by improving the efficiency of agricultural production and implementing the sustainable development agenda to contribute to the goal of "zero hunger" [331,332]. Although it is still facing technical, economic and social challenges, with the continuous updating of the application of RS technology and ML in the field of agriculture, the potential to promote the modernization and sustainable development of agricultural industry will be further tapped.

5. Conclusions

The integrated application of RS technology and ML algorithm can indeed promote the development and progress of PA. In some agricultural fusion applications, the most widely used RS data source is hyperspectral data, and its application proportion is more than 30%. In addition, the rapid development of UAV remote sensing, accounting for about 24%, is expected to shine in PA in the future. The most widely used ML algorithm is support vector machine, accounting for more than 20%, followed by random forest algorithm, accounting for about 18%. It is worth noting that the rapid development trend of DL algorithms is expected to further promote the development of PA. Monitoring and identification of PA, pest detection, land / soil management and crop yield prediction are still the main aspects of the comprehensive application of RS combined with ML. The challenges of RS and ML algorithm fusion mainly include high-quality RS data acquisition and processing, poor model interpretation and generalization ability, uncertainty of integration development and so on. The future development will mainly focus on promoting agricultural intelligence and automation, strengthening international cooperation and sharing and sustainable transformation of achievements.

] Thank you for pointing this out. We agree with this comment. Therefore, We have revised [According to your request, we have revised and retouched the content and structure of the full text to make the statement and the whole text clearer.]

Response to Reviewer Comments

1. Summary

Thank you very much for taking the time to review this manuscript. We gave a detailed reply below and highlighted the content.

2. Point-by-point response to Comments and Suggestions for Authors

Comments 1: First sentence, "Under the background of the rapid change of global climate, the increasingly tense 32 geopolitical situation and the rising risk of local war, agricultural practice is facing un-33 precedented uncertainties and challenges.” needs a citation to support this.

Response 1: [In the context of rapid global climate change, agricultural practice is facing unprecedented uncertainties and challenges, such as climate warming, sea-level rise, drought and flood and other extreme hydroclimatic frequent occurrence [1-4].] Thank you for pointing this out. We agree with this comment. Therefore, We have revised the title. [We removed the description of "local war" to avoid unnecessary controversy and provided references to the uncertainties and challenges facing global climate change and agriculture. (10.1016/j.scitotenv.2024.174289,10.3389/fenvs.2023.1304845, 10.1016/j.jenvman.2024.121375, and10.1109/jstars.2024.3380514）]

Comments 2: Of the 44 citations in the Introduction, several are older than 10 years. I suggest limiting the Introduction to scholarship no older than 10 years to better defend the contemporary need for the study. The reader gets lost in such an exhaustive Introduction that could better reinforce the modern day need for the inquiry.

Response 2: [According to your opinion, we have updated the references, and the time is strictly limited to 10 years, in order to better defend the contemporary demand for this research.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised the title. [Ensure that all citations are from the past 10 years, and the updated literature can be found in the references section.]

Comments 3: The Discussion and Conclusions are not supported by any of the provided 336 citations. This section is the weaker portion of the manuscript. I suggest careful examination of what literature the authors have to support each line in both the Discussion and Conclusions. 

Response 3: [1. Current challenges

Multi-source RS data acquisition and processing: although the current RS data sources are rich and diverse, including free and open source medium and low resolution MODIS data, and the access is relatively convenient, high-quality, high-resolution RS data is still scarce. This is mainly due to the fact that RS data are easily affected by meteorological factors, equipment factors, terrain factors and other factors in the process of RS data collection, which makes it difficult to guarantee the quality of RS data [308]. At the same time, the acquisition of RS data is also faced with many challenges, such as data accessibility, timeliness, integrity, data privacy and imperfect RS industry chain, which greatly restrict the development of agricultural RS [309]. In addition, the processing process of RS data is a highly specialized technical task, and each link needs profound technical details and professional operators to carry out accurately. For example, if data preprocessing, multi-source data fusion, data interpretation and application are not handled properly, it may not only damage the accuracy and reliability of data, but also lead to unnecessary cost and time investment [310]. The establishment of RS database can not only provide rich and high-quality data resources, meet the urgent needs of ML for a large number of data, but also achieve data standardization, rapid retrieval and scientific research sharing. However, the existing agricultural RS database is still in the stage of development, facing some problems, such as low degree of data standardization, limited scale of data set, uneven data quality and so on. However, with the continuous progress and innovation of technology, these problems may be solved step by step. For example, the application of multi-source RS fusion data and higher resolution sensors [312,313]. The interpretability and generalization ability of ML models are still two major challenges that restrict its development. Generally speaking, the interpretability of ML models is easier to explain than deep learning models, which is determined by the complexity of the model structure [314]. Although some newer models may lead to improvements in accuracy, the process of understanding and accepting the model will also face challenges for agricultural practitioners. For this reason, in the application of precision agriculture, we can choose models with intuitive decision-making process, such as decision trees or rule-based models, which are often easier to be understood and accepted by agricultural practitioners because of their relatively simple structure. it can not only promote the transparency of model decision-making, but also accelerate the transfer of knowledge from technical experts to first-line producers, which is very important to improve the science and efficiency of agricultural practice [315]. In addition, providing information about the features, variables and algorithms that affect the results of the model is also a way to enhance the interpretability of the model, such as sensitivity analysis of model parameters to obtain the calibration of key parameters, or parameter weights based on prior knowledge [316]. Under the action of many factors, such as crop growth environment, varieties, soil conditions and climatic conditions, the model shows insufficient extraction of fine features when facing new data [317]. Usually through data enhancement, model integration and the introduction of regularization technology to optimize the generalization ability of the model, it can effectively prevent the model from fitting and maintain stable performance in the face of new data [318].

2. prospects for the future

With the application of intelligence and automation technology in precision agriculture, many problems existing in traditional agricultural models, such as insufficient data sets, inaccurate analysis and untimely decision-making, have been solved. This is due to the use of high-precision multi-source RS data, further enhanced data preprocessing to improve data quality, the expansion of data sample diversity and the integration of expert knowledge, all of which improve the accuracy of intelligent decision-making [310320]. In addition, the development of smart agricultural equipment and the training and advocacy of farmers have lowered the technical threshold and contributed to the widespread dissemination of these technologies [321]. The intelligent fusion of multi-source RS data effectively solves many agricultural data problems, such as ensuring the consistency of data formats, optimizing processing speed, improving the stability of algorithms, and enhancing the generalization ability and interpretability of the model. thus reducing the uncertainty in the whole process of agricultural production [322]. RS and ML are expected to achieve cross-border integration with advanced technologies such as the Internet of things (IoT), human-computer interaction visualization, data assimilation and blockchain [323324]. This will further promote the development of comprehensive monitoring, yield prediction and disease monitoring in precision agriculture, and provide more accurate, efficient and sustainable solutions for agricultural production. In the context of global connectivity, international cooperation and data sharing mechanisms are strengthening day by day, and precision agriculture applications integrating RS and ML should go beyond geographical restrictions [325]. Different countries and regions should work together to share more accurate RS data and smarter ML algorithms to address global agricultural issues such as climate change, food safety and other challenges [319326]. In addition, ML algorithms are closely combined with geoscience RS technology, automatic machinery and intelligent robots are used to realize the intelligence and precision of field management, and experts in many fields such as agricultural economics, ecology and physical science are integrated at the same time to form a set of comprehensive agricultural management system [321327328]. They can not only improve crop yield and quality, but also effectively reduce the use of chemical fertilizers and pesticides, speed up the transformation of agricultural achievements, protect the ecological environment, and promote the green transformation of agriculture.  It is worth noting that it is a very important step to ensure that the research results of precision agriculture can really benefit farmers and realize the transformation from theoretical knowledge to scientific and technological practice [328329]. We can form a bottom-up driving force by establishing cooperative relations with agricultural research institutions and cooperating with agricultural enterprises, using their market channels and technical support capabilities to popularize new technologies, encourage farmers to adopt new technologies through policy guidance and support, and encourage farmers' groups to participate in the application of new technologies [330]. In addition, combining this science and technology with the United Nations Sustainable Development goals (SDGs) can not only enhance the scope of science, but also promote the development of practice from a broader perspective, for example, by improving the efficiency of agricultural production and implementing the sustainable development agenda to contribute to the goal of "zero hunger" [331332]. Although it is still facing technical, economic and social challenges, with the continuous updating of the application of RS technology and ML in the field of agriculture, the potential to promote the modernization and sustainable development of agricultural industry will be further tapped.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [We recommend that we carefully review and update the literature that supports each line in the discussions and conclusions, and ensure that the literature is within the past 3 years.]

Comments 4: Finally, what is also missing in the Discussion and Conclusions is the extent and how the results of ML will be transferred (diffusion, knowledge transfer, etc.) to farmers from institutions, industry, or government agencies. This aspect of how the science presented here enhances "practice' is sorely vacant. A connectivity to one or more of the sustainable development goals would also elevate the science and its impact on practice in a broader since than solely ML researchers and practitioners.

Response 4: [ 1. Current challenges

Multi-source RS data acquisition and processing: although the current RS data sources are rich and diverse, including free and open source medium and low resolution MODIS data, and the access is relatively convenient, high-quality, high-resolution RS data is still scarce. This is mainly due to the fact that RS data are easily affected by meteorological factors, equipment factors, terrain factors and other factors in the process of RS data collection, which makes it difficult to guarantee the quality of RS data [308]. At the same time, the acquisition of RS data is also faced with many challenges, such as data accessibility, timeliness, integrity, data privacy and imperfect RS industry chain, which greatly restrict the development of agricultural RS [309]. In addition, the processing process of RS data is a highly specialized technical task, and each link needs profound technical details and professional operators to carry out accurately. For example, if data preprocessing, multi-source data fusion, data interpretation and application are not handled properly, it may not only damage the accuracy and reliability of data, but also lead to unnecessary cost and time investment [310]. The establishment of RS database can not only provide rich and high-quality data resources, meet the urgent needs of ML for a large number of data, but also achieve data standardization, rapid retrieval and scientific research sharing. However, the existing agricultural RS database is still in the stage of development, facing some problems, such as low degree of data standardization, limited scale of data set, uneven data quality and so on. However, with the continuous progress and innovation of technology, these problems may be solved step by step. For example, the application of multi-source RS fusion data and higher resolution sensors [312,313]. The interpretability and generalization ability of ML models are still two major challenges that restrict its development. Generally speaking, the interpretability of ML models is easier to explain than deep learning models, which is determined by the complexity of the model structure [314]. Although some newer models may lead to improvements in accuracy, the process of understanding and accepting the model will also face challenges for agricultural practitioners. For this reason, in the application of precision agriculture, we can choose models with intuitive decision-making process, such as decision trees or rule-based models, which are often easier to be understood and accepted by agricultural practitioners because of their relatively simple structure. it can not only promote the transparency of model decision-making, but also accelerate the transfer of knowledge from technical experts to first-line producers, which is very important to improve the science and efficiency of agricultural practice [315]. In addition, providing information about the features, variables and algorithms that affect the results of the model is also a way to enhance the interpretability of the model, such as sensitivity analysis of model parameters to obtain the calibration of key parameters, or parameter weights based on prior knowledge [316]. Under the action of many factors, such as crop growth environment, varieties, soil conditions and climatic conditions, the model shows insufficient extraction of fine features when facing new data [317]. Usually through data enhancement, model integration and the introduction of regularization technology to optimize the generalization ability of the model, it can effectively prevent the model from fitting and maintain stable performance in the face of new data [318].

2. prospects for the future

With the application of intelligence and automation technology in precision agriculture, many problems existing in traditional agricultural models, such as insufficient data sets, inaccurate analysis and untimely decision-making, have been solved. This is due to the use of high-precision multi-source RS data, further enhanced data preprocessing to improve data quality, the expansion of data sample diversity and the integration of expert knowledge, all of which improve the accuracy of intelligent decision-making [310320]. In addition, the development of smart agricultural equipment and the training and advocacy of farmers have lowered the technical threshold and contributed to the widespread dissemination of these technologies [321]. The intelligent fusion of multi-source RS data effectively solves many agricultural data problems, such as ensuring the consistency of data formats, optimizing processing speed, improving the stability of algorithms, and enhancing the generalization ability and interpretability of the model. thus reducing the uncertainty in the whole process of agricultural production [322]. RS and ML are expected to achieve cross-border integration with advanced technologies such as the Internet of things (IoT), human-computer interaction visualization, data assimilation and blockchain [323324]. This will further promote the development of comprehensive monitoring, yield prediction and disease monitoring in precision agriculture, and provide more accurate, efficient and sustainable solutions for agricultural production. In the context of global connectivity, international cooperation and data sharing mechanisms are strengthening day by day, and precision agriculture applications integrating RS and ML should go beyond geographical restrictions [325]. Different countries and regions should work together to share more accurate RS data and smarter ML algorithms to address global agricultural issues such as climate change, food safety and other challenges [319326]. In addition, ML algorithms are closely combined with geoscience RS technology, automatic machinery and intelligent robots are used to realize the intelligence and precision of field management, and experts in many fields such as agricultural economics, ecology and physical science are integrated at the same time to form a set of comprehensive agricultural management system [321327328]. They can not only improve crop yield and quality, but also effectively reduce the use of chemical fertilizers and pesticides, speed up the transformation of agricultural achievements, protect the ecological environment, and promote the green transformation of agriculture.  It is worth noting that it is a very important step to ensure that the research results of precision agriculture can really benefit farmers and realize the transformation from theoretical knowledge to scientific and technological practice [328329]. We can form a bottom-up driving force by establishing cooperative relations with agricultural research institutions and cooperating with agricultural enterprises, using their market channels and technical support capabilities to popularize new technologies, encourage farmers to adopt new technologies through policy guidance and support, and encourage farmers' groups to participate in the application of new technologies [330]. In addition, combining this science and technology with the United Nations Sustainable Development goals (SDGs) can not only enhance the scope of science, but also promote the development of practice from a broader perspective, for example, by improving the efficiency of agricultural production and implementing the sustainable development agenda to contribute to the goal of "zero hunger" [331332]. Although it is still facing technical, economic and social challenges, with the continuous updating of the application of RS technology and ML in the field of agriculture, the potential to promote the modernization and sustainable development of agricultural industry will be further tapped.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [As you have mentioned, we have included in our discussions and conclusions details of how the results of machine learning will be transferred (diffusion, knowledge transfer, etc.) to provide services to farmers from institutions, industries, or government agencies. It also strengthens the influence of how to strengthen the connection between this aspect of "practice" and sustainable development goals on practice.]

Comments 5: I greatly appreciate the over 330 citations to better connect the scholarship in the Web of Science, if published. However, I go back to the quality of this extensive list adding scientific and practical value to the manuscript. I do strongly believe, besides the methodology, citations should be limited to only 10 years old. 

Response 5: [We are very grateful for your review, as you mentioned to add scientific and practical value to the manuscript, we have updated the citation should be limited to 10 years.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.

Response to Reviewer Comments

1. Summary

Thank you very much for taking the time to review this manuscript. We gave a detailed reply below and highlighted the content.

2. Point-by-point response to Comments and Suggestions for Authors

Comments 1: 1 – The definition of the keywords of the title are not well defined in the manuscript

Response 1: [Precision agriculture is an agricultural practice way to realize accurate management and decision support of agricultural production process by means of modern information technology. Remote sensing is a technology that can obtain the information of the earth's surface without physical contact. It uses sensors to capture and record electromagnetic radiation signals reflected, emitted or scattered from the earth's surface from a long distance, and then continuously identify, measure and analyze the characteristics of the target objects located on, above or even below the earth's surface by analyzing these signals. Machine learning is a data analysis method that allows computer systems to automatically learn patterns and rules from data without explicit programming. Researchers tend to use machine learning as an integrated framework for feature collection and classification, prediction, or decision support.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised the title. [We have redefined the title keywords in the manuscript.]

Comments 2: 2 – In the current format, the authors associated Remote Sensing with Orbital Remote Sensing, which is not correct. Please explain the meaning of remote sensing and their different levels.

Response 2: [Remote sensing is a technology that can obtain the information of the earth's surface without physical contact. It uses sensors to capture and record electromagnetic radiation signals reflected, emitted or scattered by the surface from a long distance, and then analyze these signals to obtain information about the characteristics of the surface. Orbital Remote Sensing (Orbital remote Sensing): refers to the collection of data in Earth orbit by satellite. This is the most common remote sensing method, which can provide global coverage and is suitable for large-scale monitoring tasks. Aerial Remote Sensing (aerial remote sensing): using aircraft, drones and other aircraft to collect data at lower altitudes. Terrestrial Remote Sensing (ground remote sensing): data collection through ground sensors or vehicle vehicles.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised the title.

Comments 3: 3 – Can the authors please explain the difference between data and information? Based on that, please carefully check for the correct terminology use through the manuscript.

Response 3: [Data refers to the original facts, numbers, symbols, measurements or observations, which are usually unprocessed or preliminarily processed. Information is the meaningful content extracted from the data, which reveals the meaning or relationship behind the data through data analysis and interpretation. The relationship between data and information is as follows:  Data is the basis of information, and information is data that has been processed and interpreted. Data can be structured (such as numbers in a table) or unstructured (such as text, images). Information is a conclusion or insight drawn through data analysis, which can be quantitative or qualitative.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [We re-examined the correct use of data and information terms in the manuscript]

Comments 4: 4 - The authors have mentioned “ML technology”, “ML”, “advanced algorithms”, “deep learning”and “intelligent ML”. Can you please describe each of their meaning and their differences? After that, please make proper changes throughout the manuscript.

Response 4: [ML technology refers to the collection of technologies that use machine learning principles, methods and tools to solve practical problems. It covers the whole process from data collection, preprocessing, modeling to evaluation and improvement. Compared with "ML", it focuses more on the implementation and application of technology, including algorithms, software tools, hardware support and so on. ML is the abbreviation of machine learning. ML is a method of data analysis, which allows computer systems to automatically learn patterns and rules from data without explicit programming. Researchers tend to use machine learning as an integrated framework for feature collection and classification, prediction, or decision support. Compared with "ML technology", it focuses more on methodology and principle, and is the basis of the implementation of "ML technology". Advanced algorithms refer to algorithms that perform well, efficiently and accurately in solving complex problems. In the field of machine learning, this includes many complex optimization algorithms, classification algorithms, clustering algorithms and so on. Compared with "ML", it focuses more on the complexity and performance of the algorithm itself; compared with "deep learning", it does not necessarily refer to neural network algorithms, but a wider range of algorithms. Deep learning is a sub-field of machine learning, which uses deep neural network model to learn the representation and characteristics of data. This model captures the complex structures and relationships in the data through multi-layer nonlinear processing units. Intelligent ML refers to machine learning systems with a higher level of intelligence, which can be learned and improved adaptively and have stronger generalization ability, explanation and interaction. Compared with "ML", it refers to a specific machine learning technology; compared with "advanced algorithm", it focuses more on the construction and application of neural network model.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.

Comments 5: 5 – To be considered as a review paper, it is necessary to provide a deeper analysis about the references used. In the current format, it is provided only partial results of the references without a critical evaluation of their study. Due to that, it is not possible to have a discussion nor current challenges nor prospects for the future nor conclusions that can actually bring new insights based on the state of the art.

Response 5: [1. Current challenges

Multi-source RS data acquisition and processing: although the current RS data sources are rich and diverse, including free and open source medium and low resolution MODIS data, and the access is relatively convenient, high-quality, high-resolution RS data is still scarce. This is mainly due to the fact that RS data are easily affected by meteorological factors, equipment factors, terrain factors and other factors in the process of RS data collection, which makes it difficult to guarantee the quality of RS data [308]. At the same time, the acquisition of RS data is also faced with many challenges, such as data accessibility, timeliness, integrity, data privacy and imperfect RS industry chain, which greatly restrict the development of agricultural RS [309]. In addition, the processing process of RS data is a highly specialized technical task, and each link needs profound technical details and professional operators to carry out accurately. For example, if data preprocessing, multi-source data fusion, data interpretation and application are not handled properly, it may not only damage the accuracy and reliability of data, but also lead to unnecessary cost and time investment [310]. The establishment of RS database can not only provide rich and high-quality data resources, meet the urgent needs of ML for a large number of data, but also achieve data standardization, rapid retrieval and scientific research sharing. However, the existing agricultural RS database is still in the stage of development, facing some problems, such as low degree of data standardization, limited scale of data set, uneven data quality and so on. However, with the continuous progress and innovation of technology, these problems may be solved step by step. For example, the application of multi-source RS fusion data and higher resolution sensors [312,313]. The interpretability and generalization ability of ML models are still two major challenges that restrict its development. Generally speaking, the interpretability of ML models is easier to explain than deep learning models, which is determined by the complexity of the model structure [314]. Although some newer models may lead to improvements in accuracy, the process of understanding and accepting the model will also face challenges for agricultural practitioners. For this reason, in the application of precision agriculture, we can choose models with intuitive decision-making process, such as decision trees or rule-based models, which are often easier to be understood and accepted by agricultural practitioners because of their relatively simple structure. it can not only promote the transparency of model decision-making, but also accelerate the transfer of knowledge from technical experts to first-line producers, which is very important to improve the science and efficiency of agricultural practice [315]. In addition, providing information about the features, variables and algorithms that affect the results of the model is also a way to enhance the interpretability of the model, such as sensitivity analysis of model parameters to obtain the calibration of key parameters, or parameter weights based on prior knowledge [316]. Under the action of many factors, such as crop growth environment, varieties, soil conditions and climatic conditions, the model shows insufficient extraction of fine features when facing new data [317]. Usually through data enhancement, model integration and the introduction of regularization technology to optimize the generalization ability of the model, it can effectively prevent the model from fitting and maintain stable performance in the face of new data [318].

2. prospects for the future

With the application of intelligence and automation technology in precision agriculture, many problems existing in traditional agricultural models, such as insufficient data sets, inaccurate analysis and untimely decision-making, have been solved. This is due to the use of high-precision multi-source RS data, further enhanced data preprocessing to improve data quality, the expansion of data sample diversity and the integration of expert knowledge, all of which improve the accuracy of intelligent decision-making [310320]. In addition, the development of smart agricultural equipment and the training and advocacy of farmers have lowered the technical threshold and contributed to the widespread dissemination of these technologies [321]. The intelligent fusion of multi-source RS data effectively solves many agricultural data problems, such as ensuring the consistency of data formats, optimizing processing speed, improving the stability of algorithms, and enhancing the generalization ability and interpretability of the model. thus reducing the uncertainty in the whole process of agricultural production [322]. RS and ML are expected to achieve cross-border integration with advanced technologies such as the Internet of things (IoT), human-computer interaction visualization, data assimilation and blockchain [323324]. This will further promote the development of comprehensive monitoring, yield prediction and disease monitoring in precision agriculture, and provide more accurate, efficient and sustainable solutions for agricultural production. In the context of global connectivity, international cooperation and data sharing mechanisms are strengthening day by day, and precision agriculture applications integrating RS and ML should go beyond geographical restrictions [325]. Different countries and regions should work together to share more accurate RS data and smarter ML algorithms to address global agricultural issues such as climate change, food safety and other challenges [319326]. In addition, ML algorithms are closely combined with geoscience RS technology, automatic machinery and intelligent robots are used to realize the intelligence and precision of field management, and experts in many fields such as agricultural economics, ecology and physical science are integrated at the same time to form a set of comprehensive agricultural management system [321327328]. They can not only improve crop yield and quality, but also effectively reduce the use of chemical fertilizers and pesticides, speed up the transformation of agricultural achievements, protect the ecological environment, and promote the green transformation of agriculture.  It is worth noting that it is a very important step to ensure that the research results of precision agriculture can really benefit farmers and realize the transformation from theoretical knowledge to scientific and technological practice [328329]. We can form a bottom-up driving force by establishing cooperative relations with agricultural research institutions and cooperating with agricultural enterprises, using their market channels and technical support capabilities to popularize new technologies, encourage farmers to adopt new technologies through policy guidance and support, and encourage farmers' groups to participate in the application of new technologies [330]. In addition, combining this science and technology with the United Nations Sustainable Development goals (SDGs) can not only enhance the scope of science, but also promote the development of practice from a broader perspective, for example, by improving the efficiency of agricultural production and implementing the sustainable development agenda to contribute to the goal of "zero hunger" [331332]. Although it is still facing technical, economic and social challenges, with the continuous updating of the application of RS technology and ML in the field of agriculture, the potential to promote the modernization and sustainable development of agricultural industry will be further tapped.]. Thank you for pointing this out. We agree with this comment. Therefore, We have revised.

Comments 6: 6 – Overall, the manuscript structure must be improved, there are several paragraphs and sentences that do not present cohesion and key words are missing their definition or were misused. Please carefully check for the MDPI guidelines, typos and redundant word and expressions throughout the manuscript.

Response 6: [We have improved the structure of the entire manuscript to ensure the coherence between paragraphs and sentences, and to check the definition of keywords and the correctness of word expressions. Pay special attention to the cohesion between sentences and the coherence of paragraphs, while checking the use of keywords to ensure compliance with MDPI Press's writing and formatting guidelines.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.

Comments 7: Keywords: please use keywords different from the title and relevant to the manuscript

Response 7: [We have updated the keywords to make them different from the title and have practical significance. Agricultural monitoring; pest detection, land management; yield prediction; Integrated application; Agricultural sustainability] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.

Comments 8: Lines 13-15 – The authors made a partial definition of precision agriculture (PA) “which combines remote sensing technology and machine learning algorithm” and have concluded stating that “provides a new possibility for the sustainable development of precision agriculture”. However, I raise the question about PA definition. Are the authors aware of the definition? Please write the definition of PA with proper references. I would encourage the authors to check the definition given by the International Society of Precision Agriculture.

Response 8: [With the global population growth, resource shortage and climate change, the contradiction between human and land has become increasingly prominent, these factors constitute a major challenge to the traditional agricultural model. There is an urgent need for a new mode of agricultural production that can achieve efficient, environmental protection and sustainable development. Precision agriculture is an agricultural practice way to realize accurate management and decision support of agricultural production process by using modern information technology.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [We give a clear definition of precision agriculture and ensure the coherence of the logic within the paragraph by adjusting the sentence order and adding transition sentences.]

Comments 9: Please carefully check for the guidelines about the correct use of abbreviations and acronyms through the manuscript including Tables and Figures. Check for typos.

Response 9: [We carefully examine the manuscripts (including tables and graphics) for guidelines on the correct use of acronyms and acronyms. Correction of Remote Sensing (RS), Machine Learning (ML) and Precision Agriculture (PA), etc.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.

Comments 10: The structure must be improved. In the current format, there are sentences without cohesion within a paragraph (e.g., Lines, 44-47, Lines 95-98, Lines 98-108, and many others)

Response 10: [For example, Yomo et al found in 2023 that the maximum likelihood algorithm based on Landsat-8 images is used to classify land use and land cover, and by using multi-layer perceptron-Markov chain modeling method, the results show that the overall accuracy (Kappa coefficient) is as high as 92% [9]. This study shows that the accuracy of agricultural monitoring and recognition can be significantly improved by integrating advanced RS technology and ML algorithm. For example, European Sentinel-2, ENVISAT MERIS, French SPOT satellite and NOAA AVHRR satellite data, India's Hyperion, China's GF series data and HJ remote sensing data have been widely used [33-35]. It is worth mentioning that the emergence of unmanned aerial vehicle (UAV) marks a new era of RS. UAV is a kind of unmanned small aircraft, which is often used to carry RS equipment for aerial data acquisition. It can provide more abundant and comprehensive spectral, spatial and temporal resolution data, vegetation height data and multi-angle observation, and has the characteristics of high efficiency, convenience, low cost and strong adaptability [36]. There have been many successful cases in crop classification, weed detection and vegetation monitoring, which prove the feasibility of UAV in precision agriculture [37]. For example, Marques et al have overcome the limitation of limited spectral coverage based on UAV in 2024, especially in low light, fog or smoke conditions to achieve real-time, efficient and distributed accurate monitoring [38]. Bah et al used drone images to detect weeds in the field in 2017 with an accuracy of more than 90% [39]. Yang et al used UAV image information to identify rice lodging based on decision tree algorithm in 2017, with an overall accuracy of 96.17% [40].] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [We re-adjust the structure of the article and the logic of the paragraph, and ensure the coherence of the article by adjusting the sentence order and adding transitional sentences.]

Comments 11: Lines 113-115 – I have the concern that ML were misinterpreted in those lines. Can the authors define ML?

Response 11: [ML is a data analysis method that allows computer systems to automatically learn patterns and rules from data without explicit programming. Researchers tend to use ML as an integrated framework for feature collection and classification, prediction or decision support.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.

Comments 12: Figures present lack of resolution and their caption are not according to the guidelines.

Response 12: [We have revised the resolution of the chart again, and its title has been revised in accordance with the MDPI guidelines.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.

Comments 13: Remote sensing: Please explain the meaning of remote sensing and their different levels. The authors have associated remote sensing to orbital remote sensing and they can´t be used as synonyms.

[Remote sensing is a technology that can obtain the information of the earth's surface without physical contact. It uses sensors to capture and record electromagnetic radiation signals reflected, emitted or scattered by the surface from a long distance, and then analyze these signals to obtain information about the characteristics of the surface. Orbital Remote Sensing (Orbital remote Sensing): refers to the collection of data in Earth orbit by satellite. This is the most common remote sensing method, which can provide global coverage and is suitable for large-scale monitoring tasks. Aerial Remote Sensing (aerial remote sensing): using aircraft, drones and other aircraft to collect data at lower altitudes. Terrestrial Remote Sensing (ground remote sensing): data collection through ground sensors or vehicle vehicles.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.

Comments 14: Lines 181-192 – The authors wrote about some studies conducted with remote sensing data, however, more detail about the study are needed, for example, the crop, location, type of sensor,and others.

Response 14: [For example, in the health assessment of urban trees using WorldView-3 satellite RS data, although the overall growth dynamics of trees can be monitored, there are some limitations in capturing the specific details of the dynamics of growth time in different parts of trees [62]. However, in another study, the use of Landsa-8 RS images with a spatial resolution of 10 to 30 meters provided a promising solution for disease detection in mixed forests in southern China [63]. In other studies for the detection of plant diseases and pests infecting vegetation, the detection accuracy does not seem to be satisfactory based on visible light (780nm) data [64]. In a 2023 study by Zhu et al., although the use of drone RS can confirm the importance of red light bands and adjacent bands, it has not achieved the desired results in the investigation of plant diseases and pests invading vegetation [65].] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.

Comments 15: Line 252-253 – I would suggest to avoid using “local sayings” since not all the readers can understand it, unless you add the proper reference.

Response 15: [In recent years, a series of emerging algorithms and technologies have emerged, such as deep learning (DL), intelligent optimization, neural networks, computer vision and data enhancement. These technologies not only inject strong impetus into the field of ML, but also provide rich opportunities for all stages of agriculture. They enable agricultural practitioners to respond more effectively to challenges and achieve set goals [].] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.[ We have replaced “some emerging algorithms and technologies, such as deep learning (DL), intelligent optimization, neural networks, computer vision and data enhancement, springing up like bamboo shoots after a spring rain.” The expression of local sayings]

Comments 16: The authors use references from the animal domain, specifically the wild animals, which might not be the target of manuscript. Please carefully check for the use of relevant references for the manuscript content.

Response 16: [We delete the relevant expressions and references "Animal and Plant count (Kellenberger et al., 2018), and some scholars skillfully combine CNN with recurrent neural network (RNN) to achieve accurate monitoring and counting of wild animals. After color space processing technology, mathematical morphology and other processing, the final accuracy is more than 90% (Barbedo et al., 2020). "] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.[ We carefully examined the use of references related to the content of the manuscript, and updated the references in the field of animals, especially wild animals.]

Comments 17: Lines 352-357 – The authors affirmed that it is possible to determine the degree of damage and type of disease from satellite sensors. Could the authors give more details about that? In the state of art, I don´t think that satellite data can actually determine the previous mentioned factors accurately.

Response 17: [Mahanta et al. acquired rich spectral features of vegetation based on a variety of sensor devices, and used ML models to identify spectral feature patterns related to specific diseases, and finally realized the assessment of the health degree of insect invasion of forests, which greatly improved the detection efficiency.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.[ As you mentioned, it is still technically challenging to accurately identify the specific degree of damage or type of disease directly from satellite data, and it is necessary to combine other multispectral and hyperspectral RS and ML models.]

Comments 18: Lines 363 – 428 – What are the main differences among the categories? In the current format, it is difficult to understand and identify the reason of having these five categories.

[Crop diseases and insect pests are not only one of the core factors affecting plant yield and quality, but also one of the main causes of crop damage. As emphasized by the United Nations Food and Agriculture Organization, the damage caused by diseases and insect pests to agro-ecosystems should not be underestimated. Unfortunately, in early agricultural production, the problem of diseases and insect pests is often marginalized and not given enough attention, resulting in huge economic losses [148]. Therefore, the implementation of efficient and accurate detection of plant diseases and insect pests not only plays a vital role in ensuring the health of agro-ecosystem, improving crop yield and quality, and reducing economic losses, but also plays a key role in the development of precision agriculture. in recent years, it has attracted the attention of many scholars [149]. The premise to achieve this goal is to accurately detect and classify diseases and insect pests, and to distinguish different types and estimated quantities, in order to implement accurate pest prevention and control strategies. Traditional pest monitoring mainly depends on the manual identification of insect experts or technicians, this method is not only subjective and labor-intensive, but also not practical in large-scale applications [150]. However, with the popularity of sensors and embedded devices on the Internet, the combination of RS and ML has opened up a new way for the detection of diseases and insect pests in modern agriculture. As shown in figure 7, Mahanta and others have realized the assessment of the health degree of insects invading forests based on a variety of sensor devices, which has greatly improved the detection efficiency. Most diseases and insect pests have the characteristics of concealment, latent, infectivity and uncertainty, which undoubtedly increase the difficulty and cost of control and bring great challenges to agricultural production [153154]. However, it is worth noting that RS data from satellite sensors show that plants affected by the disease can be distinguished in a relatively short period of time by their spectral characteristics different from healthy plants [155]. The picture shows the process of pest monitoring based on different types of sensors. More importantly, through further use of ML for analysis, we can not only determine the degree of damage, but also accurately identify the type of disease [156157]. In the early detection of diseases and insect pests, many researchers tend to use traditional methods, that is, to establish empirical statistical models between diseases and insect pests and related factors (such as environment, climate, soil and vegetation index). In order to achieve effective monitoring of diseases and insect pests. These methods include multiple linear regression, partial least square regression, support vector regression and random forest regression. For example, Ebrahimi et al use support vector regression to detect parasites in crop canopy images, which greatly improves the detection accuracy. In addition, through the comprehensive use of advanced image processing techniques such as image segmentation [161], feature extraction [162], target detection [163-166] and classification, researchers can solve complex problems in plant disease detection more accurately and efficiently [167]. Image segmentation technology is used to distinguish normal and abnormal leaves in RS images, while feature extraction is to extract meaningful information from segmented regions, such as color, texture, shape, etc. [168169]. They provide a more detailed and accurate analysis method for disease detection [170171]. For example, studies by Zhang et al have shown that TinySegformer model can provide a robust and practical solution for large-scale agricultural pest detection because of its high efficiency, accuracy and lightweight. The interactive segmentation method based on GrabCut proposed by Lu et al., which is applied to field RS, can quickly extract locust images from various parts. Barbedo et al. proposed an automatic detection algorithm for wheat scab based on hyperspectral technology. the algorithm has more than 91% classification accuracy and shows excellent robustness under a variety of complex factors such as shape, direction and shadow. Mumtaz et al. combined with optical RS, image processing and depth learning methods to accurately detect and grade wheat rust. Bao et al. proposed a RS method of UAV based on DDMA-YOLO, which can not only reduce the workload and time consumption of pest detection, but also effectively improve the detection efficiency. Many studies have shown that the image fusion method can greatly enhance the accuracy of vegetation disease detection by fusing image information from different sensors or multi-stage processing. For example, some scholars apply deep learning technology to the fusion of RGB images and segmented images, and develop a multi-head DenseNet architecture. After the strict verification of the public data set and the application of 50% discount cross-validation technology, the method shows excellent performance, and all the evaluation indicators have reached a very high level. for example, the average accuracy, recall, accuracy and F1 score reached 98.17%, 98.17%, 98.16% and 98.12% respectively [182]. Based on the multi-source fusion images of UAV and visible light, Ma et al successfully constructed a variety of ML models, which significantly improved the accuracy of cotton Vickers wilt detection. It is worth mentioning that some researchers have adopted the improved deep learning algorithm framework for plant disease detection, and achieved remarkable results [184-187]. Dong et al. creatively proposed an effective scale-aware network architecture (ESA-Net) based on low-cost RS images. After strict verification, ESA-Net showed excellent performance in plant disease detection, and achieved strong competitive results. Amarathunga et al proposed a new architecture based on visual converter (ViT), which integrates attention mechanisms driven by domain knowledge and effectively improves the accuracy of detection and identification of microscopic diseases and insect pests (such as thrips) at the species level. Ye et al designed an end-to-end automatic disease detection framework based on multi-scale MA-UNet model and single-phase image based on UAV aerial photography data and Landsat 8 satellite RS markers, which greatly improved the efficiency and accuracy of disease monitoring.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.[ We reshaped the structure and presentation of this part.]

Comments 19: Lines 430-434 – It looks like that it is missing a reference.

Response 19: [As the cornerstone of human survival and development, land not only carries the key mission of agricultural production, provides us with food to maintain our livelihood, but also is an indispensable key prerequisite for ensuring human well-being [60191]. Therefore, the management and analysis of soil and land resources is particularly urgent and important. In the field of soil monitoring and management, traditionally, we rely on field survey methods to obtain the spatial distribution data of soil groups [191].] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [We added a reference to lines 430-434, 10.1016/j.compag.2022.106844 and 10.1016/j.catena.2019.104259]

Comments 20: Lines 594 – 595 – The authors stated that crop models are prone to over-fitting, but what about“ML”?

Response 20: [However, crop models are not perfect. They may be limited in large-scale applications, errors are easy to accumulate, and there are some problems such as over-fitting. Similarly, ML models may encounter fitting problems in the training process, especially in the case of small data sets or improper feature selection. Fortunately, ML and data assimilation provide new solutions to the problems in crop models and ML.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.[When describing the limitations of the crop model, it is clearly pointed out that the ML model may also encounter the problem of fitting. This paper emphasizes the way to solve these problems by combining RS data and crop model, and with the help of ML optimization.]

Comments 21: Lines 635-638 – meaningless sentence and redundant word in the same sentence;

Response 20: [The interpretability and generalization ability of ML models are still two major challenges that restrict its development. Generally speaking, the interpretability of ML is easier to explain than DL, which is determined by the complexity of the model structure [314]. Although some newer models may lead to improvements in accuracy, the process of understanding and accepting the model will also face challenges for agricultural practitioners. For this reason, in the application of precision agriculture, we can choose models with intuitive decision-making process, such as decision trees or rule-based models, which are often easier to be understood and accepted by agricultural practitioners because of their relatively simple structure. it can not only promote the transparency of model decision-making, but also accelerate the transfer of knowledge from technical experts to first-line producers, which is very important to improve the science and efficiency of agricultural practice [315]. In addition, providing information about the features, variables and algorithms that affect the results of the model is also a way to enhance the interpretability of the model, such as sensitivity analysis of model parameters to obtain the calibration of key parameters, or parameter weights based on prior knowledge [316]. Under the action of many factors, such as crop growth environment, varieties, soil conditions and climatic conditions, the model shows insufficient extraction of fine features when facing new data [317]. Usually through data enhancement, model integration and the introduction of regularization technology to optimize the generalization ability of the model, it can effectively prevent the model from fitting and maintain stable performance in the face of new data [318].] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.[ We modified the meaningless sentences in lines 635-638 and the extra words in the same sentence, and reorganized the structure and expression of this paragraph.]

Comments 22: Lines 639-642 – I have the concern about the meaning of ML and DP conceived by the authors.

Response 20: [The interpretability and generalization ability of ML models are still two major challenges that restrict its development. Generally speaking, the interpretability of ML is easier to explain than DL, which is determined by the complexity of the model structure [314]. Although some newer models may lead to improvements in accuracy, the process of understanding and accepting the model will also face challenges for agricultural practitioners. For this reason, in the application of precision agriculture, we can choose models with intuitive decision-making process, such as decision trees or rule-based models, which are often easier to be understood and accepted by agricultural practitioners because of their relatively simple structure. it can not only promote the transparency of model decision-making, but also accelerate the transfer of knowledge from technical experts to first-line producers, which is very important to improve the science and efficiency of agricultural practice [315]. In addition, providing information about the features, variables and algorithms that affect the results of the model is also a way to enhance the interpretability of the model, such as sensitivity analysis of model parameters to obtain the calibration of key parameters, or parameter weights based on prior knowledge [316]. Under the action of many factors, such as crop growth environment, varieties, soil conditions and climatic conditions, the model shows insufficient extraction of fine features when facing new data [317]. Usually through data enhancement, model integration and the introduction of regularization technology to optimize the generalization ability of the model, it can effectively prevent the model from fitting and maintain stable performance in the face of new data [318].] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.[ We modified the meaning of ML and DP in line 639-642, reorganized the structure of the paragraph, and introduced references.]

Comments 23: Lines 526-531 – How can the authors make a conclusion in the results section before the discussion section?

Response 23: [We have learned that there are differences in the impact of different ML algorithms and RS data from different sources on the results in the operational assessment of crop status and yield. Multi-spectral and medium-resolution RS technology, represented by MODIS data, is widely used in early crop yield prediction, and shows potential uses.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.[ We revised this part of the narrative.]

Response to Reviewer Comments

1. Summary

Response to Reviewer Comments

1. Summary

Comments 1: 1 - The authors describe concepts and examples of the application of Precision Farming in the methodology. The methodology should not be done like this; this section of the article should include how the authors proceeded to obtain the results described. This problem still persists in relation to the first review.

Response 1: [The methodology section you mentioned does need to describe more clearly how we get the results of our research. We have revised this part to better reflect our research process. The following is the revised part:

Abstract: With global population growth, resource shortages and climate change, traditional agricultural models face major challenges. Precision agriculture (PA), as a way to realize accurate management and decision support of agricultural production process by using modern information technology, is becoming one of the effective ways to solve these challenges. In particular, the combination of remote sensing technology and machine learning algorithm brings new possibilities for precision agriculture. However, there are relatively few comprehensive and systematic reviews on the integrated application of these two technologies. For this reason, this study makes a systematic literature search on the databases of Web of Science, Scopus, Google Scholar and PubMed, and analyzes the integrated application of remote sensing technology and machine learning algorithm in precision agriculture in recent 10 years. The study found that: (1) because of their characteristics, different types of remote sensing data have significant differences in meeting the needs of precision agriculture, in which hyperspectral remote sensing is the most widely used, accounting for more than 30%. The application of UAV remote sensing has the most potential, accounting for about 24%, and showing an upward trend. (2) Machine learning algorithm has obvious advantages in promoting the development of precision agriculture, in which support vector machine algorithm is the most widely used, accounting for more than 20%, followed by random forest algorithm, accounting for about 18%. In addition, this study also discusses the main challenges faced at present, such as the difficult problems of acquisition and processing of high-quality remote sensing data, model interpretation and generalization ability, and looks forward to the future development trend. such as promoting agricultural intelligence and automation, strengthening international cooperation and sharing and sustainable transformation of achievements. To sum up, this study can provide new ideas and reference for remote sensing combined with machine learning in promoting the development of precision agriculture.

1. Introduction

In the context of rapid global climate change, agricultural practice is facing unprecedented uncertainties and challenges, such as climate warming, sea-level rise, drought and flood and other extreme hydroclimatic frequent occurrence [1-4]. At the same time, the global population is expected to reach 8.7 billion by 2030 and climb to 9.7 billion by 2050, which undoubtedly puts tremendous pressure on global food production [5]. However, it is gratifying that in recent years, with the increase of investment in science and technology and agricultural research, the development of PA has achieved certain results [6,7]. This progress not only changes the traditional mode of agricultural production, but also aims to optimize agricultural inputs (seeds, water resources, chemicals) through the application of advanced technologies (such as remote sensing, machine learning algorithms, agricultural robots, etc.), effectively manage crop variability, maintain or even increase yields, and cleverly avoid potential losses, thereby improving the efficiency and profitability of agricultural systems [8]. For example, Yomo et al found in 2023 that the maximum likelihood algorithm based on Landsat-8 remote sensing images is used to classify land use and land cover, and by using multi-layer perceptron-Markov chain modeling method, the results show that the overall accuracy (Kappa coefficient) is as high as 92% [9]. This study shows that the accuracy of agricultural monitoring and recognition can be significantly improved by integrating advanced remote sensing technology and machine learning algorithm. In addition, another study shows that the integrated learning random forest classifier is used to study the progressive lodging sensitive characteristics of rice types based on multi-spectral (444-842 nm) fusion Unmanned Aerial Vehicle technology, with an overall accuracy of 96.1% [10]. These examples prove the effective application of advanced technologies and algorithms in PA. Although PA has many advantages, limitations such as information accuracy, large amount of data, operational complexity and high initial cost cannot be ignored [11]. Therefore, for most countries in the world, it is very necessary to actively promote the coordinated development of PA, remote sensing (RS) and machine learning (ML), ensure agricultural production safety, and strictly abide by the food safety red line.

In recent years, people have made multi-dimensional and deep exploration and efforts in the development of PA, which is mainly reflected in the research and application of new technologies, covering many key links such as personnel training, policy support and so on [12,13]. The aim is to overcome the shortcomings of traditional agriculture, such as time-consuming and laborious, improper use of resources, unstable crop yield and environmental pollution [14,15]. In this context, it is actually a complex and critical challenge to monitor crop growth and conditions in different locations and environments in real time, accurately and at multiple scales, and to use data with different time resolutions to meet a variety of purposes. In fact, it is a complex and critical challenge to respond quickly to extreme events according to changing climate conditions [16,17]. Fortunately, RS technology has developed rapidly in agriculture, forestry, hydrology, environmental protection and other fields because of its unique advantages (such as synchronization, timeliness, spatio-temporal continuity and large-scale observation ability) [18-22]. RS is a technology that can obtain the information of the earth's surface without physical contact. It uses sensors to capture and record electromagnetic radiation signals reflected, emitted or scattered from the earth's surface from a long distance, and then continuously identify, measure and analyze the characteristics of target objects located on, above or even below the earth's surface by analyzing these signals [23]. This not only greatly improves the efficiency of agricultural information acquisition, but also provides strong support for dealing with agriculture in extreme weather, so that crop managers can take timely measures to reduce the impact of disasters and ensure the safety and stability of agricultural production [24,25]. In addition, with the maturity of RS inversion algorithms (such as linear regression, PROSAIL physical model, neural network), inversion data sets based on RS images have also appeared, such as inversion product data sets based on MODIS images, Landsat images and Sentinel-2 images: water quality and water environment elements inversion products, vegetation parameter inversion products, land surface temperature inversion products and soil parameter inversion products. It provides reliable and rich data sources for agricultural RS related research [26-28].

It is well known that most agricultural RS data are information provided by visible light and near infrared radiation reflected (or transmitted) by plants, which are measured according to wavelength, that is, spectral reflectance [29]. According to the change of vegetation, the spectral data commonly used in PA include visible light (400 nm), near infrared (700 nm) and short-wave infrared (1300 nm) [30,31]. In addition, multi-spectral remote sensing and hyperspectral remote sensing to meet different needs have also been proved to be effective means of plant phenotypic analysis, crop index acquisition and stress monitoring [32]. For example, European Sentinel-2, ENVISAT MERIS, French SPOT satellite and NOAA AVHRR satellite data, India's Hyperion, China's GF series data and HJ remote sensing data have been widely used [33-35]. It is worth mentioning that the emergence of unmanned aerial vehicle (UAV) marks a new era of RS. UAV is a kind of unmanned small aircraft, which is often used to carry remote sensing equipment for aerial data acquisition. It can provide more abundant and comprehensive spectral, spatial and temporal resolution data, vegetation height data and multi-angle observation, and has the characteristics of high efficiency, convenience, low cost and strong adaptability [36]. There have been many successful cases in crop classification, weed detection and vegetation monitoring, which prove the feasibility of UAV in PA [37]. For example, Marques et al have overcome the limitation of limited spectral coverage based on UAV in 2024, especially in low light, fog or smoke conditions to achieve real-time, efficient and distributed accurate monitoring [38]. Bah et al used drone images to detect weeds in the field in 2017 with an accuracy of more than 90% [39]. Yang et al used UAV image information to identify rice lodging based on decision tree algorithm in 2017, with an overall accuracy of 96.17% [40].

As the core means of dealing with agricultural remote sensing information, ML model has been widely used and deeply studied in recent years. ML is a data analysis method that allows computer systems to automatically learn patterns and rules from data without explicit programming. Researchers tend to use ML as an integrated framework for feature collection and classification, prediction, or decision support [42]. With the improvement of big data's computing power, many classical algorithms have been optimized and improved, and new models and methods continue to emerge [43]. Common ML methods include decision tree (DT), support vector machine (SVM) and logical regression (LR). The core of these methods is to find optimized statistical information ways, so as to automatically and efficiently solve practical problems such as classification and regression [44]. In addition, the convolution neural network (CNN) method based on ML has unique advantages in the field of image processing. It can automatically extract deep features from images and achieve accurate classification or recognition tasks [45]. Because of its unique data expression ability, these technologies can learn and extract valuable information automatically, thus effectively avoiding the complexity and subjectivity brought by traditional methods, and greatly improve the efficiency and generalization of processing multi-platform RS data [46]. It is with these advantages that ML has attracted more attention from agricultural researchers and experts, and listed it as the engine factor for the development of PA [47-51].

In view of the major challenges posed by global population growth, resource shortage and climate change to traditional agricultural models, the purpose of this study is to explore how to promote the development of precision agriculture through the integrated application of remote sensing technology and machine learning algorithms. In order to achieve this goal, based on the keywords "remote sensing" (RS), "machine learning" (ML) and "precision agriculture" (PA), we preliminarily searched the literatures in the past 10 years by using databases such as Web of Science, Scopus, Google Scholar and PubMed. More than 12000 related research articles were selected and quantitatively analyzed (as shown in figure 1). We have observed an overall upward trend in the number of related publications over the past 10 years (as shown in figure 1 (a)). In order to ensure the comprehensiveness and depth of the study, on the basis of preliminary search, we also combine key words including "agricultural monitoring", "detection of diseases and insect pests", "land use and management", "yield prediction" and "sustainable development of agriculture". After rigorous screening, the peer-reviewed papers published in 2014-2024 were finally identified, involving agricultural science, environmental science and related cross-cutting fields, and it can be seen that the number of research papers is also on the rise (as shown in figure 1 (b)). In addition, from the perspective of international cooperation and regional distribution, researchers in China, the United States, Brazil and other countries have made significant contributions to the application of RS technology and ML in the field of precision agriculture. At the same time, there is an obvious imbalance in the spatial distribution of these studies, and there are great differences in research contributions among different regions (as shown in figure 2). Therefore, through in-depth analysis and summary of the existing research results, it is very important to systematically summarize the application status of RS and ML in precision agriculture, and to discuss the current challenges and possible future development directions in this field.

] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [First of all, we use Web of Science, Scopus, Google Scholar and PubMed databases for systematic literature retrieval to identify the research related to remote sensing technology, machine learning algorithms and precision agriculture in the past 10 years. In addition, keywords were screened: keywords "remote sensing" (RS), "machine learning" (ML) and "precision agriculture" (PA), and other relevant keywords were added on the basis of the preliminary search, such as "agricultural monitoring", "pest detection", "land use and management", "yield forecasting" and "sustainable agricultural development". The selected literatures are analyzed in detail, including the statistics of the application of different types of remote sensing data and machine learning algorithms, and their application in precision agriculture is evaluated.]

Comments 2: 2 - The authors included part of the description of the methodology in the abstract, but this is not repeated in the methodology of the article. The abstract should contain parts of the article and not show any inconsistency between what is described in it and what will be found in the article. 

Response 1: [Thank you for pointing out our consistency in the summary and methodology section. We have adjusted the summary to be consistent with the methodology section:

Abstract: With global population growth, resource shortages and climate change, traditional agricultural models face major challenges. Precision agriculture (PA), as a way to realize accurate management and decision support of agricultural production process by using modern information technology, is becoming one of the effective ways to solve these challenges. In particular, the combination of remote sensing technology and machine learning algorithm brings new possibilities for precision agriculture. However, there are relatively few comprehensive and systematic reviews on the integrated application of these two technologies. For this reason, this study makes a systematic literature search on the databases of Web of Science, Scopus, Google Scholar and PubMed, and analyzes the integrated application of remote sensing technology and machine learning algorithm in precision agriculture in recent 10 years. The study found that: (1) because of their characteristics, different types of remote sensing data have significant differences in meeting the needs of precision agriculture, in which hyperspectral remote sensing is the most widely used, accounting for more than 30%. The application of UAV remote sensing has the most potential, accounting for about 24%, and showing an upward trend. (2) Machine learning algorithm has obvious advantages in promoting the development of precision agriculture, in which support vector machine algorithm is the most widely used, accounting for more than 20%, followed by random forest algorithm, accounting for about 18%. In addition, this study also discusses the main challenges faced at present, such as the difficult problems of acquisition and processing of high-quality remote sensing data, model interpretation and generalization ability, and looks forward to the future development trend. such as promoting agricultural intelligence and automation, strengthening international cooperation and sharing and sustainable transformation of achievements. To sum up, this study can provide new ideas and reference for remote sensing combined with machine learning in promoting the development of precision agriculture.

1. Introduction

In the context of rapid global climate change, agricultural practice is facing unprecedented uncertainties and challenges, such as climate warming, sea-level rise, drought and flood and other extreme hydroclimatic frequent occurrence [1-4]. At the same time, the global population is expected to reach 8.7 billion by 2030 and climb to 9.7 billion by 2050, which undoubtedly puts tremendous pressure on global food production [5]. However, it is gratifying that in recent years, with the increase of investment in science and technology and agricultural research, the development of PA has achieved certain results [6,7]. This progress not only changes the traditional mode of agricultural production, but also aims to optimize agricultural inputs (seeds, water resources, chemicals) through the application of advanced technologies (such as remote sensing, machine learning algorithms, agricultural robots, etc.), effectively manage crop variability, maintain or even increase yields, and cleverly avoid potential losses, thereby improving the efficiency and profitability of agricultural systems [8]. For example, Yomo et al found in 2023 that the maximum likelihood algorithm based on Landsat-8 remote sensing images is used to classify land use and land cover, and by using multi-layer perceptron-Markov chain modeling method, the results show that the overall accuracy (Kappa coefficient) is as high as 92% [9]. This study shows that the accuracy of agricultural monitoring and recognition can be significantly improved by integrating advanced remote sensing technology and machine learning algorithm. In addition, another study shows that the integrated learning random forest classifier is used to study the progressive lodging sensitive characteristics of rice types based on multi-spectral (444-842 nm) fusion Unmanned Aerial Vehicle technology, with an overall accuracy of 96.1% [10]. These examples prove the effective application of advanced technologies and algorithms in PA. Although PA has many advantages, limitations such as information accuracy, large amount of data, operational complexity and high initial cost cannot be ignored [11]. Therefore, for most countries in the world, it is very necessary to actively promote the coordinated development of PA, remote sensing (RS) and machine learning (ML), ensure agricultural production safety, and strictly abide by the food safety red line.

In recent years, people have made multi-dimensional and deep exploration and efforts in the development of PA, which is mainly reflected in the research and application of new technologies, covering many key links such as personnel training, policy support and so on [12,13]. The aim is to overcome the shortcomings of traditional agriculture, such as time-consuming and laborious, improper use of resources, unstable crop yield and environmental pollution [14,15]. In this context, it is actually a complex and critical challenge to monitor crop growth and conditions in different locations and environments in real time, accurately and at multiple scales, and to use data with different time resolutions to meet a variety of purposes. In fact, it is a complex and critical challenge to respond quickly to extreme events according to changing climate conditions [16,17]. Fortunately, RS technology has developed rapidly in agriculture, forestry, hydrology, environmental protection and other fields because of its unique advantages (such as synchronization, timeliness, spatio-temporal continuity and large-scale observation ability) [18-22]. RS is a technology that can obtain the information of the earth's surface without physical contact. It uses sensors to capture and record electromagnetic radiation signals reflected, emitted or scattered from the earth's surface from a long distance, and then continuously identify, measure and analyze the characteristics of target objects located on, above or even below the earth's surface by analyzing these signals [23]. This not only greatly improves the efficiency of agricultural information acquisition, but also provides strong support for dealing with agriculture in extreme weather, so that crop managers can take timely measures to reduce the impact of disasters and ensure the safety and stability of agricultural production [24,25]. In addition, with the maturity of RS inversion algorithms (such as linear regression, PROSAIL physical model, neural network), inversion data sets based on RS images have also appeared, such as inversion product data sets based on MODIS images, Landsat images and Sentinel-2 images: water quality and water environment elements inversion products, vegetation parameter inversion products, land surface temperature inversion products and soil parameter inversion products. It provides reliable and rich data sources for agricultural RS related research [26-28].

It is well known that most agricultural RS data are information provided by visible light and near infrared radiation reflected (or transmitted) by plants, which are measured according to wavelength, that is, spectral reflectance [29]. According to the change of vegetation, the spectral data commonly used in PA include visible light (400 nm), near infrared (700 nm) and short-wave infrared (1300 nm) [30,31]. In addition, multi-spectral remote sensing and hyperspectral remote sensing to meet different needs have also been proved to be effective means of plant phenotypic analysis, crop index acquisition and stress monitoring [32]. For example, European Sentinel-2, ENVISAT MERIS, French SPOT satellite and NOAA AVHRR satellite data, India's Hyperion, China's GF series data and HJ remote sensing data have been widely used [33-35]. It is worth mentioning that the emergence of unmanned aerial vehicle (UAV) marks a new era of RS. UAV is a kind of unmanned small aircraft, which is often used to carry remote sensing equipment for aerial data acquisition. It can provide more abundant and comprehensive spectral, spatial and temporal resolution data, vegetation height data and multi-angle observation, and has the characteristics of high efficiency, convenience, low cost and strong adaptability [36]. There have been many successful cases in crop classification, weed detection and vegetation monitoring, which prove the feasibility of UAV in PA [37]. For example, Marques et al have overcome the limitation of limited spectral coverage based on UAV in 2024, especially in low light, fog or smoke conditions to achieve real-time, efficient and distributed accurate monitoring [38]. Bah et al used drone images to detect weeds in the field in 2017 with an accuracy of more than 90% [39]. Yang et al used UAV image information to identify rice lodging based on decision tree algorithm in 2017, with an overall accuracy of 96.17% [40].

As the core means of dealing with agricultural remote sensing information, ML model has been widely used and deeply studied in recent years. ML is a data analysis method that allows computer systems to automatically learn patterns and rules from data without explicit programming. Researchers tend to use ML as an integrated framework for feature collection and classification, prediction, or decision support [42]. With the improvement of big data's computing power, many classical algorithms have been optimized and improved, and new models and methods continue to emerge [43]. Common ML methods include decision tree (DT), support vector machine (SVM) and logical regression (LR). The core of these methods is to find optimized statistical information ways, so as to automatically and efficiently solve practical problems such as classification and regression [44]. In addition, the convolution neural network (CNN) method based on ML has unique advantages in the field of image processing. It can automatically extract deep features from images and achieve accurate classification or recognition tasks [45]. Because of its unique data expression ability, these technologies can learn and extract valuable information automatically, thus effectively avoiding the complexity and subjectivity brought by traditional methods, and greatly improve the efficiency and generalization of processing multi-platform RS data [46]. It is with these advantages that ML has attracted more attention from agricultural researchers and experts, and listed it as the engine factor for the development of PA [47-51].

In view of the major challenges posed by global population growth, resource shortage and climate change to traditional agricultural models, the purpose of this study is to explore how to promote the development of precision agriculture through the integrated application of remote sensing technology and machine learning algorithms. In order to achieve this goal, based on the keywords "remote sensing" (RS), "machine learning" (ML) and "precision agriculture" (PA), we preliminarily searched the literatures in the past 10 years by using databases such as Web of Science, Scopus, Google Scholar and PubMed. More than 12000 related research articles were selected and quantitatively analyzed (as shown in figure 1). We have observed an overall upward trend in the number of related publications over the past 10 years (as shown in figure 1 (a)). In order to ensure the comprehensiveness and depth of the study, on the basis of preliminary search, we also combine key words including "agricultural monitoring", "detection of diseases and insect pests", "land use and management", "yield prediction" and "sustainable development of agriculture". After rigorous screening, the peer-reviewed papers published in 2014-2024 were finally identified, involving agricultural science, environmental science and related cross-cutting fields, and it can be seen that the number of research papers is also on the rise (as shown in figure 1 (b)). In addition, from the perspective of international cooperation and regional distribution, researchers in China, the United States, Brazil and other countries have made significant contributions to the application of RS technology and ML in the field of precision agriculture. At the same time, there is an obvious imbalance in the spatial distribution of these studies, and there are great differences in research contributions among different regions (as shown in figure 2). Therefore, through in-depth analysis and summary of the existing research results, it is very important to systematically summarize the application status of RS and ML in precision agriculture, and to discuss the current challenges and possible future development directions in this field.

] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [We have ensured that the method description in the summary matches the methodology section to eliminate any potential inconsistencies. I hope these changes will meet your requirements.]

Comments 3: 3- The text is still confusing and very descriptive, and does not fit in with what a review should be. It is not clear what the aim of the review is and how it is organized. The text would need to be completely rewritten.  

Response 1: [Thank you for your valuable comments on our research. You pointed out that our text is still confusing and over-descriptive and does not conform to the structure and purpose of the review. We fully understand your concerns and have thoroughly rewritten the summary and introduction based on your feedback, and adjusted the structure of the article to ensure that it expresses the purpose and organizational structure of the research more clearly:

Abstract: With global population growth, resource shortages and climate change, traditional agricultural models face major challenges. Precision agriculture (PA), as a way to realize accurate management and decision support of agricultural production process by using modern information technology, is becoming one of the effective ways to solve these challenges. In particular, the combination of remote sensing technology and machine learning algorithm brings new possibilities for PA. However, there are relatively few comprehensive and systematic reviews on the integrated application of these two technologies. For this reason, this study makes a systematic literature search on the databases of Web of Science, Scopus, Google Scholar and PubMed, and analyzes the integrated application of remote sensing technology and machine learning algorithm in PA in recent 10 years. The study found that: (1) because of their characteristics, different types of remote sensing data have significant differences in meeting the needs of PA, in which hyperspectral remote sensing is the most widely used, accounting for more than 30%. The application of UAV remote sensing has the most potential, accounting for about 24%, and showing an upward trend. (2) Machine learning algorithm has obvious advantages in promoting the development of PA, in which support vector machine algorithm is the most widely used, accounting for more than 20%, followed by random forest algorithm, accounting for about 18%. In addition, this study also discusses the main challenges faced at present, such as the difficult problems of acquisition and processing of high-quality remote sensing data, model interpretation and generalization ability, and looks forward to the future development trend. such as promoting agricultural intelligence and automation, strengthening international cooperation and sharing and sustainable transformation of achievements. To sum up, this study can provide new ideas and reference for remote sensing combined with machine learning in promoting the development of PA.

Keywords: Agricultural monitoring; Disease and pest detection, Land use and management; Yield prediction; Agricultural sustainable development

1. Introduction

In the context of rapid global climate change, agricultural practice is facing unprecedented uncertainties and challenges, such as climate warming, sea-level rise, drought and flood and other extreme hydroclimatic frequent occurrence [1-4]. At the same time, the global population is expected to reach 8.7 billion by 2030 and climb to 9.7 billion by 2050, which undoubtedly puts tremendous pressure on global food production [5]. However, it is gratifying that in recent years, with the increase of investment in science and technology and agricultural research, the development of PA has achieved certain results [6,7]. This progress not only changes the traditional mode of agricultural production, but also aims to optimize agricultural inputs (seeds, water resources, chemicals) through the application of advanced technologies (such as remote sensing, machine learning algorithms, agricultural robots, etc.), effectively manage crop variability, maintain or even increase yields, and cleverly avoid potential losses, thereby improving the efficiency and profitability of agricultural systems [8]. For example, Yomo et al found in 2023 that the maximum likelihood algorithm based on Landsat-8 remote sensing images is used to classify land use and land cover, and by using multi-layer perceptron-Markov chain modeling method, the results show that the overall accuracy (Kappa coefficient) is as high as 92% [9]. This study shows that the accuracy of agricultural monitoring and recognition can be significantly improved by integrating advanced remote sensing technology and machine learning algorithm. In addition, another study shows that the integrated learning random forest classifier is used to study the progressive lodging sensitive characteristics of rice types based on multi-spectral (444-842 nm) fusion Unmanned Aerial Vehicle technology, with an overall accuracy of 96.1% [10]. These examples prove the effective application of advanced technologies and algorithms in PA. Although PA has many advantages, limitations such as information accuracy, large amount of data, operational complexity and high initial cost cannot be ignored [11]. Therefore, for most countries in the world, it is very necessary to actively promote the coordinated development of PA, remote sensing (RS) and machine learning (ML), ensure agricultural production safety, and strictly abide by the food safety red line.

In recent years, people have made multi-dimensional and deep exploration and efforts in the development of PA, which is mainly reflected in the research and application of new technologies, covering many key links such as personnel training, policy support and so on [12,13]. The aim is to overcome the shortcomings of traditional agriculture, such as time-consuming and laborious, improper use of resources, unstable crop yield and environmental pollution [14,15]. In this context, it is actually a complex and critical challenge to monitor crop growth and conditions in different locations and environments in real time, accurately and at multiple scales, and to use data with different time resolutions to meet a variety of purposes. In fact, it is a complex and critical challenge to respond quickly to extreme events according to changing climate conditions [16,17]. Fortunately, RS technology has developed rapidly in agriculture, forestry, hydrology, environmental protection and other fields because of its unique advantages (such as synchronization, timeliness, spatio-temporal continuity and large-scale observation ability) [18-22]. RS is a technology that can obtain the information of the earth's surface without physical contact. It uses sensors to capture and record electromagnetic radiation signals reflected, emitted or scattered from the earth's surface from a long distance, and then continuously identify, measure and analyze the characteristics of target objects located on, above or even below the earth's surface by analyzing these signals [23]. This not only greatly improves the efficiency of agricultural information acquisition, but also provides strong support for dealing with agriculture in extreme weather, so that crop managers can take timely measures to reduce the impact of disasters and ensure the safety and stability of agricultural production [24,25]. In addition, with the maturity of RS inversion algorithms (such as linear regression, PROSAIL physical model, neural network), inversion data sets based on RS images have also appeared, such as inversion product data sets based on MODIS images, Landsat-8 images and Sentinel-2 images: water quality and water environment elements inversion products, vegetation parameter inversion products, land surface temperature inversion products and soil parameter inversion products. It provides reliable and rich data sources for agricultural RS related research [26-28].

It is well known that most agricultural RS data are information provided by visible light and near infrared radiation reflected (or transmitted) by plants, which are measured according to wavelength, that is, spectral reflectance [29]. According to the change of vegetation, the spectral data commonly used in PA include visible light (400 nm), near infrared (700 nm) and short-wave infrared (1300 nm) [30,31]. In addition, multi-spectral remote sensing and hyperspectral remote sensing to meet different needs have also been proved to be effective means of plant phenotypic analysis, crop index acquisition and stress monitoring [32]. For example, European Sentinel-2, ENVISAT MERIS, French SPOT satellite and NOAA AVHRR satellite data, India's Hyperion, China's GF series data and HJ remote sensing data have been widely used [33-35]. It is worth mentioning that the emergence of unmanned aerial vehicle (UAV) marks a new era of RS. UAV is a kind of unmanned small aircraft, which is often used to carry remote sensing equipment for aerial data acquisition. It can provide more abundant and comprehensive spectral, spatial and temporal resolution data, vegetation height data and multi-angle observation, and has the characteristics of high efficiency, convenience, low cost and strong adaptability [36]. There have been many successful cases in crop classification, weed detection and vegetation monitoring, which prove the feasibility of UAV in PA [37]. For example, Marques et al have overcome the limitation of limited spectral coverage based on UAV in 2024, especially in low light, fog or smoke conditions to achieve real-time, efficient and distributed accurate monitoring [38]. Bah et al used UAV images to detect weeds in the field in 2017 with an accuracy of more than 90% [39]. Yang et al used UAV image information to identify rice lodging based on decision tree algorithm in 2017, with an overall accuracy of 96.17% [40].

As the core means of dealing with agricultural remote sensing information, ML model has been widely used and deeply studied in recent years. ML is a data analysis method that allows computer systems to automatically learn patterns and rules from data without explicit programming. Researchers tend to use ML as an integrated framework for feature collection and classification, prediction, or decision support [42]. With the improvement of big data's computing power, many classical algorithms have been optimized and improved, and new models and methods continue to emerge [43]. Common ML methods include decision tree (DT), support vector machine (SVM) and logical regression (LR). The core of these methods is to find optimized statistical information ways, so as to automatically and efficiently solve practical problems such as classification and regression [44]. In addition, the convolution neural network (CNN) method based on ML has unique advantages in the field of image processing. It can automatically extract deep features from images and achieve accurate classification or recognition tasks [45]. Because of its unique data expression ability, these technologies can learn and extract valuable information automatically, thus effectively avoiding the complexity and subjectivity brought by traditional methods, and greatly improve the efficiency and generalization of processing multi-platform RS data [46]. It is with these advantages that ML has attracted more attention from agricultural researchers and experts, and listed it as the engine factor for the development of PA [47-51].

In view of the major challenges posed by global population growth, resource shortage and climate change to traditional agricultural models, the purpose of this study is to explore how to promote the development of PA through the integrated application of RS technology and ML algorithms. In order to achieve this goal, based on the keywords "remote sensing" (RS), "machine learning" (ML) and "precision agriculture" (PA), we preliminarily searched the literatures in the past 10 years by using databases such as Web of Science, Scopus, Google Scholar and PubMed. More than 12000 related research articles were selected and quantitatively analyzed (as shown in figure 1). We have observed an overall upward trend in the number of related publications over the past 10 years (as shown in figure 1 (a)). In order to ensure the comprehensiveness and depth of the study, on the basis of preliminary search, we also combine key words including "agricultural monitoring", "detection of diseases and insect pests", "land use and management", "yield prediction" and "sustainable development of agriculture". After rigorous screening, the peer-reviewed papers published in 2014-2024 were finally identified, involving agricultural science, environmental science and related cross-cutting fields, and it can be seen that the number of research papers is also on the rise (as shown in figure 1 (b)). In addition, from the perspective of international cooperation and regional distribution, researchers in China, the United States, Brazil and other countries have made significant contributions to the application of RS technology and ML in the field of PA. At the same time, there is an obvious imbalance in the spatial distribution of these studies, and there are great differences in research contributions among different regions (as shown in figure 2). Therefore, through in-depth analysis and summary of the existing research results, it is very important to systematically summarize the application status of RS and ML in PA, and to discuss the current challenges and possible future development directions in this field.

Figure 1. The changing trend of peer-reviewed papers published in the past 10 years based on keyword retrieval over time. (a) Based on the databases of Web of Science, Scopus, Google Scholar and PubMed, we searched 12000 papers published in the past 10 years; (b)The changing trend of peer-reviewed papers published in agricultural science, environmental science and related cross-fields in the past 10 years based on keywords.

Figure 2. The geographical distribution of precision agriculture research based on the integration of remote sensing technology and ML the color depth directly reflects the number of research.

2. Remote Sensing Technology and Machine Learning method

2.1 Remote sensing data in precision agriculture

There is no doubt that the application of RS technology in agriculture has greatly promoted agricultural reform. This technology enables us to collect global data on the earth's surface remotely on a regular basis, providing unprecedented convenience for agricultural production and management [52-54]. Through a variety of sensors, we can directly or indirectly obtain almost all the key elements of agricultural practice, from crop growth to soil moisture monitoring, pest and pest early warning to yield prediction. At the same time, the wide geographical coverage and diversified resolution of RS technology also provide valuable data support for agricultural production and management [55]. As shown in figure 3, remote sensing satellites with different resolutions play different key roles in different PA practices, and rely on different characteristics and advantages to comprehensively serve the specific needs of PA from many angles [56,57]. With the continuous updating and upgrading of remote sensing sensors, agricultural managers and practitioners will continue to benefit from the in-depth application of RS technology, for example, RS data show high practicability and effectiveness in evaluating and monitoring agricultural practice [58,59].

In general, when obtaining remote sensing data, the value of RS images with appropriate resolution, band, reliable quality and cost-effectiveness can be maximized if they are selected according to specific agricultural problems. For example, using daily 10m NDVI data from Sentinel-2 images can quickly, efficiently and accurately monitor the flowering date of apples, and then provide technical reference for accurate classification and growth trend prediction of fruit trees [62]. In another study, the use of Landsat-8 images with a spatial resolution of 10 to 30 meters provided a promising solution for disease detection in mixed forests in southern China [63]. In other studies for the detection of plant diseases and pests infecting vegetation, the detection accuracy does not seem to be satisfactory based on visible light (780nm) data. In a 2023 study by Zhu et al., although the use of UAV technology can confirm the importance of red-light bands and adjacent bands, it did not achieve the desired results in the investigation of plant diseases and pests invading vegetation [65]. However, it is gratifying that multi-spectral remote sensing data with rich bands and a wide range of wavelengths can capture subtle changes in infected plants affected by diseases and insect pests, thus showing excellent ability in early pest detection [66]. In their research in 2024, Ren et al used the characteristics of UAV to obtain crop growth status quickly and accurately in small and medium-sized areas [67]. By assimilating remote sensing data with WOFOST model effectively by Kalman filter algorithm, the accuracy of yield simulation of different processing schemes is significantly improved, and more accurate and reliable yield prediction information is provided for agricultural producers.

In the practical application of PA, according to different requirements and application scenarios, commonly used RS data sources include hyperspectral, multispectral, thermal infrared remote sensing, LiDAR remote sensing, SAR remote sensing and UAV technology and so on. As shown in figure 4, the application of various RS data sources in PA is shown in detail, including the time distribution and proportion of RS data sources in PA. These informations undoubtedly provide valuable reference for agricultural managers and practitioners, not only help them have a more comprehensive and in-depth understanding of the characteristics and applicability of various RS technologies, but also provide strong support for them to make scientific and reasonable decisions in practical work.

Figure 3. Based on the comprehensive application framework of different remote sensing satellites in precision agriculture.

Figure 4. Remote sensing data commonly used in precision agriculture. (a) the distribution characteristics of different types of remote sensing data with time, and marked with different types of colors; (b) based on the literature retrieved in this paper, the statistical analysis of the proportion of all kinds of remote sensing data sources.

2.2 Overview of ML algorithms in Precision Agriculture

The concept of machine learning (ML) can usually be traced back to Alan Turing's classic research article published in 1950, that is, the possibility that machines can exhibit behaviors similar to human intelligence [68,69]. This concept continued to develop in the following decades and gradually became a vital branch of computer science. The core principle of ML is to automatically learn and sum up the rules in the input data, and realize the accurate prediction or classification of unknown data by extracting key features and constructing mapping functions [70]. In addition, as the core component of artificial intelligence, ML gives computer systems the ability to perform a variety of tasks efficiently, and continues to promote the innovation and development of intelligent technology [71]. Generally speaking, ML mainly contains three elements, namely: model, objective function and optimization algorithm. The model explains the correlation between input and output and the meaning and range of the parameters, the objective function measures the difference between the model prediction and the actual results, and the optimization algorithm minimizes or maximizes the objective function by iteratively adjusting the parameters. as a result, the best model parameters are obtained [72,73]. According to different types of learning, ML can be divided into four main categories: supervised learning, unsupervised learning, semi-supervised learning and reinforcement learning [74,75]. As shown in table 1, the applications of the four main categories of algorithms in PA and their scope of application are listed and described in detail.

Table 1. Common Machine Learning algorithms and references in the Application Field of Precision Agriculture.

Figure 5. The distribution of the most commonly used machine learning algorithms is obtained based on the keywords "machine learning" and "precision agriculture".

Recent studies have shown that with the improvement of computing performance and the enhancement of massive data sets, ML has shown strong application capabilities in many fields, especially in the field of PA [120,121]. In particular, a series of emerging algorithms and technologies such as deep learning (DL), intelligent optimization, neural networks, computer vision and data enhancement continue to emerge. It has not only injected strong impetus into the field of ML, but also provided rich opportunities for all stages of agriculture. They enable agricultural practitioners to respond more effectively to challenges and achieve set goals [122]. As shown in figure 5, the frequency distribution of the algorithm is obtained by searching the keywords "ML" and "PA". From the chart, we can see that ML is widely used in the field of PA, and support vector machine algorithm has the highest frequency, accounting for more than 20%, followed by random forest algorithm, accounting for about 18%.

In addition, in the specific application of PA, different algorithms have given full play to their own advantages, and achieved a series of encouraging results. For example, Sladojevic et al proposed a new plant leaf disease detection and classification model based on deep convolution neural network. The model can accurately identify 13 different plant diseases and effectively distinguish plant leaves from the surrounding environment, which provides a powerful tool for plant health monitoring [123]. Li et al have made remarkable progress in the field of vegetable disease detection. They propose a lightweight network improvement algorithm based on YOLOv5s. The algorithm effectively eliminates external interference and significantly enhances the ability of multi-scale feature extraction, thus improving the scope and performance of disease detection [124]. Ashwinkuma et al. developed a convolution neural network based on the optimal mobile network, which is used to automatically detect and classify plant leaf diseases. The experimental results show that the CNN model performs well, the maximum accuracy is 0.985, the recall rate is 0.9892, the accuracy is 0.987 and the Kappa coefficient is 0.985 [125]. Yu et al use DL target detection technology to extract image feature information through complex network structure to achieve non-destructive recognition of crop diseases. Compared with the traditional method, this technique has higher recognition accuracy, faster detection speed and good stability in the visible light range [126]. Ang et al. creatively used Landsat-8 time series satellite images, combined with ML and normalized difference vegetation index (NDVI), successfully developed a new method, which effectively proved its value in yield prediction [127]. Aydin et al tested gradient lifting methods such as XGBoost, LightGBM and CatBoost for soil sample classification, and achieved high classification accuracy of up to 90%. Compared with previous studies, the prediction accuracy has been significantly improved [128].

3. Integrated Application of remote Sensing Technology and Machine Learning method

3.1 Agricultural monitoring and identification

Recent scientific research shows that the integration of RS technology and ML methods has brought remarkable progress to the application of agricultural monitoring and identification. RS technology can efficiently obtain crop planting area, growth status and other important information, while ML technology can accurately detect targets and extract features from these rich RS data, so as to achieve fine identification and classification of crops. With the continuous integration and development of these two technologies, the related agricultural problems, such as horticulture line detection [129], crop detection and classification [130,131], vegetation distribution [132,133], will usher in a new turning point. For example, Zhao et al improved the standardized precipitation evapotranspiration index (SPEI) by integrating RS data, and the results show that the new SPEI can greatly enhance the capacity of agricultural drought monitoring [134]. Lyu et al used EO-1 Hyperion images, combined with multi-terminal spectral mixing analysis (MESMA) and fully constrained least square pixel mixing (FCLS) techniques, to successfully identify typical vegetation species and improve the accuracy of grassland degradation monitoring [135]. Xiao et al fused Sentinel-2 and MODIS RS images using the enhanced spatio-temporal adaptive reflection fusion model (ESTARFM), and then accurately obtained the spatial distribution of irrigated rice fields by random forest (RF) algorithm. Based on the Penman-Monteith model and making full use of the daily observation data of the meteorological station, the dynamic monitoring of water resources in the critical irrigation period has been realized, and remarkable results have been achieved [136]. This application increases the feasibility of spatio-temporal fusion of multi-source RS data and makes it possible to continuously monitor the irrigation dynamics of paddy fields on a large scale.

In addition, it is very important to grasp the planting and distribution of crops on a large scale in a timely and efficient manner. Although scholars have done a lot of research on the basis of low and medium resolution RS, due to the widespread existence of mixed pixels and the lack of red edge bands, these techniques are difficult to effectively identify small plots of farmland, resulting in unsatisfactory recognition accuracy [137,138]. However, the research of Guo et al in this field has brought new breakthroughs. Using GF-6 WFV images, they constructed several decision tree models, which not only efficiently obtained the information of crop planting area and its spatial distribution, but also significantly improved the accuracy of image recognition [139]. In their latest research, Zhang et al used GF-1 RS images, combined with advanced multi-scale segmentation algorithms, to improve the accuracy of identifying forest types in Engebe ecological demonstration areas, and used nearest neighbor classification and random forest (RF) classification respectively, and compared the recognition results. The results showed that the effect of random forest classification was better, and the Kappa coefficients obtained in two consecutive years were 0.92 and 0.90 [99] respectively. Through a lot of research, people have reached a consensus: ML algorithms, including random forest (RF), support vector machine (SVM), artificial neural network (ANN), decision tree (DT) and so on, have greater potential in agricultural monitoring and recognition, and can significantly improve the efficiency and accuracy of monitoring and recognition [140-147].

3.2 Stress detection of diseases and insect pests

Crop diseases and insect pests are not only one of the core factors affecting plant yield and quality, but also one of the main causes of crop damage. As emphasized by the United Nations Food and Agriculture Organization, the damage caused by diseases and insect pests to agro-ecosystems should not be underestimated. Unfortunately, in early agricultural production, the problem of diseases and insect pests is often marginalized and not given enough attention, resulting in huge economic losses [148]. Therefore, the implementation of efficient and accurate detection of plant diseases and insect pests not only plays a vital role in ensuring the health of agro-ecosystem, improving crop yield and quality, and reducing economic losses, but also plays a key role in the development of PA. in recent years, it has attracted the attention of many scholars [149]. The premise to achieve this goal is to accurately detect and classify diseases and insect pests, and to distinguish different types and estimated quantities, in order to implement accurate pest prevention and control strategies. Traditional pest monitoring mainly depends on the manual identification of insect experts or technicians, this method is not only subjective and labor-intensive, but also not practical in large-scale applications [150]. However, with the popularity of sensors and embedded devices on the Internet, the combination of RS and ML has opened up a new way for the detection of diseases and insect pests in modern agriculture. Mahanta et al obtained rich spectral features of vegetation based on a variety of sensor devices, and used ML models to identify spectral patterns related to specific diseases. Finally, the evaluation of the health degree of insects invading the forest was realized, and the detection efficiency was greatly improved [152].

Most diseases and insect pests have the characteristics of concealment, latent, infectivity and uncertainty, which undoubtedly increase the difficulty and cost of control and bring great challenges to agricultural production [153,154]. However, it is worth noting that RS data from satellite sensors show that plants affected by the disease can be distinguished in a relatively short period of time by their spectral characteristics different from healthy plants [155]. The figure 6 shows the process of pest monitoring based on different types of sensors. More importantly, through further use of ML for analysis, we can not only determine the degree of damage, but also accurately identify the type of disease [156,157]. In the early detection of diseases and insect pests, many researchers tend to use traditional methods, that is, to establish empirical statistical models between diseases and insect pests and related factors (such as environment, climate, soil and vegetation index) [158,159]. In order to achieve effective monitoring of diseases and insect pests. These methods include multiple linear regression, partial least square regression, support vector regression and random forest regression. For example, Ebrahimi et al use support vector regression to detect parasites in crop canopy images, which greatly improves the detection accuracy [160].

Figure 6. Based on IDS map, MODIS map, Landsat-8 map and hyperspectral response technology, the evaluation process of detecting insect infestation forest health was realized.

In addition, through the comprehensive use of advanced image processing techniques such as image segmentation [161], feature extraction [162], target detection [163-166] and classification, researchers can solve complex problems in plant disease detection more accurately and efficiently [167]. Image segmentation technology is used to distinguish normal and abnormal leaves in RS images, while feature extraction is to extract meaningful information from segmented regions, such as color, texture, shape, etc. [168,169]. They provide a more detailed and accurate analysis method for disease detection [170,171]. For example, studies by Zhang et al have shown that TinySegformer model can provide a robust and practical solution for large-scale agricultural pest detection because of its high efficiency, accuracy and lightweight [172]. The interactive segmentation method based on GrabCut proposed by Lu et al., which is applied to field RS, can quickly extract locust images from various parts [173]. Barbedo et al. proposed an automatic detection algorithm for wheat scab based on hyperspectral technology. the algorithm has more than 91% classification accuracy and shows excellent robustness under a variety of complex factors such as shape, direction and shadow [174]. Mumtaz et al. combined with optical RS, image processing and depth learning methods to accurately detect and grade wheat rust [175]. Bao et al. proposed a RS method of UAV based on DDMA-YOLO, which can not only reduce the workload and time consumption of pest detection, but also effectively improve the detection efficiency [176].

Many studies have shown that the image fusion method can greatly enhance the accuracy of vegetation disease detection by fusing image information from different sensors or multi-stage processing [177-181]. For example, some scholars apply DL technology to the fusion of RGB images and segmented images, and develop a multi-head DenseNet architecture. After the strict verification of the public data set and the application of 50% discount cross-validation technology, the method shows excellent performance, and all the evaluation indicators have reached a very high level. for example, the average accuracy, recall, accuracy and F1 score reached 98.17%, 98.17%, 98.16% and 98.12% respectively [182]. Based on the multi-source fusion images of UAV and visible light, Ma et al successfully constructed a variety of ML models, which significantly improved the accuracy of cotton Vickers wilt detection [183].

It is worth mentioning that some researchers have adopted the improved DL algorithm framework for plant disease detection, and achieved remarkable results [184-187]. Dong et al. creatively proposed an effective scale-aware network architecture (ESA-Net) based on low-cost RS images [188]. After strict verification, ESA-Net showed excellent performance in plant disease detection, and achieved strong competitive results. Amarathunga et al proposed a new architecture based on visual converter (ViT), which integrates the attention mechanism driven by domain knowledge and effectively improves the accuracy of micro-pest detection and recognition at the species level [189]. Ye et al designed an end-to-end automatic disease detection framework based on multi-scale MA-UNet model and single-phase image based on UAV aerial photography data and Landsat-8 satellite RS markers, which greatly improved the efficiency and accuracy of disease monitoring [190].

3.3 Management and analysis of soil and land

As the cornerstone of human survival and development, land not only carries the key mission of agricultural production, provides us with food to maintain our livelihood, but also is an indispensable key prerequisite for ensuring human well-being [60,191]. Therefore, the management and analysis of soil and land resources is particularly urgent and important. In the field of soil monitoring and management, traditionally, we rely on field survey methods to obtain the spatial distribution data of soil groups [191]. However, these methods have many shortcomings, such as long monitoring period, high cost, complex operation procedures, many subjective judgment factors, and relatively limited accuracy [192]. Therefore, using traditional methods for soil monitoring is not only time-consuming and labor-consuming, but also may be difficult to meet the needs of modern soil management for accuracy and efficiency [16]. With its more accurate, richer and more professional characteristics, RS has brought revolutionary changes to soil monitoring and management activities. It provides multi-temporal images, enabling us to fully capture dynamic changes in land and soil characteristics [193-195]. In addition, RS has a wide range of data sources with large amounts of information and high accuracy, providing unprecedented possibilities for accurate assessment of soil conditions [196]. The use of advanced ML technology can achieve efficient and accurate processing and analysis of RS data, so as to realize the automation of data processing and feature extraction. It is very important to improve the efficiency and accuracy of soil and land management [197-199].

A survey found that the application of various types of RS data provides convenience and opportunities for soil management [200]. At the same time, in different RS soil applications, multi-spectral RS is the most widely used in soil [201]. Duan et al used the mean value of reflectivity and entropy texture parameters extracted from Landsat- 8 image, combined with MLC, SVM, ANN and RF ML, to identify soil groups in depth, and achieved good results. Zhou et al proposed a general ML method based on spatio-temporal constraints by using Sentinel-1 and Sentinel-2 data in 2024 [202]. Through verification, its accuracy and practicability have been fully affirmed [203]. Musasa et al made a detailed review of soil problems in arid environments in 2023, clearly pointing out that the Landsat-8 satellite mission plays an indispensable role in promoting soil assessment and monitoring [204]. In addition, in view of the significant challenges such as insufficient information acquisition and limited measurement accuracy in the early soil moisture monitoring technology [205], the introduction of ML technology is a revolutionary change, which greatly makes up for these deficiencies [206,207]. In addition, high-resolution data show significant applicability in soil applications, especially in soil resource estimation and mapping [208-211]. Moreover, UAV shows great potential in soil analysis and evaluation, and many studies have fully proved its effectiveness in practical application. For example, Bertalan et al., through the mapping of soil moisture based on UAVs, deeply revealed the spatial heterogeneity of soil moisture and provided strong support for PA [212]. Menzies Pluer et al used UAV to draw the spatial distribution model of farmland soil characteristics and nutrient concentration, which provided a novel and low-cost method and new idea for soil management [213]. In addition, scholars also pointed out that the combination of UAV data fusion and ML is very important for accurate field estimation of soil texture [214-216]. At the same time, in many studies on the integrated application of RS and ML in soil management, we found that the discussion of soil organic carbon and salinity is also an eye-catching direction [217-221].

As one of the key factors of global ecological change, land use / land cover (LULC) has a far-reaching impact on the balance of ecosystem and the sustainable development of human society [222]. It represents the different ways in which human beings maximize the use of land resources and deal with related resources, and is very important for land management and analysis [223]. Therefore, in the research field of land management and analysis, we pay special attention to the temporal and spatial distribution of LULC and its applications. It goes without saying that the application of ML to RS data is of great significance for efficient and accurate land management and analysis [224]. On the one hand, traditional land management and analysis methods are often time-consuming and costly, and it is difficult to provide up-to-date information on various land use / land cover changes [225]. On the other hand, with its strong data acquisition and processing ability, RS can extract high-resolution multispectral information covering large areas that are difficult to access in real time, making land management and classification more cost-effective and time-saving [226,227]. Figure 7 shows the whole process of determining farming patterns using Landsat-8 and MODIS RS data, greatly improving the efficiency of agricultural practices [228]. In recent years, with the continuous development of ML technology, it is becoming more and more popular in mapping, analysis and land spatio-temporal analysis of LULC changes using RS data [229,230]. Examples of land management and analysis based on different ML methods include: random forest [100,101,231], support vector machine [102,103,232], decision tree [90,91,233], maximum likelihood classification [234,235], artificial neural network [97,99,236], convolution neural network [237,242], hybrid multiple model [243,246].

Figure 7. A general process for determining tillage patterns and cultivated / non-cultivated land areas based on multi-source remote sensing data.

3.4 Prediction and decision-making of crop yield

Determining crop yield information plays an indispensable role in crop field management, and crop yield prediction is one of the important cornerstones to ensure food security [247,248]. Traditional crop yield prediction methods usually involve destructive sampling, which not only wastes a lot of human and material resources in practical application, but also is inefficient and cannot meet the needs of the development of modern PA [249]. In order to overcome this bottleneck, we conducted an in-depth and systematic review of the literature, which covered many aspects, such as RS data sources, biological and abiotic factors, physical and chemical parameters, modeling methods and so on. the aim is to provide a more accurate and efficient crop yield prediction scheme and provide strong support for the sustainable development of yield prediction and decision-making.

We have learned that there are differences in the applicability and accuracy of operational assessment of crop status and yield based on different ML algorithms and RS data from different sources. Multi-spectral and medium-resolution RS represented by MODIS data are widely used in early crop yield prediction, and show potential uses [250,251]. Hyperspectral data have unique advantages in prediction, especially data from Landsat-8 satellites and hyperspectral imagers. Related studies have shown that they show great potential in yield prediction of crops such as citrus, wheat, corn, sugarcane and so on [252-256]. In addition, airborne LiDAR and high spatial and temporal resolution images are more suitable for crop yield prediction in fine abundance models [257-260]. A number of studies have shown that UAV data provide accurate and efficient support for PA prediction, especially in crop yield estimation accuracy and phenotypic analysis [261-267]. As shown in figure 8, Yang et al predicted maize yield based on UAV multispectral images combined with ML technology, revealing the great potential of UAV in yield prediction [268].

Figure 8. Prediction Framework of Maize yield based on UAV Image and Machine Learning.

In addition, as an integral part of PA practice, yield forecasting usually does not exist in isolation, but is the result of the interweaving and interaction of climate, soil, water, diseases and insect pests, management and other factors. For example, in an in-depth study, Anwar et al revealed that Australian wheat yields are extremely sensitive to climatic factors [269]. Bai et al. made it clear that assessing the impact of extreme weather on crop production is a key prerequisite for exploring agronomic measures to address climate change, and that fluctuations in climate variables closely related to crop production can have a profound impact on regional and global food production [270]. The importance of soil as a key factor affecting crop yield cannot be ignored. By combining RS data with ML, we can evaluate soil properties more accurately and, taking into account cost-effectiveness and time-benefit, achieve accurate prediction of crop yield [271,272]. Fry et al discussed the spatial variability between field soil properties and soybean yield and found that there was a significant correlation between different soil properties and changes in soybean yield, mainly affected by soil texture and organic carbon content in topsoil (the first 20cm) rather than surface topography [273]. In order to explore the actual effect of water on yield, Zain et al failed to consider the adaptability of the model, which led to adverse results [274]. In another study, Wang et al. developed an accurate polynomial function model, which can effectively adapt to the characteristics of irrigation and application in different areas, provides a scientific guidance strategy for water and fertilizer management, and realizes the accurate prediction of crop yield combined with advanced ML [275]. In addition, the impact of diseases and insect pests and management on yield estimates is also of concern [276-278].

In recent years, the combination of RS technology and ML for crop yield estimation and decision-making has become a research direction with great potential and prospect. The integration of this field not only improves the accuracy and efficiency of crop yield estimation, but also provides strong technical support for the fine management of agricultural production [265]. In this process, the selection of crop physical and chemical parameters is particularly important, which is directly related to the accuracy and reliability of the yield prediction model. Commonly used physiochemical parameters include vegetation coverage (FVC) [279], photosynthetically active radiation absorption (FPAR) [280-282], evapotranspiration (ET) [283-285], leaf area index (LAI) [251,286,287], chlorophyll content [288-290], and various vegetation indices (VIs). Such as normalized difference vegetation index (NDVI) [291-293] and enhanced vegetation index (EVI) [294]. These physical and chemical parameters and indexes are not only widely used in actual agricultural production, but also closely related to yield estimation.

In addition, the crop yield prediction model is also constantly adapting to a variety of new situation changes. For example, although the early traditional ground survey methods and sampling statistics methods based on empirical knowledge have experienced a lot of research and practice, they cannot meet the needs of improving the accuracy of production prediction and reducing costs [295]. With the application of crop growth model and data assimilation model, the yield prediction accuracy will be greatly improved [296]. For example, Zhang et al and Kheir et al have made yield predictions based on APSIM crop model, and achieved remarkable results [297,298]. In addition, the WOFOST model also performs well in crop yield prediction, and a number of studies have revealed its potential use in forecasting [299,300]. The SAFY model provides a new perspective and idea for the estimation of crop yield in a large area [301]. However, crop models are not perfect. They may be limited in large-scale applications, errors are easy to accumulate, and there are some problems such as over-fitting [302]. Similarly, ML models may encounter fitting problems in the training process, especially in the case of small data sets or improper feature selection. Fortunately, ML and data assimilation methods provide new solutions to the problems in crop models and ML [303]. By combining RS data and crop model, and with the help of ML optimization, we can not only make up for the shortcomings of the model in some aspects, but also significantly improve the prediction accuracy and enhance the applicability. This innovative method is gradually being widely concerned and favored by researchers [304-306].

4. Discussion

4.1 Current challenges

Acquisition and processing of multi-source RS data: Although the current RS data sources are rich and diverse, including free and open-source medium and low-resolution MODIS data, and the access is relatively convenient, high-quality, high-resolution RS data is still scarce [307]. This is mainly due to the fact that RS data are easily affected by meteorological factors, equipment factors, terrain factors and other factors in the process of RS data collection, which makes it difficult to guarantee the quality of RS data [308]. At the same time, the acquisition of RS data is also faced with many challenges, such as data accessibility, timeliness, integrity, data privacy and imperfect RS industry chain, which greatly restrict the development of agricultural RS [309]. In addition, the processing process of RS data is a highly specialized technical task, and each link needs profound technical details and professional operators to carry out accurately. For example, if data preprocessing, multi-source data fusion, data interpretation and application are not handled properly, it may not only damage the accuracy and reliability of data, but also lead to unnecessary cost and time investment [310]. It is worth noting that the establishment of RS database can not only provide rich and high-quality data resources, meet the urgent needs of ML for a large number of data, but also achieve data standardization, rapid retrieval and scientific research sharing. For example, the existing problems in agricultural RS, such as low degree of data standardization, limited scale of data sets and uneven data quality, are expected to be solved gradually with the continuous improvement of the database [60]. At the same time, with the continuous progress and innovation of RS technology in the future, the application of multi-source RS fusion data and higher resolution sensors may bring new agricultural changes [312,313].

Interpretability and generalization of the model: The interpretability and generalization ability of ML models are still two major challenges that restrict its development. Generally speaking, the interpretability of ML models is easier to explain than DL models, which is determined by the complexity of the model structure [314]. Although some newer models may lead to improvements in accuracy, the process of understanding and accepting the model will also face challenges for agricultural practitioners. For this reason, in the application of PA, we can choose models with intuitive decision-making process, such as decision trees or rule-based models, which are often easier to be understood and accepted by agricultural practitioners because of their relatively simple structure. it can not only promote the transparency of model decision-making, but also accelerate the transfer of knowledge from technical experts to first-line producers, which is very important to improve the science and efficiency of agricultural practice [315]. In addition, providing information about the features, variables and algorithms that affect the results of the model is also a way to enhance the interpretability of the model, such as sensitivity analysis of model parameters to obtain the calibration of key parameters, or parameter weights based on prior knowledge [316]. Under the action of many factors, such as crop growth environment, varieties, soil conditions and climatic conditions, the model shows insufficient extraction of fine features when facing new data [317]. Usually through data enhancement, model integration and the introduction of regularization technology to optimize the generalization ability of the model, it can effectively prevent the model from fitting and maintain stable performance in the face of new data [318].

4.2 prospects for the future

Trend of intelligence and automation: With the application of intelligence and automation technology in PA, many problems existing in traditional agricultural models, such as insufficient data sets, inaccurate analysis and untimely decision-making, have been solved. This is due to the use of high-precision multi-source RS data, further enhanced data preprocessing to improve data quality, the expansion of data sample diversity and the integration of expert knowledge, all of which improve the accuracy of intelligent decision-making [310,320]. In addition, the development of smart agricultural equipment and the training and advocacy of farmers have lowered the technical threshold and contributed to the widespread dissemination of these technologies [321]. The intelligent fusion of multi-source RS data effectively solves many agricultural data problems, such as ensuring the consistency of data formats, optimizing processing speed, improving the stability of algorithms, and enhancing the generalization ability and interpretability of the model. Thus reducing the uncertainty in the whole process of agricultural production [322]. RS and ML are expected to achieve cross-border integration with advanced technologies such as the Internet of things (IoT), human-computer interaction visualization, data assimilation and blockchain [323,324]. This will further promote the development of comprehensive monitoring, yield prediction and disease monitoring in PA, and provide more accurate, efficient and sustainable solutions for agricultural production.

Data sharing and multidisciplinary interaction: In the context of global connectivity, international cooperation and data sharing mechanisms are strengthening day by day, and PA applications integrating RS and ML should go beyond geographical restrictions [325]. Different countries and regions should work together to share more accurate RS data and smarter ML algorithms to address global agricultural issues such as climate change, food safety and other challenges [319,326]. In addition, ML algorithms are closely combined with geoscience RS technology, automatic machinery and intelligent robots are used to realize the intelligence and precision of field management, and experts in many fields such as agricultural economics, ecology and physical science are integrated at the same time to form a set of comprehensive agricultural management system [321,327,328]. They can not only improve crop yield and quality, but also effectively reduce the use of chemical fertilizers and pesticides, speed up the transformation of agricultural achievements, protect the ecological environment, and promote the green transformation of agriculture. It is worth noting that it is a very important step to ensure that the research results of PA can really benefit farmers and realize the transformation from theoretical knowledge to scientific and technological practice [328,329]. We can form a bottom-up driving force by establishing cooperative relations with agricultural research institutions and cooperating with agricultural enterprises, using their market channels and technical support capabilities to popularize new technologies, encourage farmers to adopt new technologies through policy guidance and support, and encourage farmers' groups to participate in the application of new technologies [330]. In addition, combining this science and technology with the United Nations Sustainable Development goals (SDGs) can not only enhance the scope of science, but also promote the development of practice from a broader perspective, for example, by improving the efficiency of agricultural production and implementing the sustainable development agenda to contribute to the goal of "zero hunger" [331,332]. Although it is still facing technical, economic and social challenges, with the continuous updating of the application of RS technology and ML in the field of agriculture, the potential to promote the modernization and sustainable development of agricultural industry will be further tapped.

5. Conclusions

The integrated application of RS technology and ML algorithm can indeed promote the development and progress of PA. In some agricultural fusion applications, the most widely used RS data source is hyperspectral data, and its application proportion is more than 30%. In addition, the rapid development of UAV remote sensing, accounting for about 24%, is expected to shine in PA in the future. The most widely used ML algorithm is support vector machine, accounting for more than 20%, followed by random forest algorithm, accounting for about 18%. It is worth noting that the rapid development trend of DL algorithms is expected to further promote the development of PA. Monitoring and identification of PA, pest detection, land / soil management and crop yield prediction are still the main aspects of the comprehensive application of RS combined with ML. The challenges of RS and ML algorithm fusion mainly include high-quality RS data acquisition and processing, poor model interpretation and generalization ability, uncertainty of integration development and so on. The future development will mainly focus on promoting agricultural intelligence and automation, strengthening international cooperation and sharing and sustainable transformation of achievements.

] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [Thank you again for your valuable comments on our research. We also fully understand your concern, and all of us authors have revised it on the basis of our common understanding of your suggestion. We sincerely hope to meet your suggestion. Thank you again.]

Response to Reviewer Comments

1. Summary

Thank you very much for taking the time to review this manuscript. We gave a detailed reply below and highlighted the content.

2. Point-by-point response to Comments and Suggestions for Authors

Comments 1: Again, the vast amount of references do not convey how the science presented in this manuscript improves practice. Items I mentioned in my first review, including providing specific details were not addressed in the Discussion and Conclusions. 

Response 1: [.1 Current challenges

4.1.1 Acquisition and processing of multi-source RS data

Although there are free and open source medium and low resolution data resources such as MODIS, which provide the basis for scientific research and application, the scarcity of high-quality and high-resolution RS data is still a significant problem. This scarcity mainly stems from the multiple complexities in the process of data acquisition, including the unpredictability of meteorological conditions, the limitations of equipment performance and the impact of complex terrain on the signal. These factors work together on data quality and increase the difficulty and cost of data acquisition. The accessibility, real-time, integrity and privacy protection of RS data are also important factors restricting the sustainable development of agricultural RS [307,308]. In addition, for the processing of RS data, its highly specialized and technology-intensive characteristics can not be ignored. Data preprocessing, multi-source data fusion, and subsequent data interpretation and application, each link requires fine technical operation and profound professional knowledge. Improper handling will not only reduce the accuracy and reliability of data, but also lead to unnecessary waste of resources and loss of efficiency [309,310]. For example, Zhao et al. proposed a framework for robust classification of multi-view RS images under the condition of missing data, which effectively reduces the practical cost. This research not only improves the classification accuracy, but also reduces the uncertainty and deviation in the process of data processing, and helps to achieve more efficient agricultural management decision-making [311].

 It is worth noting that the construction of RS database system, as one of the effective strategies to alleviate the above problems, its importance has become increasingly prominent. For example, studies have shown that the development of a strong and publicly accessible RS water quality database system can effectively improve the efficiency of water resources management and monitoring [312]. In addition, it can not only provide rich and standardized data resources to meet the needs of ML models for large amounts of data, but also promote rapid data retrieval and scientific research sharing, and inject new vitality into agricultural RS research and application. At the same time, with the continuous improvement of the database system, the long-standing problems in the field of agricultural RS, such as insufficient data standardization, limited scale of data sets, different data quality and so on, are expected to be solved gradually [313]. The research shows that the application of multi-source RS fusion technology and high-resolution sensors may bring revolutionary changes to the field of agricultural RS. The progress of these technologies will greatly enrich the data dimensions, improve the monitoring accuracy, and provide strong data support for the practice of PA based on ML [314]. For example, by combining multi-source RS technology, Joshi et al have achieved rapid and large-scale early warning and accurate control of wheat diseases and insect pests [315]. Based on hyperspectral reflectance and satellite multispectral images, Wu et al significantly improved the accuracy of wheat grain water content estimation and successfully reduced the risk of grain loss and additional drying cost [316].

4.1.2 Interpretability and generalization of the model

Generally speaking, ML model is easier to explain than DL model, which is determined by the complexity of the model structure. Although some newer models may improve accuracy, it will also be a challenge for agricultural practitioners to understand and accept these models. For example, although the DL model performs well in many tasks, because of its black box nature, it may be difficult for agricultural workers to understand the decision logic behind it [317]. Therefore, in the application of PA, such as DT or rule-based model, because of its relatively simple structure, it is often easier for agricultural practitioners to understand and accept. For example, Marin et al use multiple decision tree models to quickly understand the types of vegetation diseases and spatially continuous monitoring, thus improving the efficiency of decision-making. In addition, the interpretable model can not only promote the transparency of agricultural decision-making, but also accelerate the knowledge transfer from technical experts to front-line producers, which is very important to improve the science and efficiency of agricultural practice. For example, in a study of precision irrigation, the use of a rule-based ML model greatly promoted water resources management and agricultural policy decision-making. Usually providing information about the features, variables, and algorithms that affect the results of the model is an effective way to enhance the interpretability of the model. For example, Hao et al realized the high-precision prediction of wheat yield by analyzing the key variables of APSIM-Wheat model through Sobol sensitivity analysis. Under the influence of many factors, such as crop growth environment, varieties, soil conditions and climatic conditions, the model shows insufficient ability to extract fine features in the face of new data. It is often necessary to optimize the generalization ability of the model through data enhancement, model integration and the introduction of regularization technology, so as to effectively prevent overfitting and maintain stable performance in the face of new data. For example, Fawakher et al. used a data enhancement strategy for improved segmentation to significantly improve the accuracy of the model for crop and weed segmentation. In addition, regularization technology can help reduce the complexity of the model, thus improving the generalization ability of the model. For example, by applying L2 regularization in the process of model training, the dependence of the model on training data can be effectively reduced and its performance in the face of new data can be more stable [322].

4.2 prospects for the future

4.2.1 Trend of intelligence and automation

With the application of intelligent and automation technology in PA, many problems existing in traditional agricultural models, such as insufficient data set, inaccurate analysis and untimely decision-making, have been solved. This is due to the use of high-precision multi-source RS data, further enhancement of data preprocessing to improve data quality, expansion of data sample diversity and integration of expert knowledge, all of which improve the accuracy of intelligent decision-making. For example, based on expert dialogue and multi-standard decision-making technology, Goodridge et al proposed an expert system that can intelligently diagnose plant diseases, which significantly improves the detection accuracy of plant diseases. In addition, the development of intelligent agricultural equipment and the promotion of farmers' training have lowered the technical threshold and promoted the wide application of new technologies. A study points out that by developing easy-to-operate smart agricultural equipment, farmers can more easily master new technologies and save water and reduce labour demand by promoting the cultivation of different crop types. Improve the profitability of each farm [324]. 

The intelligent fusion of multi-source RS data effectively solves many problems of agricultural data, such as ensuring the consistency of data format, optimizing processing speed, improving the stability of algorithm, and enhancing the generalization ability and interpretability of the model. as a result, the uncertainty in the whole process of agricultural production is reduced. For example, Zhou et al enhanced the generalization ability through the deep migration learning classification model constructed by integrating ground sensor data and UAV data, and realized the intelligent RS recognition of corn straw type [325]. In the future, RS technology and ML are expected to achieve cross-border integration with advanced technologies such as Internet of things (IoT), human-computer interaction visualization, data assimilation and blockchain. This will further promote comprehensive monitoring, yield forecasting and disease surveillance in PA and provide more accurate, efficient and sustainable solutions for agricultural production. For example, intelligent systems that use the Internet of things to track and schedule accurate irrigation have been studied to help farmers plan irrigation effectively and make informed decisions [326].

4.2.2 Data sharing and multidisciplinary interaction:

In the context of global connectivity, international cooperation and data sharing mechanisms are constantly strengthening, and the application of PA should go beyond geographical restrictions. Different countries and regions should work together to share more accurate RS data and smarter ML algorithms to address global agricultural issues such as climate change, food safety and other challenges [327]. In addition, ML algorithm is closely combined with RS technology, and automatic machinery and intelligent robots are used to realize the intelligence and refinement of field management. at the same time, experts in many fields such as agricultural economy, ecology and physics are integrated to form a set of comprehensive agricultural management system. For example, a cloud-based intelligent irrigation system was introduced in one study to optimize irrigation water use through comprehensive big data collection, storage and analysis, significantly promoting informed decisions on water resources management [328]. In addition, strengthening the cooperative research of RS and ML can not only effectively reduce the use of chemical fertilizers and pesticides, accelerate the transformation of agricultural scientific and technological achievements, protect the ecological environment, and promote the green transformation of agriculture, but also is very important to promote the research results of PA to really benefit farmers and realize the transformation from theoretical knowledge to scientific and technological practice. For example, they can use their market channels and technical support capabilities to popularize new technologies by establishing partnerships with agricultural research institutions and working with agribusinesses. Encourage farmers to adopt new technologies through policy guidance and support, and participate in the application of new technologies through peasant groups [329]. In addition, combining this science and technology with the United Nations Sustainable Development goals (SDGs) can not only broaden the scope of science, but also promote the development of practice from a broader perspective, for example, by improving agricultural production efficiency and implementing the sustainable development agenda to contribute to the "zero hunger" goal [330]. Although it still faces technical, economic and social challenges, with the continuous updating of the application of RS technology and ML in agriculture, its potential to promote the modernization and sustainable development of agricultural industry will be further tapped.

5. Conclusions

The comprehensive application of RS and ML algorithm can not only promote the development and progress of PA, but also provide a possible solution to the challenges of global population growth, resource shortage and climate change. In some fusion applications of PA, there are significant differences in different types of RS data, among which hyperspectral RS is the most widely used, accounting for more than 30%, while the application of UAV technology has the most potential, accounting for about 24%. It is expected to play a more important role in PA in the future. In addition, the most widely used ML algorithm is SVM, accounting for more than 20%, followed by RF algorithm, accounting for about 18%. It is worth noting that in the future, the rapid development trend of DL integrated platform, multimodal fusion algorithm, cloud computing and edge computing is expected to further promote the progress of PA. Monitoring and identification of crop growth status, pest detection, land or soil management and crop yield prediction are still the main aspects of the comprehensive application of RS technology and ML. However, in obtaining and processing high-quality RS data, improving the interpretability and generalization of the model, and the uncertainty of integration development, we still need to further explore new algorithms and technologies to promote interdisciplinary cooperation and the integration of multi-domain knowledge. Further promote the intelligence and automation of PA, and develop more intelligent agricultural robots, automation equipment and expert systems to promote the sustainable development of PA.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [Thank you very much for your advice and guidance, and we have revised the discussion and conclusions to put more emphasis on how to translate scientific research into practical application throughout the paragraph. The impact of technological progress on agricultural production efficiency and sustainability is emphasized. Some complex sentence structures are simplified to make the content easier to read. More emphasis is placed on the practical application value of the research results and its positive impact on agricultural practice. We earnestly hope to meet the requirements. Thank you again for your guidance.]

Comments 2: The Discussion and Conclusions and the vast majority of citations not being relevant to this manuscript are its primary and very LARGE weaknesses for scientists and practitioners hoping to make use of the science to improve their circumstances. 

Response 2: [4. Discussion

4.1 Current challenges

4.1.1 Acquisition and processing of multi-source RS data

Although there are free and open source medium and low resolution data resources such as MODIS, which provide the basis for scientific research and application, the scarcity of high-quality and high-resolution RS data is still a significant problem. This scarcity mainly stems from the multiple complexities in the process of data acquisition, including the unpredictability of meteorological conditions, the limitations of equipment performance and the impact of complex terrain on the signal. These factors work together on data quality and increase the difficulty and cost of data acquisition. The accessibility, real-time, integrity and privacy protection of RS data are also important factors restricting the sustainable development of agricultural RS [307,308]. In addition, for the processing of RS data, its highly specialized and technology-intensive characteristics can not be ignored. Data preprocessing, multi-source data fusion, and subsequent data interpretation and application, each link requires fine technical operation and profound professional knowledge. Improper handling will not only reduce the accuracy and reliability of data, but also lead to unnecessary waste of resources and loss of efficiency [309,310]. For example, Zhao et al. proposed a framework for robust classification of multi-view RS images under the condition of missing data, which effectively reduces the practical cost. This research not only improves the classification accuracy, but also reduces the uncertainty and deviation in the process of data processing, and helps to achieve more efficient agricultural management decision-making [311].

 It is worth noting that the construction of RS database system, as one of the effective strategies to alleviate the above problems, its importance has become increasingly prominent. For example, studies have shown that the development of a strong and publicly accessible RS water quality database system can effectively improve the efficiency of water resources management and monitoring [312]. In addition, it can not only provide rich and standardized data resources to meet the needs of ML models for large amounts of data, but also promote rapid data retrieval and scientific research sharing, and inject new vitality into agricultural RS research and application. At the same time, with the continuous improvement of the database system, the long-standing problems in the field of agricultural RS, such as insufficient data standardization, limited scale of data sets, different data quality and so on, are expected to be solved gradually [313]. The research shows that the application of multi-source RS fusion technology and high-resolution sensors may bring revolutionary changes to the field of agricultural RS. The progress of these technologies will greatly enrich the data dimensions, improve the monitoring accuracy, and provide strong data support for the practice of PA based on ML [314]. For example, by combining multi-source RS technology, Joshi et al have achieved rapid and large-scale early warning and accurate control of wheat diseases and insect pests [315]. Based on hyperspectral reflectance and satellite multispectral images, Wu et al significantly improved the accuracy of wheat grain water content estimation and successfully reduced the risk of grain loss and additional drying cost [316].

4.1.2 Interpretability and generalization of the model

Generally speaking, ML model is easier to explain than DL model, which is determined by the complexity of the model structure. Although some newer models may improve accuracy, it will also be a challenge for agricultural practitioners to understand and accept these models. For example, although the DL model performs well in many tasks, because of its black box nature, it may be difficult for agricultural workers to understand the decision logic behind it [317]. Therefore, in the application of PA, such as DT or rule-based model, because of its relatively simple structure, it is often easier for agricultural practitioners to understand and accept. For example, Marin et al use multiple decision tree models to quickly understand the types of vegetation diseases and spatially continuous monitoring, thus improving the efficiency of decision-making. In addition, the interpretable model can not only promote the transparency of agricultural decision-making, but also accelerate the knowledge transfer from technical experts to front-line producers, which is very important to improve the science and efficiency of agricultural practice. For example, in a study of precision irrigation, the use of a rule-based ML model greatly promoted water resources management and agricultural policy decision-making. Usually providing information about the features, variables, and algorithms that affect the results of the model is an effective way to enhance the interpretability of the model. For example, Hao et al realized the high-precision prediction of wheat yield by analyzing the key variables of APSIM-Wheat model through Sobol sensitivity analysis. Under the influence of many factors, such as crop growth environment, varieties, soil conditions and climatic conditions, the model shows insufficient ability to extract fine features in the face of new data. It is often necessary to optimize the generalization ability of the model through data enhancement, model integration and the introduction of regularization technology, so as to effectively prevent overfitting and maintain stable performance in the face of new data. For example, Fawakher et al. used a data enhancement strategy for improved segmentation to significantly improve the accuracy of the model for crop and weed segmentation. In addition, regularization technology can help reduce the complexity of the model, thus improving the generalization ability of the model. For example, by applying L2 regularization in the process of model training, the dependence of the model on training data can be effectively reduced and its performance in the face of new data can be more stable [322].

4.2 prospects for the future

4.2.1 Trend of intelligence and automation

With the application of intelligent and automation technology in PA, many problems existing in traditional agricultural models, such as insufficient data set, inaccurate analysis and untimely decision-making, have been solved. This is due to the use of high-precision multi-source RS data, further enhancement of data preprocessing to improve data quality, expansion of data sample diversity and integration of expert knowledge, all of which improve the accuracy of intelligent decision-making. For example, based on expert dialogue and multi-standard decision-making technology, Goodridge et al proposed an expert system that can intelligently diagnose plant diseases, which significantly improves the detection accuracy of plant diseases. In addition, the development of intelligent agricultural equipment and the promotion of farmers' training have lowered the technical threshold and promoted the wide application of new technologies. A study points out that by developing easy-to-operate smart agricultural equipment, farmers can more easily master new technologies and save water and reduce labour demand by promoting the cultivation of different crop types. Improve the profitability of each farm [324]. 

The intelligent fusion of multi-source RS data effectively solves many problems of agricultural data, such as ensuring the consistency of data format, optimizing processing speed, improving the stability of algorithm, and enhancing the generalization ability and interpretability of the model. as a result, the uncertainty in the whole process of agricultural production is reduced. For example, Zhou et al enhanced the generalization ability through the deep migration learning classification model constructed by integrating ground sensor data and UAV data, and realized the intelligent RS recognition of corn straw type [325]. In the future, RS technology and ML are expected to achieve cross-border integration with advanced technologies such as Internet of things (IoT), human-computer interaction visualization, data assimilation and blockchain. This will further promote comprehensive monitoring, yield forecasting and disease surveillance in PA and provide more accurate, efficient and sustainable solutions for agricultural production. For example, intelligent systems that use the Internet of things to track and schedule accurate irrigation have been studied to help farmers plan irrigation effectively and make informed decisions [326].

4.2.2 Data sharing and multidisciplinary interaction:

In the context of global connectivity, international cooperation and data sharing mechanisms are constantly strengthening, and the application of PA should go beyond geographical restrictions. Different countries and regions should work together to share more accurate RS data and smarter ML algorithms to address global agricultural issues such as climate change, food safety and other challenges [327]. In addition, ML algorithm is closely combined with RS technology, and automatic machinery and intelligent robots are used to realize the intelligence and refinement of field management. at the same time, experts in many fields such as agricultural economy, ecology and physics are integrated to form a set of comprehensive agricultural management system. For example, a cloud-based intelligent irrigation system was introduced in one study to optimize irrigation water use through comprehensive big data collection, storage and analysis, significantly promoting informed decisions on water resources management [328]. In addition, strengthening the cooperative research of RS and ML can not only effectively reduce the use of chemical fertilizers and pesticides, accelerate the transformation of agricultural scientific and technological achievements, protect the ecological environment, and promote the green transformation of agriculture, but also is very important to promote the research results of PA to really benefit farmers and realize the transformation from theoretical knowledge to scientific and technological practice. For example, they can use their market channels and technical support capabilities to popularize new technologies by establishing partnerships with agricultural research institutions and working with agribusinesses. Encourage farmers to adopt new technologies through policy guidance and support, and participate in the application of new technologies through peasant groups [329]. In addition, combining this science and technology with the United Nations Sustainable Development goals (SDGs) can not only broaden the scope of science, but also promote the development of practice from a broader perspective, for example, by improving agricultural production efficiency and implementing the sustainable development agenda to contribute to the "zero hunger" goal [330]. Although it still faces technical, economic and social challenges, with the continuous updating of the application of RS technology and ML in agriculture, its potential to promote the modernization and sustainable development of agricultural industry will be further tapped.

5. Conclusions

The comprehensive application of RS and ML algorithm can not only promote the development and progress of PA, but also provide a possible solution to the challenges of global population growth, resource shortage and climate change. In some fusion applications of PA, there are significant differences in different types of RS data, among which hyperspectral RS is the most widely used, accounting for more than 30%, while the application of UAV technology has the most potential, accounting for about 24%. It is expected to play a more important role in PA in the future. In addition, the most widely used ML algorithm is SVM, accounting for more than 20%, followed by RF algorithm, accounting for about 18%. It is worth noting that in the future, the rapid development trend of DL integrated platform, multimodal fusion algorithm, cloud computing and edge computing is expected to further promote the progress of PA. Monitoring and identification of crop growth status, pest detection, land or soil management and crop yield prediction are still the main aspects of the comprehensive application of RS technology and ML. However, in obtaining and processing high-quality RS data, improving the interpretability and generalization of the model, and the uncertainty of integration development, we still need to further explore new algorithms and technologies to promote interdisciplinary cooperation and the integration of multi-domain knowledge. Further promote the intelligence and automation of PA, and develop more intelligent agricultural robots, automation equipment and expert systems to promote the sustainable development of PA.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [Thank you very much for your advice and guidance, and we have revised the discussion and conclusion section to ensure that the discussion and conclusion section revolves more closely around the topic of this article, and the relevant literature cited is more appropriate. At the same time, in the whole paragraph, we pay more attention to how to transform scientific research into practical application, and reorganize the structure of the article. We earnestly hope to meet the requirements. Thank you again for your guidance.]

Response to Reviewer Comments

1. Summary

Thank you very much for taking the time to review this manuscript. We gave a detailed reply below and highlighted the content.

2. Point-by-point response to Comments and Suggestions for Authors

Comments 1: I don’t think that “RS”, “PA” and “ML” were the main searching parameters. Please double check if the authors used only the acronyms. In addition, more details are required for the searching parameters, such as specific years, field of study, etc.

Response 1: [In view of this, firstly, based on the key words "remote sensing" (RS), "machine learning" (ML) and "precision agriculture" (PA), this study makes a preliminary search of the literature in the past 10 years by using databases such as Web of Science, Scopus, Google Scholar and PubMed, and selects more than 12000 related research articles, and makes a quantitative analysis of these articles, as shown in figure 1. We have observed an overall upward trend in the number of publications over the past decade (as shown in figure 1 (a)). Then, in order to ensure the comprehensiveness and depth of the study, we further refine the literature retrieval. On the basis of the preliminary search, we also combined the key words including "agricultural monitoring", "pest detection", "land use and management", "yield forecast" and "agricultural sustainable development". After strict screening, it was finally confirmed that peer-reviewed papers had been published between 2014 and 2024, involving agricultural science, environmental science and related cross-fields. (as shown in figure 1 (b)).] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [As you have pointed out, we not only use "RS", "PA" and "ML" as the keywords for preliminary search, but also based on the keywords "remote sensing" (RS), "machine learning" (ML) and "precision agriculture" (PA). We use databases such as Web of Science, Scopus, Google Scholar and PubMed to conduct a preliminary search of the literature, covering articles published between 2014 and 2024. It is mainly concentrated in agricultural science, environmental science and related cross-fields.]

Comments 2: The authors wrote about “over the past two decades” (line 140), however after that it is stated “published in the past 10 years” (line 145). Please double check the information.

Response 2: [In view of this, firstly, based on the key words "remote sensing" (RS), "machine learning" (ML) and "precision agriculture" (PA), this study makes a preliminary search of the literature in the past 10 years by using databases such as Web of Science, Scopus, Google Scholar and PubMed, and selects more than 12000 related research articles, and makes a quantitative analysis of these articles, as shown in figure 1. We have observed an overall upward trend in the number of publications over the past decade (as shown in figure 1 (a)). Then, in order to ensure the comprehensiveness and depth of the study, we further refine the literature retrieval. On the basis of the preliminary search, we also combined the key words including "agricultural monitoring", "pest detection", "land use and management", "yield forecast" and "agricultural sustainable development". After strict screening, it was finally confirmed that peer-reviewed papers had been published between 2014 and 2024, involving agricultural science, environmental science and related cross-fields. (as shown in figure 1 (b)).] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [I have unified the time range as "nearly 10 years" to ensure the consistency of the text.]

Comments 3: What is the premise to be considered “high quality paper”? Please define it in the text with details, so it can be reproducible. (FYI: be careful on the metrics used, because some journals in the past were not considered “good” and today they are).

Response 3: [In view of this, firstly, based on the key words "remote sensing" (RS), "machine learning" (ML) and "precision agriculture" (PA), this study makes a preliminary search of the literature in the past 10 years by using databases such as Web of Science, Scopus, Google Scholar and PubMed, and selects more than 12000 related research articles, and makes a quantitative analysis of these articles, as shown in figure 1. We have observed an overall upward trend in the number of publications over the past decade (as shown in figure 1 (a)). Then, in order to ensure the comprehensiveness and depth of the study, we further refine the literature retrieval. On the basis of the preliminary search, we also combined the key words including "agricultural monitoring", "pest detection", "land use and management", "yield forecast" and "agricultural sustainable development". After strict screening, it was finally confirmed that peer-reviewed papers had been published between 2014 and 2024, involving agricultural science, environmental science and related cross-fields. (as shown in figure 1 (b)).] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [In order to make the presentation of the paper more reasonable, as you suggested, "some journals were not considered" good "in the past, but today they are considered" good ". Therefore, we have deleted the description of" high quality paper ".]

Comments 4: Figure with low resolution (please improve it)

Response 4:

[] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [We regenerated the charts to ensure that they have a higher resolution]

Comments 5: Check for typos (e.g., “ Landsa-8”)

Response 5: [However, in another study, the use of Landsat-8 images with a spatial resolution of 10 to 30 meters provided a promising solution for disease detection in mixed forests in southern China [63]]. Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [We noticed spelling errors in the text, for example, "Landsa-8" should be corrected to "Landsat-8". We will examine the full text carefully, correct all spelling mistakes and make sure that the correct terminology is used.]

Comments 6: Please avoid using references that are not compatible with both PA and agricultural practices (e.g.,“health assessment of urban trees”)

Response 6: [For example, using daily 10m NDVI data from Sentinel-2 images can quickly, efficiently and accurately monitor the flowering date of apples, and then provide technical reference for accurate classification and growth trend prediction of fruit trees [62].] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [We revise references that are incompatible with PA and agricultural practices (e.g. "Health Assessment of Urban trees"), New literature: 10.1016/j.compag.2024.109260.]

Comments 7: Check for the correct use of terminology (e.g., what is a “drone RS”?)

Response 7: [In a 2023 study by Zhu et al., although the use of Unmanned Aerial Vehicle technology can confirm the importance of red-light bands and adjacent bands, it has not achieved the desired results in the investigation of plant diseases and pests invading vegetation [65].] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.[We examined the use of the term and revised the expression "drone RS"]

Comments 8: Did the authors copy any part of the Figure 3 from another reference? In positive case, please add proper references.

Response 8: [Figure 3 is our original, but as you mentioned, the composition elements in figure 3 come from our corn plot shooting and public data.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.

Comments 9: Figure 4 – image is too stretched to the vertical and resolution must be improved (look at the bottom of “A”).

Response 9:

[] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.[ We reconstructed figure 4 and improved the resolution.]

Comments 10: Please check if it is suitable to use the “Materials and Methods” (M&M) section name, because in the current format, it doesn’t look like a M&M section

Response 10: [Remote Sensing Technology and Machine Learning Methods] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [As you mentioned, we have revised the name of the second part "Materials and Methods" to Remote Sensing Technology and Machine Learning Methods.]

Comments 11: Table 1 – there are missing application for each “model name”. Table must be carefully completed, probably it happened due to the lack of references retrieved based on the searching parameters.

Response 11: [

] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.

Comments 12: Figure must appear after it is announced in the text. (e.g., Figure 6)

Response 12: [As shown in figure 6, Mahanta et al obtained rich spectral features of vegetation based on a variety of sensor devices, and used machine learning models to identify spectral patterns related to specific diseases. Finally, the evaluation of the health degree of in-sects invading the forest was realized, and the detection efficiency was greatly im-proved [152].] Thank you for pointing this out. We agree with this comment. Therefore, We have revised.[Figure 6 has been published in the text]

Comments 13: Lines 576-582 – these lines are antagonistic. Initially, the authors affirms that it can “achieve data standardization” and, in the end, the authors stated the same problem “low degree of data standardization”. Based on what it is written, these lines can be considered confusing to the reader.

Response 13: [The establishment of remote sensing database can not only provide rich and high-quality data resources, meet the urgent needs of ML for a large number of data, but also achieve data standardization, rapid retrieval and scientific research sharing. For example, the existing problems in agricultural remote sensing, such as low degree of data standardization, limited scale of data sets, uneven data quality and so on, are expected to be solved gradually with the continuous improvement of the database. In the future, with the continuous progress and innovation of RS technology, the application of multi-source remote sensing fusion data and higher resolution sensors may bring new agricultural changes [312，313].] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [We have revised the statement of this division to make the logic more reasonable and the expression clearer.]

Response to Reviewer Comments

1. Summary

Thank you very much for taking the time to review this manuscript. We gave a detailed reply below and highlighted the content.

2. Point-by-point response to Comments and Suggestions for Authors

Comments 1: The authors have extensively reworked the article, making it now at a more suitable stage for publication.

Response 1: [First of all, on behalf of all the authors, I would like to express our highest respect to you. Thank you very much for your guidance and assistance with our manuscript. At the same time, we are very fortunate to receive your recognition, and we will continue to work hard. Thank you again]

Comments 2: Some adjustments, such as correcting phrases and terms, are still necessary.

Response 2: [“For example, Marques et al. have overcome the limitation of limited spectral coverage based on UAV in 2024, especially in low light, fog or smoke conditions to achieve real-time, efficient and distributed accurate monitoring [38]. Bah et al. used UAV images to detect weeds in the field in 2017 with an accuracy of more than 90% [39]. Yang et al. used UAV image information to identify rice lodging based on decision tree algorithm in 2017, with an overall accuracy of 96.17% [40].”,’’ In their research in 2024, Ren et al. used the characteristics of UAV to obtain crop growth status quickly and accurately in small and medium-sized areas [67]. By assimilating remote sensing data with WOFOST model effectively by Kalman filter algorithm, the accuracy of yield simulation of different processing schemes is significantly improved, and more accurate and reliable yield prediction information is provided for agricultural producers.”,” Recent scientific research shows that the integration of RS technology and ML methods has brought remarkable progress to the application of agricultural monitoring and identification. RS technology can efficiently obtain crop planting area, growth status and other important information, while ML technology can accurately detect targets and extract features from these rich RS data, so as to achieve fine identification and classification of crops. With the continuous integration and development of these two technologies, the related agricultural problems, such as horticulture line detection [129], crop detection and classification [130,131], vegetation distribution [132,133], will usher in a new turning point. For example, Zhao et al. improved the standardized precipitation evapotranspiration index (SPEI) by integrating RS data, and the results show that the new SPEI can greatly enhance the capacity of agricultural drought monitoring [134]. Lyu et al. used EO-1 Hyperion images, combined with multi-terminal spectral mixing analysis (MESMA) and fully constrained least square pixel mixing (FCLS) techniques, to successfully identify typical vegetation species and improve the accuracy of grassland degradation monitoring [135]. Xiao et al. fused Sentinel-2 and MODIS RS images using the enhanced spatio-temporal adaptive reflection fusion model (ESTARFM), and then accurately obtained the spatial distribution of irrigated rice fields by random forest (RF) algorithm. Based on the Penman-Monteith model and making full use of the daily observation data of the meteorological station, the dynamic monitoring of water resources in the critical irrigation period has been realized, and remarkable results have been achieved [136]. This application increases the feasibility of spatio-temporal fusion of multi-source RS data and makes it possible to continuously monitor the irrigation dynamics of paddy fields on a large scale. ”,” In their latest research, Zhang et al. used GF-1 RS images, combined with advanced multi-scale segmentation algorithms, to improve the accuracy of identifying forest types in Engebe ecological demonstration areas, and used nearest neighbor classification and random forest (RF) classification respectively, and compared the recognition results. The results showed that the effect of random forest classification was better, and the Kappa coefficients obtained in two consecutive years were 0.92 and 0.90 respectively [99]. Through a lot of research, people have reached a consensus: ML algorithms, including RF, support vector machine (SVM), artificial neural network (ANN), decision tree (DT) and so on, have greater potential in agricultural monitoring and recognition, and can significantly improve the efficiency and accuracy of monitoring and recognition [140-147].”,” Mahanta et al. obtained rich spectral features of vegetation based on a variety of sensor devices, and used ML models to identify spectral patterns related to specific diseases. Finally, the evaluation of the health degree of insects invading the forest was realized, and the detection efficiency was greatly improved [152].”,” It is worth mentioning that some researchers have adopted the improved DL algorithm framework for plant disease detection, and achieved remarkable results [184-187]. Dong et al. creatively proposed an effective scale-aware network architecture (ESA-Net) based on low-cost RS images [188]. After strict verification, ESA-Net showed excellent performance in plant disease detection, and achieved strong competitive results. Amarathunga et al. proposed a new architecture based on visual converter, which integrates the attention mechanism driven by domain knowledge and effectively improves the accuracy of micro-pest detection and recognition at the species level [189]. Ye et al. designed an end-to-end automatic disease detection framework based on multi-scale MA-UNet model and single-phase image based on UAV aerial photography data and Landsat-8 satellite RS markers, which greatly improved the efficiency and accuracy of disease monitoring [190].”,” A survey found that the application of various types of RS data provides convenience and opportunities for soil management [200]. At the same time, in different RS soil applications, multi-spectral RS is the most widely used in soil [201]. Duan et al. used the mean value of reflectivity and entropy texture parameters extracted from Landsat- 8 image, combined with MLC, SVM, ANN and RF ML, to identify soil groups in depth, and achieved good results. Zhou et al. proposed a general ML method based on spatio-temporal constraints by using Sentinel-1 and Sentinel-2 data in 2024 [202]. Through verification, its accuracy and practicability have been fully affirmed [203]. Musasa et al. made a detailed review of soil problems in arid environments in 2023, clearly pointing out that the Landsat-8 satellite mission plays an indispensable role in promoting soil assessment and monitoring [204]. In addition, in view of the significant challenges such as insufficient information acquisition and limited measurement accuracy in the early soil moisture monitoring technology [205], the introduction of ML technology is a revolutionary change, which greatly makes up for these deficiencies [206,207]. In addition, high-resolution data show significant applicability in soil applications, especially in soil resource estimation and mapping [208-211]. Moreover, UAV shows great potential in soil analysis and evaluation, and many studies have fully proved its effectiveness in practical application. For example, Bertalan et al., through the mapping of soil moisture based on UAVs, deeply revealed the spatial heterogeneity of soil moisture and provided strong support for PA [212]. Menzies Pluer et al. used UAV to draw the spatial distribution model of farmland soil characteristics and nutrient concentration, which provided a novel and low-cost method and new idea for soil management [213]. In addition, scholars also pointed out that the combination of UAV data fusion and ML is very important for accurate field estimation of soil texture [214-216]. At the same time, in many studies on the integrated application of RS and ML in soil management, we found that the discussion of soil organic carbon and salinity is also an eye-catching direction [217-221].”,” In addition, the crop yield prediction model is also constantly adapting to a variety of new situation changes. For example, although the early traditional ground survey methods and sampling statistics methods based on empirical knowledge have experienced a lot of research and practice, they cannot meet the needs of improving the accuracy of production prediction and reducing costs [295]. With the application of crop growth model and data assimilation model, the yield prediction accuracy will be greatly improved [296]. For example, Zhang et al. and Kheir et al. have made yield predictions based on APSIM crop model, and achieved remarkable results [297,298]. In addition, the WOFOST model also performs well in crop yield prediction, and a number of studies have revealed its potential use in forecasting [299,300]. The SAFY model provides a new perspective and idea for the estimation of crop yield in a large area [301]. However, crop models are not perfect. They may be limited in large-scale applications, errors are easy to accumulate, and there are some problems such as over-fitting [302]. Similarly, ML models may encounter fitting problems in the training process, especially in the case of small data sets or improper feature selection. Fortunately, ML and data assimilation methods provide new solutions to the problems in crop models and ML [303]. By combining RS data and crop model, and with the help of ML optimization, we can not only make up for the shortcomings of the model in some aspects, but also significantly improve the prediction accuracy and enhance the applicability. This innovative method is gradually being widely concerned and favored by researchers [304-306].”] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [According to your opinion, we have re-examined the full text, especially on the sentence structure, words and format, to ensure that all the terms used are accurate, in order to meet your requirements. Thank you again for your guidance.]

Response to Reviewer Comments

1. Summary

Thank you very much for taking the time to review this manuscript. We gave a detailed reply below and highlighted the content.

2. Point-by-point response to Comments and Suggestions for Authors

Comments 1: Please specific the date range instead of using "general term without a reference", by doing that the searching process can be reproducible.

Response 1: [In view of the major challenges posed by global population growth, resource shortage and climate change to traditional agricultural models, the purpose of this study is to explore how to promote the development of PA by integrating RS technology and ML algorithm. In order to achieve this goal, based on the keywords such as "remote sensing (RS)", "machine learning (ML)" and "precision agriculture (PA)", we use databases such as Web of Science, Scopus, Google Scholar and PubMed to search the related literature from 2014 to 2024. We selected more than 12000 related research articles and made a quantitative analysis (as shown in figure 1). The results of the analysis show that the number of related publications showed an overall upward trend during the decade 2014-2024 (as shown in figure 1 (a)). In order to ensure the comprehensiveness and depth of the study, on the basis of preliminary search, we also combine the keywords such as "agricultural monitoring", "detection of diseases and insect pests", "land use and management", "yield prediction" and "agricultural sustainable development" for further screening. After a rigorous screening process, more than 330 peer-reviewed papers published between 2014 and 2024 related to agricultural science, environmental science and related cross-disciplines were identified, from which it can be seen that the number of research papers is also increasing year by year (as shown in figure 1 (b)). In addition, from the perspective of international cooperation and regional distribution, researchers in China, the United States, Brazil and other countries have made significant contributions to the application of remote sensing and machine learning in precision agriculture. However, at the same time, we also note that there is an obvious imbalance in the spatial distribution of these studies, and there are great differences in research contributions among different regions (as shown in figure 2). Therefore, through the in-depth analysis and summary of the existing research results, systematically sort out the application status of remote sensing technology and machine learning in precision agriculture, and discuss the current challenges and possible future development direction in this field. It has important theoretical and practical significance.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [Thank you very much for your careful review and valuable comments on this study. According to your feedback, I have revised the relevant statements in the article to ensure that they are more specific and clear, so as to enhance the repeatability and transparency of the research. It is hoped that such a modification will make the time range of the search process clearer, and other researchers can reproduce our search and analysis process according to this specific time period.]

Comments 2: Check for typos throughout the manuscript (e.g., "key words" or "keywords"?)

Response 2: [“In view of the major challenges posed by global population growth, resource shortage and climate change to traditional agricultural models, the purpose of this study is to explore how to promote the development of PA by integrating RS technology and ML algorithm. In order to achieve this goal, based on the keywords such as "remote sensing (RS)", "machine learning (ML)" and "precision agriculture (PA)", we use databases such as Web of Science, Scopus, Google Scholar and PubMed to search the related literature from 2014 to 2024. We selected more than 12000 related research articles and made a quantitative analysis (as shown in figure 1). The results of the analysis show that the number of related publications showed an overall upward trend during the decade 2014-2024 (as shown in figure 1 (a)). In order to ensure the comprehensiveness and depth of the study, on the basis of preliminary search, we also combine the keywords such as "agricultural monitoring", "detection of diseases and insect pests", "land use and management", "yield prediction" and "agricultural sustainable development" for further screening. After a rigorous screening process, more than 330 peer-reviewed papers published between 2014 and 2024 related to agricultural science, environmental science and related cross-disciplines were identified, from which it can be seen that the number of research papers is also increasing year by year (as shown in figure 1 (b)). In addition, from the perspective of international cooperation and regional distribution, researchers in China, the United States, Brazil and other countries have made significant contributions to the application of remote sensing and machine learning in precision agriculture. However, at the same time, we also note that there is an obvious imbalance in the spatial distribution of these studies, and there are great differences in research contributions among different regions (as shown in figure 2). Therefore, through the in-depth analysis and summary of the existing research results, systematically sort out the application status of remote sensing technology and machine learning in precision agriculture, and discuss the current challenges and possible future development direction in this field. It has important theoretical and practical significance. ”,’’ Figure 1. The changing trend of peer-reviewed papers published in the past 10 years based on keywords retrieval over time. (a) Based on the databases of Web of Science, Scopus, Google Scholar and PubMed, we searched 12000 papers published in the past 10 years; (b)The changing trend of peer-reviewed papers published in agricultural science, environmental science and related cross-fields in the past 10 years based on keywords.’’,” Figure 5. The distribution of the most commonly used machine learning algorithms is obtained based on the keywords "machine learning" and "precision agriculture".

”,” Recent studies have shown that with the improvement of computing performance and the enhancement of massive data sets, ML has shown strong application capabilities in many fields, especially in the field of PA [120,121]. In particular, a series of emerging algorithms and technologies such as deep learning (DL), intelligent optimization, neural networks, computer vision and data enhancement continue to emerge. It has not only injected strong impetus into the field of ML, but also provided rich opportunities for all stages of agriculture. They enable agricultural practitioners to respond more effectively to challenges and achieve set goals [122]. As shown in figure 5, the frequency distribution of the algorithm is obtained by searching the keywords "ML" and "PA". From the chart, we can see that ML is widely used in the field of PA, and support vector machine algorithm has the highest frequency, accounting for more than 20%, followed by random forest algorithm, accounting for about 18%. ”] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [Thank you very much for your careful review and valuable comments on this study. According to your feedback, we carefully examined the entire manuscript and finally made a unified use of "keywords" in the entire manuscript.]

Comments 3: Figure 2 are still presented in low resolution. Please add a high resolution image with at least 300 dpi. In addition, there is a need to change the color bar scale ("the color depth directly reflects the number of research" - the reader can't understand the meaning of that nor have any number in the mind to think of)

Response 3: [

] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [Thank you very much for your careful review and valuable comments on this study. Based on your feedback, we rebuilt the low-resolution images in Figure 2 as high-resolution images. And the details of the composition in the diagram are clearly described. In addition, we modify the misleading expression to "the change in color reflects the difference in the number of studies."]

Comments 4: Please check for the correct use of an acronym or abbreviation (there is a need to write in the complete form in the first appearance)

Response 4: [“Precision agriculture (PA), as a way to realize accurate management and decision support of agricultural production process by using modern information technology, is becoming one of the effective ways to solve these challenges. ”, ” For example, Yomo et al. found in 2023 that the maximum likelihood algorithm based on Landsat-8 remote sensing images is used to classify land use and land cover, and by using multi-layer perceptron-Markov chain modeling method, the results show that the overall accuracy (Kappa coefficient) is as high as 92% [9].”,’’ Therefore, for most countries in the world, it is very necessary to actively promote the coordinated development of PA, remote sensing (RS) and machine learning (ML), ensure agricultural production safety, and strictly abide by the food safety red line. ’’,’’ It is worth mentioning that the emergence of unmanned aerial vehicle (UAV) marks a new era of RS. UAV is a kind of unmanned small aircraft, which is often used to carry remote sensing equipment for aerial data acquisition.’’,” Yang et al. used UAV image information to identify rice lodging based on decision tree (DT), algorithm in 2017, with an overall accuracy of 96.17% [40].”,’’ As the core means of dealing with agricultural remote sensing information, ML model has been widely used and deeply studied in recent years. ML is a data analysis method that allows computer systems to automatically learn patterns and rules from data without explicit programming. Researchers tend to use ML as an integrated framework for feature collection and classification, prediction, or decision support [42]. With the improvement of big data's computing power, many classical algorithms have been optimized and improved, and new models and methods continue to emerge [43]. Common ML methods include DT, support vector machine (SVM) and logical regression (LR). The core of these methods is to find optimized statistical information ways, so as to automatically and efficiently solve practical problems such as classification and regression [44]. In addition, the convolution neural network (CNN) method based on ML has unique advantages in the field of image processing. It can automatically extract deep features from images and achieve accurate classification or recognition tasks [45]. Because of its unique data expression ability, these technologies can learn and extract valuable information automatically, thus effectively avoiding the complexity and subjectivity brought by traditional methods, and greatly improve the efficiency and generalization of processing multi-platform RS data [46]. It is with these advantages that ML has attracted more attention from agricultural researchers and experts, and listed it as the engine factor for the development of PA [47-51].’’,” Recent studies have shown that with the improvement of computing performance and the enhancement of massive data sets, ML has shown strong application capabilities in many fields, especially in the field of PA [120,121]. In particular, a series of emerging algorithms and technologies such as deep learning (DL), intelligent optimization, neural networks, computer vision and data enhancement continue to emerge. It has not only injected strong impetus into the field of ML, but also provided rich opportunities for all stages of agriculture. They enable agricultural practitioners to respond more effectively to challenges and achieve set goals [122]. As shown in figure 5, the frequency distribution of the algorithm is obtained by searching the keywords "ML" and "PA". From the chart, we can see that ML is widely used in the field of PA, and SVM algorithm has the highest frequency, accounting for more than 20%, followed by random forest（RF） algorithm, accounting for about 18%. ”,” In addition, it is very important to grasp the planting and distribution of crops on a large scale in a timely and efficient manner. Although scholars have done a lot of research on the basis of low and medium resolution RS, due to the widespread existence of mixed pixels and the lack of red edge bands, these techniques are difficult to effectively identify small plots of farmland, resulting in unsatisfactory recognition accuracy [137,138]. However, the research of Guo et al. in this field has brought new breakthroughs. Using GF-6 WFV images, they constructed several DT models, which not only efficiently obtained the information of crop planting area and its spatial distribution, but also significantly improved the accuracy of image recognition [139]. In their latest research, Zhang et al. used GF-1 RS images, combined with advanced multi-scale segmentation algorithms, to improve the accuracy of identifying forest types in Engebe ecological demonstration areas, and used nearest neighbor classification and RF classification respectively, and compared the recognition results. The results showed that the effect of RF classification was better, and the Kappa coefficients obtained in two consecutive years were 0.92 and 0.90 respectively [99]. Through a lot of research, people have reached a consensus: ML algorithms, including RF, SVM, artificial neural network (ANN), decision tree (DT) and so on, have greater potential in agricultural monitoring and recognition, and can significantly improve the efficiency and accuracy of monitoring and recognition [140-147].”,” In addition, in the specific application of PA, different algorithms have given full play to their own advantages, and achieved a series of encouraging results. For example, Sladojevic et al. proposed a new plant leaf disease detection and classification model based on deep CNN. The model can accurately identify 13 different plant diseases and effectively distinguish plant leaves from the surrounding environment, which provides a powerful tool for plant health monitoring [123]. Li et al. have made remarkable progress in the field of vegetable disease detection. They propose a lightweight network improvement algorithm based on YOLOv5s. The algorithm effectively eliminates external interference and significantly enhances the ability of multi-scale feature extraction, thus improving the scope and performance of disease detection [124]. Ashwinkuma et al. developed a CNN based on the optimal mobile network, which is used to automatically detect and classify plant leaf diseases. The experimental results show that the CNN model performs well, the maximum accuracy is 0.985, the recall rate is 0.9892, the accuracy is 0.987 and the Kappa coefficient is 0.985 [125]. Yu et al. use DL target detection technology to extract image feature information through complex network structure to achieve non-destructive recognition of crop diseases. Compared with the traditional method, this technique has higher recognition accuracy, faster detection speed and good stability in the visible light range [126]. Ang et al. creatively used Landsat-8 time series satellite images, combined with ML and normalized difference vegetation index (NDVI), successfully developed a new method, which effectively proved its value in yield prediction [127]. Aydin et al. tested gradient lifting methods such as XGBoost, LightGBM and CatBoost for soil sample classification, and achieved high classification accuracy of up to 90%. Compared with previous studies, the prediction accuracy has been significantly improved [128].”,” Through a lot of research, people have reached a consensus: ML algorithms, including RF, SVM, DT and so on, have greater potential in agricultural monitoring and recognition, and can significantly improve the efficiency and accuracy of monitoring and recognition [140-147].”,” A survey found that the application of various types of RS data provides convenience and opportunities for soil management [200]. At the same time, in different RS soil applications, multi-spectral RS is the most widely used in soil [201]. Duan et al. used the mean value of reflectivity and entropy texture parameters extracted from Landsat- 8 image, combined with MLC, SVM, ANN and RF ML, to identify soil groups in depth, and achieved good results. Zhou et al. proposed a general ML method based on spatio-temporal constraints by using Sentinel-1 and Sentinel-2 data in 2024 [202]. Through verification, its accuracy and practicability have been fully affirmed [203].”,” Examples of land management and analysis based on different ML methods include: RF [100,101,231], SVM [102,103,232], DT [90,91,233], maximum likelihood classification [234,235], ANN [97,99,236], CNN [237,242], hybrid multiple model [243,246]. ”,” Commonly used physiochemical parameters include vegetation coverage (FVC) [279], photosynthetically active radiation absorption (FPAR) [280-282], evapotranspiration (ET) [283-285], leaf area index (LAI) [251,286,287], chlorophyll content [288-290], and various vegetation indices (VIs). Such as normalized difference vegetation index (NDVI) [291-293] and enhanced vegetation index (EVI) [294]. These physical and chemical parameters and indexes are not only widely used in actual agricultural production, but also closely related to yield estimation.”] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [Thank you very much for your careful review and valuable comments on this study. Based on your feedback, we carefully examined every page of the article, especially those that might contain abbreviations or abbreviations, and for each abbreviation or acronym, we found where it first appeared. Make sure that the full name is given in that location and that the abbreviation or acronym is attached in parentheses. And avoid mixing full names and abbreviations / abbreviations in the same article.]

Comments 5: Table 1 - check for typos and standard writing (e.g., use of "," or ";")

Response5: [

] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [Thank you very much for your careful review and valuable comments on this study. Based on your feedback, we carefully examined some spelling errors (typos) and improper use of punctuation in Table 1, and we revised the statement in the table in order to use it correctly.]

Comments 6: Figure 6 is not complete.

Response 6: [] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [We have perfected the interpretation of the picture to make it more clear, complete and logical. “Figure 6 shows the pest monitoring process based on different remote sensing data such as IDS maps, MODIS data, Landsat-8 data and drones.”]

Comments 7: Figure 7 needs to have higher resolution. It is not possible to visualize the content of the image.

Response 7: [

] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [We reconstructed figure 7 in order to achieve the required resolution. Thank you again.]

Comments 8: Please carefully check if the use of PA is correct and fit the content. Lines 668-669 "Monitoring and identification of PA [...]". PA is not something to be monitored and identified.

Response 8: [The integrated application of RS technology and ML algorithm can indeed promote the development and progress of PA. In some agricultural fusion applications, the most widely used RS data source is hyperspectral data, and its application proportion is more than 30%. In addition, the rapid development of UAV technology, accounting for about 24%, is expected to play an important role in PA in the future. The most widely used ML algorithm is SVM, accounting for more than 20%, followed by RF algorithm, accounting for about 18%. It is worth noting that the rapid development trend of DL algorithms is expected to further promote the progress of PA. Monitoring and identification of crop growth status, pest detection, land / soil management and crop yield prediction are still the main aspects of the comprehensive application of RS combined with ML. However, the challenges faced by the integration of RS and ML algorithms mainly include the acquisition and processing of high-quality RS data, the improvement of model interpretation and generalization ability, and the uncertainty of integration development. The future development direction is mainly focused on promoting the intelligence and automation of agriculture, strengthening international cooperation and sharing and sustainable transformation of achievements.] Thank you for pointing this out. We agree with this comment. Therefore, We have revised. [We revise this part of the statement to make it more clear and correct. Thank you again for your correction.]