Open AccessSystematic Review

Machine Learning in Sustainable Agriculture: Systematic Review and Research Perspectives

Juan Botero-Valencia

^1,*,†

Vanessa García-Pineda

^1,†

Alejandro Valencia-Arias

^2,†

Jackeline Valencia

Erick Reyes-Vera

Mateo Mejia-Herrera

and

Ruber Hernández-García

^4,5,*

Grupo Sistemas de Control y Robótica, Faculty of Engineering, Instituto Tecnológico Metropolitano—ITM, Medellin 050034, Colombia

Vicerrectoría de Investigación e Innovación, Universidad Arturo Prat, Santiago 1110939, Chile

Instituto de Investigación de Estudios de la Mujer, Universidad Ricardo Palma, Lima 15039, Peru

⁴

Laboratory of Technological Research in Pattern Recognition–LITRP, Facultad de Ciencias de la Ingeniería, Universidad Católica del Maule, Talca 3480112, Chile

⁵

Department of Computing and Industries, Facultad de Ciencias de la Ingeniería, Universidad Católica del Maule, Talca 3480112, Chile

Authors to whom correspondence should be addressed.

^†

These authors contributed equally to this work.

Agriculture 2025, 15(4), 377; https://doi.org/10.3390/agriculture15040377

Submission received: 22 January 2025 / Revised: 4 February 2025 / Accepted: 9 February 2025 / Published: 11 February 2025

(This article belongs to the Special Issue Innovations in Precision Farming for Sustainable Agriculture)

Download

Browse Figures

Versions Notes

Abstract

Machine learning (ML) has revolutionized resource management in agriculture by analyzing vast amounts of data and creating precise predictive models. Precision agriculture improves agricultural productivity and profitability while reducing costs and environmental impact. However, ML implementation faces challenges such as managing large volumes of data and adequate infrastructure. Despite significant advances in ML applications in sustainable agriculture, there is still a lack of deep and systematic understanding in several areas. Challenges include integrating data sources and adapting models to local conditions. This research aims to identify research trends and key players associated with ML use in sustainable agriculture. A systematic review was conducted using the PRISMA methodology by a bibliometric analysis to capture relevant studies from the Scopus and Web of Science databases. The study analyzed the ML literature in sustainable agriculture between 2007 and 2025, identifying 124 articles that meet the criteria for certainty assessment. The findings show a quadratic polynomial growth in the publication of articles on ML in sustainable agriculture, with a notable increase of up to 91% per year. The most productive years were 2024, 2022, and 2023, demonstrating a growing interest in the field. The study highlights the importance of integrating data from multiple sources for improved decision making, soil health monitoring, and understanding the interaction between climate, topography, and soil properties with agricultural land use and crop patterns. Furthermore, ML in sustainable agriculture has evolved from understanding weather data to integrating advanced technologies like the Internet of Things, remote sensing, and smart farming. Finally, the research agenda highlights the need for the deepening and expansion of predominant concepts, such as deep learning and smart farming, to develop more detailed and specialized studies and explore new applications to maximize the benefits of ML in agricultural sustainability.

Keywords:

deep learning; neural networks; precision farming; Internet of Things; PRISMA 2020

1. Introduction

Machine learning (ML) algorithms have transformed resource management in industries, including agriculture, by analyzing vast amounts of data and creating precise predictive models. These technologies enable the identification and analysis of complex patterns in data from various sources, such as remote sensors and agricultural equipment, which facilitates informed and data-driven decision making [1]. In sustainable agriculture, ML is used to improve the accuracy in predicting the nutritional indices of crops, which is essential for the efficient application of fertilizers and the management of plant health [2]. Also, it has been shown that using unmanned aerial vehicles (UAVs) and ML techniques together allows the prediction of the nitrogen nutrition index in rice [3], which not only optimizes the use of resources but also contributes to the reduction of negative environmental impacts associated with the excessive use of fertilizers [4].

Precision agriculture improves agricultural productivity and profitability while reducing costs and environmental impact, with one of its pillars being crop-specific management, which adjusts inputs according to field variability. In this context, artificial intelligence and ML are used to analyze geospatial data for optimizing input utilization [5]. In the context of precision agriculture and environmental protection, using ML and smart materials to detect and separate herbicides like glyphosate is gaining relevance. Thus, combining ML with advanced detection technologies offers new opportunities to address these challenges effectively and sustainably [6].

Moreover, integrating multi-source and multi-temporal data through ML techniques has also proven effective in improving the mapping of crop types. Sun et al. [7] found that the use of remote sensing data and ML algorithms can increase the accuracy in identifying different kinds of crops in subtropical agricultural regions, marking a fundamental advancement for agricultural planning and the implementation of sustainable practices that adapt to the specific conditions of each region [7]. Integrating these technologies with Internet of Things (IoT) devices, including sensors, cameras, and irrigation tools, significantly improves the management and development of sustainable agriculture [8]. Thus, agricultural automation is becoming a key factor for the economy, creating a new trend and growing focus of research [9].

In this way, sustainable agriculture and ML have been used in various applications to improve the accuracy and efficiency of agricultural practices. A relevant example is the use of ML algorithms, such as decision trees, support vector machines (SVMs), and multilayer perceptron (MLP) neural networks, to classify the spacing between plants in soybean fields using images captured by unmanned aerial vehicles (UAVs). In this case, the SVM and MLP models demonstrated superior performance compared to the decision tree, achieving high precision and recall, highlighting the applicability of these methods for precise monitoring of soybean crops during their growth stage [10].

Another notable ML application in sustainable agriculture was a study on estimating aboveground biomass (AGB) in soybean crops using features derived from UAV images. This study evaluated different ML algorithms, such as decision trees, Random Forests, neural networks, and Extreme Gradient Boosting (XGBoost), and found that the fusion of spectral, textural, and structural UAV features significantly improved the accuracy of AGB estimates. In particular, the ensemble learning (EL) model outperformed the other algorithms in predicting AGB, showing consistent and accurate performance across various crop growth stages, demonstrating the importance of integrating multi-source data for precision agriculture [11].

Furthermore, ML has shown its potential in early disease detection in plants, as evidenced by the case of virus identification in tomato plants. Using Raman spectroscopy combined with ML techniques, it was possible to discriminate between infected and healthy plants in the early days after virus inoculation, achieving an average accuracy of between 90% and 95%. This approach not only improved early infection detection but also enabled robust validation through data preprocessing and discriminant analysis models. This advancement highlights how ML can be key to efficient disease management in crops, supporting more sustainable agricultural practices [12].

Similarly, using ML combined with the IoT is transforming smart agriculture. Recently, researchers have explored how these technologies can automate various agricultural tasks, such as monitoring soil conditions [13], weather [14], irrigation management [15], and pest control [16]. These applications reduce farmers’ workloads, improve the precision and effectiveness of agricultural practices, and contribute to sustainability [17]. Additionally, the combination of ML with UAVs has significantly enhanced crop yield estimation for crops such as wheat. A recent study optimized an ML-based model using UAV datasets and hyperparameter tuning, achieving notable improvements in metrics like RMSE, MAE, and R². The Random Forest and XGBoost algorithms outperformed other models, and a statistical test confirmed their superiority. Furthermore, through Shapley additive explanations, key variables such as the chlorophyll index and wheat height were identified, contributing to the model’s increased accuracy. This optimized approach demonstrated 7–12% greater accuracy compared to traditional linear models, highlighting ML’s potential to advance precision agriculture [18].

In addition, ML enables early prediction of crop yields at the field level, which is critical for agricultural planning and management, providing farmers with valuable tools to make informed decisions and maximize productivity [19]. Integrating geographic information systems (GISs) and ML facilitates more sustainable and efficient management of agricultural lands [20]. The prediction of cropland suitability based on GIS and ML techniques represents a novel practice for sustainable agricultural production [5,21]. These technologies allow for identifying optimal crop areas, optimizing resource use, and minimizing environmental impact.

In high-productivity regions, such as northwest India, interpretable ML methods have proven helpful in explaining the variability in crop yields in the field. These techniques allow farmers to understand better the factors influencing crop yields, which is essential for optimizing agricultural practices and improving sustainability [22]. Also, optical, structural, and thermal measurements from unmanned aerial vehicles have been integrated to model the fraction of PAR absorption (fAPAR) in winter wheat, providing key information for crop management and decision making [23].

Similarly, the implementation of ML techniques has shown successful results in various tasks, such as identifying areas suitable for agriculture. A notable example took place in Colombia, where a neural network-based model trained with high-resolution satellite images from the Sentinel-2 system predicted the most appropriate areas for agriculture with 85% accuracy in regions like Barichara and El Socorro in Santander. This system has significantly optimized the traditionally lengthy and costly process of identifying agricultural lands and could be used nationwide to support decision making on land use for specific crops [24].

Another successful ML application in precision agriculture can be seen in using remote sensing platforms, such as drones and ground robots, to provide detailed measurements of crop conditions. These technologies, combined with the use of soil sensors to measure properties like moisture and nutrients, enable more precise application of agricultural inputs, resulting in greater efficiency and lower costs. For instance, ML-powered targeted spraying systems can apply fertilizers and pesticides only where needed, reducing waste and environmental impact (GAO-24-105962) [25].

In Peru, the National Institute of Agricultural Innovation (INIA), under the Ministry of Agrarian Development and Irrigation (MIDAGRI), has successfully implemented precision agriculture through the use of drones with multispectral cameras. This technology allows detailed crop monitoring, promoting conservation and improving quality and yield. During an internship organized by INIA, technicians, and specialists were trained in using tools such as sensors and geospatial data analysis, aiming to foster innovation in agricultural management and strengthen technology transfer to producers [26].

Additionally, at the international level, ML solutions using remote sensing data have been successfully implemented to optimize agricultural management. A notable example involves using Sentinel-2 satellite images and the Cropland Data Layer (CDL) to predict crop yields. This approach enables large-scale predictive analysis, facilitating real-time decision making and improving the accuracy of crop maps. Training courses on ML and remote sensing, such as those organized by NASA ARSET, equip participants with the tools needed to train predictive algorithms on geospatial datasets, thereby contributing to effective agricultural monitoring [27].

As reported in the literature, the application of ML in sustainable agriculture has become increasingly important in recent years due to its ability to improve the precision and efficiency of agricultural practices [28]. However, implementing ML in agriculture faces several challenges, including the management of large volumes of data and the need for adequate infrastructure [29]. A significant challenge lies in integrating and analyzing extensive agricultural data, requiring advanced technologies and a deep comprehension of farming practices and crop biology. Additionally, the accuracy and reliability of predictive models depend significantly on the data quality and representativeness [30]. Despite significant advances in ML applications for sustainable agriculture, important gaps still exist, and there remains a notable absence of deep and systematic comprehension in several areas [31].

Although ML applications in crop production have demonstrated promising outcomes, the integration of various data sources, the adaptation of specific models to local conditions, addressing the variability in model accuracy, and the development of more interpretable and robust algorithms are challenging research areas [31,32]. On the other hand, in a systematic review on ML-driven remote sensing applications in agriculture, the need for more studies focused on the validation and scalability of these models in different agroecological contexts was identified [33].

Consequently, the application of ML in sustainable agriculture has grown significantly, driven by technological advancements and the need for more efficient and sustainable farming practices. In this context, precision agriculture has greatly benefited from using algorithms such as decision trees, support vector machines (SVMs), neural networks, and data fusion techniques, enabling farmers to make informed decisions about crop management. However, despite this growth, several gaps remain in the systematic understanding of its impact and potential.

This systematic literature review aims to identify research trends in the use of ML in sustainable agriculture, with a particular focus on the most relevant applications, current challenges, and future opportunities. Previous and current trends are explored not only from a technological perspective but also regarding the methodological advancements that can facilitate the scalability of the proposed solutions. The present review aims to answer the following questions:

Which years have seen the most interest in using ML in sustainable agriculture?
What type of growth does the number of scientific articles on the use of ML in sustainable agriculture present?
What are the main research references on using ML in sustainable agriculture?
What is the evolution of topics in the scientific literature on the use of ML in sustainable agriculture?
What are the main topic clusters on using ML in sustainable agriculture?
What are the growing and emerging keywords in the research field of ML use in sustainable agriculture?
Which topics are emerging as key players in designing a research agenda on the use of ML in sustainable agriculture?

Therefore, the main contributions of this study are threefold: (i) analyzing current research trends on the use of ML in sustainable agriculture, (ii) identifying the most relevant and emerging areas of research, and (iii) evaluating the persistent challenges affecting the large-scale integration of these technologies. Unlike previous reviews, this study provides a detailed overview of the main thematic clusters shaping the direction of research, highlighting theoretical and practical implications, and discussing current research limitations and gaps.

The rest of the paper is organized as follows. Section 2 presents the methods and techniques used to perform the bibliometric analysis, including the source databases consulted, the inclusion and exclusion criteria, and the selection and collection processes. Section 3 provides a results overview, analyzing publication trends, temporal evolution, and the geographical distribution of the related research. Section 4 analyzes the results in the context of the existing literature, highlighting the theoretical and practical implications and research limitations and perspectives. Finally, the Conclusions summarize the main findings of the study in Section 5.

2. Methodology

The bibliometric analysis carried out in this study performed an initial exploration using secondary sources. We conducted the study by performing a systematic review following the PRISMA 2020 statement [34], the process diagram of which is presented in Figure 1. Applying these parameters ensures transparency in the collection, selection, and synthesis of the relevant literature on machine learning in sustainable agriculture. Thus, the following sections describe the different processes followed in the methodology to collect and analyze the related literature.

2.1. Eligibility Criteria

The inclusion criteria were based on keywords and titles as the main metadata, explicitly focusing on articles with the combination of “machine learning” and “sustainable agriculture” with reported outcomes of interest and published in English. These criteria allowed for identifying the most relevant and representative studies in the field. Additionally, we considered different citing methods to provide comprehensive coverage of the existing literature.

The exclusion process was carried out in three distinct phases, as described by PRISMA 2020. In the first phase, all records with erroneous indexing were excluded to ensure the accuracy of the analyzed data. Secondly, we excluded all documents to which full-text access was unavailable, but this stage only applies to systematic literature reviews. Finally, we discarded documents with incomplete indexing or from conference proceedings to maintain a coherent and high-quality dataset. We did not restrict the study location or dataset size.

2.2. Information Sources

We conducted electronic searches for eligible studies on the Scopus and Web of Science databases, both recognized as the leading academic databases due to their extensive coverage and high quality of indexed content. These databases provide access to a large number of peer-reviewed articles and are considered essential for conducting bibliometric analyses, providing a relevant and reliable collection of academic publications [35]. In addition, we performed a ‘snowball’ search to identify additional studies by exploring the reference lists of all eligible reports. Thus, we made a forward citation tracking using Google Scholar to screen further eligible studies.

2.3. Search Strategy

As a search strategy, we designed two specialized search equations to perform the search in the two selected databases. These equations were formulated to meet the defined inclusion criteria and adapt to the specific search characteristics of each database. Therefore, we ensured that the queries conducted in Scopus and Web of Science effectively captured relevant studies on using machine learning in sustainable agriculture. In the following, we detail the search equations for each database:

Scopus database: TITLE-ABS (“machine learning” AND (“sustainable agriculture” OR agriculture OR “agricultural production”)) AND (TITLE-ABS-KEY (“crop* management”)).
Web of Science database: TS = ((“machine learning”) AND (“sustainable agriculture” OR agriculture OR “agricultural production”)) AND (TS = (“crop* management”)).

2.4. Data Management

We used Microsoft Excel^® (version 16.93.1) as a data management tool to extract, store, and process the information collected from each database, allowing us to manage the collected bibliometric data efficiently. Additionally, the research team used the free software VOSviewer^® (version 1.6.20) to create bibliometric maps, facilitating the visualization of relationships and trends in the literature [36]. To obtain an accurate and clear presentation of the results, we generated graphs of the different bibliometric indicators using both VOSviewer^® and Microsoft Excel^®.

2.5. Selection Process

According to the PRISMA 2020 statement, it is crucial to mention whether an internally derived automatic classifier was used to assist in the selection process, as well as performing internal or external validation to understand the risk of lost studies or incorrect classifications [34]. In the present study, we used Microsoft Excel^® automation tools as an internal tool. The researchers developed the internal selection tool and applied it independently in implementing the inclusion and exclusion criteria, aiming to reduce the risk of lost studies or incorrect classifications through the convergence of results.

2.6. Data Collection Process

It is also important to specify the methods used to collect data from the reports, specifying how many reviewers collected data, whether they worked independently, and, if applicable, details of the automation tools used in the process [34]. We used Microsoft Excel^® as an automated tool for the data collection process, with all researchers participating as reviewers in terms of validation and working independently. In addition, the data confirmation process was performed collectively until an absolute convergence of results was achieved, ensuring the accuracy and consistency of the data collected.

We imported into Microsoft Excel^® the identified articles from database searches and reference list checking, saving the primary metadata of each one, such as titles, authors, abstracts, keywords, and source titles, among others. Therefore, we removed duplicate reports and those citations that met the exclusion criteria. Finally, we retrieved the full text of all potentially eligible studies.

2.7. Data Items

We listed and defined all outcomes for which data were collected, specifying that all outcomes consistent with each domain result in each study were selected, including all relevant measures, time points, and analyses. If not all outcomes were collected, we detailed the methods to decide which results to include. Furthermore, all other variables for which data were searched, such as participant and intervention characteristics and funding sources, were listed and defined. Finally, we described the assumptions about any missing or unclear information.

We searched for all articles that met the research objective, satisfying the specialized search equation designed for each database; this included all articles mentioning the use of machine learning in sustainable agriculture. However, if there was missing or unclear information, we excluded it under the category of “non-relevant texts” as they do not allow us to understand the knowledge foundation on the topic adequately, thus ensuring consistency with the purpose and scope of the research.

2.8. Study Risk of Bias Assessment

As all the authors performed the data collection process, we conducted the bias risk assessment using the automated tool Microsoft Excel^® in the same manner. All authors independently applied the bias evaluation, which guaranteed the quality and integrity of the obtained results and ensured a detailed and unbiased review of the included studies. Each researcher recorded details and justifications for the risk of bias concerns. All authors discussed discrepancies or uncertainty about judgments of bias risk to achieve a consistent consensus.

2.9. Effect Measures

The risk ratio or the mean difference is commonly used to specify the effect measures of a successful outcome in the synthesis or presentation of the results. Contrary to primary research, it is worth highlighting that the present study is based on secondary data sources. Given the research questions being investigated, we examined the included reports using Microsoft Excel^® for data management and analysis along three dimensions: (1) number of publications, (2) number of citations, and (3) temporal usage of each keyword. Additionally, existing nodes were evaluated to determine thematic association using the VOSviewer^® software.

2.10. Synthesis Methods

In the present research, we used the bibliometric indicators of quantity, quality, and structure, as suggested in [37]. These indicators were applied automatically using Microsoft Excel^® to all documents that passed the three exclusion phases. The methodology used not only ensured precision but also provided a robust framework for evaluating bibliometric data. Structuring the analysis in this way significantly enhances our ability to visualize results clearly and draw meaningful insights.

2.11. Reporting Bias Assessment

It is crucial to consider possible biases in the terms used during the search and data collection to assess the risk of bias due to missing results that may arise from reporting biases. In the present study, we observed a bias for certain synonyms found in the IEEE thesaurus, which is reflected in the inclusion criteria, the search strategy, and the data collection. Additionally, documents with incomplete indexing and conference proceedings were excluded, as they might omit valuable information for knowledge construction, reported outcomes, or trial protocols. We considered these factors to minimize the risk of bias and ensure the integrity and comprehensiveness of the synthesis conducted.

2.12. Certainty Assessment

The assessment of certainty or confidence in the body of evidence was approached differently from primary studies that examine certainty individually. Instead of assessing the certainty of each single outcome separately, we conducted an overall assessment based on the independent application of inclusion and exclusion criteria and the definition of bibliometric indicators. We classified the certainty of evidence for each outcome as high, moderate, low, or very low. In addition, we noted possible biases defined in the methodological design. The study’s limitations are mentioned in the discussion phase to ensure a comprehensive assessment of the quality of, and confidence in, the body of evidence obtained.

3. Results

In this section, the findings obtained from the bibliometric analysis are presented. We examine the results of identifying and selecting articles, analyzing publication trends, temporal evolution, and thematic distribution. It provides an overview of the development and relevance of research in this area, highlighting the main areas of interest and the most significant contributions.

Figure 2 depicts the overall methodological processes following the PRISMA guide. The initial part of the process of this bibliometric study consisted of identifying records by applying a specific search strategy to each of the selected information sources, followed by removing duplicate records. Subsequently, we applied the three exclusion phases described earlier. Finally, a total of 124 articles that met the established criteria were included in the analysis, which form the basis of this bibliometric study on the use of machine learning in sustainable agriculture.

3.1. Publication Growth, Main Authors, Journals, and Countries

As shown in Figure 3, we observed a quadratic polynomial growth in the publication of articles on the use of ML in sustainable agriculture, with a notable increase of 91.49%. This growth can be attributed to advancements in artificial intelligence models and increased awareness of the importance of sustainable agriculture. The years with the highest number of publications were 2024, 2022, and 2023, demonstrating a growing and sustained interest in the field, which reflects the relevance and necessity of research for improving sustainable agriculture through ML technologies.

Among the most representative articles, due to the number of citations obtained in the years of highest productivity, is the publication by Diaz-Gonzalez et al. [38], who conducted a review of recent research on crop yield prediction using estimates of chemical, physical, and biological indicators of soil quality. These studies integrate various ML techniques to analyze data obtained through remote sensing systems. In 2023, Raman et al. [39] proposed an integrated method to transform agriculture by combining robots, the Internet of Things, and environmental protection, using convolutional neural networks (CNNs) with cameras on autonomous robotic platforms to identify types of crops and weeds for focused and efficient treatment. Recently, Attri et al. [40] published an article presenting a literature review on the ML algorithms used in agriculture.

Thus, the increase in publications during the mentioned years may be linked to the greater integration of ML in agriculture, reflected in the implementation of various technological solutions, ranging from agronomic data analysis to the automation of agricultural processes. The role of advanced ML techniques in improving yield prediction and crop health monitoring is explored in Section 4.3 and Section 4.8. Moreover, the need to address emerging challenges stemming from climate change and the demand for sustainable food production has driven research in this area, integrating ML techniques to optimize crop management and enhance the sector’s sustainability [38,39].

On the other hand, the growing trend in publications may also reflect the impact of inter-institutional collaboration and the use of data platforms such as Scopus and Web of Science, which facilitate greater visibility and dissemination of advancements in this field. The support of government institutions and international organizations promoting research on technological solutions for agriculture has also played a crucial role in boosting productivity in article publication.

In terms of the main authors, three groups of authors were identified, as shown in Figure 4. The first group includes those who stand out in both productivity and impact. Benos et al. [1] conducted a literature review on ML in agriculture and found that crop management was the central focus in the scientific literature. The second group comprises authors who, although they have a low scientific productivity index, position themselves as references due to their high impact. Researchers like Pearson S and Busato P [31] have as their only development work with Bochtis, which has 1593 citations. Finally, the third group of authors is characterized by their high scientific productivity, although they do not have many citations, with Miao YX [4] being the most notable.

Figure 5 shows the leading journals in terms of the number of publications and citations, where three groups were also identified. The first group consists of journals known for their high productivity and impact, such as Computers and Electronics in Agriculture and Remote Sensing. The second group consists of journals positioned as benchmarks due to their high impact, although they have a low scientific productivity index, e.g., Sensors. Lastly, the third group of journals is characterized by their high scientific productivity, although they do not have a high number of citations, with Agronomy being the most notable.

For the most relevant countries, we recognized two distinctive groups based on criteria such as research output, impact, and contributions to the field of ML in sustainable agriculture, as shown in Figure 6. The first group includes the countries that stand out in productivity and impact, led by China and India, which exhibit high scientific output and significant influence in the field. The second group includes countries with lower scientific productivity indexes but who position themselves as leaders due to their high impact, e.g., the United Kingdom, Greece, and the United States.

3.2. Topic Evolution and Main Keywords

Furthermore, we investigated the evolution of topics related to the use of ML in sustainable agriculture in the literature by analyzing the most used keyword in each year of research between 2007 and 2025. Figure 7 identifies key themes and trends per year to understand the progression of research in the field. In 2007, the first year of the analysis, the emergence of concepts including weather data stands out, which refers to incorporating historical meteorological data through weather prediction applications utilizing sophisticated algorithms [41]. Between 2011 and 2018, terms related to the development of algorithms began to emerge for identifying and classifying data related to agricultural production and management [42,43], namely, Random Forest and time-series analysis. In recent years, topics such as the IoT, artificial intelligence, remote sensing, and smart farming have been predominant, reflecting recent research trends in this field.

It is important to note that the evolution of keywords reflects not only the growth of ML in agriculture but also the progressive integration of technology across different phases of the production chain. In the early years (2007–2010), the most repeated terms were primarily related to the use of meteorological data for weather prediction, as shown by the prominence of “Weather Data” in 2007. At that time, research focused on developing models to forecast weather conditions and consequently improve crop planning. However, as computational power and the accuracy of predictive models improved, terms like “Random Forest” and “Time Series Analysis” emerged between 2011 and 2018. This reflects a transition toward developing more sophisticated algorithms for analyzing agricultural data. This shift was driven by an increasing need to understand complex patterns in agricultural production and to manage resources efficiently, enabling more precise predictions of crop yields, disease detection, and irrigation planning.

In recent years (2019–2025), there has been a predominance of terms like “Internet of Things”, "Artificial Intelligence”, “Remote Sensing”, and “Smart Farming”, highlighting the convergence of advanced technologies in sustainable agriculture. These terms indicate a clear shift toward integrating intelligent solutions, where ML is combined with other emerging technologies to enhance the efficiency and sustainability of agricultural systems. The rise of the “Internet of Things” stems from its ability to connect sensors, devices, and platforms, allowing for real-time, precise management of variables such as weather, soil moisture, and crop health. Meanwhile, the inclusion of “Artificial Intelligence” reflects the growing interest in leveraging autonomous systems and advanced algorithms to make data-driven decisions from large datasets. The relevance of “Remote Sensing” has also increased, as this technology has become crucial for non-invasive crop monitoring using satellite or drone imagery. Lastly, “Smart Farming” encapsulates the convergence of all these technologies, emphasizing a holistic approach to optimizing agricultural systems and promoting more sustainable, precise, and efficient farming practices.

A keywords co-occurrence network is presented in Figure 8 through a total of seven thematic clusters. These clusters play a crucial role in illustrating the interconnectedness of various concepts and themes, providing valuable insights into the relationships between different research areas in sustainable agriculture and machine learning. The blue cluster is the most prominent, composed of terms such as deep learning, crop management, crop disease detection, CNN, image processing, and resource allocation, which together refer to works related to the use of machine learning applications in agricultural production systems and crop management [31]. These terms also refer to using deep learning and image processing in detecting plant diseases [44]. The green cluster includes terms such as smart agriculture, data mining, sensors, IoT, and crop recommendation, referring to machine learning for data mining and analysis in precision agriculture [45]. The network identifies other clusters in light blue, yellow, orange, purple, and red, representing different elements of conceptual affinity in the field of study.

The present research proposes a Cartesian plane in Figure 9 that measures the frequency of keyword usage on the x-axis and the relevance of usage on the y-axis, allowing for observing four different quadrants. This plane provides a clear view of how keywords are distributed and evolve in the studied research topic. Each quadrant represents distinct categories of keyword usage, influencing the interpretation of relevance and frequency in ML applications for sustainable agriculture. Quadrant I comprises consolidated and growing concepts such as deep learning, smart farming, remote sensing, and artificial intelligence, identifying current trends and predicting future research directions in this field. In quadrant II, we observe infrequent but highly relevant keywords considered emerging, like neural networks, data mining, Internet of Things, and yield prediction. Quadrant III contains unconsolidated concepts, while quadrant IV contains declining concepts.

3.3. Representative State-of-the-Art Approaches on ML for Sustainable Agriculture

The term IoT primarily refers to using IoT technology to manage, maintain, and develop agricultural systems. Rodríguez et al. [46] implement an intelligent agriculture system based on a three-layer architecture (agricultural perception, edge computing, and data analysis) to monitor climatic variables, estimate coffee production, and configure IoT infrastructure. Edge computing mitigates connectivity challenges in agricultural IoT by enabling local data processing, reducing costs, and supporting real-time decision making. Applications include estimating coffee production through a three-layer system architecture [46], detecting corn leaf diseases using TinyML [47], managing irrigation with microcontrollers [48], and classifying olive varieties by using CNNs on ESP32-S3 [49].

In [46], a system based on a three-layer architecture is proposed. In the first layer, known as perception agriculture, technologies such as Libelium and Ómicron are analyzed in terms of price and the number of devices supported. In the second layer, corresponding to edge computing, interpolation algorithms, and machine learning techniques are examined for outlier removal. Experiments were conducted using DBSCAN, Isolation Forest (IF), and SVM, with the best results obtained using IF, achieving 99% precision. The approach combined cubic splines to replace outliers, resulting in an RMSE of 0.085. In the data analysis layer, different ML methods were tested to estimate coffee production based on a series of variables previously established by expert coffee growers. The estimation was performed using five models: Decision Tree Regressor, artificial neural network, XGBoost, Support Vector Regression, and Random Forest. The performance was compared based on metrics such as RMSE, MAE, and RSE, with XGBoost delivering the best results (RMSE: 0.008, MAE: 0.032, and RSE: 0.585).

Additionally, a system based on TinyML integrated into an Arduino Nano 33 BLE Sense microcontroller has been developed for maize leaf disease detection and classification through images using an advanced CNN. The datasets used for this purpose were sourced from Kaggle, Harvard Dataverse, and Mendeley. Tests for detecting maize plant diseases were conducted using Edge Impulse and TensorFlow Lite, two leading platforms, analyzing ease of use, flexibility, scalability, and reliability. Both platforms were executed with TinyML on the Arduino Nano 33 BLE Sense. The results show that TensorFlow Lite presents a slightly higher level of complexity in usability but also allows for the development of more tailored solutions with specific features and access to a broader variety of models. On the other hand, Edge Impulse is somewhat easier to use but has certain limitations due to its reliance on third-party services. The accuracy results were 97% for TensorFlow Lite and 96% for Edge Impulse. However, despite similar metrics, edge impulse demonstrated lower computational cost, reflected in the system latency: 76.48 ms for Edge Impulse compared to 84.99 ms for TensorFlow Lite [47].

In [48], a system was implemented that combines humidity, temperature, soil moisture, and lighting level sensors to manage plant irrigation using a Keras classifier, an open-source library for machine learning and deep learning, and Edge Impulse for database creation. The system was executed on a Wio Terminal microcontroller, based on the ATSAMD51 with an ARM Cortex-M4F core, leveraging the Edge Impulse framework for data collection and training. Similarly, in [49], a system was developed using a CNN model to classify the quality and category of olive fruits, thereby enhancing the efficiency of the post-harvesting process for olives. This system employed TinyML on an ESP32-S3 microcontroller, achieving an accuracy of 95% in classifying four categories. The term remote sensing refers to using multi-temporal and multi-source remote sensing data to improve the monitoring of different types of crops [7]. Smart farming mainly focuses on the role of smart agriculture through the collection of information on the productive management of crops, for example, by applying different technologies such as IoT, drones, machine learning, 3D mapping, and virtual and augmented reality, among others, for the precise and efficient understanding and management of the food supply chain [50,51].

One of the most widely used techniques in agricultural tasks is the set of machine learning methods known as deep learning. These models function similarly to the human brain, using neural networks to learn patterns from large datasets. This type of model has gained significant traction in the field of computer vision due to its ability to identify and classify patterns from images and videos. These features have been extensively exploited in agricultural tasks, such as detecting diseases or pests in plants, estimating growth stages, classifying plant varieties, or monitoring harvestable fruits—often using images as the model’s input data. This is supported by studies such as [52,53,54,55,56].

However, given the numerical nature of problems like production estimation or remote sensing, such tasks are often treated as regression problems. In this context, convolutional neural networks (CNNs) have shown promising results. Additionally, for tasks such as estimating groundwater availability, some authors report better outcomes using the Normalized Difference Vegetation Index (NDVI), which indicates vegetation quantity and greenness and correlates with the presence of water. The NDVI, however, requires spectral images to analyze the amount of red and infrared light reflected by vegetation [52].

On the other hand, numerical estimation models are commonly used to predict a specific state or variable necessary for controlling crop growth or ensuring appropriate crop rotation to maintain soil health. Applications in this category include irrigation estimation, supplemental lighting, productivity forecasting, and fertilizer management to optimize crop production [57]. In these cases, there is a trend toward using computationally less expensive techniques than those described earlier, aiming to integrate predictive models into economic control systems, such as microcontrollers, to address challenges like energy consumption and connectivity often encountered in rural areas. Common techniques in this area include Naive Bayes, Decision Trees, regressions such as Random Forest Regression or Linear Regression, Random Forest, support vector machines (SVMs), and XGBoost [46,58,59,60]. These approaches achieve varying results depending on the application or crop type, with accuracies exceeding 95% in most cases [61].

4. Discussion

In the following, we discuss the obtained results in detail to analyze the practical implications of the findings and the limitations of the study. The main research gaps are also identified, and future research directions are proposed, highlighting key areas that require greater attention to advance the knowledge and application of ML in sustainable agriculture.

4.1. Growth of Scientific Publications on the Use of Machine Learning in Sustainable Agriculture

In the analysis of scientific publications on the use of ML in sustainable agriculture, the years with the highest research activity were 2024, 2022, and 2023. These years were crucial for advancing ML applications in agricultural management and environmental monitoring, reflecting a growing and sustained interest in this study area.

In 2024, the application of ML in crop management stands out [40], providing a comprehensive view of how these technologies are transforming agriculture. The research addresses various applications, including pest management, crop yield prediction, and resource use optimization, highlighting the importance of integrating data from multiple sources to improve decision making in the agricultural field. The work also emphasizes the need for advanced and precise tools to tackle the challenges of modern agriculture and enhance the sustainability and efficiency of the agricultural sector.

In 2022, Diaz-Gonzalez et al. [38] reviewed the application of ML and remote sensing techniques to estimate soil indicators. The study is significant because it addresses the need to monitor and manage soil health, a crucial factor for agricultural sustainability. The authors highlight how the combination of remote sensing data and ML algorithms can provide accurate estimates of soil properties, such as moisture and fertility, which is essential for efficient crop planning and management, demonstrating the potential of these technologies to improve the accuracy and effectiveness of environmental monitoring and agricultural management.

Lastly, in 2023, the interaction between climate, topography, and soil properties with agricultural land use and crop patterns was investigated using remote sensing data and ML methods [13]. The study is relevant because it integrates multiple environmental factors to understand better how they affect land use and agricultural practices. The results provide a solid foundation for developing more sustainable and adaptive agricultural management strategies, highlighting the importance of considering a wide range of variables in agricultural planning.

The significant growth in the scientific literature on the use of ML in sustainable agriculture is clearly observable in the data on the annual number of publications. As shown in Figure 3, scientific output has experienced a remarkable 91.49% increase in recent years, with particular focus on the years 2022, 2023, and 2024. This growth can be explained by the advancement and integration of innovative technologies such as remote sensing and the Internet of Things (IoT) in agriculture. For example, the prominent research in 2024 on crop management using ML highlights the transformation of the agricultural sector, promoting more sustainable and efficient practices. Such studies are crucial, as the publication figures reflect an increase in investment in the development of precise tools to optimize the use of agricultural resources, confirming the growing interest of the scientific community in the field.

4.2. Leading Researchers, Journals, and Countries in the Use of Machine Learning in Sustainable Agriculture

We identified several key authors who have stood out both for their productivity and the impact of their publications. Bochtis et al. have contributed significantly through studies on farm logistics optimization and AI-based automation, impacting crop planning and precision agriculture strategies [1]. This study has been fundamental in consolidating the position of this work as a reference, as it offers a detailed view of how ML technologies are being implemented in various areas of agriculture, from crop management to resource optimization.

The publications by Pearson and Busato highlight their impact, evidenced by the high number of citations received. Their most notable work is a review published in 2018 [31], which examined the use of ML in agriculture, covering a wide range of applications and techniques. The review has been widely cited due to the clarity of presentation of the benefits and challenges of using ML in the agricultural sector, resulting in a work that remains an influential reference for subsequent studies in this field.

On the other hand, Miao YX has excelled in scientific productivity, with one of his most noteworthy publications being the study on predicting the nitrogen nutrition index in rice using machine learning based on remote sensing with unmanned aerial vehicles [4]. The study is significant because it demonstrates how advanced ML techniques can improve the accuracy and efficiency of nutrient management in rice crops, a crucial aspect of sustainable agriculture. Miao YX’s high frequency of publications and the relevance of his research have contributed to his recognition as a relevant author in ML applications for agriculture.

Similarly, we identified the most relevant key journals that have stood out for their productivity and impact on the field. Among them, the Computers and Electronics in Agriculture journal emerges as one of the most influential due to its focus on integrating advanced technologies in agriculture. A notable study published in this journal is the review on using support vector machines in precision agriculture [62], which deeply explains how these techniques can optimize various agricultural practices. The publication has been fundamental for advancing knowledge in the area, offering a theoretical and practical framework for applying ML in improving agricultural productivity and sustainability.

Another highly relevant journal is Remote Sensing, which is known for its high impact and productivity in the field of remote sensing applied to agriculture. A significant example of its contribution is the study titled Automated Weed Mapping using Random Forest-OBIA Algorithm, introducing an automatic algorithm based on Random Forest-OBIA that enables the early mapping of weeds using UAV images [63]. The work demonstrated how remote sensing and ML techniques can be effectively combined to improve crop management, allowing for more precise identification and control of weeds, which is essential for agricultural sustainability.

The Sensors journal has also been highlighted for its impact, especially in integrating data from multiple sources and times to improve the mapping of crop types in subtropical regions. The study exemplifies how using multiple sensor data and temporal analysis can provide a more detailed and precise understanding of crop dynamics, which is vital for the efficient and sustainable management of agricultural resources [7].

Lastly, the Agronomy journal is recognized for its high productivity in publications related to sustainable agriculture. A notable study published in this journal uses a novel approach based on GIS and ML to predict the suitability of cropland [20], promoting sustainable agricultural production and providing new perspectives on how emerging technologies improve decision making in agricultural management.

Emphasizing the most important countries in the field of research, China has made significant contributions to the field, excelling in the use of ML regression algorithms for retrieving the leaf area index of cotton using the spectral bands of Sentinel-2 [64], which demonstrates the potential of these technologies to improve precision in precision agriculture. Furthermore, India has contributed valuable studies on applying ML techniques in agricultural production, reviewing various applications ranging from yield prediction to crop management and highlighting the importance of these technologies in improving agricultural productivity and sustainability [32].

On the other hand, a review on ML use in agriculture has been widely cited in the United Kingdom, reflecting its relevance in the scientific community. The work has contributed to consolidating knowledge and has provided a foundation for future research [31]. In addition, Greece has also played a crucial role, with studies that provide an updated review on ML applications in agriculture, highlighting various applications and their potential benefits [1].

Finally, the United States has been a leading country in terms of impact thanks to its contributions to genomic resources for crop improvement and sustainable agriculture. A relevant study addresses the use of genomic resources in plant breeding, highlighting the importance of these technologies to face contemporary and future agricultural challenges [65].

The analysis of leading authors in the field of ML and sustainable agriculture, supported by the information in Figure 4, highlights the prominence of certain researchers such as Bochtis D, Pearson S, and Busato P, who have been instrumental in consolidating the theoretical and applied foundations in this field. According to citation data, Bochtis D stands out for his contributions to reviews on ML applications in agriculture, reinforcing his influence in the scientific literature. In particular, the 2021 review conducted by Bochtis et al. has been pivotal in establishing a comprehensive framework for ML applications in agriculture, with a high citation rate that underscores its impact on advancing knowledge about the use of these technologies to enhance agricultural sustainability. These citation figures reflect the relevance of these authors’ work and the global recognition of their research in the sector.

4.3. Evolution of Topics in the Use of Machine Learning in Sustainable Agriculture

In the early years of research on using ML in sustainable agriculture, the concept of weather data was crucial. This initial perspective allowed researchers to better understand and predict weather conditions, which is fundamental for effective crop management. In one of the pioneering studies in this field, the authors developed the CAMEL model, an intelligent computational model for agro-meteorological data [66]. This work established the foundation for integrating weather data into ML applications, demonstrating how precise weather prediction can positively influence agricultural decision making and resource optimization. Over time, the conceptual approach has expanded to include advanced technologies such as IoT, artificial intelligence, remote sensing, and smart farming, reflecting an evolution towards more integrated and sophisticated solutions in sustainable agriculture.

Smart farming gained relevance in 2021 due to its ability to integrate advanced technologies such as agricultural drones and IoT into the food supply chain. Discussions focused on how the application of agricultural drones and IoT can enhance understanding of the food supply chain, especially in the post-COVID-19 context. The technology allows for greater efficiency in food production and distribution, contributing to the sustainability and resilience of the global food system [50].

In 2022, remote sensing was the predominant concept due to its ability to provide robust indicators of crop growth in different stages of development and under various water and nutrient management conditions. Authors have demonstrated that indicators based on unmanned aerial vehicles (UAVs) are reliable for different water and nutrient management practices [67]. However, they vary between crop development stages, allowing for precise and real-time monitoring of crops and facilitating more informed and efficient management of agricultural resources.

The focus on artificial intelligence (AI) during 2023 reflects its growing importance in the optimization of various agricultural practices. AI facilitates the analysis of large volumes of agricultural data, allowing for more precise and effective management of natural resources and improving crop yields by identifying patterns and accurate predictions [68]. Thus, they highlighted the benefits, challenges, and trends in AI in agriculture, emphasizing how this technology can revolutionize decision making, improve the efficiency of agricultural processes, and increase sustainability.

Lastly, we identified that advanced concepts such as the IoT, artificial intelligence, remote sensing, and smart farming have gained significant relevance. In 2024, the most studied concept was the IoT due to its ability to improve precision in agriculture by predicting crop diseases and personalized recommendations. During this period, researchers have demonstrated how an IoT-based system can efficiently predict crop diseases and provide recommendations on specific crops [69], optimizing resource use and improving agricultural productivity. This integration of the IoT in agriculture allows for continuous monitoring and a quick response to changing agricultural environmental conditions.

The thematic evolution of the use of machine learning in sustainable agriculture has followed an interesting trajectory, reflected in the progression of key terms used in the scientific literature over the years. According to the information in Figure 7, in 2024 the most cited term was “Internet of Things” (IoT), highlighting its growing relevance in agriculture. This increase reflects how the IoT has enabled greater precision in crop monitoring and efficient resource management. In 2023, “Artificial Intelligence” (AI) also gained significant prominence, underscoring the crucial role of AI in optimizing agricultural processes and enhancing sustainability in agricultural production. These thematic shifts and their correlation with the rise in publications, as observed in Figure 3, demonstrate the rapid evolution of approaches and technologies applied in sustainable agriculture, paving the way for more integrated and sophisticated solutions.

4.4. Topic-Based Co-Occurrence Clusters on the Use of Machine Learning in Sustainable Agriculture

The analysis of the keyword co-occurrence network shows several thematic clusters and their significant conceptual affinities, as shown in Figure 8. The main cluster, in blue, consists of keywords such as deep learning, crop management, crop disease detection, CNN, image processing, and resource allocation. This reflects an emphasis on using advanced machine learning techniques for crop management and disease detection, using CNNs and image processing to improve resource allocation in agriculture. In some studies, machine learning techniques in agroclimatic research are reviewed [70]. Some authors explore climate-dependent crop management through data modeling and emphasize the importance of these approaches [71]. In contrast, others demonstrate the practical application of these technologies in spikelet counting and crop yield optimization through quantum computing [72].

The second most significant cluster, in green, consists of keywords such as smart agriculture, data mining, sensors, IoT, and crop recommendation. This cluster highlights the integration of smart technologies and the use of IoT sensors to optimize agriculture. Data mining and data-based crop recommendations are central themes in this cluster. Among the most relevant works, research has been conducted on optimizing onion crop management through a smart agriculture framework with IoT sensors and cloud technology [73]. Also, a comparative analysis of machine learning techniques to predict crop production capacity was carried out in [74]. Additionally, an efficient IoT-based system for predicting crop diseases and crop recommendations was introduced in [69], emphasizing the importance of these emerging technologies in precision agriculture.

The thematic clusters identified in the keyword co-occurrence analysis, represented in Figure 6, demonstrate a concentrated focus on areas such as “Deep Learning”, “Crop Management”, and “Crop Disease Detection”. These clusters reflect a growing trend towards the use of advanced techniques, such as convolutional neural networks (CNNs) and image processing, in crop management and resource optimization. As studies on “Smart Agriculture” and “IoT” continue to gain popularity, as seen in the most cited papers of 2023 and 2024 in [69], the literature shows a clear preference for technological solutions that integrate multiple systems to address the challenges of sustainable agriculture. The findings suggest that interdisciplinary collaboration and innovative approaches, reflected in the most relevant papers, have been key to advancing research in this field.

4.5. Frequency and Conceptual Relevance of the Use of Machine Learning in Sustainable Agriculture

The analysis of frequency and conceptual relevance depicted in Figure 9 explores the emerging and consolidated concepts in the field, focusing on key technologies such as neural networks, data mining, Internet of Things (IoT), deep learning, smart farming, remote sensing, and artificial intelligence. These technologies are transforming traditional agricultural practices by enabling more precise, data-driven decision-making processes and efficient resource management, ultimately contributing to more sustainable and productive farming systems. We focus our discussion on quadrants I and II, which comprise the most current topics related to ML-based approaches for sustainable agriculture.

In quadrant I, emerging and consolidated concepts in the field of study were identified, including deep learning, smart farming, remote sensing, and artificial intelligence. Deep learning has emerged as a fundamental technique for estimating crop yields by leveraging large volumes of climatic data and irrigation information. This technology allows for detailed and precise complex data analysis, facilitating high-precision yield predictions and efficient water resource management. It is revolutionizing how agricultural practices are managed by providing advanced tools for decision making based on extensive and multifaceted data [75]. Smart farming, or intelligent agriculture, is another key concept that integrates advanced technologies such as IoT, unmanned aerial vehicles, and augmented reality. Combining these technologies allows for more precise and efficient crop management, improving agricultural productivity by enabling real-time monitoring and optimizing resource management [51]. Remote sensing continues to be a crucial tool for modern agriculture, providing valuable data on crop status and environmental conditions. The integration of optical, structural, and thermal measurements through unmanned aerial vehicles allows for precise modeling of the leaf area index of winter wheat, which is essential for assessing crop growth and adjusting management practices accordingly [23].

Alternatively, quadrant II shows emerging concepts in ML applied to sustainable agriculture, among which are neural networks, data mining, and IoT. Neural networks are revolutionizing modern agricultural challenges, particularly in plant phenotyping, allowing for the efficient analysis of large volumes of data and improving the precision in identifying and characterizing phenotypic traits. The application of neural networks in sunflower phenotyping is highlighted in [76], showing significant advantages in terms of processing power and machine learning, enabling researchers and farmers to make informed decisions based on complex and multi-source data. Data mining is another key concept that is transforming crop management by allowing the analysis of large datasets to extract useful patterns and predict trends [77]. The application of data analysis in crop management provides a comprehensive and data-driven vision for decision making, optimizing agricultural yields, and improving sustainability. The ability to process and analyze large volumes of data enables the identification of critical factors affecting agricultural production and the development of more effective management strategies. Finally, the IoT is emerging as a crucial technology in smart agriculture, with the IoT-Agro system, implemented in coffee farms in Colombia, highlighting its ability to monitor and manage agricultural resources efficiently. The IoT facilitates real-time data collection, allowing farmers to quickly respond to changing environmental conditions and improve the precision in the application of agricultural inputs. This technology not only optimizes production but also contributes to sustainability by reducing waste and improving resource management [46].

In the last few years, artificial intelligence and related technologies have been leading in integrating and analyzing data from different sources for crop management and phenotyping. Therefore, AI-based solutions are a current topic in analyzing remote sensing images for crop management, such as in strawberry production, providing advanced tools for monitoring and optimizing agricultural practices [78].

4.6. Classification of Keywords According to Their Function in the Use of Machine Learning in Sustainable Agriculture

From the identification of related approaches, Table 1 presents a detailed classification of the leading emerging and growing keywords in the field of ML use in sustainable agriculture according to their specific function. It allows for identifying and analyzing the key characteristics and applications, providing a clear view of how these keywords are integrated into the research field. This classification helps researchers and professionals in the agricultural sector to identify emerging areas of interest and practical applications of ML technologies in sustainable agriculture, facilitating further innovations and effective solutions. Beyond serving as a taxonomy of ML applications, this classification highlights how different techniques complement each other in addressing key agricultural challenges. For instance, deep learning and remote sensing are frequently combined for advanced crop monitoring, while IoT-based approaches enable real-time data collection to enhance precision agriculture. The implementation of ML-based predictive models, integrating multispectral and IoT data, has improved agricultural planning, reducing climate uncertainty and optimizing production.

4.7. Theoretical Implications

The analysis of the frequency of publications per year reveals the growth dynamics in research. The progressive growth in the number of publications reflects a growing interest and expansion in using ML techniques to address current challenges in sustainable agriculture. This pattern suggests that integrating advanced technologies in agricultural management is increasingly recognized and valued, driving more significant investment in research and development.

Analyzing key theoretical references, including influential authors and journals, helps grasp the primary contributions to the field. Authors like Bochtis, Pearson, and Miao have been fundamental in developing theory and disseminating key concepts in the application of ML in agriculture. High-impact journals, such as Sensors and Remote Sensing, have played a crucial role in publishing relevant research, highlighting the importance of these channels in the theoretical evolution of the field. The influence of these authors and publications underscores the centrality of some approaches and methodologies in developing new theories and applications. For instance, Bochtis has significantly contributed to the optimization of agricultural logistics through AI-based automation [1,31,87,88], while Pearson’s work in remote sensing has shaped methodologies for integrating satellite imagery into precision farming [89,90,91,92]. Miao’s research has provided key advancements in real-time monitoring systems using IoT and AI, allowing for dynamic adjustments in agricultural management [4]. These contributions form the foundation for new frameworks that integrate hybrid ML models, multimodal data analysis, and self-learning adaptive systems tailored for various agricultural contexts.

The thematic evolution, observed through the shift in focus from meteorological data to more advanced concepts such as the IoT and AI, indicates a growing sophistication in the methodologies employed. Integrating emerging technologies and advanced concepts reflects an evolution towards more global and multidisciplinary models, which enable more precise and efficient agricultural management. Notably, the convergence of ML with geospatial analysis [93], UAV-based crop monitoring [11,35], and automated sensing networks [94,95] has enabled the development of real-time adaptive models that can predict climate-related risks and optimize irrigation strategies. For example, integrating ML models with high-resolution UAV-derived multispectral data has demonstrated significant improvements in soybean yield estimation [93]. The study highlights that textural features extracted using gray-level co-occurrence matrices can outperform traditional vegetation indices in predictive accuracy. Additionally, combining UAV-based spectral bands, canopy height, and advanced regression models, such as Cubist and Random Forest, achieved R² values up to 0.89, establishing a robust framework for high-throughput phenotyping. Furthermore, leveraging ensemble learning models, particularly XGBoost and a stacking-based ensemble approach, has significantly improved the robustness of biomass estimation across different soybean growth stages [11]. The study demonstrated that integrating UAV-derived spectral, structural, and textural features enhanced model accuracy, achieving an R² of 0.85.

Analyzing the co-occurrence of keywords and identifying emerging and established keywords offer a comprehensive view of evolving research areas. Emerging concepts, such as neural networks and data mining, and growing concepts like deep learning and smart agriculture, highlight the most outstanding innovation and development areas. This analysis allows for identifying the main trends and current approaches in research, as well as areas that are gaining relevance in the scientific literature. A deeper exploration of the theoretical underpinnings of these research trends would strengthen the analysis. It is essential to understand how these developments have influenced current models, introduced new methods, or improved analytical frameworks to fully appreciate their impact. For example, the combination of IoT and AI [96,97], Big Data [98], or ML [99] in agricultural studies has broadened data-driven decision making and revolutionized predictive modeling and automation in precision agriculture. Likewise, deep learning and data mining techniques have enabled more robust and adaptable analytical strategies. Explicitly linking these research directions to their theoretical contributions will provide a more comprehensive understanding of their importance in the evolution of the field.

Finally, research gaps are essential to guide future research efforts. The thematic, geographical, interdisciplinary, and temporal gaps indicate areas lacking sufficient knowledge or research. These gaps provide opportunities to address unresolved problems and explore new directions in using ML for sustainable agriculture. Identifying and addressing these gaps will advance theoretical knowledge and improve practical applications and the effectiveness of the proposed solutions in sustainable agricultural management. For example, while ML has demonstrated success in yield prediction and disease detection, gaps remain in its application to smallholder farms due to limited access to digital infrastructure. Further research is needed to develop cost-effective, low-power ML models such as TinyML that can operate in resource-constrained environments [100]. Additionally, the integration of ML with agroecological models remains underexplored despite its potential to enhance biodiversity conservation and soil health monitoring.

4.8. Practical Implications

The practical implications highlight the need to adapt research and development strategies to emerging and growing trends. The thematic evolution from a focus on “Weather Data” to more advanced concepts such as “Internet of Things”, “Artificial Intelligence”, “Remote Sensing”, and “Smart Farming” reveals a significant change toward the integration of more sophisticated technologies and methodologies. This implication suggests that future research and applications should incorporate these advanced concepts to effectively address contemporary challenges in sustainable agriculture.

Examining thematic clusters shows a strong conceptual connection among terms like “Deep Learning”, “Crop Management”, “Crop Disease Detection”, “Convolutional Neural Networks”, “Image Processing”, and “Resource Allocation”. These terms reflect key areas where ML can offer innovative and effective solutions. For example, Navrozidis et al. demonstrated the effectiveness of ML-based hyperspectral data analysis for early disease detection in olive trees in 2023, achieving high classification accuracy with Random Forest and XGBoost models [88]. This approach reinforces the necessity for researchers and professionals to focus on developing and applying deep learning techniques to improve crop management, disease detection, and resource allocation, thereby optimizing agricultural productivity and sustainability.

The identification of emerging terms such as “Neural Networks”, “Data Mining”, “Internet of Things”, and “Yield Prediction” suggests that these concepts are gaining relevance and should be prioritized in the development of new technologies and methodologies. Recent advancements in UAV-based ML models have demonstrated the potential of high-resolution multispectral data for soybean yield prediction, significantly improving accuracy compared to traditional vegetation indices [93]. Incorporating these emerging technologies can improve the accuracy of yield prediction and enable more efficient crop management by integrating data from various sources, such as IoT sensors. Accordingly, projects and practical applications should explore and apply these emerging technologies to maximize their benefits.

On the other hand, the growing concepts such as “Deep Learning”, “Smart Farming”, “Remote Sensing”, and “Artificial Intelligence” reflect consolidated and expanding areas in research and innovation on machine learning in agriculture. The work presented by Sharma and Shivandu in 2024 emphasizes the transformative potential of AI and IoT in precision agriculture, detailing how smart farming technologies can automate critical processes such as fertilization, irrigation, and pest management, leading to substantial reductions in resource use and environmental impact [94]. These concepts are proving effective in solving complex problems and improving agricultural processes. In practical terms, this means that investments in research and development should focus on improving and expanding the applications of these technologies to address current and future challenges in sustainable agriculture comprehensively.

In addition to these dimensions, bibliometrics suggests several additional practical implications. First, identifying research gaps in areas such as the intersection of ML with agricultural sustainability suggests opportunities for new studies that can fill these gaps. For example, the integration of emerging technologies in regions with limited resources could improve the resilience of agricultural systems in the face of climate change; second, the analysis of emerging and growing keywords indicates that future research should focus on developing practical applications for these technologies, such as crop yield prediction systems based on deep learning and smart agricultural management platforms that integrate IoT and remote sensing.

Studies have demonstrated that deep learning-based crop yield prediction models [101,102] significantly enhance forecasting accuracy compared to traditional methods, leading to better resource allocation and planning [103]. Additionally, research on IoT-integrated smart farming platforms has shown improvements in precision agriculture by enabling real-time monitoring and automated decision making [94,95]. Similarly, remote sensing applications have been proven effective in optimizing irrigation and detecting crop diseases, contributing to more sustainable agricultural practices [99]. These findings highlight the tangible benefits of integrating these technologies, reinforcing the argument that investments in research and development should focus on expanding their practical applications to address current and future agricultural challenges.

Finally, researchers and agricultural professionals should consider implementing AI- and IoT-based systems to improve the efficiency and sustainability of agricultural practices. The work by Tagarakis and Bochtis [87] discusses how integrating robotics with advanced sensing technologies has led to a closed-loop interaction between AI, sensing, and automation, revolutionizing precision agriculture. Real-time data analysis platforms and smart sensors enable precise resource management, such as optimizing irrigation and nutrient application while allowing for quick responses to changing environmental conditions. These innovations not only improve crop yields but also contribute to reducing agriculture’s environmental footprint, reinforcing the importance of sustained investment in digital agriculture technologies.

4.9. Limitations and Research Gaps

The reliance on selected databases can introduce biases in the coverage and representation of the literature, as these sources may not include all relevant studies, especially those published in less indexed journals or lower-visibility conferences. Likewise, the use of tools such as Microsoft Excel^® and VOSviewer^® for bibliometric analysis, although widely accepted, may be limited by the accuracy and comprehensiveness of the obtained data and the analytical and visualization capabilities of these tools.

Additionally, the methodology used to define quantity, quality, and structure indicators may not fully capture the heterogeneity and complexity of the field under investigation. For example, categorizing keywords in terms of their growth or emergence may be influenced by the subjective selection of terms and the analysis methodology used, which could affect the interpretation of thematic trends and research gaps. These limitations underscore the need for a critical and complementary evaluation of the results, considering the inclusion of additional sources and the use of alternative methodologies to obtain a more comprehensive and nuanced view of the current and future state of the field.

Table 2 summarizes the main research gaps in the field of ML use in sustainable agriculture, highlighting the areas that require greater attention and development in future research. The related gaps have been divided into four categories: topic, geographical, interdisciplinary, and temporal. Regarding topic-related gaps, there is a pressing need to understand how different ML techniques can be integrated effectively and their long-term impact on agricultural sustainability, particularly across diverse agroclimatic conditions. Geographically, there remains limited research in underdeveloped and developing regions and a lack of comparative studies between different climatic zones. The interdisciplinary domain highlights significant gaps in collaboration between disciplines, notably in integrating behavioral sciences and socioeconomic knowledge with ML studies. Temporal gaps reveal the necessity for more longitudinal studies and research analyzing the evolution and trends in ML usage in agriculture. Each of these gaps is accompanied by specific research questions that could guide future investigations, such as examining barriers to ML adoption in underdeveloped regions, exploring interdisciplinary approaches to complex agricultural challenges, and evaluating the long-term impacts of ML technologies on agricultural sustainability.

4.10. Research Perspectives

Figure 10 illustrates the research agenda derived from analyzing topics in machine learning applied to sustainable agriculture, where several key techniques and approaches have emerged as critical areas for future investigation. The research perspectives comprise diverse technologies, including artificial intelligence, remote sensing, deep learning, data mining, precision farming, convolutional neural networks, Big Data, Sentinel-2 satellite applications, data analytics, and disease detection. Each of these technologies plays a unique and complementary role in advancing sustainable agricultural practices, with their integration offering promising opportunities for improving crop management and resource efficiency, among others.

Despite advancements in the use of machine learning models to predict agricultural yields, several limitations still affect their implementation in sustainable agriculture. While models such as GBM and ensemble methods show high predictive accuracy, their lack of interpretability and “black-box” nature hinder their adoption by non-expert farmers and policymakers. Furthermore, current models rely on geographically limited datasets, which restricts their applicability in different agroecological contexts and does not incorporate socio-cultural and political factors that may influence agricultural decisions. On the other hand, the high computational cost of these advanced models represents another barrier, especially in resource-limited settings, making it necessary to explore lighter and more efficient techniques for real-time implementation [110].

Another study proposes a metamodel for a connected farm based on the ISO/IS 19440-2020 standard [111], which conceptualizes the farm as an integrated entity, incorporating digital technologies in key areas such as processes, goals, resources, and the digital perspective. This approach enables the optimization of agricultural operations, improves resource efficiency, and reduces environmental impact through technologies such as data analytics, remote sensing, and automation. However, significant challenges remain, such as limited access to high-quality data and the integration of the Internet of Things (IoT), which hinder the implementation of smart agricultural systems in resource-limited environments. Additionally, interoperability between digital platforms and devices must improve to maximize the effectiveness of these technological solutions [112].

Moreover, despite promising results obtained with machine learning algorithms used in processing IoT data, such as the soil moisture meter with a 99% effectiveness rate, there are significant areas for improvement. The current system, limited by a small number of sensors and data processing methods, could benefit from additional sensors to monitor critical environmental factors such as soil moisture and nutrients, as well as from the refinement of algorithms, including deep learning techniques. Furthermore, although the models showed high accuracy, scalability, and adaptability in different agricultural environments, along with overcoming barriers such as the lack of awareness and training for farmers, remain key challenges for the effective adoption of these technologies in the real world [113].

AI is a popular component in the field of ML for sustainable agriculture, facilitating automation and improving decision making. The ability of AI to analyze large volumes of data and make accurate predictions allows for significant optimization in crop management, early disease detection, and adaptation to changing climatic conditions. Currently, AI is used to develop recommendation systems and predictive models that enhance the efficiency of agricultural practices and promote sustainability. Future research should explore integrating advanced AI techniques with technologies like IoT and computer vision to develop resilient and adaptive agricultural systems. Furthermore, it is crucial to study the effects of AI on waste reduction and productivity enhancement at field and regional scales and assess its feasibility in diverse socioeconomic and geographical settings.

Remote sensing has revolutionized agricultural monitoring by providing detailed data on crop conditions, land use, and environmental conditions from an aerial and satellite perspective. This method enables ongoing and accurate evaluation of natural resources and crop health, making management more informed and efficient. Integrating satellite images and drones with ML algorithms has proven effective for crop classification, yield estimation, and identifying problematic areas. To further explore this area, it is recommended to investigate how emerging remote sensing technologies can be combined with ML techniques to improve the resolution and accuracy of the data obtained. It would also be valuable to examine the applicability of remote sensing in precision agriculture contexts in regions with technological limitations or limited economic resources. Additionally, investigating how remote sensing data can be integrated with predictive models to anticipate and mitigate the impacts of climate change on sustainable agriculture would be a promising line of research.

Deep learning has proven to be a powerful tool in sustainable agriculture by enabling the creation of complex predictive models capable of identifying patterns and making precise classifications from large datasets. This technique is applied in crop disease detection, species classification, and yield prediction, offering unprecedented capacity to address complex agricultural problems and improve the sustainability of agricultural practices. Forthcoming research must study how deep learning models can be adapted and optimized to work with multidimensional data from different sources, such as satellite images, weather data, and field sensors. Moreover, exploring the integration of deep learning techniques with other emerging technologies, such as the Internet of Things (IoT) and cloud computing, could offer innovative new solutions for specific challenges in sustainable agriculture. It is also relevant to evaluate the impact of these models on decision making and practical implementation in various agricultural contexts.

Data mining is a fundamental technique in using ML in sustainable agriculture, as it allows for extracting patterns and valuable knowledge from large volumes of agricultural data. This methodology is crucial for analyzing historical and real-time data related to crop performance, soil conditions, and climate, facilitating the identification of trends and informed decision making. The ability of data mining to reveal hidden inputs and make accurate predictions significantly contributes to improving the efficiency and sustainability of agricultural practices. Further investigation is recommended to examine how advanced data mining techniques can be integrated with other analytical and visualization tools to improve complex and multidimensional data interpretation. Similarly, exploring the application of data mining in integrating data from different sources, such as IoT sensors and remote sensing platforms, could offer new perspectives on resource management and performance optimization. It is also important to analyze how these techniques can be adapted to specific contexts, such as regions with particular climate challenges or emerging agricultural systems.

Precision farming has transformed the field of sustainable agriculture by enabling more precise and efficient management of agricultural resources. Using ML-based technologies, precision agriculture optimizes the application of inputs such as water, fertilizers, and pesticides, leading to greater efficiency and sustainability in resource use. This methodology allows farmers to make decisions based on detailed and field-specific data, reducing environmental impact and improving productivity. To advance in this area, it is suggested to study how precision farming techniques can be improved by integrating emerging technologies, such as drones and advanced sensors. It is also valuable to explore how precision agriculture can be adapted to different crops and agricultural systems, considering variations in soil, climate, and available resources. Furthermore, the economic and social impact of implementing precision farming practices should be analyzed, evaluating their viability in different contexts and regions.

Convolutional neural networks have established themselves as a powerful technique in ML-based solutions to sustainable agriculture, especially in image interpretation and detection of complex patterns. CNNs are particularly effective in tasks such as crop image classification, disease detection, and yield estimation thanks to their ability to extract spatial and temporal features from visual data. Their application allows for more precise and efficient crop management, improving agricultural sustainability. The research agenda should focus on optimizing CNNs to work with multisensory and multimodal data, combining visual information with climate and soil data. In addition, how CNNs can adapt to different types of images and resolutions should be investigated, considering the specific challenges of each agricultural region. Evaluating the impact of CNNs on the accuracy and effectiveness of sustainable agricultural practices, as well as their integration with other emerging technologies, could offer new opportunities to improve crop management and performance.

Big Data also plays a crucial role by providing a large volume of data to extract valuable information and make informed decisions in the context of sustainable agriculture. The ability to process large amounts of agricultural data, including information on weather conditions, crop yields, and management practices, allows for a deeper understanding of the patterns and trends that affect agricultural production. Big Data analysis facilitates identifying correlations and predicting future events, contributing to a more efficient and sustainable management of agricultural resources. To ensure that Big Data remains relevant in sustainable agriculture, future studies could focus on developing new methodologies for integrating and analyzing data from various sources, such as field sensors and satellite monitoring platforms. Similarly, research could explore how to improve data processing and storage techniques to manage the growing amount of generated information efficiently. The advancement in ML algorithms specific to Big Data analysis in agricultural contexts could enable more precise prediction of patterns and trends, ensuring that Big Data remains a fundamental component in decision making and optimizing sustainable agricultural practices.

The Sentinel-2 satellites, part of the Copernicus program of the European Space Agency, have revolutionized agricultural monitoring by providing high-resolution images of the Earth’s surface with regular acquisition frequency. These satellite images are essential for assessing crop conditions, water management, and detecting land-use changes, allowing for continuous and detailed monitoring of agricultural conditions. Using Sentinel-2 data in combination with ML techniques facilitates the identification of issues such as crop diseases and nutritional deficiencies, improving the ability to intervene in a timely and precise manner. Integrating Sentinel-2 images with data from other sources, such as ground sensors or images obtained by drones, suggests new research perspectives to improve the accuracy and utility of the analyses. Additionally, exploring new ML applications to interpret satellite images and develop more robust predictive models more effectively can increase the relevance of Sentinel-2 in sustainable agricultural management. Research could also focus on optimizing algorithms to process large volumes of satellite data and adapting these technologies to specific agricultural contexts.

Data analytics is fundamental in the use of ML for sustainable agriculture, as it allows for the extraction of valuable insights from large agricultural datasets. Data analytics techniques facilitate the identification of patterns and the evaluation of the effectiveness of different agricultural practices, from crop management to resource use. This deep analysis capability contributes to decision optimization, performance improvement, and environmental impact reduction, thus supporting more efficient and sustainable agriculture. To further develop data analytics solutions, it is recommended to investigate how advancements in data analysis techniques can be applied to new areas and emerging challenges in sustainable agriculture. The development of analytical tools integrates data from multiple sources, such as IoT sensors and remote monitoring platforms, to provide a more comprehensive and accurate view of agricultural conditions. Also, the research could focus on creating adaptive analytical models for climate patterns change and agricultural practices, ensuring agricultural decision making and innovation in this regard.

Disease detection is a critical application of machine learning in sustainable agriculture, as it allows for the early and accurate identification of crop diseases through data analysis and image interpretation. This capability is essential for implementing timely and effective interventions, minimizing the impact of diseases on crop yields, and reducing the need for extensive chemical treatments. Precise disease detection improves the overall health of crops and contributes to more sustainable agricultural practices. Future studies could combine advanced machine learning techniques with new sensing and monitoring technologies like drones and on-field sensors. Additional research could focus on developing more precise and adaptive detection algorithms to identify a broader range of diseases in different types of crops and environmental conditions. Additionally, integrating disease detection data with other agricultural management tools could provide a more comprehensive view and enable a more effective response to disease outbreaks.

5. Conclusions

The presented bibliometric study identified research trends and variables regarding the use of machine learning in sustainable agriculture. The analysis of machine learning applications in this field revealed several key findings related to the research questions addressed. The comprehensive analysis and discussion provide a robust foundation for understanding the current state and future directions of ML applications in sustainable agriculture, offering valuable insights for researchers, professionals, and policymakers working to advance this crucial field. Unlike previous reviews, this study provides a more structured examination of thematic clusters, highlighting the dominant research trajectories and identifying knowledge gaps that require further exploration.

The years 2022–2024 have shown the most significant research activity in this field, demonstrating a quadratic polynomial growth in published articles and indicating sustained interest and expansion in this area. Notably, prominent researchers such as Bochtis D., Pearson S., and Busato P. have made fundamental contributions to the literature. At the same time, journals like Computers and Electronics in Agriculture and Remote Sensing have emerged as leading publication sources. India and China have distinguished themselves as the most productive countries in this research topic. The concentration of research output in select countries indicates that the global adoption of this technology is uneven. In many developing countries, infrastructure limitations, restricted access to high-quality agricultural datasets, and insufficient research funding impede the extensive use of ML-based solutions. Moreover, differences in farmers’ technology expertise and digital literacy impede the successful incorporation of these advances into practical agricultural applications.

On the other hand, the thematic evolution of scientific production shows a significant shift from an initial focus on weather data to a more comprehensive approach incorporating the IoT, artificial intelligence, and remote sensing technologies. This evolution indicates a transition towards a more advanced and technological integration in agricultural research, reflecting a transformation to the new demands and capabilities of the sector. Furthermore, the analysis suggests that ML is increasingly used for real-time monitoring, automated decision making, and predictive analytics, indicating a paradigm shift towards data-driven agriculture. Regarding the most relevant keywords, the growing and consolidated concepts are deep learning, smart farming, remote sensing, and artificial intelligence. At the same time, the emerging terms include neural networks, data mining, and IoT. This distinction suggests a dynamic field with established methodologies, which is also continuously incorporating new technologies and approaches. In addition, the analysis of thematic clusters shows a high degree of conceptual affinity, indicating that the current literature emphasizes the advanced application of ML for crop management and improvement using image processing techniques and resource optimization. The integration of UAV-based imaging, hyperspectral analysis, and geospatial ML techniques has become a major focus, reinforcing the role of high-resolution remote sensing in precision agriculture. Despite these advances, challenges related to interoperability between different ML platforms, standardization of data formats, and the need for higher computational efficiency remain pressing issues that require further investigation.

Finally, the research agenda for the future of the topic underlines the deepening and expansion of predominant concepts, such as deep learning and smart farming, to develop more detailed and specialized studies. Identifying these concepts as protagonists emphasizes the need to further explore their applications and effects in sustainable agriculture, which is essential for advancing the understanding and application of ML in this sector. Future research should prioritize the development of explainable AI models to improve interpretability for end-users, particularly farmers with limited technical expertise. Additionally, efforts should focus on deploying scalable and adaptable ML solutions to diverse agricultural settings, ensuring inclusivity in the adoption of these technologies. Addressing the economic and environmental implications of ML-driven agricultural practices is also critical to maximizing sustainability and minimizing unintended consequences.

Author Contributions

Conceptualization, J.B.-V., V.G.-P. and E.R.-V.; methodology, J.B.-V., V.G.-P., A.V.-A., J.V., E.R.-V. and M.M.-H.; software, M.M.-H.; validation, J.B.-V., V.G.-P., A.V.-A., J.V., E.R.-V. and M.M.-H.; formal analysis, J.B.-V., V.G.-P., E.R.-V. and R.H.-G.; investigation, J.B.-V., V.G.-P., A.V.-A., J.V., E.R.-V. and M.M.-H.; resources, J.B.-V., E.R.-V. and R.H.-G.; data curation, V.G.-P., A.V.-A., J.V. and M.M.-H.; writing—original draft preparation, J.B.-V., V.G.-P., A.V.-A., J.V., E.R.-V. and M.M.-H.; writing—review and editing, J.B.-V., E.R.-V. and R.H.-G.; visualization, V.G.-P. and M.M.-H.; supervision, J.B.-V., E.R.-V. and R.H.-G.; project administration, J.B.-V. and E.R.-V.; funding acquisition, J.B.-V. and R.H.-G. All authors have read and agreed to the published version of the manuscript.

Funding

This work has been carried out under contract RC130-2024, corresponding to project code 10922, titled “Diversificación de fuentes de proteínas para uso alimentario mediante el empleo de terrazas de cultivo aeropónicas o hidropónicas, integradas con sistemas automatizados, inteligencia artificial y energía renovable para la creación de comunidades autosostenibles”.

Data Availability Statement

The data and materials that supported this bibliometric study will be publicly available and can be accessed in the Zenodo data repository openly available in: Botero Valencia, J., García-Pineda, V., Valencia-Arias, A., Valencia, J., Reyes-Vera, E., Mejía-Herrera, M., & Hernández-García, R. (2025). Machine Learning in Sustainable Agriculture: Bibliometric Analysis and Research Perspectives [Data set]. Zenodo. https://doi.org/10.5281/zenodo.14847122 (accessed on 10 February 2025).

Acknowledgments

R.H.-G. and J.B.-V. acknowledge the support of the ANID Vinculación Internacional FOVI230126 project and BRECS.NET for contributing to academic collaboration support that gave origin to this work. The authors thank Universidad Católica del Maule for supporting the APC payment.

Conflicts of Interest

The authors declare no conflicts of interest.

References

Benos, L.; Tagarakis, A.C.; Dolias, G.; Berruto, R.; Kateris, D.; Bochtis, D. Machine learning in agriculture: A comprehensive updated review. Sensors 2021, 21, 3758. [Google Scholar] [CrossRef]
Sai Sharvesh, R.; Suresh Kumar, K.; Raman, C.J. An Accurate Plant Disease Detection Technique Using Machine Learning. EAI Endorsed Trans. Internet Things 2024, 10, 1–9. [Google Scholar] [CrossRef]
Lobo, A.D.; Shetty, S.; Rai, V.; Naik, S.C.; Badiger, M.; Singh, C. Revolutionizing Agriculture: A Comprehensive Review of UAVs, Artificial Intelligence, and IoT Integration for Precision Farming. In Drone Applications for Industry 5.0; IGI Global: Hershey, PA, USA, 2024; pp. 398–418. [Google Scholar]
Zha, H.; Miao, Y.; Wang, T.; Li, Y.; Zhang, J.; Sun, W.; Feng, Z.; Kusnierek, K. Improving unmanned aerial vehicle remote sensing-based rice nitrogen nutrition index prediction with machine learning. Remote Sens. 2020, 12, 215. [Google Scholar] [CrossRef]
Ali, A.; Hassan, M.U.; Kaul, H.P. Broad Scope of Site-Specific Crop Management and Specific Role of Remote Sensing Technologies Within It—A Review. J. Agron. Crop. Sci. 2024, 210, e12732. [Google Scholar] [CrossRef]
Mazuryk, J.; Klepacka, K.; Kutner, W.; Sharma, P.S. Glyphosate separating and sensing for precision agriculture and environmental protection in the era of smart materials. Environ. Sci. Technol. 2023, 57, 9898–9924. [Google Scholar] [CrossRef]
Sun, C.; Bian, Y.; Zhou, T.; Pan, J. Using of multi-source and multi-temporal remote sensing data improves crop-type mapping in the subtropical agriculture region. Sensors 2019, 19, 2401. [Google Scholar] [CrossRef] [PubMed]
Pukrongta, N.; Taparugssanagorn, A.; Sangpradit, K. Enhancing Crop Yield Predictions with PEnsemble 4: IoT and ML-Driven for Precision Agriculture. Appl. Sci. 2024, 14, 3313. [Google Scholar] [CrossRef]
Sridhar, H.; Divyashree, V.; Keerthana, B.; Sushmitha, D. Design and demonstration of IoT and machine learning based smart irrigation system. AIP Conf. Proc. 2024, 3111, 030009. [Google Scholar]
de Souza, F.L.P.; Dias, M.A.; Setiyono, T.D.; Campos, S.; Shiratsuchi, L.S.; Tao, H. Identification of soybean planting gaps using machine learning. Smart Agric. Technol. 2025, 10, 100779. [Google Scholar] [CrossRef]
Da, H.; Li, Y.; Xu, L.; Wang, S.; Hu, L.; Hu, Z.; Wei, Q.; Zhu, R.; Chen, Q.; Xin, D.; et al. Advancing soybean biomass estimation through multi-source UAV data fusion and machine learning algorithms. Smart Agric. Technol. 2025, 10, 100778. [Google Scholar] [CrossRef]
Orecchio, C.; Sacco Botto, C.; Alladio, E.; D’Errico, C.; Vincenti, M.; Noris, E. Non-invasive and early detection of tomato spotted wilt virus infection in tomato plants using a hand-held Raman spectrometer and machine learning modelling. Plant Stress 2025, 15, 100732. [Google Scholar] [CrossRef]
Liu, J.; Yang, K.; Tariq, A.; Lu, L.; Soufan, W.; El Sabagh, A. Interaction of climate, topography and soil properties with cropland and cropping pattern using remote sensing data and machine learning methods. Egypt. J. Remote Sens. Space Sci. 2023, 26, 415–426. [Google Scholar] [CrossRef]
Botero-Valencia, J.; Mejia-Herrera, M.; Pearce, J.M. Low cost climate station for smart agriculture applications with photovoltaic energy and wireless communication. HardwareX 2022, 11, e00296. [Google Scholar] [CrossRef]
Ranjan, P.; Garg, R.; Rai, J.K. Artificial intelligence applications in soil & crop management. In Proceedings of the 2022 IEEE Conference on Interdisciplinary Approaches in Technology and Management for Social Innovation (IATMSI), Gwalior, India, 21–23 December 2022; IEEE: Piscataway, NJ, USA, 2022; pp. 1–5. [Google Scholar]
Depuru, S.; Amala, K.; Supriya, P.; Reddy, A.B.; Gireesh, R.S. VGG-16 Technique to Reduce the Global Food Crises and Enhance the Crop Yields: Deep Learning Approaches. In Proceedings of the 2024 3rd International Conference on Applied Artificial Intelligence and Computing (ICAAIC), Salem, India, 5–7 June 2024; IEEE: Piscataway, NJ, USA, 2024; pp. 596–599. [Google Scholar]
Sen, A.; Roy, R.; Dash, S.R. Smart farming using machine learning and IoT. In Agricultural Informatics: Automation Using the IoT and Machine Learning; Wiley-Scrivener: Austin, TX, USA, 2021; pp. 13–34. [Google Scholar]
Khodjaev, S.; Bobojonov, I.; Kuhn, L.; Glauben, T. Optimizing machine learning models for wheat yield estimation using a comprehensive UAV dataset. Model. Earth Syst. Environ. 2025, 11, 15. [Google Scholar] [CrossRef]
Leukel, J.; Zimpel, T.; Stumpe, C. Machine learning technology for early prediction of grain yield at the field scale: A systematic review. Comput. Electron. Agric. 2023, 207, 107721. [Google Scholar] [CrossRef]
Radočaj, D.; Jurišić, M. GIS-based cropland suitability prediction using machine learning: A novel approach to sustainable agricultural production. Agronomy 2022, 12, 2210. [Google Scholar] [CrossRef]
Tanveer, M.U.; Munir, K.; Raza, A.; Almutairi, M.S. Novel artificial intelligence assisted Landsat-8 imagery analysis for mango orchard detection and area mapping. PLoS ONE 2024, 19, e0304450. [Google Scholar] [CrossRef]
Nayak, H.S.; Silva, J.V.; Parihar, C.M.; Krupnik, T.J.; Sena, D.R.; Kakraliya, S.K.; Jat, H.S.; Sidhu, H.S.; Sharma, P.C.; Jat, M.L.; et al. Interpretable machine learning methods to explain on-farm yield variability of high productivity wheat in Northwest India. Field Crop. Res. 2022, 287, 108640. [Google Scholar] [CrossRef]
Lv, Z.; Meng, R.; Man, J.; Zeng, L.; Wang, M.; Xu, B.; Gao, R.; Sun, R.; Zhao, F. Modeling of winter wheat fAPAR by integrating Unmanned Aircraft Vehicle-based optical, structural and thermal measurement. Int. J. Appl. Earth Obs. Geoinf. 2021, 102, 102407. [Google Scholar] [CrossRef]
Ministerio de Agricultura. Inteligencia Artificial (IA) Identifica Zonas Aptas Para La Agricultura En Colombia. Available online: https://www.agronet.gov.co/Noticias/Paginas/Inteligencia-artificial-(IA)-identifica-zonas-aptas-para-la-agricultura-en-Colombia.aspx (accessed on 28 January 2025).
U.S. Government Accountability Office. Precision Agriculture: Benefits and Challenges for Technology Adoption and Use. Available online: https://www.gao.gov/products/gao-24-105962 (accessed on 28 January 2025).
Gobierno de Perú. MIDAGRI Fortalece Transferencia De Tecnologías De Agricultura De Precisión Para Mejorar La Calidad De La Producción Agrícola. Available online: https://www.gob.pe/institucion/inia/noticias/1048980-midagri-fortalece-transferencia-de-tecnologias-de-agricultura-de-precision-para-mejorar-la-calidad-de-la-produccion-agricola (accessed on 28 January 2025).
National Aeronautics and Space Administration. Available online: https://forum.earthdata.nasa.gov/viewtopic.php?t=5139 (accessed on 28 January 2025).
Goldenits, G.; Mallinger, K.; Raubitzek, S.; Neubauer, T. Current applications and potential future directions of reinforcement learning-based Digital Twins in agriculture. Smart Agric. Technol. 2024, 8, 100512. [Google Scholar] [CrossRef]
Vinayaka; Prasad, P.R.C. AI-Enhanced Remote Sensing Applications in Indian Sugarcane Research: A Comprehensive Review. Sugar Tech 2024, 26, 609–628. [Google Scholar] [CrossRef]
Cravero, A.; Pardo, S.; Sepúlveda, S.; Muñoz, L. Challenges to use machine learning in agricultural big data: A systematic literature review. Agronomy 2022, 12, 748. [Google Scholar] [CrossRef]
Liakos, K.G.; Busato, P.; Moshou, D.; Pearson, S.; Bochtis, D. Machine learning in agriculture: A review. Sensors 2018, 18, 2674. [Google Scholar] [CrossRef] [PubMed]
Mishra, S.; Mishra, D.; Santra, G.H. Applications of machine learning techniques in agricultural crop production: A review paper. Indian J. Sci. Technol. 2016, 9, 1–14. [Google Scholar] [CrossRef]
Pokhariyal, S.; Patel, N.; Govind, A. Machine Learning-Driven Remote Sensing Applications for Agriculture in India—A Systematic Review. Agronomy 2023, 13, 2302. [Google Scholar] [CrossRef]
Page, M.J.; McKenzie, J.E.; Bossuyt, P.M.; Boutron, I.; Hoffmann, T.C.; Mulrow, C.D.; Shamseer, L.; Tetzlaff, J.M.; Akl, E.A.; Brennan, S.E.; et al. The PRISMA 2020 statement: An updated guideline for reporting systematic reviews. BMJ 2021, 372, n71. [Google Scholar] [CrossRef] [PubMed]
Khosravi, H.; Shafie, M.R.; Hajiabadi, M.; Raihan, A.S.; Ahmed, I. Chatbots and ChatGPT: A bibliometric analysis and systematic review of publications in Web of Science and Scopus databases. Int. J. Data Min. Model. Manag. 2024, 16, 113–147. [Google Scholar] [CrossRef]
Van Eck, N.; Waltman, L. Software survey: VOSviewer, a computer program for bibliometric mapping. Scientometrics 2010, 84, 523–538. [Google Scholar] [CrossRef] [PubMed]
Durieux, V.; Gevenois, P.A. Bibliometric indicators: Quality measurements of scientific publication. Radiology 2010, 255, 342–351. [Google Scholar] [CrossRef]
Diaz-Gonzalez, F.A.; Vuelvas, J.; Correa, C.A.; Vallejo, V.E.; Patino, D. Machine learning and remote sensing techniques applied to estimate soil indicators—Review. Ecol. Indic. 2022, 135, 108517. [Google Scholar] [CrossRef]
Raman, R.; Parvathy, P.R.; Sapra, P.; Sonule, V.B.; Murugan, S. Robotic Weed Control and Biodiversity Preservation: IoT Solutions for Sustainable Farming. In Proceedings of the 2023 7th International Conference on Electronics, Communication and Aerospace Technology (ICECA), Coimbatore, India, 22–24 November 2023; IEEE: Piscataway, NJ, USA, 2023; pp. 112–117. [Google Scholar]
Attri, I.; Awasthi, L.K.; Sharma, T.P. Machine learning in agriculture: A review of crop management applications. Multimed. Tools Appl. 2024, 83, 12875–12915. [Google Scholar] [CrossRef]
Acang, J.P.A.; Contillo, G.L.; Bucao, D.S.; Alibuyog, N.R.; Pascual, R.J.; Galacgac, E.S.; Manuel, M.B.; Dadoy, K.C.A. Mitigating Garlic Insect Pests and Diseases Using Knowledge-Driven Decision Support System. In Proceedings of the 2023 IEEE 15th International Conference on Humanoid, Nanotechnology, Information Technology, Communication and Control, Environment, and Management (HNICEM), Coron, Philippines, 19–23 November 2023; IEEE: Piscataway, NJ, USA, 2023; pp. 1–5. [Google Scholar]
Dutta, S.; Chakraborty, S.; Goswami, R.; Banerjee, H.; Majumdar, K.; Li, B.; Jat, M. Maize yield in smallholder agriculture system—An approach integrating socio-economic and crop management factors. PLoS ONE 2020, 15, e0229100. [Google Scholar] [CrossRef] [PubMed]
Maseko, S.; van der Laan, M.; Tesfamariam, E.; Delport, M.; Otterman, H. Evaluating machine learning models and identifying key factors influencing spatial maize yield predictions in data intensive farm management. Eur. J. Agron. 2024, 157, 127193. [Google Scholar] [CrossRef]
Kini, A.S.; Reddy, P.K.; Pai, S.N. Techniques of deep learning and image processing in plant leaf disease detection: A review. Int. J. Electr. Comput. Eng. (IJECE) 2023, 13, 3029–3040. [Google Scholar] [CrossRef]
Kinaneva, D.; Hristov, G.; Georgiev, G.; Zahariev, P. Machine Learning Algorithms for Data Mining and Predictive Analytics in Precision Agriculture. In Proceedings of the 2024 47th MIPRO ICT and Electronics Convention (MIPRO), Opatija, Croatia, 20–24 May 2024; IEEE: Piscataway, NJ, USA, 2024; pp. 132–139. [Google Scholar]
Rodríguez, J.P.; Montoya-Munoz, A.I.; Rodriguez-Pabon, C.; Hoyos, J.; Corrales, J.C. IoT-Agro: A smart farming system to Colombian coffee farms. Comput. Electron. Agric. 2021, 190, 106442. [Google Scholar] [CrossRef]
Gookyi, D.A.N.; Wulnye, F.A.; Arthur, E.A.E.; Ahiadormey, R.K.; Agyemang, J.O.; Agyekum, K.O.B.O.; Gyaang, R. TinyML for smart agriculture: Comparative analysis of TinyML platforms and practical deployment for maize leaf disease identification. Smart Agric. Technol. 2024, 8, 100490. [Google Scholar] [CrossRef]
Tsoukas, V.; Gkogkidis, A.; Kakarountas, A. A TinyML-based System For Smart Agriculture. In Proceedings of the 26th Pan-Hellenic Conference on Informatics, New York, NY, USA, 25–27 November 2022; pp. 207–212. [Google Scholar] [CrossRef]
Hayajneh, A.M.; Batayneh, S.; Alzoubi, E.; Alwedyan, M. TinyML Olive Fruit Variety Classification by Means of Convolutional Neural Networks on IoT Edge Devices. AgriEngineering 2023, 5, 2266–2283. [Google Scholar] [CrossRef]
Dutta, P.K.; Mitra, S. Application of agricultural drones and IoT to understand food supply chain during post COVID-19. In Agricultural Informatics: Automation Using the IoT and Machine Learning; Wiley-Scrivener: Austin, TX, USA, 2021; pp. 67–87. [Google Scholar]
Ponnusamy, V.; Natarajan, S. Precision agriculture using advanced technology of IoT, unmanned aerial vehicle, augmented reality, and machine learning. In Smart Sensors for Industrial Internet of Things: Challenges, Solutions and Applications; Springer: Cham, Switzerland, 2021; pp. 207–229. [Google Scholar]
El Sakka, M.; Ivanovici, M.; Chaari, L.; Mothe, J. A Review of CNN Applications in Smart Agriculture Using Multimodal Data. Sensors 2025, 25, 472. [Google Scholar] [CrossRef]
Dolatabadian, A.; Neik, T.X.; Danilevicz, M.F.; Upadhyaya, S.R.; Batley, J.; Edwards, D. Image-based crop disease detection using machine learning. Plant Pathol. 2024, 74, 18–38. [Google Scholar] [CrossRef]
Waters, E.K.; Chen, C.C.M.; Rahimi Azghadi, M. Sugarcane health monitoring with satellite spectroscopy and machine learning: A review. Comput. Electron. Agric. 2025, 229, 109686. [Google Scholar] [CrossRef]
Wang, D.; Zhao, M.; Li, Z.; Xu, S.; Wu, X.; Ma, X.; Liu, X. A survey of unmanned aerial vehicles and deep learning in precision agriculture. Eur. J. Agron. 2025, 164, 127477. [Google Scholar] [CrossRef]
Castillo-Girones, S.; Munera, S.; Martínez-Sober, M.; Blasco, J.; Cubero, S.; Gómez-Sanchis, J. Artificial Neural Networks in Agriculture, the core of artificial intelligence: What, When, and Why. Comput. Electron. Agric. 2025, 230, 109938. [Google Scholar] [CrossRef]
Sharma, R.; Kamble, S.S.; Gunasekaran, A.; Kumar, V.; Kumar, A. A systematic literature review on machine learning applications for sustainable agriculture supply chain performance. Comput. Oper. Res. 2020, 119, 104926. [Google Scholar] [CrossRef]
Hafiyya, R.M.; Jemsheer Ahmed, P.; Nabhan, P.A.; Melethil, S.; Mohammed, N.; Muhammed Murshid, A.P. AI-Enhanced Precision Crop Rotation Management for Sustainable Agriculture. In Proceedings of the International Conference on E-Mobility, Power Control and Smart Systems: Futuristic Technologies for Sustainable Solutions, ICEMPS 2024, Thiruvananthapuram, India, 8–20 April 2024; pp. 1–6. [Google Scholar] [CrossRef]
Haval, A.M.; Rahman, F. Application of machine learning techniques and the Internet of Things for smart, sustainable agriculture. BIO Web Conf. 2024, 82, 05021. [Google Scholar] [CrossRef]
El-Kenawy, E.S.M.; Alhussan, A.A.; Khodadadi, N.; Mirjalili, S.; Eid, M.M. Predicting Potato Crop Yield with Machine Learning and Deep Learning for Sustainable Agriculture; Springer: Dordrecht, The Netherlands, 2024. [Google Scholar] [CrossRef]
Joshi, T.; Sehgal, H.; Puri, S.; Karnika; Mahapatra, T.; Joshi, M.; Deepa, P.R.; Sharma, P.K. ML-based technologies in sustainable agro-food production and beyond: Tapping the (semi) arid landscape for bioactives-based product development. J. Agric. Food Res. 2024, 18, 101350. [Google Scholar] [CrossRef]
Kok, Z.H.; Shariff, A.R.M.; Alfatni, M.S.M.; Khairunniza-Bejo, S. Support vector machine in precision agriculture: A review. Comput. Electron. Agric. 2021, 191, 106546. [Google Scholar] [CrossRef]
De Castro, A.I.; Torres-Sánchez, J.; Peña, J.M.; Jiménez-Brenes, F.M.; Csillik, O.; López-Granados, F. An automatic random forest-OBIA algorithm for early weed mapping between and within crop rows using UAV imagery. Remote Sens. 2018, 10, 285. [Google Scholar] [CrossRef]
Mao, H.; Meng, J.; Ji, F.; Zhang, Q.; Fang, H. Comparison of machine learning regression algorithms for cotton leaf area index retrieval using Sentinel-2 spectral bands. Appl. Sci. 2019, 9, 1459. [Google Scholar] [CrossRef]
Thudi, M.; Palakurthi, R.; Schnable, J.C.; Chitikineni, A.; Dreisigacker, S.; Mace, E.; Srivastava, R.K.; Satyavathi, C.T.; Odeny, D.; Tiwari, V.K.; et al. Genomic resources in plant breeding for sustainable agriculture. J. Plant Physiol. 2021, 257, 153351. [Google Scholar] [CrossRef] [PubMed]
Leung, C.K.S.; Mateo, M.A.; Nadler, A.J. CAMEL: An intelligent computational model for agro-meteorological data. In Proceedings of the 2007 International Conference on Machine Learning and Cybernetics, Hong Kong, China, 19–22 August 2007; IEEE: Piscataway, NJ, USA, 2007; Volume 4, pp. 1960–1965. [Google Scholar]
Zhu, W.; Rezaei, E.E.; Nouri, H.; Sun, Z.; Li, J.; Yu, D.; Siebert, S. UAV-based indicators of crop growth are robust for distinct water and nutrient management but vary between crop development phases. Field Crop. Res. 2022, 284, 108582. [Google Scholar] [CrossRef]
Oliveira, R.C.d.; Silva, R.D.d.S.e. Artificial intelligence in agriculture: Benefits, challenges, and trends. Appl. Sci. 2023, 13, 7405. [Google Scholar] [CrossRef]
Sravanthi, G.; Moparthi, N.R. An efficient IoT based crop disease prediction and crop recommendation for precision agriculture. Clust. Comput. 2024, 27, 5755–5782. [Google Scholar] [CrossRef]
Tamayo-Vera, D.; Wang, X.; Mesbah, M. A Review of Machine Learning Techniques in Agroclimatic Studies. Agriculture 2024, 14, 481. [Google Scholar] [CrossRef]
Kaur, N.; Gupta, V. Climate dependent crop management through data modeling. In Proceedings of the Data Analytics and Management: ICDAM 2021, Polkowice, Poland, 26 June 2021; Springer: Berlin/Heidelberg, Germany, 2022; Volume 2, pp. 739–767. [Google Scholar]
Alkhudaydi, T.; De la Lglesia, B. Counting spikelets from infield wheat crop images using fully convolutional networks. Neural Comput. Appl. 2022, 34, 17539–17560. [Google Scholar] [CrossRef]
Khan, A.; Hassan, M.; Shahriyar, A.K. Optimizing onion crop management: A smart agriculture framework with IoT sensors and cloud technology. Appl. Res. Artif. Intell. Cloud Comput. 2023, 6, 49–67. [Google Scholar]
Jain, K.; Choudhary, N. Comparative analysis of machine learning techniques for predicting production capability of crop yield. Int. J. Syst. Assur. Eng. Manag. 2022, 13, 583–593. [Google Scholar] [CrossRef]
Alibabaei, K.; Gaspar, P.D.; Lima, T.M. Crop yield estimation using deep learning based on climate big data and irrigation scheduling. Energies 2021, 14, 3004. [Google Scholar] [CrossRef]
Centorame, L.; Gasperini, T.; Ilari, A.; Del Gatto, A.; Foppa Pedretti, E. An overview of machine learning applications on plant phenotyping, with a focus on sunflower. Agronomy 2024, 14, 719. [Google Scholar] [CrossRef]
Chergui, N.; Kechadi, M.T. Data analytics for crop management: A big data view. J. Big Data 2022, 9, 123. [Google Scholar] [CrossRef]
Zheng, C.; Abd-Elrahman, A.; Whitaker, V. Remote sensing and machine learning in crop phenotyping and management, with an emphasis on applications in strawberry farming. Remote Sens. 2021, 13, 531. [Google Scholar] [CrossRef]
Chodey, M.D.; Noorullah Shariff, C. Neural network-based pest detection with k-means segmentation: Impact of improved dragonfly algorithm. J. Inf. Knowl. Manag. 2021, 20, 2150040. [Google Scholar] [CrossRef]
Jyothi, K.D.; Sekhar, M.; Kumar, S. Applications of Statistical Machine Learning Algorithms in Agriculture Management Processes. In Proceedings of the 2021 6th International Conference on Signal Processing, Computing and Control (ISPCC), Solan, India, 7–9 October 2021; IEEE: Piscataway, NJ, USA, 2021; pp. 237–241. [Google Scholar]
Prabha, C.; Pathak, A. Enabling technologies in smart agriculture: A way forward towards future fields. In Proceedings of the 2023 International Conference on Advancement in Computation & Computer Technologies (InCACCT), Gharuan, India, 5–6 May 2023; IEEE: Piscataway, NJ, USA, 2023; pp. 821–826. [Google Scholar]
Dhaliwal, D.S.; Williams, M.M. Sweet corn yield prediction using machine learning models and field-level data. Precis. Agric. 2024, 25, 51–64. [Google Scholar] [CrossRef]
Papageorgiou, E.I.; Markinos, A.T.; Gemtos, T.A. Fuzzy cognitive map based approach for predicting yield in cotton crop production as a basis for decision support system in precision agriculture application. Appl. Soft Comput. 2011, 11, 3643–3657. [Google Scholar] [CrossRef]
Manoj, T.; Makkithaya, K.; Narendra, V. A federated learning-based crop yield prediction for agricultural production risk management. In Proceedings of the 2022 IEEE Delhi Section Conference (DELCON), New Delhi, India, 11–13 February 2022; IEEE: Piscataway, NJ, USA, 2022; pp. 1–7. [Google Scholar]
Palaparthi, A.; Ramiya, A.; Ram, H.; Mishra, D. Classification of Horticultural Crops in High Resolution Multispectral Imagery Using Deep Learning Approaches. In Proceedings of the 2023 International Conference on Machine Intelligence for GeoAnalytics and Remote Sensing (MIGARS), Hyderabad, India, 27–29 January 2023; IEEE: Piscataway, NJ, USA, 2023; Volume 1, pp. 1–4. [Google Scholar]
Latha, S.B.; Gundavarapu, M.R.; Kumar, N.; Parameswari, D.L.; Reddy, B. Technology for kisan samanvayam: Nutrition intelligibility of groundnut plant using IoT-ML framework. Int. J. Recent Innov. Trends Comput. Commun. 2023, 11, 273–282. [Google Scholar] [CrossRef]
Tagarakis, A.; Bochtis, D. Sensors and Robotics for Digital Agriculture. Sensors 2023, 23, 7255. [Google Scholar] [CrossRef] [PubMed]
Navrozidis, I.; Pantazi, X.; Lagopodi, A.; Bochtis, D.; Alexandridis, T. Application of Machine Learning for Disease Detection Tasks in Olive Trees Using Hyperspectral Data. Remote Sens. 2023, 15, 5683. [Google Scholar] [CrossRef]
Cho, Y.J.; Gaju, O.; Govindan, V.; Lloyd, R.; Pearson, S.; Valluru, R. Canopy architectural and spectral phenotypes improve the prediction of radiation use efficiency in wheat canopies. TechRxiv 2023. [Google Scholar] [CrossRef]
Asiminari, G.; Moysiadis, V.; Kateris, D.; Busato, P.; Wu, C.; Achillas, C.; Sørensen, C.; Pearson, S.; Bochtis, D. Integrated Route-Planning System for Agricultural Robots. AgriEngineering 2024, 6, 657–677. [Google Scholar] [CrossRef]
Qi, C.; Gao, J.; Pearson, S.; Harman, H.; Chen, K.; Shu, L. Tea chrysanthemum detection under unstructured environments using the TC-YOLO model. Expert Syst. Appl. 2022, 193, 116473. [Google Scholar] [CrossRef]
Gong, L.; Yu, M.; Jiang, S.; Cutsuridis, V.; Pearson, S. Deep Learning Based Prediction on Greenhouse Crop Yield Combined TCN and RNN. Sensors 2021, 21, 4537. [Google Scholar] [CrossRef] [PubMed]
Alabi, T.R.; Abebe, A.T.; Chigeza, G.; Fowobaje, K.R. Estimation of soybean grain yield from multispectral high-resolution UAV data with machine learning models in West Africa. Remote Sens. Appl. Soc. Environ. 2022, 27, 100782. [Google Scholar] [CrossRef]
Sharma, K.; Shivandu, S.K. Integrating artificial intelligence and Internet of Things (IoT) for enhanced crop monitoring and management in precision agriculture. Sens. Int. 2024, 5, 100292. [Google Scholar] [CrossRef]
Mohammed, M.E.; Munir, M. Chapter 3-Towards smart farming: Applications of artificial intelligence and internet of things in precision agriculture. In Hyperautomation in Precision Agriculture; Singh, S., Sood, V., Srivastav, A.L., Ampatzidis, Y., Eds.; Academic Press: Cambridge, MA, USA, 2025; pp. 27–37. [Google Scholar] [CrossRef]
Devi, N.; Sarma, K.K.; Laskar, S. Design of an intelligent bean cultivation approach using computer vision, IoT and spatio-temporal deep learning structures. Ecol. Inform. 2023, 75, 102044. [Google Scholar] [CrossRef]
Kim, T.H.; Alzubi, A.A. AI-Enhanced Precision Irrigation in Legume Farming: Optimizing Water Use Efficiency. Legume Res. Int. J. 2024, 47, 1382–1389. [Google Scholar] [CrossRef]
San Emeterio de la Parte, M.; Martínez-Ortega, J.F.; Hernández Díaz, V.; Lucas Martínez, N. Big Data and precision agriculture: A novel spatio-temporal semantic IoT data management framework for improved interoperability. J. Big Data 2023, 10, 52. [Google Scholar] [CrossRef]
Ferraz, M.A.J.; Barboza, T.O.C.; Arantes, P.d.S.; Von Pinho, R.G.; Santos, A.F.d. Integrating Satellite and UAV Technologies for Maize Plant Height Estimation Using Advanced Machine Learning. AgriEngineering 2024, 6, 20–33. [Google Scholar] [CrossRef]
Botero-Valencia, J.; Reyes-Vera, E.; Ospina-Rojas, E.; Prieto-Ortiz, F. A Portable Tool for Spectral Analysis of Plant Leaves That Incorporates a Multichannel Detector to Enable Faster Data Capture. Instruments 2024, 8, 24. [Google Scholar] [CrossRef]
Oikonomidis, A.; Catal, C.; Kassahun, A. Hybrid deep learning-based models for crop yield prediction. Appl. Artif. Intell. 2022, 36, 2031822. [Google Scholar] [CrossRef]
Ahmed, M.S.; Tazwar, M.T.; Khan, H.; Roy, S.; Iqbal, J.; Rabiul Alam, M.G.; Hassan, M.R.; Hassan, M.M. Yield Response of Different Rice Ecotypes to Meteorological, Agro-Chemical, and Soil Physiographic Factors for Interpretable Precision Agriculture Using Extreme Gradient Boosting and Support Vector Regression. Complexity 2022, 2022, 5305353. [Google Scholar] [CrossRef]
Akintuyi, O.B. Adaptive AI in precision agriculture: A review: Investigating the use of self-learning algorithms in optimizing farm operations based on real-time data. Res. J. Multidiscip. Stud. 2024, 7, 16–30. [Google Scholar] [CrossRef]
Behera, S.; Menon, D.; Shenoy, G.V.; Suresh, J.M. Suggestion of Appropriate Crops Based on Rainfall and Underground Water Analysis. In Proceedings of the 2023 3rd International Conference on Intelligent Technologies (CONIT), Hubli, India, 23–25 June 2023; IEEE: Piscataway, NJ, USA, 2023; pp. 1–6. [Google Scholar]
Chaudhary, B.; Kumar, V. Emerging technological frameworks for the sustainable agriculture and environmental management. Sustain. Horizons 2022, 3, 100026. [Google Scholar] [CrossRef]
Karan, R.; Ramkumar, M.; Yokesh, M. Revolutionizing Tamilnadu Agriculture: AI-Powered Crop Yield Forecasting and Disease Prediction for Oil Seed Crops. In Proceedings of the 2024 1st International Conference on Cognitive, Green and Ubiquitous Computing (IC-CGU), Bhubaneswar, India, 1–2 March 2024; IEEE: Piscataway, NJ, USA, 2024; pp. 1–10. [Google Scholar]
Sihi, D.; Dari, B.; Kuruvila, A.P.; Jha, G.; Basu, K. Explainable machine learning approach quantified the long-term (1981–2015) impact of climate and soil properties on yields of major agricultural crops across conus. Front. Sustain. Food Syst. 2022, 6, 847892. [Google Scholar] [CrossRef]
Tamagno, S.; Eagle, A.J.; McLellan, E.L.; van Kessel, C.; Linquist, B.A.; Ladha, J.K.; Lundy, M.E.; Pittelkow, C.M. Predicting nitrate leaching loss in temperate rainfed cereal crops: Relative importance of management and environmental drivers. Environ. Res. Lett. 2022, 17, 064043. [Google Scholar] [CrossRef]
Suwanlee, S.R.; Keawsomsee, S.; Pengjunsang, M.; Homtong, N.; Prakobya, A.; Borgogno-Mondino, E.; Sarvia, F.; Som-ard, J. Monitoring Agricultural Land and Land Cover Change from 2001–2021 of the Chi River Basin, Thailand Using Multi-Temporal Landsat Data Based on Google Earth Engine. Remote Sens. 2023, 15, 4339. [Google Scholar] [CrossRef]
Abbasi, M.; Váz, P.; Silva, J.; Martins, P. Machine learning approaches for predicting maize biomass yield: Leveraging feature engineering and comprehensive data integration. Sustainability 2025, 17, 256. [Google Scholar] [CrossRef]
ISO 19440:2020; Enterprise Modelling and Architecture—Constructs for Enterprise Modelling. International Organization for Standardization: Geneva, Switzerland, 2020.
Rabhi, L.; Jabir, B.; Falih, N.; Afraites, L.; Bouikhalene, B. Digital transformation metamodel in smart farming: Crop classification prediction based on recurrent neural network. Foods Raw Mater. 2024, 13, 107–118. [Google Scholar] [CrossRef]
Bouni, M.; Hssina, B.; Douzi, K.; Douzi, S. Integrated IoT approaches for crop recommendation and yield-prediction using machine-learning. IoT 2024, 5, 634–649. [Google Scholar] [CrossRef]

Figure 1. Methodological process conducted for the systematic review following the PRISMA 2020 statement.

Figure 2. PRISMA flow diagram. Own elaboration based on Scopus and Web of Science.

Figure 3. Publications per year. Own elaboration based on Scopus and Web of Science.

Figure 4. Leading researchers in terms of number of publications and number of citations. Different groups of authors are identified with different circle colors. Own elaboration based on Scopus and Web of Science.

Figure 5. Main journals in terms of number of publications and number of citations. Different groups of journals are identified with different circle colors. Own elaboration based on Scopus and Web of Science.

Figure 6. Main countries in terms of number of publications and number of citations. Different groups of countries are identified with different circle colors. Own elaboration based on Scopus and Web of Science.

Figure 7. Topic evolution per year from 2007. Own elaboration based on Scopus and Web of Science.

Figure 8. Keywords co-occurrence network. Own elaboration based on Scopus and Web of Science.

Figure 9. Cartesian plane of keywords’ relevance and frequency. Different groups of keywords are identified with different circle colors. Own elaboration based on Scopus and Web of Science.

Figure 10. Research agenda based on studied topics. Own elaboration based on Scopus and Web of Science.

Table 1. Classification of keywords according to their function, applications, and characteristics. Own elaboration based on Scopus and Web of Science.

Keyword	Tools	Applications	Characteristics	References
Neural Networks	Artificial neural networks	Detection of crop diseases	Predictive models	[79]
Data Mining	Data mining algorithms	Pattern analysis in agricultural data	Discovery of hidden patterns	[45]
Internet of Things	IoT sensors, connected devices	Real-time monitoring of crop conditions	Automated management and control	[46,80,81]
Yield Prediction	Predictive performance models	Prediction of crop yields	Data-driven estimations	[82,83,84]
Deep Learning	Deep neural networks	Crop image classification	Image recognition and processing	[44,84,85]
Smart Farming	Drones, sensors, management systems	Precision agriculture and resource management	Optimization of agricultural practices	[17,51,86]
Remote Sensing	Satellite images, UAVs	Assessment of crop conditions at a distance	Detailed visual data	[7,78]
Artificial Intelligence	AI algorithms, intelligent systems	Automation and optimization of agricultural processes	Smart decision making	[1,31]

Table 2. Research gaps to be addressed in future research to advance the effective and sustainable application of ML in agriculture.

Category	Identified Gaps	Justification	Future Research Questions	References
Topic Gaps	1. Integration of different machine learning techniques. 2. Long-term impact of machine learning technologies on agricultural sustainability. 3. Adaptability of machine learning to various agroclimatic and socioeconomic conditions.	The integration of techniques can improve the precision and efficiency of agricultural practices. The long-term impact is crucial for evaluating the sustainability and ongoing effectiveness of these technologies. Adaptability is key to global applicability.	1. How can different machine learning techniques be integrated to optimize sustainable agriculture? 2. What is the long-term impact of machine learning technologies on agricultural sustainability? 3. How can machine learning adapt to diverse agroclimatic and socioeconomic conditions?	[28,81,104,105,106,107,108]
Geographical Gaps	1. Limited research in underdeveloped and developing regions. 2. Lack of comparative studies between different climatic zones. 3. Lack of research in the specific agricultural contexts of each country.	Underdeveloped and developing regions face unique challenges that require tailored solutions. Comparative studies can identify the best practices applicable in different climates. Specific contexts can offer effective local solutions.	1. What are the barriers and opportunities for the use of machine learning in agriculture in underdeveloped regions? 2. How does the effectiveness of machine learning techniques vary across different climatic zones? 3. What specific adaptations are required to implement machine learning in the agriculture of each country?	[71,104,106,109]
Interdisciplinary Gaps	1. Limited collaboration between disciplines. 2. Limited integration of behavioral sciences and socioeconomic knowledge in machine learning studies. 3. Lack of interdisciplinary approaches to address complex challenges in agriculture.	Interdisciplinary collaboration can provide holistic and effective solutions to agricultural problems. Integrating behavioral sciences and socioeconomics can improve the adoption and effectiveness of machine learning technologies. Interdisciplinary approaches are essential to address complex challenges.	1. How can interdisciplinary collaboration be improved in research on machine learning in agriculture? 2. How can behavioral sciences and socioeconomics be integrated into machine learning studies? 3. What interdisciplinary approaches are most effective for addressing complex agricultural challenges?	[81]
Temporal Gaps	1. Need for longitudinal studies. 2. Lack of research analyzing the evolution and trends in machine learning usage in agriculture. 3. Insufficiency of historical studies comparing the advancement of machine learning technologies in different periods.	Longitudinal studies are necessary to evaluate the long-term impact and evolution of technologies. Analyzing usage trends can identify emerging developments and changes in agricultural practices. Historical studies can provide a perspective on technological progress.	1. What is the long-term impact of using Machine Learning in agriculture? 2. How have the trends in the use of machine learning in agriculture evolved in recent years? 3. How does the advancement of machine learning technologies compare across different historical periods?	[1,31,32]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

© 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).

Share and Cite

MDPI and ACS Style

Botero-Valencia, J.; García-Pineda, V.; Valencia-Arias, A.; Valencia, J.; Reyes-Vera, E.; Mejia-Herrera, M.; Hernández-García, R. Machine Learning in Sustainable Agriculture: Systematic Review and Research Perspectives. Agriculture 2025, 15, 377. https://doi.org/10.3390/agriculture15040377

AMA Style

Botero-Valencia J, García-Pineda V, Valencia-Arias A, Valencia J, Reyes-Vera E, Mejia-Herrera M, Hernández-García R. Machine Learning in Sustainable Agriculture: Systematic Review and Research Perspectives. Agriculture. 2025; 15(4):377. https://doi.org/10.3390/agriculture15040377

Chicago/Turabian Style

Botero-Valencia, Juan, Vanessa García-Pineda, Alejandro Valencia-Arias, Jackeline Valencia, Erick Reyes-Vera, Mateo Mejia-Herrera, and Ruber Hernández-García. 2025. "Machine Learning in Sustainable Agriculture: Systematic Review and Research Perspectives" Agriculture 15, no. 4: 377. https://doi.org/10.3390/agriculture15040377

APA Style

Botero-Valencia, J., García-Pineda, V., Valencia-Arias, A., Valencia, J., Reyes-Vera, E., Mejia-Herrera, M., & Hernández-García, R. (2025). Machine Learning in Sustainable Agriculture: Systematic Review and Research Perspectives. Agriculture, 15(4), 377. https://doi.org/10.3390/agriculture15040377

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Menu