AI Algorithms in the Agrifood Industry: Application Potential in the Spanish Agrifood Context

Abstract

This research explores the prospective implementations of artificial intelligence (AI) algorithms within the agrifood sector, focusing on the Spanish context. AI methodologies, encompassing machine learning, deep learning, and neural networks, are increasingly integrated into various agrifood sectors, including precision farming, crop yield forecasting, disease diagnosis, and resource management. Utilizing a comprehensive bibliometric analysis of scientific literature from 2020 to 2024, this research outlines the increasing incorporation of AI in Spain and identifies the prevailing trends and obstacles associated with it in the agrifood industry. The findings underscore the extensive application of AI in remote sensing, water management, and environmental sustainability. These areas are particularly pertinent to Spain’s diverse agricultural landscapes. Additionally, the study conducts a comparative analysis between Spain and global research outputs, highlighting its distinctive contributions and the unique challenges encountered within its agricultural sector. Despite the considerable opportunities presented by these technologies, the research identifies key limitations, including the need for enhanced digital infrastructure, improved data integration, and increased accessibility for smaller agricultural enterprises. The paper also outlines future research pathways aimed at facilitating the integration of AI in Spain’s agriculture. It addresses cost-effective solutions, data-sharing frameworks, and the ethical and societal implications inherent to AI deployment.

1. Introduction

Artificial intelligence (AI) is increasingly acknowledged as a pivotal catalyst within the agrifood sector, addressing imperative issues related to productivity enhancement, resource optimization, and environmental stewardship. Specifically, AI algorithms, encompassing machine learning, deep learning, and optimization techniques, are being used to refine agricultural practices such as crop yield prediction, disease identification, and water resource management. In the context of Spain, where agriculture plays a substantial role in the national economy and is influenced by a variety of climatic conditions, the integration of AI presents significant opportunities for enhancing both operational efficiency and sustainability within the sector.

This systematic analysis, underpinned by AI, big data, and Data Mining methodologies, examines the prospective applications of AI algorithms in the Spanish agrifood industry. The paper delivers an exhaustive review of AI’s current and future contributions to this domain. It begins with a comprehensive overview of cutting-edge AI algorithms that are currently used or have the potential for application in the global agrifood industry. This highlights methodologies such as supervised and unsupervised learning, neural networks, and reinforcement learning, among others. Subsequently, the methodology section elucidates the data acquisition and analytical tools used. For that purpose, it emphasizes the bibliometric approach and big data techniques used to research AI trends and patterns within the agrifood sector from 2020 to 2024.

Section 4 offers an in-depth analysis of AI applications within Spain’s agricultural landscape, concentrating on areas such as remote sensing, crop monitoring, and precision agriculture. Additionally, the study identifies prominent trends in AI adoption across various agricultural sub-sectors, underscoring the leading role of larger enterprises in driving AI-driven innovations in contrast to smaller farms, which frequently encounter barriers to implementing these technologies. The discussion segment addresses the inherent limitations and obstacles associated with AI integration in Spain. In particular, emphasis is made on the need for enhanced digital infrastructure, improved data integration, and greater accessibility for small to medium-sized agricultural enterprises.

Ultimately, the paper outlines future research pathways aimed at fostering the adoption of AI for agriculture in Spain. This includes the development of scalable AI solutions tailored to Spain’s diverse agricultural environments and contributions toward achieving sustainability goals. The insights and recommendations provided are intended to guide policymakers, researchers, and industry stakeholders in accelerating AI integration within Spanish agriculture, thereby ensuring its extensive impact on the sector’s productivity and environmental management.

2. State of the Art: AI Algorithms Applied or Applicable in the Global Agrifood Industry

Artificial intelligence (AI) algorithms, encompassing a diverse array of methodologies such as supervised and unsupervised learning, neural network-based approaches, reinforcement learning, and other emergent technologies are driving profound transformations within the global agrifood sector. These sophisticated computational techniques are being deployed to surmount challenges that previously seemed insurmountable. In turn, they enhance operational efficiency and foster sustainability across the entire agrifood value chain. AI applications in this domain are multifaceted. They can go from precision crop monitoring and pest detection in agricultural practices to advanced predictive modeling and resource optimization during the industrial phases of production.

This section undertakes a thorough review of AI algorithms that, based on an exhaustive analysis of pertinent academic literature, exhibit the highest potential for application in the agrifood industry or are already being actively used. The selection of these specific algorithms and methodological families are informed by the foundational insights presented in “Artificial Intelligence: A Modern Approach” [1], a groundbreaking work that outlines the fundamental principles and state-of-the-art advancements in AI. While this review is not intended to encompass an exhaustive survey of all existing AI algorithms—a task more suited to a comprehensive monograph—it highlights proven key methodologies within the agrifood sector. These include, but are not limited to, machine learning, deep learning, and optimization techniques. Each of them has demonstrated its efficacy in addressing targeted challenges within agriculture and food processing.

The integration of these AI algorithms into the agrifood industry addresses immediate operational challenges and contributes to long term sustainability goals by promoting efficient resource utilization and reducing environmental impacts. As AI technologies continue to evolve, their applications within the agrifood sector are expected to expand, driving further innovations and enhancing the resilience of agricultural systems globally.

2.1. Supervised Learning: A Subset of the Machine Learning Family

Supervised learning constitutes a foundational approach within the realm of machine learning, wherein models are trained using labeled datasets to perform predictions or classifications. As presented by Christopher M. Bishop [2], these algorithms establish a functional mapping between inputs and outputs based on exemplary input-output pairs. This methodology is widely used across various domains, including this research subject, where its applications are increasingly yielding substantial advancements.

In the realm of supervised learning, the architecture of models, the quality of data sources, and the achieved accuracy rates are pivotal factors influencing performance. A notable example is the application of convolutional neural networks (CNNs) in image classification tasks, such as handwritten digit recognition using the MNIST dataset. The MNIST dataset comprises 60,000 training images and 10,000 testing images of 28 × 28 pixel grayscale handwritten digits. In a recent study [3], researchers employed an ensemble approach involving three CNN models with varying kernel sizes (3 × 3, 5 × 5, and 7 × 7). Each model consisted of multiple convolutional layers followed by a single fully connected layer, utilizing batch normalization and ReLU activation functions, without pooling layers. Data augmentation techniques, including rotation and translations were applied to enhance the training dataset. This ensemble achieved a remarkable accuracy of 99.87% on the MNIST test set, underscoring the significance of selecting appropriate model architectures and utilizing high-quality data sources in supervised learning applications. In the agricultural sector, supervised learning plays an instrumental role in the detection of pests and diseases through the analysis of aerial or drone-acquired imagery. Models trained on annotated images possess the capability to autonomously identify afflicted plants, thereby enabling prompt interventions and mitigating potential crop losses. For instance, convolutional neural networks (CNNs), a specialized subset of supervised learning algorithms, demonstrate exceptional efficacy in image classification tasks and are increasingly being applied to agricultural image datasets [4]. The integration of machine learning with advanced imaging techniques, such as hyperspectral and multispectral imaging, facilitates the detection of not only visible diseases but also early-stage infections that may elude human observation [5].

A pertinent application of supervised learning in agriculture is exemplified by the detection of early blight in tomato crops utilizing CNNs. Mohanty et al. [6] developed a deep learning model that achieved high accuracy in discerning between healthy and diseased leaves, thereby empowering farmers to undertake timely remedial actions. Similarly, in the Spanish context, supervised learning has been used to monitor olive groves for indications of Xylella fastidiosa infection, a bacterium that poses a severe threat to olive production [7]. Furthermore, supervised learning methodologies have been extensively used in yield prediction models. By analyzing historical data encompassing weather patterns, soil quality, and previous yields, these models can forecast future crop productivity with considerable precision [8].

Moreover, classification algorithms within the supervised learning framework are indispensable for automating the sorting and grading of agricultural products, such as fruits and vegetables. By training models on labeled datasets that encapsulate product attributes (e.g., color, size, and texture), these algorithms can categorize produce into various quality grades, thereby streamlining post-harvest processing [9]. For instance, Cubero et al. [10] devised an automated system employing machine vision to classify citrus fruits based on external defects, thereby enhancing efficiency within packing facilities.

Advancements in supervised learning are also being leveraged in precision farming, where they support decision-making processes through the real-time analysis of sensor data. Models trained on soil and environmental conditions can furnish customized recommendations for irrigation, fertilizer application, and pesticide use, thereby ensuring optimal crop growth while minimizing environmental impacts [11]. Additionally, recent studies demonstrate that supervised models, when trained on environmental datasets, can be adapted to assess spatiotemporal contamination patterns in water bodies. By incorporating hydrological and ecological variables, these models offer refined insights into the distribution of heavy metal pollution and its ecological risks in impounded lakes [12].

An illustrative example of supervised learning in the agrifood sector is the detection of late blight disease in potato crops using image-based classification models. Suarez Baron et al. [13] applied convolutional neural networks (CNN) and support vector machines (SVM) to identify infected potato plants based on field-acquired images. Through pre-processing and feature extraction, their model achieved high classification accuracy, with an AUC of 0.97 using CNN and 0.87 with SVM. While this study was conducted in Colombia, its methodological framework demonstrates a transferable potential for similar implementations in Spanish agricultural systems, emphasizing early disease detection through supervised learning models.

These developments underscore the transformative potential of supervised learning in enhancing agricultural productivity and sustainability. By enabling precise monitoring, early disease detection, and optimized resource management, supervised learning significantly contributes to the advancement of modern agricultural practices.

2.2. Unsupervised Learning: A Subset of the Machine Learning Family

Unsupervised learning algorithms, engineered to uncover intrinsic patterns within unlabeled datasets, are meticulously examined in “The Elements of Statistical Learning” [14] by Hastie et al. These algorithms are particularly efficient in environments where explicit labels or predefined categories are absent, thereby proving invaluable for identifying underlying structures within intricate data landscapes. A notable example is the application of autoencoders for anomaly detection in image data. Autoencoders are neural networks designed to learn efficient coding of input data, typically consisting of an encoder that maps the input into a latent space and a decoder that reconstructs the original input from this representation. For instance, in a study on unsupervised feature learning, a deep trans-layer network was employed, achieving 99.45% accuracy on the MNIST dataset, which comprises 60,000 training images and 10,000 testing images of 28 × 28 pixel grayscale handwritten digits [15].

In the agrifood industry, unsupervised learning methodologies are adeptly applied to segment cultivation zones based on attributes such as soil properties, temperature variations, and humidity levels. This data-driven segmentation facilitates more precise agricultural management by enabling targeted interventions tailored to each specific zone. For instance, clustering techniques such as k-means and hierarchical clustering are effectively used to partition fields into smaller, more manageable sections, thereby optimizing the allocation of resources such as water and fertilizers. Through these applications, unsupervised learning algorithms contribute to the promotion of sustainable practices, enhancement of productivity, reduction of waste, and overall improvement of operational efficiency across diverse farmland areas. Their ability to process extensive datasets and adapt to evolving conditions further underscore their practicality and reliability in the context of precision agriculture.

Within the Spanish agrifood sector, unsupervised learning has been instrumental in optimizing wheat crops. Gómez-Candón et al. [16] used clustering algorithms to segment wheat groves based on canopy characteristics derived from aerial imagery, thereby enabling differentiated harvesting strategies. Also applicable to the Spanish agrilocal sector is the research by Lin et al. [17] which explored clustering techniques in multitemporal crop phenology monitoring, developing a model for detecting growth stages in rice fields through satellite data—an approach that is both adaptable to other crops and scalable for smaller plots in data-limited environments.

Additional studies exemplify the extensive applications of clustering and unsupervised learning in agriculture. For example, a recent hybrid model proposed by Chai et al. [18] integrates self-supervised learning with forecasting techniques to enhance resource management in complex environments, with applications to energy-efficient processes in agricultural facilities. This integration underscores a synergy between advanced computational approaches and sustainable practices, emphasizing the role of hybrid learning in optimizing resource allocation and minimizing environmental impacts within the agrifood sector.

An illustrative implementation of unsupervised learning relevant to the Spanish agrifood context is the segmentation and classification of tree crowns in overlapping orchards, as explored by Moussaid et al. [19] Although conducted in Morocco, the study’s focus on citrus orchards within a Mediterranean-like climate mirrors the agricultural conditions of Spanish regions such as Valencia or Murcia. This reference was selected from the dataset obtained through a targeted SCOPUS search, reflecting the closest geographic and methodological relevance identified within the available academic literature. Using satellite imagery from the Mohammed VI satellite, their model successfully segmented orchard rows and classified trees into categories based on canopy size, achieving an F-measure score of 0.93. This demonstrates the potential applicability of similar machine learning frameworks for orchard management in Spain, enabling precise monitoring of tree health and facilitating sustainable agricultural practices through data-driven decision-making.

These advancements underscore the transformative potential of unsupervised learning in fostering sustainable and efficient agricultural practices. By enabling precise segmentation, anomaly detection, and resource optimization, unsupervised learning not only addresses current challenges but also paves the way for future innovations in the agrifood industry.

2.3. Support Vector Machines (SVMs): A Supervised Learning Algorithm

Support vector machines (SVMs) constitute a prominent class of supervised learning models extensively used for both classification and regression tasks. SVMs function by identifying the optimal hyperplane that segregates data points into distinct categories. This methodological approach renders SVMs exceptionally efficient at managing complex datasets, particularly when integrated with kernel methods, which empower the model to capture intricate non-linear relationships [20]. Owing to their robustness in handling noisy datasets and their adaptability through kernel techniques, SVMs have proven highly effective in the detection and identification of pests and plant diseases. By scrutinizing spectral data or imagery obtained from sensors and drones, SVMs can discern between healthy and afflicted crops with remarkable precision. This enhanced classification capability facilitates more accurate and targeted interventions in crop health management. A pioneering and enduring example demonstrating the resilience of SVMs in agricultural applications is presented by Camps-Valls and Bruzzone [21]. Nearly two decades ago, their study established a foundational benchmark by showcasing the efficacy of SVMs in classifying hyperspectral images of agricultural regions. By adeptly handling the high-dimensional and noisy data inherent in such imaging techniques, the authors illustrated that SVMs achieved superior classification accuracy compared to traditional classifiers, successfully differentiating between various crop types and soil conditions. Since then, SVMs have been extensively used for similar purposes in numerous studies. Thereby, contributing to more efficient resource allocation and enhanced overall agricultural productivity.

In contemporary applications, Pal and Mather [22] leveraged SVMs for land cover classification utilizing remotely sensed data, a critical task for monitoring agricultural landscapes and orchestrating resource distribution. Additionally, SVMs have been applied to classify vineyard conditions based on leaf spectral signatures, thereby supporting precision viticulture [23]. Building on this, El Jamaoui et al. conducted a case study in Corvera, Murcia, employing SVM-based models for spatial prediction of soil organic carbon in Mediterranean agricultural landscapes. Their approach effectively captured spatial variability, outperforming conventional interpolation techniques and enabling more accurate soil mapping. This highlights the adaptability of SVMs in precision agriculture, particularly in regions such as Spain and southern Europe, where soil heterogeneity plays a critical role in determining crop yields [24].

In the field of agriculture, support vector machines (SVMs) have been effectively applied to classify stress conditions in pepper seedlings [25]. Researchers cultivated two-week-old pepper seedlings under varying environmental conditions, including different temperatures (20, 25, and 30°C), light intensities (50, 250, and 450 µmol m⁻² s⁻¹), and photoperiods (8/16, 10/14, and 16/8 h). Daily images of the seedling canopies were captured using a low-cost RGB camera over a two-week period. From these images, 18 color features, 9 texture features using the gray-level co-occurrence matrix (GLCM), and 1 morphological feature were extracted. A sequential feature selection (SFS) method was applied to identify the optimal features for the classification model. The SVM model, evaluated with the receiver operating characteristic (ROC) curve and confusion matrix, achieved a classification accuracy of 85%. This classification approach enables real-time stress monitoring, allowing growers to optimize environmental conditions and enhance seedling growth. These implementations underscore the versatility and effectiveness of SVMs in diverse agricultural contexts. Further expanding these applications, Dhal et al. [26] introduced an SVM-based model for smart irrigation control in turfgrass systems. This innovation exemplifies how SVMs enable early intervention in vegetation management within agricultural settings, thereby enhancing the viability of small-scale farms with limited resources. Similarly, Xiao et al. [27] developed an SVM framework for the early prediction of maize resistance to herbicides, optimizing herbicide application based on specific crop conditions. This approach aligns with sustainable resource management practices and addresses prevalent herbicide resistance issues in sensitive agricultural regions. Collectively, these advancements underscore the foundational role of SVMs in advancing agricultural technology, promoting sustainability, and enhancing productivity across diverse farming environments.

2.4. Genetic Algorithms: An Optimization Algorithm Inspired by Evolutionary Biology

Genetic algorithms (GAs) represent a sophisticated class of optimization techniques inspired by the principles of natural selection and genetics. These algorithms simulate evolutionary processes to resolve complex problems by generating a population of potential solutions and iteratively applying selection, crossover, and mutation operators to evolve solutions with enhanced fitness across successive generations [28].

Within the agrifood industry, GAs are optimally used to determine the most efficient production schedules in food manufacturing facilities, such as optimizing the sequence of processing tasks in dairy plants. By considering factors such as processing times, resource constraints, and delivery deadlines, GAs can generate optimized schedules that bolster operational efficiency, reduce production costs, and ensure the timely delivery of fresh products. This application not only increases productivity, but also contributes to waste reduction and increased customer satisfaction within the agrifood supply chain.

For instance, Liu YC et al. [29] applied genetic algorithms to optimize the scheduling of batch processing in food production, resulting in significant cost savings. In the Spanish context, GAs have been used to optimize the energy costs associated with irrigation networks, encompassing both drip and sprinkler systems. The escalating energy expenses linked to these systems, primarily due to pressure demands, have demanded innovative solutions. Spanish researchers have demonstrated that effective management of irrigation schedules through GAs can substantially reduce energy expenditures by approximately 6–7%. Traditionally, irrigation optimization required adherence to stringent minimum pressure requirements, which posed computational challenges. However, by reorganizing irrigation schedules and integrating daily volume requirements at each hydrant, this approach not only minimizes energy consumption but also ensures adequate pressure at critical hydrants [30]. The Spanish implementation achieves remarkable efficiency through a multi-objective approach, coupled with the parallel evaluation of objective functions, thereby accelerating the optimization process by an order of magnitude. This advancement facilitates daily irrigation optimization, thereby fostering greater sustainability in agricultural practices across Spain.

In agricultural applications, genetic algorithms (GAs) have been effectively utilized to optimize predictive models for crop yield forecasting. A notable recent study by Mahmud et al. [31] presents a hybrid model combining GAs with a random forest (RF) classifier. The model was trained on a dataset encompassing various features such as soil composition, weather conditions, and historical crop yields. The GA component optimized the hyperparameters of the RF classifier, enhancing its predictive performance. This integrated approach achieved an impressive accuracy rate of 99.3% in predicting the outcomes for 22 different crops. The study underscores the efficacy of combining genetic algorithms with machine learning techniques to enhance predictive accuracy in agricultural applications.

Another notable example of genetic algorithms (GAs) applied in the Spanish agrifood sector are optimizing irrigation networks, as explored by Perea et al. [32]. Their study, conducted at the University of Castilla-La Mancha in Spain, proposed a decision support system based on GAs for managing irrigation schedules in agricultural fields. The approach integrated variable-speed pumping technologies to reduce water and energy consumption while ensuring optimal irrigation coverage. By dynamically adjusting water delivery schedules, the system achieved significant improvements in resource efficiency, demonstrating how GAs can address pressing sustainability challenges in water-scarce agricultural regions.

These developments underscore the transformative potential of genetic algorithms in enhancing operational efficiencies and promoting sustainable practices within the agrifood sector. By enabling optimized scheduling and resource management, GAs contribute significantly to reducing costs, minimizing waste, and improving overall productivity, thereby supporting the industry’s sustainability objectives.

2.5. Convolutional Neural Networks (CNNs): A Specialized Deep Learning Architecture for Grid-Structured Data

Convolutional neural networks (CNNs) represent an advanced category of neural networks meticulously designed to process data with grid-like structures, such as images, with exceptional efficiency. These networks utilize convolutional layers to autonomously and hierarchically extract features from input data, thereby rendering them particularly efficient at tasks involving pattern recognition and image classification [33].

Li and Zhang [34] present a lightweight deep learning model using CNNs for image segmentation tasks, demonstrating its potential in agricultural applications such as precision monitoring of crop health and waste management. Although developed primarily for underwater waste segmentation, the model’s architecture is optimized for high efficiency in complex environments, which translates well to agricultural settings where resource constraints are common. This CNN approach can facilitate the detection of crop residues and contaminants, offering a scalable solution for enhancing the cleanliness and sustainability of farmland. Their work underscores the adaptability of CNNs across diverse agricultural challenges, from monitoring soil quality to supporting eco-friendly waste management practices.

In agricultural applications, convolutional neural networks (CNNs) have been effectively utilized for plant disease identification ref. [35] using a hybrid model combining CNNs with Vision Transformers. The proposed PlantXViT model comprises a lightweight structure with only 0.8 million trainable parameters, making it suitable for Internet of Things (IoT)-based smart agriculture services. The model was trained and evaluated on five publicly available datasets, achieving an average accuracy exceeding 93.55% on the Apple dataset, 92.59% on the Maize dataset, and 98.33% on the Rice dataset, even under challenging background conditions. These results underscore the efficacy of integrating CNN architectures with Vision Transformers and utilizing high-quality data sources in supervised learning applications within agriculture.

Another illustrative application of convolutional neural networks (CNNs) in an environmental context relevant to agriculture is short-term spatio-temporal forecasting of air temperatures, as investigated by García-Duarte et al. [36]. Their study used deep graph CNNs to predict meteorological variables critical for agricultural planning. The proposed approach is validated using real, open weather data from 37 meteorological stations in Spain, ensuring its applicability to regional agricultural contexts. Although focused on environmental forecasting, the method is highly relevant to precision agriculture, where accurate climatic predictions enhance irrigation and crop management strategies. This integration underscores the versatility of CNN-based models in supporting data-driven agricultural decision-making.

Given the paramount significance of the viticulture sector in various regions of Spain, the ability to efficiently process large datasets through CNNs is particularly noteworthy. These networks facilitate real-time monitoring and the seamless integration of vegetation indices with other remote sensing data, thereby improving the precision of plant health assessments and contributing to more informed agricultural decision-making in large-scale viticulture operations. For example, Kerkech et al. [37] used CNNs for the early detection of diseases in vineyards using UAV imagery, thereby enabling timely and targeted management interventions to mitigate vine diseases.

2.6. K-Means Clustering: An Unsupervised Learning Algorithm for Data Partitioning

Within the agricultural sector, k-means clustering is extensively used to delineate homogeneous zones across expansive fields, which may exhibit disparities in soil characteristics, moisture levels, or nutrient requirements. This analytical method facilitates the adoption of targeted management practices, such as precision irrigation and tailored fertilization strategies. By grouping regions with similar attributes, farmers can optimize resource allocation, reduce waste, and subsequently enhance crop yields.

Recent advancements have broadened the scope of k-means clustering applications within agriculture. For example, Chang et al. [38] used k-means clustering to monitor canopy height uniformity, a critical indicator of crop health that significantly improves growth prediction and management accuracy, especially in large-scale field settings. This application underscores the algorithm’s versatility and its capacity to contribute to comprehensive crop health assessments. Thereby, it supports more informed and effective agricultural decision-making processes. In addition, the k-means algorithm can categorize soil samples into distinct fertility classes, as mentioned by Hayatu et al. [39], and in the refinement of agricultural maps using UAV and machine learning methods, as explored by Mazzia et al. [40]. Their study, conducted in Italy, applied k-means clustering to enhance the classification of crop types from moderate resolution satellite imagery. A vineyard in Serralunga d’Alba (Northern Italy) was chosen as a case study for validation purposes, a setting with climatic and agricultural characteristics similar to viticultural regions in Spain.

These developments highlight the enduring relevance and adaptability of k-means clustering in addressing the multifaceted challenges of modern agriculture. By enabling precise segmentation and insightful data analysis, k-means clustering continues to support the evolution of sustainable and efficient agricultural practices, thereby playing a crucial role in the advancement of the agrifood industry.

2.7. Random Forest: A Supervised Ensemble Learning Algorithm Using Decision Trees

A seminal study by Belgiu and Drăguţ [41] demonstrated the superiority of random forest in land cover classification using Sentinel-2 data. Building upon their findings, subsequent research has corroborated that RF classifiers exhibit high effectiveness in remote sensing applications, particularly excelling with high-dimensional datasets such as hyperspectral imagery. The inherent resistance of random forest to overfitting, even when dealing with complex training samples, ensures a reliable classification accuracy. Moreover, the internal variable importance measures (e.g., mean decrease in accuracy) facilitate refined feature selection, thereby enhancing model efficiency and precision across various remote sensing data types, including multispectral, radar, and LiDAR sources.

Recent research further highlights the versatility of RF models in diverse agricultural contexts. For instance, Rodríguez-Díaz et al. [42] present a machine learning-based approach utilizing RF models for forecasting pest infestations in olive groves and vineyards. By providing insights into pest dynamics, the model aids farmers in implementing timely interventions, thus promoting sustainable pest control practices. This research underscores the value of random forest in the agricultural sector, where predictive accuracy and model transparency are key for effective field management. T. Zhou et al. [43] investigate the effects of managed aquifer recharge (MAR) on the behavior of pesticides within agricultural soils, employing RF algorithms to model pesticide transport and retention, understanding complex soil-water interactions, and providing crucial insights for sustainable water resource management and pollution mitigation in agricultural landscapes.

Random forest models have been effectively utilized too for crop yield prediction [44], implementing a random forest model to forecast crop yields. The dataset encompassed variables such as rainfall, crop type, meteorological conditions, area, production, and yield. The random forest model achieved a maximum accuracy of 98.96%, a mean absolute error of 1.97, a root mean square error of 2.45, and a standard deviation of 1.23. These results underscore the efficacy of random forest models in agricultural yield prediction, highlighting the importance of selecting appropriate model architectures and utilizing high-quality data sources in supervised learning applications.

Another notable example of random forest application in the Spanish agricultural context is the estimation of forest above-ground biomass, as investigated by Tanase et al. [45]. Their study used random forest models to estimate biomass in Mediterranean forests ecosystems using remote sensing data from multispectral and LiDAR sensors. The researchers developed and tested a comprehensive framework to map forest above-ground biomass, canopy cover, and forest height across two ecologically significant regions in peninsular Spain: the Mediterranean biome in the Madrid region and the temperate biome in the Basque Country. This approach demonstrated high accuracy in biomass prediction, providing essential insights into forest management and environmental monitoring. The study underscores how random forest can support sustainable land management practices through data-driven decision-making and precise environmental assessment.

2.8. Reinforcement Learning: A Paradigm of Autonomous Learning Focused on Decision-Making Through Interaction with an Environment

Reinforcement Learning (RL) constitutes a distinct paradigm within machine learning, emphasizing the decision-making processes of agents operating within dynamic environments to maximize cumulative rewards. Richard S. Sutton and Andrew G. Barto provide a comprehensive introduction to this methodology in “Reinforcement Learning: An Introduction”, wherein they elaborate on its foundational principles [46] and delve into practical applications and solution methods in subsequent chapters.

Recent advancements in RL have catalyzed innovative applications within the agricultural sector, offering adaptive solutions that optimize resource management and enhance production efficiency, involving the utilization of RL in crop yield prediction, where real-time environmental data informs cultivation decisions, thereby fostering agricultural practices that exhibit greater resilience to climate variability [47]. These studies underscore RL’s potential not only to automate but also to personalize agricultural tasks, such as monitoring and irrigation control, thereby increasing the sustainability of contemporary farming systems and presenting a Deep Recurrent Q-Network (DRQN) model that integrates a Recurrent Neural Network (RNN) with Q-Learning. The model processes sequentially data related to environmental factors, soil conditions, water availability, and crop parameters. The RNN component captures temporal dependencies in the data, while the Q-Learning algorithm constructs a crop yield prediction environment based on the input parameters. A linear layer maps the RNN outputs to Q-values, enabling the reinforcement learning agent to predict crop yields by minimizing error and maximizing forecast accuracy. The model was trained and evaluated on a dataset comprising various agronomic factors, achieving an accuracy of 93.7% in crop yield prediction.

Expanding beyond agriculture, recent research on RL demonstrates its versatile applications across a spectrum of fields, including healthcare. For instance, Chen et al. [48] developed an RL-based model that integrates crop data to optimize resource allocation, dynamically adapting to varying agricultural conditions. This model enhances farming decision-making by aligning resource distribution with real-time data inputs. Similarly, Yang et al. [49] explored the application of RL for irrigation control, incorporating deep learning techniques to adjust water usage based on environmental factors. This model effectively reduces water waste, thereby promoting sustainable agricultural practices.

In agriculture, Gautron et al. [50] demonstrate RL, including multi-armed bandits, as an effective machine learning method for sequential decision-making in uncertain conditions. While RL is widely recognized for successes such as creating the world’s best Go player, its application potential spans various fields, including crop management decision support systems (DSS). In these systems, RL offers an interactive, action-oriented, and context-sensitive approach to managing sequential crop operations under uncertainty. A review shows a limited but expanding body of work with promising applications for real-world, human-centered DSS in agriculture, calling for combined efforts from RL and agronomy experts to advance theoretical and practical development in this area.

Collectively, these studies illustrate the expansive usefulness of RL in providing tailored, efficient, and sustainable solutions. By enabling both automation and personalization of critical tasks, RL significantly contributes to the advancement of operational efficiencies and the achievement of sustainability goals within diverse industries.

2.9. Natural Language Processing (NLP): A Subfield of AI Focused on the Interaction Between Computers and Human Language

In recent years, natural language processing (NLP) has undergone a profound transformation, emerging as one of the most groundbreaking advancements in artificial intelligence (AI). This evolution holds significant implications for its application within the agrifood sector. Despite the transformative potential of NLP, its benefits remain underexplored in the existing literature on AI applications in agriculture, as evidenced by the reviewed articles and referenced academic works.

NLP’s capacity to analyze and interpret vast volumes of unstructured textual data—ranging from weather reports and scientific publications to market trend analyses—introduces an innovative dimension of insight for the agrifood industry. By extracting pertinent information, NLP facilitates more informed decision-making processes that can significantly influence agricultural practices, supply chain optimization, and food security management. The integration of NLP into AI-driven systems offers considerable advantages, including enhanced predictive capabilities for food pricing, climate variability, and pest detection.

Recent scholarly investigations into NLP algorithms demonstrate their extensive applicability across diverse fields. For instance, Nguyen et al. [51] developed an NLP-based system designed to analyze the composition of life sciences articles, thereby automating the extraction of biological and chemical information from unstructured text. This methodology enhances data accessibility in biomedical research by converting extensive textual information into structured datasets. Similarly, Hernández-Romero et al. [52] developed an NLP-based system designed to analyze the composition of life sciences articles, thereby automating the extraction of biological and chemical information from unstructured text. This methodology enhances data accessibility in biomedical research by converting extensive textual information into structured datasets. In the context of Spain, Domínguez et al. [53] determined that NLP, particularly through sentiment analysis of social media posts, serves as an instrumental tool for evaluating agricultural and rural policies, such as the European Common Agricultural Policy (CAP). Their study elucidates how the analysis of public sentiment on platforms such as Twitter (currently X) uncovers citizen concerns and priorities in areas including environmental policies, food safety, and funding allocation. The findings suggest that incorporating NLP into the policymaking process effectively complements traditional evaluation methods, facilitating a more responsive approach to public opinion and improving policy alignment with societal needs. Furthermore, Ortego-Antón [54] explored the application of NLP in linguistic analysis, specifically in translating descriptors within the agrifood sector. This application supports the internationalization of Spanish food products by generating culturally accurate translations, thereby enhancing market accessibility and competitiveness.

2.10. Fuzzy Logic: A Mathematical Approach for Handling Uncertainty and Imprecision in Decision-Making Systems

Fuzzy logic provides a robust and versatile framework for decision-making in environments marked by uncertainty and imprecision, rendering it exceptionally suitable for applications where traditional binary logic falls short. In the agricultural domain, fuzzy logic plays a critical role in optimizing systems such as automatic irrigation and fertilization. These systems must navigate uncertain variables, including fluctuating soil moisture levels, varying temperatures, and nutrient concentrations. By facilitating reasoning based on degrees of truth rather than strict binary classifications, fuzzy logic enables these systems to execute more nuanced and adaptive adjustments. Consequently, this enhances resource efficiency and promotes the adoption of improved agricultural practices, particularly within the realm of precision farming. Here, the judicious use of water and fertilizers is paramount for achieving both sustainability and productivity. These foundational concepts are comprehensively examined in “Fuzzy Logic with Engineering Applications” [55].

A practical implementation of fuzzy logic is exemplified by the deployment of a Fuzzy Logic Controller (FLC) for managing the microclimate within agricultural greenhouses. This application addresses the complexity introduced by numerous perturbations and a multitude of input parameters that influence the internal climate [56]. Additionally, Jamroen et al. [57] developed a self-tuning adaptive control system grounded in fuzzy logic for sustainable greenhouse management. Their system dynamically adjusts climate variables, thereby ensuring optimal growth conditions while simultaneously minimizing resource consumption. Similarly, Keswani et al. [58] used fuzzy logic to enhance irrigation practices in precision agriculture. By utilizing weather-dependent, zone-specific irrigation strategies, their approach significantly improves water use efficiency. Furthermore, Stitou et al. [59] used fuzzy logic to enhance irrigation practices in precision agriculture. By utilizing weather-dependent, zone-specific irrigation strategies, their approach significantly improves water use efficiency.

In agricultural applications, fuzzy logic has been effectively utilized for plant disease detection [60] with a hybrid model combining Adaptive Neuro-Fuzzy Inference System (ANFIS) with convolutional neural networks (CNNs). The model was trained on a dataset comprising images of diseased and healthy plant leaves. The architecture integrates fuzzy logic principles to handle uncertainty and imprecision in image data, enhancing the model’s robustness. The proposed model achieved an accuracy score of 84.78% without using Local Binary Pattern (LBP) features, while its precision, recall, and F1 scores were 89.59%, 90.45%, and 89.53%, respectively. These results underscore the efficacy of incorporating fuzzy logic into CNN architectures for plant disease detection, highlighting the importance of selecting appropriate model architectures and utilizing high-quality data sources in agricultural applications.

2.11. Particle Swarm Optimization (PSO): An Evolutionary Optimization Algorithm Inspired by Collective Behaviors in Nature

PSO is a nature-inspired algorithm designed to solve complex optimization problems by emulating the social behavior of organisms, such as bird flocking or fish schooling. PSO operates by utilizing a group of particles that explore the search space collaboratively, with each particle adjusting its position based on both its individual experiences and the shared knowledge of the swarm. Through this iterative process, the algorithm converges efficiently toward optimal solutions. In the agrifood industry, PSO proves particularly useful in addressing logistical and resource management challenges. For instance, it can be used to optimize transportation routes, reducing fuel consumption and delivery times, or to improve the distribution of water resources for irrigation. Moreover, PSO has been applied to enhance energy efficiency in agricultural facilities by optimizing operational parameters, thereby contributing to a significant reduction in operational costs. These applications of PSO contribute to more sustainable and cost-effective practices across the sector, as discussed in “Swarm Intelligent” [61]. Recent advancements in particle swarm optimization (PSO) algorithms showcase their versatility across various sectors, from energy management to precision agriculture. For instance, Boudjerda et al. [62] applied a sine-cosine PSO variant to optimize reservoir operations, achieving efficient water distribution while accounting for environmental constraints. This study exemplifies how PSO can adapt complex systems to real-world limitations in resource management. In the field of agricultural automation, Singh et al. [63] developed a PSO-based robotic arm for precision harvesting, particularly in cotton farming. Their model enhances in-field performance by optimizing movement paths to reduce resource use and operational time. Another notable application by Jawad et al. [64] utilizes PSO in wireless sensor networks to improve path-loss models, which strengthens connectivity in remote agricultural settings. Granata et al. [65] explored the application of PSO algorithms for advanced evapotranspiration forecasting in Mediterranean agricultural settings. Their research, conducted in central Italy, combined PSO with feature selection methods to improve the accuracy of water demand predictions. By optimizing evapotranspiration models, the study provided a scalable framework for managing water resources in regions prone to water scarcity, demonstrating the algorithm’s capacity to enhance agricultural sustainability through precise environmental monitoring.

These studies underscore PSO’s effectiveness in refining operational efficiency and adapting to diverse real-world challenges.

2.12. Autoregressive Time Series Models (ARIMA): A Class of Time Series Models for Analyzing and Forecasting Temporal Patterns

ARIMA models are a class of time-series forecasting techniques that combine autoregression, differencing, and moving averages to analyze data points that change over time. ARIMA models are particularly effective for handling data with trends or seasonality, making them suitable for a wide range of forecasting applications. These models rely on historical data to make predictions about future values, accounting for both temporal dependencies and random fluctuations. The detailed methodology and theoretical foundation of ARIMA models are comprehensively discussed in “Time Series Analysis: Forecasting and Control” [66].

In the agrifood sector, ARIMA models have several valuable applications. They are commonly used to forecast weather patterns, which are critical for agricultural planning and risk management. Additionally, ARIMA models can predict product demand, enabling more efficient supply chain management and reducing waste. Crop yield forecasting across different seasons is another key application, allowing for better resource allocation and strategic decision-making in planting and harvesting. By providing accurate predictions of such essential variables, ARIMA models support long term planning and improve operational efficiency in agriculture. For example, ARIMA models have been effectively utilized for forecasting crop growth parameters [67] employing ARIMA models to predict canopy cover in cotton fields.

Recent applications of Autoregressive Time Series Models, particularly the ARIMA model, highlight their effectiveness in forecasting complex environmental and economic patterns. For example, Jariwala and Agnihotri [68] utilize ARIMA to model drought occurrences in semi-arid regions, providing insights into rainfall variability and improving water resource management. In the energy sector, Zahoor et al. [69] employ ARIMA models to forecast energy demand trends based on historical consumption data, aiding in resource allocation and policy planning for sustainable development. Additionally, Hachimi et al. [70] apply ARIMA within precision agriculture, using it for accurate weather predictions that optimize planting and harvesting schedules. These studies underscore ARIMA’s adaptability in handling time-series data, offering robust predictive capabilities for varied applications in environmental management and resource planning.

2.13. Hierarchical Clustering Algorithms: A Class of Unsupervised Learning Methods for Creating Nested Clusters Based on Data Similarity

This method of clustering builds a hierarchy of nested clusters by either merging smaller clusters into larger ones (agglomerative) or splitting larger clusters into smaller ones (divisive). This approach allows for a flexible grouping structure, where clusters can be examined at different levels of granularity. Hierarchical clustering does not require a predefined number of clusters, which makes it particularly useful when the number of distinct groups in the data is unknown. The detailed methods and applications of hierarchical clustering are thoroughly explained in “Data Mining: Concepts and Techniques” [71].

In agriculture, hierarchical clustering is used to structure data pertaining to crops, soil properties, and climate conditions. For instance, this technique is applied to categorize regions based on similarities in soil composition, thereby facilitating more targeted and efficient land-use planning. Moreover, hierarchical clustering groups crops according to their growth patterns and environmental requirements, optimizing the allocation of resources such as irrigation and fertilization. By identifying inherent groupings within agricultural data, hierarchical clustering aids in decision-making processes that enhance resource efficiency and improve overall agricultural management strategies.

Recent studies on hierarchical clustering algorithms reveal their versatility across diverse scientific disciplines. For instance, Imran et al. [72] applied hierarchical clustering to analyze plant responses under stress conditions, particularly salinity, facilitating the grouping of genetic traits and improving breeding strategies for stress resilience. In a different field, Cheng et al. [73] utilized hierarchical clustering in food safety, applying it to detect chemical compounds in food samples. This method enhances detection accuracy by grouping similar chemical profiles, supporting better regulatory practices. Additionally, Schnack et al. [74] explored hierarchical clustering to analyze trends in sustainable agricultural technologies. This study allowed researchers to classify and evaluate emerging technologies based on performance metrics.

In agricultural applications, hierarchical clustering algorithms have been effectively utilized for crop classification [75], presenting a model that integrates multiple data sources to enhance classification accuracy. The researchers employed a dataset comprising 7.4 million agricultural parcels in France and the Netherlands, associating each parcel with time-series data from Sentinel-2 and Landsat 8 satellite observations, as well as biophysical variables such as Leaf Area Index (LAI) and Fraction of Absorbed Photosynthetically Active Radiation (FAPAR). The model utilizes a hierarchical fusion approach to combine satellite imagery, crop rotation patterns, and local crop distribution data. Performance assessments at different levels of semantic aggregation—ranging from 151 to 8 crop types or groups—yielded accuracy rates between 85% and 95%, depending on the dataset and aggregation level. These results underscore the efficacy of hierarchical clustering algorithms in agricultural applications, highlighting the importance of integrating diverse data sources to achieve high precision in crop classification tasks.

A relevant application of hierarchical clustering in a Mediterranean agricultural context is the enhancement of crop yield prediction accuracy, as explored by Bregaglio et al. [76]. Their study, conducted in Italy, applied hierarchical clustering to group agricultural fields based on phenological and environmental data derived from remote sensing sources. By segmenting fields into clusters with similar growth conditions, the researchers improved yield forecasting models, enabling more precise crop management strategies. This demonstrates the potential of hierarchical clustering to optimize agricultural planning through data-driven decision-making.

These applications highlight hierarchical clustering’s capability to manage complex datasets, offering insights that advance research in environmental resilience, food safety, and technological innovation.

2.14. Generative Adversarial Networks (GANs): A Class of Deep Learning Architectures for Generating Synthetic Data Through Adversarial Training

Generative Adversarial Networks (GANs) consist of two neural networks: a generator and a discriminator. These networks engage in a process of competition, where the generator creates synthetic data instances, and the discriminator attempts to distinguish between real and generated data. Through this adversarial process, the generator improves its ability to produce increasingly realistic data. This innovative approach to data generation is introduced by Goodfellow et al. in “Deep Learning” [77] and has since become a cornerstone in fields requiring synthetic data generation.

In the agrifood sector, GANs are used to address challenges related to data scarcity, particularly in scenarios where the collection of real-world data is either costly or limited, especially at large scales. For instance, GANs are used to simulate new climate conditions or pest infestation scenarios, generating additional training data to enhance the robustness of predictive models. These synthetic datasets facilitate more accurate forecasting and decision-making in large-scale agricultural operations, where environmental variability significantly influences crop yields and resource management.

Recent studies on Generative Adversarial Networks (GANs) showcase their transformative potential in agriculture and related fields. For instance, Hu et al. [78] utilized GANs to enhance the prediction accuracy of starch content in kudzu through advanced image synthesis, enabling non-destructive quality assessment in food production. Similarly, Fawakherji et al. [79] applied shape and style GANs to generate multispectral data for crop protection. As a result, they provided an innovative data augmentation approach that supports more accurate pest and disease control and monitoring. In the realm of plant disease diagnostics, Khare et al. [80] developed a GAN-based augmentation model to improve leaf disease classification, facilitating faster and more reliable crop management.

Morales-García et al. [81] explored the application of Generative Adversarial Networks (GANs) for generating synthetic agricultural data. Their study evaluated the effectiveness of GAN-based synthetic data for improving the accuracy of predictive models in agricultural monitoring tasks. Conducted in Spain, the research highlighted how GANs could supplement real-world datasets, reducing data scarcity issues common in agricultural research. By generating realistic synthetic data, their approach supports enhanced model training, improving forecasting accuracy in agricultural management. These examples illustrate GANs efficacy in generating high-quality synthetic data, significantly advancing image-based analysis and monitoring capabilities across agricultural applications.

Table 1 offers a synthesized view of the methods explored in Section 2, emphasizing their specific contributions to the agrifood industry while facilitating easier reference for readers. The bibliography references in the table correspond to the numbered citations in the References Section of this paper.

3. Materials and Methods

This study employs ETL (Extract, Transform, Load) methodologies to facilitate the acquisition, preprocessing, normalization, and subsequent analysis of data extracted from the SCOPUS database. The extraction phase involves the direct retrieval of the dataset from SCOPUS, ensuring comprehensive and relevant data collection. Following extraction, the transformation phase encompasses meticulous preprocessing and normalization procedures to guarantee consistency and comparability across all records. Ultimately, the normalized data was integrated into the analytical environment, enabling in-depth exploration and analysis to underpin the research of AI applications within the agrifood industry.

3.1. Data Retrieval

A comprehensive search of scholarly articles was conducted utilizing the SCOPUS database. SCOPUS was selected for its extensive coverage of peer-reviewed literature across a broad spectrum of scientific disciplines, thereby ensuring the assembly of a robust and pertinent collection of publications. A meticulously crafted search query was designed to capture a wide array of AI methodologies and their applications within the agrifood sector. This query was formulated to incorporate a diverse range of AI-related terms alongside sector-specific keywords, ensuring the inclusion of both general AI approaches and specialized techniques. The specific search query used is as follows:

TITLE-ABS-KEY((“artificial intelligence” OR “machine learning” OR “deep learning” OR “computational intelligence” OR “data mining” OR “neural networks” OR “genetic algorithms” OR “support vector machines” OR “fuzzy logic” OR “reinforcement learning” OR “convolutional neural networks” OR “random forest” OR “natural language processing” OR “particle swarm optimization” OR “time series analysis” OR “hierarchical clustering” OR “generative adversarial networks”) AND (algorithm* OR model* OR technique* OR method* OR application*) AND (“agrifood” OR “agriculture” OR “aquaculture” OR “horticulture” OR “farming” OR “food industry” OR “food processing” OR “food quality” OR “plant disease” OR “soil management” OR “pest control” OR “agricultural robotics” OR “crop yield prediction” OR “climate-smart agriculture” OR “digital agriculture” OR “food safety” OR “carbon footprint” OR “supply chain optimization” OR “waste management” OR “novel food design” OR “crop production” OR “weather forecast” OR “food design” OR “agribusiness” OR “irrigation”))

• AND PUBYEAR >2019 AND PUBYEAR <2025

• AND (LANGUAGE(English) OR LANGUAGE (Spanish))

• AND DOCTYPE(ar)

• AND SRCTYPE(j)

• AND PUBSTAGE (final)

After extraction, the data cleaning process did not involve the removal of any records, as no duplicate entries, missing values in critical fields, or irreparable formatting inconsistencies were identified in the high-quality dataset provided by SCOPUS. The only necessary adjustments included ensuring that the ‘Year’ and ‘Cited by’ columns were in numeric format for proper processing, with missing citation values replaced by zero before conversion. Additionally, extraneous whitespace in the ‘Title’ column was removed, and records with null or empty values in the ‘Year’ or ‘Title’ fields were discarded. Finally, the dataset was exported in Parquet format for improved manageability. These transformations were carried out without compromising the integrity or reliability of the dataset.

The dataset, obtained on 19 October 2024, reveals that, out of a total of 19,606 articles, 19,573 are published in English, underscoring the dominance of English as the primary language for research related to AI algorithms in the agrifood industry. In contrast, only 33 publications are in Spanish, illustrating the considerably lower volume of research output in this language. Among the English-language publications, 671 originated in Spain, indicating the country’s notable contribution to global research in this field. Furthermore, 26 of the Spanish-language articles are from countries other than Spain, further emphasizing that Spanish is infrequently used for scientific publications within Spain, where English predominates as the preferred language for disseminating research findings in this area.

3.2. Data Analysis Tools

Given the extensive scale of the dataset under examination, the bibliometric study used a comprehensive suite of advanced techniques associated with big data and machine learning, meticulously adapted for the efficient management and analysis of large-scale datasets. Central to the methodology was the construction of a binary matrix based on author keywords, systematically categorizing terms into algorithm-related and application-related keywords. Prominent algorithms, including neural networks, deep learning, random forest, and support vector machines, were focal points of the analysis and were scrutinized in relation to specific applications such as crop yield prediction, soil management, food quality, and digital agriculture.

Natural language processing (NLP) techniques were applied to the abstracts to identify and elucidate relationships between the algorithms and their respective application areas, utilizing regular expressions and NLP libraries for precise keyword mapping. Additionally, machine learning models, specifically random forest and deep learning, were used to detect patterns and predict outcomes within the dataset. Latent Dirichlet Allocation (LDA) facilitated topic modeling to uncover underlying thematic structures, while sentiment analysis using TextBlob 0.17.1 assessed the emotional tone of the abstracts, providing insights into the prevailing sentiments within the research community.

The study also used hierarchical agglomerative clustering for bibliographic coupling similarity analysis, revealing distinct clusters of related research topics based on shared references. Co-occurrence matrices were generated to explore the relationships between specific algorithms and their practical implementations in the agrifood sector. To manage and visualize the complex data, several Python libraries—including Pandas 2.2.3 for data manipulation, Matplotlib 3.10.0 for visualization, and NetworkX 3.4.2 for constructing knowledge graphs—were used.

These methodologies enabled a structured and scalable approach to extracting meaningful insights from a dataset composed of records with 47 attributes, including complex textual fields such as abstracts. The integration of knowledge graphs visualized international collaborations, while the combination of LDA and sentiment analysis provided a comprehensive understanding of both thematic and emotional dimensions of the research corpus.

In summary, the study utilized a combination of ETL processes, bibliometric analysis, NLP, and machine learning models, topic modeling, sentiment analysis, hierarchical clustering, and co-occurrence matrices, supported by Python-based tools. This multifaceted analytical framework facilitated an in-depth and nuanced evaluation of the application of AI algorithms within the Spanish agrifood sector, highlighting trends, collaborations, and emerging technological strategies.

3.3. Summary of Data Analysis Tools and Techniques Utilized

1. ETL Methodologies: Extract, Transform, Load processes for data acquisition and preprocessing.

2. Bibliometric Analysis: Evaluation of research output, collaboration networks, and publication trends.

3. Natural Language Processing (NLP): Extraction and categorization of keywords and thematic analysis.

4. Machine Learning Models: Random forests and deep learning for pattern detection and outcome prediction.

5. Latent Dirichlet Allocation (LDA): Topic modeling to identify core research themes.

6. Sentiment Analysis: Assessment of emotional tone in research abstracts using TextBlob.

7. Hierarchical Agglomerative Clustering: Clustering based on bibliographic coupling to identify related research topics.

8. Co-occurrence Matrices: Analysis of relationships between algorithms and application areas.

9. Python Libraries: Pandas, Matplotlib, NetworkX, and Regular Expressions for data processing and visualization.

10. Knowledge Graphs: Visualization of international research collaborations and networks.

These comprehensive tools and methodologies facilitated an exhaustive evaluation of the scientific literature, identifying trends, international collaborations, and predominant areas of focus in the integration of AI within the Spanish agrifood industry. The combination of quantitative and qualitative methods ensured a deep understanding of the application and challenges of AI technologies in this sector.

4. Results

The analysis of the SCOPUS dataset on AI algorithms in the agrifood industry reveals significant trends in research output over recent years. The following section presents a detailed exploration of these findings, offering insights into the growing prominence of AI in addressing challenges within the agrifood industry, with a focus on applications within the Spanish sector.

4.1. Analysis of Article Volume, Origin, and Access Model

The dataset encompasses articles published between 2020 and 2024, as illustrated in Figure 1.

The number of publications has steadily increased over the analyzed period. In 2020, there were 2220 articles published, which grew to 5259 by 2024. It is important to note that this figure for 2024 is not final, as there are still two and a half months remaining in the year. This establishes that the total number of publications may increase further before the year’s end. The rise in publication volume underscores a growing interest in the application of AI techniques within the agrifood industry, driven by increasing attention to sustainability, climate-smart agriculture, and food security. This trend is observed across both English and Spanish publications, although the latter represents a small proportion of the total output.

4.1.1. Publication by Country

The dataset on publications by the country reveals that China leads in research output related to AI applications in the agrifood industry, with a total of 4789 publications. India follows with 2851 publications, while the United States holds the third position with 1759 articles. Notably, Iran and South Korea also contribute significantly to the field, with 473 and 465 publications, respectively. These figures highlight the dominant role of Asian countries, particularly China and India, in advancing research on AI-driven innovations in agriculture. The United States also maintains a strong presence, reflecting its continued investment in agricultural technology and artificial intelligence. This geographical distribution supports the notion that the application of AI in the agrifood sector is a globally significant research area. This is particularly relevant in regions experiencing rapid technological development and agricultural transformation, as shown in Figure 2.

The distribution of publications across continents highlights the global reach of research on AI applications in the agrifood industry. Asia leads with 11,513 publications, representing 59% of the total output, underscoring its dominant role in driving innovation and research in this sector. This prominence is closely related to Asia being the most populous continent, where the growing demand for food production presents a persistent challenge. Addressing food security through technological advancements, such as AI, is critical to meeting the needs of its vast population. Europe follows with 3632 publications (19%), reflecting the continent’s strong engagement in technological advancements within agriculture, particularly through collaborations among countries such as the United Kingdom, Germany, and Spain. North America, primarily driven by the United States, accounts for 12% of the publications, indicating its continued investment in agricultural AI technologies. Meanwhile, Africa (5%) and South America (3%) have smaller but notable contributions, likely reflecting emerging interest and developing research infrastructures in these regions. Oceania, led by Australia, contributes 2% to the global research output, further illustrating its active participation despite its relatively smaller geographic footprint. These results can be seen in Table 2, highlighting the increasingly global nature of research in AI for the agrifood industry, with strong regional hubs emerging in key continents.

4.1.2. Collaboration Between Countries

A Knowledge Graph has been devised to illustrate the collaboration between countries based on the provided data. In this graph, the triples are formed by country entities, with the relationships defined by the number of collaborative research outputs between them. The size of the sphere surrounding each country in the graph is proportional to the volume of article production, while countries with more collaborations and higher research output are represented by thicker lines connecting them. This Knowledge Graph is shown in Figure 3.

The analysis of international collaborations reveals a complex and interconnected global research network, with several nations playing pivotal roles. China stands out as a major hub of research activity, with notable partnerships, such as its 78 collaborations with Saudi Arabia, highlighting its dual role in contributing substantially to the volume of AI-related publications and fostering extensive cross-border research partnerships. However, it is significant that the number of China’s collaborations is not proportional to its high volume of research output, suggesting that much of its research is conducted domestically or within limited international partnerships. These collaborations span multiple regions, indicating China’s strategic engagement in fostering AI innovations in agriculture across both developed and developing nations.

India, similarly, plays a significant role in this global network, as evidenced by its 178 partnerships with the United States. India’s collaborations with countries such as the United Kingdom (47 collaborations) and Saudi Arabia (78 collaborations) emphasize its efforts to bridge technological advancements and agricultural innovations. The United States, which maintains 178 collaborations with India and 84 with Australia, continues to serve as a central actor in AI research, particularly through partnerships with other leading nations. Notable collaborations with European countries, such as 87 partnerships with Germany and 50 with Italy, further emphasize the U.S.’s role in driving research forward in the agrifood sector through global scientific cooperation.

European nations also exhibit robust collaborative efforts. Spain, in particular, has significantly increased its participation, with 53 collaborations with the United States, 41 with Germany, and 33 with Italy, highlighting its growing influence in AI research in the agrifood sector. These partnerships emphasize Spain’s strategic role in fostering cross-border research, particularly within Europe, and contributing to the advancement of AI-driven agricultural technologies.

4.1.3. Access Model

The analysis of the access model for the articles in this dataset reveals a varied distribution of open access availability. A sizable portion of the articles, 42.2%, are categorized under an unknown access model, indicating either incomplete metadata or variability in access conditions across platforms. Among those with a defined access model, the largest group, representing 35.3% of the articles, is available under the Gold Open Access model, which offers free access to all articles immediately upon publication, typically funded by article processing charges (APCs). This growing prevalence of Gold Open Access underscores the increasing emphasis on making research widely accessible without paywall restrictions.

Hybrid Gold Open Access, accounting for 7.8% of the articles, combines open access content with subscription-based content in the same journal, where some articles are freely accessible while others remain behind a paywall. Additionally, 7.5% of the publications follow a mixed model that includes various combinations of gold, green, and hybrid access strategies. This reflects the diverse approaches used by researchers and publishers to balance open access and financial sustainability. All Open Access, where 100% of the articles in the journal are freely available, guarantees unrestricted access to all published content.

Lastly, 3.3% of the articles fall under Bronze Open Access, where content is temporarily available to the public without a long term commitment to open access by the publishers. This distribution reflects a shift towards greater openness in scientific publishing, especially in fields such as AI and agriculture, where global access to research can drive innovation and practical applications. However, the substantial number of articles with unknown or ambiguous access models points to ongoing challenges in tracking and ensuring consistent open access, highlighting the need for improved transparency and standardization in publishing practices.

4.1.4. Keywords

An analysis of the keywords in the corpus of articles (Table 3) was conducted to identify the primary research areas within the agrifood industry, specifically in relation to AI. In the global articles, the most frequently used keywords demonstrate a strong focus on advanced AI technologies within this sector. The top keywords include machine learning (3268 occurrences), deep learning (2917 occurrences), and artificial intelligence (864 occurrences), highlighting a clear emphasis on machine learning techniques and their application to agricultural processes. Additionally, terms such as precision agriculture (825 occurrences) and Agriculture (767 occurrences) further underscore the integration of AI-driven solutions aimed at enhancing efficiency and precision in farming practices.

An analysis of the keywords in the corpus of articles was conducted to identify the primary research areas within the agrifood industry, specifically in relation to AI. In the global articles, the most frequently used keywords demonstrate a strong focus on advanced AI technologies within this sector. The top keywords include machine learning (3268 occurrences), deep learning (2917 occurrences), and artificial intelligence (864 occurrences), highlighting a clear emphasis on machine learning techniques and their application to agricultural processes. Additionally, terms such as precision agriculture (825 occurrences) and Agriculture (767 occurrences) further underscore the integration of AI-driven solutions aimed at enhancing efficiency and precision in farming practices.

These patterns indicate that researchers are particularly focused on how AI can transform agriculture decision-making, optimize resource management, and improve crop yields. The prominence of machine learning and deep learning as keywords also reflects the widespread use of these methods to address complex challenges in data analysis and predictive modeling.

In contrast, the articles written in Spain reveal a somewhat different focus, with a more moderate use of AI-related terms. The leading keywords in these texts include machine learning (125 occurrences) and deep learning (62 occurrences), while also highlighting terms such as precision agriculture (50 occurrences) and artificial intelligence (40 occurrences). This indicates a stronger emphasis on the direct application of these technologies in optimizing precision farming practices. Additionally, agriculture appears 35 times, reflecting a broader interest in the agricultural context. A comparison between the two sets of articles shows that, although machine learning and deep learning are significant in both English and Spanish publications, their absolute frequency is much higher in English-language texts. However, Spanish-language articles tend to emphasize practical implementations, such as precision agriculture, suggesting that research in Spain leans more towards applying AI technologies in specific agricultural domains, while the global community focuses more on foundational AI technologies. This contrast highlights regional differences in AI research priorities within the agrifood sector.

4.1.5. Number of Publications per AI Category

An analysis of the number of publications by category was conducted to examine the distribution of research topics related to the application of artificial intelligence within the agrifood industry. As illustrated in Figure 4, the identified categories include optimization, yield prediction, quality control, disease detection, precision agriculture, climate and weather forecasting, soil and irrigation management, animal health and behavior, and autonomous systems. The category with the highest number of publications is optimization, with 4650 articles, followed by yield prediction (2285 articles), and quality control (2267 articles). These results suggest that enhancing resource efficiency and predicting agricultural outputs are central areas of focus for researchers investigating AI applications in agriculture.

Other categories, such as disease detection (1472 articles) and precision agriculture (902 articles), are also significant but receive comparatively less attention than optimization-related topics. This indicates that while advancements in AI-driven precision farming and disease management are progressing, optimization remains the dominant concern, reflecting the sector’s prioritization of operational efficiency and resource management. To categorize the articles, a Python-based method was used to analyze abstracts and keywords, using regular expressions to identify terms associated with each category, such as “optimization” for optimization and “disease detection” or “pest control” for disease detection.

The results demonstrate a wide range of AI applications in the agrifood sector, with considerable attention also directed towards climate and weather forecasting (758 articles) and soil and irrigation management (556 articles). However, AI research in areas such as animal health and behavior (398 articles) and autonomous systems (283 articles) is less prevalent, indicating that these fields are emerging but still underdeveloped. This distribution highlights the varied focus across different AI technologies and their applications, revealing both ongoing challenges and potential growth opportunities within each category.

4.1.6. Latent Topics

Latent Dirichlet Allocation (LDA), a widely used technique in natural language processing for topic modeling, was applied to uncover thematic structures within the corpus of agrifood research articles. This method facilitated the identification of primary areas of focus based on the abstracts, represented by the most frequent and pertinent terms associated with each topic. Five key topics were identified, each characterized by a distinct set of representative words, providing insight into specific areas of research emphasis within the field.

To conduct this analysis, the abstracts were processed using Term Frequency-Inverse Document Frequency (‘TF-IDF’) vectorization to highlight the most significant words. LDA was subsequently applied to detect the underlying topics, with the model configured to identify five distinct themes. The top ten most representative words for each theme were selected to maintain clarity and facilitate interpretation. Abstracts lacking substantial content were excluded to ensure the robustness of the findings.

The analysis yielded the following five topics:

• Topic 1 is characterized by terms such as ‘data’, ‘model’, ‘waste’, ‘based’, ‘proposed’, ‘energy’, and ‘location’, establishing a focus on sustainability and resource management within the agrifood sector.

• Topic 2 features terms such as ‘food’, ‘microbial’, ‘species’, ‘bacterial’, ‘milk’, ‘products’, and ‘pathogen’, indicating a concentration on microbiology and food safety research.

• Topic 3 revolves around ‘soil’, ‘model’, ‘water’, ‘models’, ‘yield’, ‘crop’, and ‘climate’, reflecting an emphasis on soil management, crop yield prediction, and environmental modeling.

• Topic 4 includes words such as ‘disease’, ‘detection’, ‘plant’, ‘classification’, ‘pathogen’, and ‘control’, highlighting a focus on plant disease detection and the application of machine learning in this domain.

• Topic 5 centers on terms such as ‘water’, ‘groundwater’, ‘carbon’, ‘economic’, ‘emission’, and ‘ecosystem’, indicating a focus on environmental impacts, particularly related to water management and carbon emissions in agriculture.

These topics not only illustrate the diversity of research areas within the agrifood industry but also reveal the deeper intersections between environmental sustainability, biological science, and technological innovations. For instance, both Topic 1 and Topic 5 emphasize a growing research interest in sustainable practices, resource efficiency, and environmental protection. This signals an increasing focus on how AI technologies can aid in addressing climate-related challenges in agriculture. Similarly, the prominence of Topic 2 and Topic 4 suggests a strong focus on food safety and plant disease detection, areas critical to maintaining global food security. The identification of these latent topics underscores the broad range of applications where AI and technological advancements are being leveraged to address both practical and strategic challenges within the agrifood sector. This analysis not only reveals current research priorities but also highlights emerging areas of innovation and potential gaps that future studies may seek to explore.

4.1.7. Co-Occurrence Matrix Between Algorithms and Application Areas

The co-occurrence matrix analysis, as shown in Figure 4, was conducted to research the relationship between specific algorithms and their corresponding areas of application within the agrifood industry. This method provides insights into the frequency with which various AI algorithms, such as deep learning, random forest, and neural networks are used in conjunction with specific agricultural tasks, thereby offering a deeper understanding of the implementation of AI within this domain. The matrix reflects the number of instances where each algorithm is associated with key areas such as crop yield prediction, soil management, pest control, food quality, and digital agriculture.

The co-occurrence matrix analysis, as shown in Figure 5, was conducted to research the relationship between specific algorithms and their corresponding areas of application within the agrifood industry. This method provides insights into the frequency with which various AI algorithms, such as deep learning, random forest, and neural networks, are used in conjunction with specific agricultural tasks, thereby offering a deeper understanding of the implementation of AI within this domain. The matrix reflects the number of instances where each algorithm is associated with key areas such as crop yield prediction, soil management, pest control, food quality, and digital agriculture.

The analysis was based on extracting author keywords from the articles, transforming them into a binary matrix, and subsequently calculating the co-occurrence of algorithm-related terms with agricultural applications. This quantitative approach enables an accurate representation of how various AI technologies are used across different research areas.

The findings reveal that deep learning is the most frequently applied algorithm, with 21 occurrences in crop yield prediction, 28 in digital agriculture, and smaller but notable instances in pest control (3) and food quality (5). By comparison, random forest appears less frequently but is still used across multiple areas, including six occurrences in crop yield prediction and four in digital agriculture. Other algorithms, such as neural networks, demonstrate limited usage, with only two occurrences in food quality and one in digital agriculture. Interestingly, genetic algorithms and support vector machines do not co-occur with any of the identified application areas, indicating either minimal usage or a lack of emphasis on these methods within the current scope of agrifood research.

4.1.8. Sentiment Polarity Analysis

The sentiment analysis of the top 100 most cited articles was conducted using the TextBlob library (Figure 5), focusing on the polarity of each article’s abstract to assess its emotional tone. Polarity scores provide a quantitative measure of the sentiment conveyed in the abstracts. The analysis revealed a spectrum of sentiment scores, with an overall trend toward neutrality.

The numerical values in Figure 5 represent the sentiment polarity scores assigned to the abstracts of the top 100 most cited articles, calculated using the TextBlob library. These scores range from −1 (very negative) to 1 (very positive) and are displayed along the x-axis, indicating the emotional tone conveyed in the articles. The y-axis represents the number of articles corresponding to each specific sentiment score, providing a visual distribution of emotional tones within the analyzed publications. The total of the y-axis values sums to 100, reflecting the complete set of articles included in the sentiment analysis.

TextBlob performs sentiment analysis using a lexicon-based approach, assigning polarity scores to predefined words and phrases. This method identifies emotional tone by detecting terms with affective connotations. Given that scientific writing should maintain objectivity, the sentiment score analysis presented here, visualized through a bar graph, enables the detection of outliers with significant deviations. Identifying emotional biases in highly cited works allows for the assessment of whether their tone aligns with the expected neutrality, ensuring compliance with impartiality standards in scientific research.

The most positively polarized article, with a sentiment score of 0.317949 and 226 citations, discusses the use of a support vector machine for modeling and simulation in manufacturing, highlighting advancements in this field. On the opposite end, the most negatively polarized article, with a sentiment score of −0.180769 and 177 citations, addresses the challenges of shallow landslide susceptibility mapping, emphasizing the damage caused to buildings and infrastructure by such natural disasters.

Regarding Spanish publications, the most positively polarized article from Spain, titled “An intelligent Edge-IoT platform for monitoring livestock and crops in a dairy farming scenario” [82], with a sentiment score of 0.140504 and 209 citations, focuses on the development of a platform for monitoring sustainable environments. This emphasis on environmental monitoring aligns with Spain’s ongoing efforts to improve water management and ensure sustainability in agriculture. Additionally, it reflects the country’s growing interest in sustainable agricultural practices, particularly in optimizing irrigation systems and reducing environmental impacts through technological innovations.

In contrast, the most negatively polarized Spanish article, titled “More than one million barriers fragment Europe’s rivers” [83], with a sentiment score of −0.021652 and 367 citations, addresses the fragmentation of Europe’s rivers by over one million barriers, highlighting the negative impacts on biodiversity. This concern is particularly relevant to Spain, where river fragmentation affects water availability for irrigation—an issue critical to the country’s agricultural sector. The coherence between these findings and Spain’s dual focus on sustainable agriculture and water management underscores the significance of addressing water-related challenges in both environmental and agricultural contexts. Moreover, the use of sentiment analysis techniques in bibliographic analysis proves valuable in uncovering underlying trends and attitudes within the research community. By evaluating the emotional tone of scientific abstracts, sentiment analysis can highlight the focus areas where optimism or concern predominates, providing deeper insights into the priorities and challenges being addressed in specific fields, such as water management and sustainable agriculture in Spain.

4.1.9. Similarity Analysis

The bibliographic coupling similarity analysis, as shown in Figure 6, was conducted on the same top 100 most cited articles used in the Sentiment Polarity Analysis. The matrix presents cosine similarity scores between pairs of articles based on the references they share, allowing for the quantification of how closely related these articles are in terms of citation patterns. Higher similarity scores indicate that two articles share a greater number of references, reflecting a strong connection in their research focus or methodologies. Given the nature of this representation, the number of labels in the figure does not correspond directly to the total number of articles analyzed. Instead, the labels represent aggregated similarity ranges, grouping articles with similar bibliographic coupling scores into broader categories. Note that the Scopus EID identifier has been used on the axes. This approach enhances the visual clarity of the heatmap by summarizing dense citation relationships into interpretable clusters.

The values in the matrix range from 0 (indicating no shared references) to 1 (indicating identical references). A general trend was observed where several articles exhibit high similarity scores, establishing that the most cited articles tend to draw from a common set of references. This pattern points to the presence of influential works and methodologies that are widely referenced within the top-cited research.

4.1.10. Hierarchical Cluster Analysis (HCA)

Using HCA, the analysis revealed a total of seven distinct clusters (Figure 7) among these articles, with each cluster representing a subgroup of research focus within the agrifood sector. To represent the data in a hierarchical component analysis (HCA) scatter plot instead of a heatmap, dimensionality reduction was applied to the similarity data, enabling cluster visualization in a 2D space. This approach is particularly useful for visually analyzing the relationship between articles based on their reference similarities and cluster associations:

• Cluster 0: This cluster contains articles primarily focused on land-use change, climate impacts, and agricultural sustainability. Many of the articles in this group heavily cite works related to environmental management and food security, reflecting a strong interest in understanding the broader impacts of agricultural practices on ecosystems.

• Cluster 1: Articles in this cluster tend to focus on machine learning applications and AI-driven innovations in precision agriculture. The common references suggest a focus on data analytics, remote sensing technologies, and the development of predictive models for improving crop yields and resource efficiency.

• Cluster 2: This group of articles is centered around soil science and crop management techniques. References commonly include works related to soil properties, irrigation strategies, and yield optimization, indicating a concentration of research on improving soil health and maximizing agricultural output.

• Cluster 3: Articles in this cluster are heavily focused on plant disease detection and pest management. The shared references point to common methodologies for monitoring crop health and preventing disease outbreaks, a critical aspect of sustainable agriculture.

• Cluster 4: This cluster covers a range of articles dealing with food quality, safety, and supply chain management. The references shared among these works often relate to methods for enhancing food processing techniques, ensuring safety standards, and improving the efficiency of agricultural supply chains.

• Cluster 5: Articles in this group frequently cite research on deep learning and neural network applications in agriculture. The cluster emphasizes cutting-edge AI technologies, such as image recognition and autonomous systems, applied to precision farming and resource optimization.

• Cluster 6: This final cluster includes articles focused on the socio-economic impacts of agricultural practices and the role of policy in shaping sustainable farming initiatives. The shared references highlight research that bridges technological advancements with economic and social considerations in the agrifood sector.

These clusters illustrate the diversity of influential research topics within the agrifood domain, providing an understanding of the key methodologies and trends driving innovation. This clustering also underscores the presence of shared knowledge and common reference points within specific subfields, such as precision agriculture, sustainability, and plant health monitoring, contributing to the advancement of agricultural science.

5. Discussion

Our analysis, rooted in a rigorous examination of scientific publications indexed in Scopus, prioritizes evidence-based findings to shed light on the applications and challenges of AI technologies in the Spanish agrifood sector. This approach ensures that our conclusions are anchored in verifiable research data rather than aspirational goals or strategic visions, such as those outlined in policy-oriented documents such as the Adra Strategic Research, Innovation, and Deployment Agenda 2025–2027 [84]. While such initiatives offer a valuable perspective on potential future directions, our focus lies in reflecting the state of the art as represented in peer-reviewed literature. This allows us to capture not only the tangible advancements made in Spain but also the limitations and challenges revealed through actual research outcomes.

The potential applications, limitations, and challenges of AI technologies in the Spanish agrifood sector can be effectively illustrated by analyzing the top-cited research articles from Spain. These works, covering fields such as remote sensing, crop yield prediction, and disease detection, demonstrate the country’s key contributions to AI in agriculture, while also highlighting specific constraints and areas for further development.

5.1. Potential Applications

Remote sensing emerges as one of the most critical applications of AI technologies within Spanish agriculture, particularly for managing water resources and monitoring crop health. For instance, several studies focus on the integration of AI with UAVs and satellite imagery to enhance the precision of crop monitoring and irrigation systems. In particular, Calera et al. [75] demonstrated how remote sensing could be used to manage water efficiency in crops through evapotranspiration modeling. Such research is particularly valuable in Spain, where efficient water use is essential due to varying climatic conditions. In addition, Castro et al. [85] developed an automated procedure that integrates UAV-acquired imagery with object-based image analysis (OBIA) and decision tree algorithms to map Cynodon dactylon infestations in cover crops within Spanish vineyards. Similarly, Peña et al. [86] utilized high-resolution UAV imagery to discriminate early-stage weeds in sunflower fields, achieving significant accuracy in weed identification.

The use of hyperspectral and multispectral imaging technologies, coupled with ML models such as random forest have gained traction in Spain. For instance, Tanase et al. [38] employed these techniques to estimate forest biomass in Mediterranean regions, demonstrating the versatility of remote sensing for both agricultural and environmental applications. These approaches can also be adapted for precision agriculture, where they facilitate the early detection of crop stress and optimize resource management.

Another promising area is crop-yield prediction. Spanish research contributes significantly to the integration of machine learning algorithms with environmental data to forecast yields under different conditions. Campos et al. [76] used a remote sensing-driven operational approach to assess water productivity and crop yields. Additionally, ARIMA models have been employed to forecast climatic variables that directly impact agricultural planning, such as rainfall and temperature, as shown by Hachimi et al. [61].

In the realm of pest and disease management, AI has proven highly effective. Rodríguez-Díaz et al. [36] developed machine learning models, including random forests, to predict pest infestations in olive groves and vineyards. This has enabled more proactive and sustainable pest control strategies, particularly in regions such as Andalucía and La Rioja, where these crops are economically vital.

Precision viticulture has also benefited from AI advancements. Kerkech et al. [32] applied convolutional neural networks (CNNs) to UAV imagery for early detection of diseases in vineyards, reducing the need for extensive manual inspections. This application is particularly relevant in Spanish wine-producing regions such as La Rioja and Ribera del Duero.

Furthermore, the optimization of irrigation systems through genetic algorithms (GAs) have demonstrated significant potential. Perea et al. [28] developed a GA-based decision support system to manage variable-speed pumping technologies, thereby reducing water and energy consumption in Castilla-La Mancha. This approach aligns with Spain’s emphasis on achieving sustainability in water-scarce regions.

These examples highlight how the integration of AI technologies continues to revolutionize Spanish agriculture, promoting sustainability, productivity, and resource efficiency across diverse agricultural landscapes.

5.2. Limitations

Despite significant advancements in AI applications, several critical limitations hinder broader adoption in Spain’s agrifood sector. One of the primary challenges is data availability and digital infrastructure. AI systems rely on extensive datasets to train models, but systematic data collection remains difficult, especially in smallholder farms where advanced sensing technologies, automated systems, and connectivity are often lacking. This data scarcity limits the implementation of precision agriculture practices that require real-time environmental and operational data.

Another major barrier is the high cost of AI deployment. Implementing technologies such as UAVs, multispectral sensors, and real-time monitoring systems demands significant capital investment. This expense is manageable for large agricultural enterprises with established market access but poses a considerable challenge for smaller farms with limited financial resources. Public and private funding initiatives could mitigate this issue by subsidizing key technologies or offering incentives for digital transformation in smaller farms.

Additionally, technical expertise and skill shortages represent another critical obstacle. AI-driven agriculture requires professionals skilled in data science, machine learning, and agricultural technology integration. However, such expertise is unevenly distributed, with rural regions often facing severe shortages of specialized technical support. Without local experts, implementing and maintaining AI-powered solutions becomes challenging. Collaboration between universities, research institutions, and agrifood companies could help address this gap by fostering specialized training programs and knowledge transfer initiatives.

Moreover, interoperability and standardization issues further complicate the large-scale adoption of AI solutions. Many AI applications are developed as standalone tools with limited compatibility across platforms and technologies. This lack of standardization hinders the seamless integration of multiple AI-powered tools into a unified agricultural management system. Developing open-source frameworks or standardized APIs could enhance compatibility and reduce implementation complexity.

Lastly, regulatory and policy limitations also play a role in restricting AI adoption. Clearer guidelines on data privacy, AI ethics, and sustainability standards specific to agriculture could foster a more supportive environment for technological innovation. Current policy frameworks often lag behind technological advancements, creating uncertainty for agribusinesses interested in AI-driven transformation.

While some research initiatives show promising advances in early disease detection through spectral reflectance, widespread adoption is still limited by the availability of specialized equipment and knowledge. This disconnect between academic research and practical field application highlights the need for stronger partnerships between research institutions and industry stakeholders. Addressing these multifaceted limitations would enable Spain’s agrifood sector to fully harness the transformative potential of AI technologies.

5.3. Challenges

The nature of scientific research, by definition, does not aim to predict the future but rather to establish the identified development patterns that guide us in understanding possible evolutions. Based on the analysis of specific data on AI applications, not only in the Spanish agrifood sector but also at a global scale—given their extrapolability across different contexts—we have identified several key challenges that must be addressed to maximize the impact of these technologies in this field. The following specific lines of research—based on the analysis of data obtained on current lines of research and documented practical results—aim to improve the adoption of AI, optimize the use of resources, and align technological advances with sustainability objectives.

1. Enhancing AI adaptability through remote sensing and UAV integration: Future research should focus on integrating multispectral data from Unmanned Aerial Vehicles (UAVs), hyperspectral imaging, and localized agro-climatic information to enhance the adaptability of AI models. Recent studies have demonstrated the potential of remote sensing for crop classification in Mediterranean environments and pest control through pesticide application. Refining these models by incorporating high-resolution soil and climate data is essential to improve their accuracy in Spain’s diverse agrifood landscapes. Additionally, adjustments to current regulations are needed to facilitate UAV usage in agricultural operations, an evident necessity in Spain given their significant practical utility.

2. Enhancing AI accessibility for small and medium-sized enterprises (SMEs): AI adoption remains concentrated in large agrifood enterprises due to high implementation costs. Future research should explore the development of open-source AI platforms and user-friendly mobile applications that enable real-time decision-making for small-scale producers. Notably, satellite imagery-based disease monitoring in olive groves has been identified as a key need, yet small-scale farmers lack access to such technologies.

3. AI-driven optimization for water and energy conservation: The ability of AI to improve resource efficiency requires further investigation. Genetic algorithms (GAs) have already demonstrated a 6–7% reduction in energy consumption through irrigation optimization in Spain. Future research should extend these applications to a broader agrifood context, focusing on adaptive water management models that dynamically respond to climate variability.

4. Integration of AI with IoT and robotics for precision agrifood management: AI-based irrigation management and IoT sensor networks have shown promise in predicting water stress levels. Future studies should refine predictive models for real-time automation of irrigation, pest control, and crop monitoring through AI and robotics integration. These advancements would strengthen precision agriculture and food processing while reducing reliance on manual labor.

5. Standardization of data collection, digital transformation, and AI accessibility in the agrifood sector: Beyond improving digital infrastructure, research should prioritize continuous training initiatives and the establishment of unified data frameworks to ensure interoperability among AI-driven solutions in the agrifood industry. Additionally, efforts should focus on simplifying access to AI technologies, particularly for small and medium-sized enterprises, by reducing bureaucratic burdens and streamlining regulatory processes. A structured approach to standardizing data collection across the sector would facilitate the seamless implementation of AI-based technologies while promoting broader adoption and practical applicability.

Addressing these five key areas of advancement, as clearly identified in our study—within the broader context of energy efficiency and environmental sustainability—will enhance the applicability, sustainability, and economic feasibility of AI in the Spanish agrifood sector. While these insights hold global relevance, their implementation in Spain will ensure both technological progress and the responsible management of critical resources.

6. Conclusions

AI algorithms have the potential to drive a profound transformation in the agrifood sector, enhancing both efficiency and sustainability. Globally, substantial advancements have been made in integrating AI into agricultural practices, particularly in precision farming, crop yield prediction, and resource management. The capacity of AI to process vast datasets, optimize decision-making, and automate complex agricultural processes paves the way for more sustainable and efficient practices, minimizing resource wastage, and enhancing productivity. Despite these advancements, considerable opportunities for growth remain underutilized, particularly within the Spanish context.

This study deliberately focuses on generic AI algorithms to illustrate their applicability across the agrifood sector, ensuring that conclusions are grounded in empirical data derived from peer-reviewed research. This evidence-based approach prevents reliance on political narratives or recommendations from consulting firms or supranational government institutions, thereby avoiding biases and providing a precise and objective representation of the current state and future potential of AI technologies in the agrifood sector, with a particular focus on the Spanish context, while still maintaining relevance in other settings due to the multifaceted nature of the sector in Spain.

In Spain, the adoption of AI technologies in the agrifood sector remains in a developmental phase. Given the country’s wide range of agricultural systems—including irrigated and rainfed farming, smallholdings, large estates, as well as tropical, Mediterranean, and continental climate crops—the findings from global studies analyzed in this research reveal more similarities than differences with the Spanish case. Studies have demonstrated promising AI applications both globally and locally, particularly in remote sensing for water management—an area of critical importance for Spain—and in disease and pest detection. The integration of AI with UAVs and satellite imagery has enabled enhanced crop monitoring and optimized irrigation strategies. However, widespread implementation remains limited, particularly among smaller farms, due to constraints related to infrastructure, data availability, costs, and, in Spain’s specific case, regulatory hurdles. These barriers hinder the large-scale adoption of AI solutions, emphasizing the need for targeted investment in digital infrastructure, the development of AI tools tailored to Spain’s agricultural conditions, and regulatory reforms that alleviate bureaucratic obstacles discouraging small-scale farmers.

A significant challenge in the Spanish agrifood sector is the disparity in AI accessibility between large agricultural enterprises and small farming operations. While larger companies possess the financial and technical resources to invest in AI-driven solutions, smaller farms often lack the capital and expertise required to deploy such technologies. Bridging this gap necessitates coordinated efforts from both public and private sectors to democratize access to AI solutions. Possible strategies include financial incentives such as subsidies, dedicated training programs, and the development of cost-effective AI tools designed specifically for small-scale agricultural enterprises.

While Spain’s diverse climate and agricultural landscapes present challenges for AI model development, they also provide unique opportunities for advancing AI applications. Adapting AI algorithms to local conditions—such as Spain’s Mediterranean climate and specific crop varieties—will be crucial to maximizing their effectiveness. Continued research and innovation, driven by interdisciplinary collaboration between academic institutions, industry stakeholders, and policymakers, will be fundamental in unlocking the full potential of AI technologies in the agrifood sector.

Although AI-driven innovations are already delivering tangible benefits to the global agrifood industry, Spain still holds substantial untapped potential in this field. The country’s agricultural diversity, combined with the increasing accessibility of AI-based solutions, creates a strong foundation for innovation. However, realizing this potential requires significant investment in infrastructure, dedicated research efforts, and the development of AI applications tailored to the specific needs of Spanish agriculture. Addressing these areas would position Spain as a leading innovator in AI applications within agriculture, improving sustainability, productivity, and overall sector efficiency.

Future research must focus on ensuring that AI technologies become more accessible, adaptable, and sustainable, thereby promoting inclusivity and resilience across the agrifood industry. By addressing these strategic priorities, researchers and stakeholders can facilitate the equitable distribution of AI-driven benefits, fostering sustainable and efficient practices across the entire agrifood sector in Spain and beyond.

Author Contributions

Conceptualization, J.A.-R. and J.B.-F.; methodology, J.A.-R. and F.-J.F.-M.; software, J.A.-R.; validation, E.M.-C., R.T.-R., and J.-I.L.-B.; formal analysis, J.-I.L.-B. and R.T.-R.; investigation, J.A.-R. and F.-J.F.-M.; resources, E.M.-C.; data curation, J.A.-R.; writing—original draft preparation, J.A.-R., F.-J.F.-M. and J.B.-F.; writing—review and editing, J.B.-F., R.T.-R. and J.-I.L.-B.; visualization, E.M.-C.; supervision, J.B.-F.; project administration, E.M.-C. All authors have read and agreed to the published version of the manuscript version of the manuscript.

Conflicts of Interest

The authors declare no conflicts of interest.

Footnotes

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.

Figures and Tables

Figure 1. Yearly trend in the publication of scientific articles on AI applications in the agrifood industry (2020–2024).

Figure 2. Global distribution of scientific publications on AI applications in the agrifood industry by country.

Figure 3. Frequency of research collaborations among the top 15 countries publishing on AI applications in the agrifood industry.

Figure 4. Keyword co-occurrence matrix mapping AI algorithm applications in the agrifood industry.

Figure 5. Sentiment polarity analysis of abstracts from the top 100 most cited scientific publications.

Figure 6. Similarity heatmap representing research overlaps between key scientific topics.

Figure 7. PCA-based clustering of scientific publications categorized by research themes and associated keywords.

Table 1

Summary of AI algorithms and their specific applications in the agrifood industry based on bibliometric analysis.

Algorithm/Technique	Core Function	Potential Applications in the Agrifood Industry	BibliographyReferences
Supervised Learning	Predictive modeling from labeled datasets	Pest detection, yield prediction, product grading	[2,3,5,6]
Unsupervised Learning	Pattern recognition in unlabeled datasets	Field segmentation, anomaly detection, soil classification	[7,10,13]
SVMs	Classification and regression using hyperplanes	Crop disease detection, vineyard monitoring	[16,17,18]
Genetic Algorithms	Optimization through evolutionary processes	Production scheduling, irrigation optimization	[22,23,24]
Convolutional Neural Networks	Feature extraction from grid-structured data	Crop health monitoring, UAV-based disease detection	[25,26,27]
K-Means Clustering	Data partitioning into homogeneous clusters	Field zoning, yield estimation	[14,28]
Random Forest	Ensemble learning with decision trees	Yield prediction, pest outbreak forecasting	[29,30,31]
Reinforcement Learning	Decision-making through interaction	Irrigation management, adaptive resource allocation	[32,33,36]
Natural Language Processing	Text analysis and language understanding	Market analysis, policy evaluation, sentiment analysis	[37,39,40]
Fuzzy Logic	Reasoning under uncertainty	Climate control, precision irrigation	[41,42,44]
Particle Swarm Optimization	Optimization through swarm-based exploration	Resource management, route optimization	[46,47,48]
ARIMA Models	Time series forecasting	Weather forecasting, demand prediction	[50,51,53]
Hierarchical Clustering	Creation of nested data clusters	Land-use planning, crop grouping	[54,55,57]
Generative Adversarial Networks	Synthetic data generation through adversarial learning	Data augmentation, crop monitoring, disease diagnostics	[58,59,60]

Table 2

Geographic distribution and percentage share of AI-related publications in the agrifood industry across continents (2020–2024).

Continent	Number of Publications	Percentage (%)
Asia	11,513	59
Europe	3632	19
North America	2285	12
Africa	949	5
South America	667	3
Oceania	499	2

Table 3

Comparative analysis of AI-related keywords in global and Spanish scientific publications: Insights from article abstracts.

Keyword (Global)	Count (Global)	Keywords (Spain)	Count (Spain)
Machine Learning	3268	Machine Learning	125
Deep Learning	2917	Deep Learning	62
Artificial Intelligence	864	Precision Agriculture	50
Precision Agriculture	825	Artificial Intelligence	40
Agriculture	767	Agriculture	35
Random Forest	726	Remote Sensing	32
Convolutional N.N	725	Computer Vision	27
Remote Sensing	603	Random Forest	27
Classification	521	Sentinel-2	20
Computer Vision	502	Convolutional N.N.	19

References

1. Russell, S.J.; Norvig, P. Artificial Intelligence: A Modern Approach; Pearson: London, UK, 2016.

2. Bishop, C.M.; Nasrabadi, N.M. Pattern Recognition and Machine Learning; Springer: Berlin/Heidelberg, Germany, 2006; Volume 4.

3. An, S.; Lee, M.; Park, S.; Yang, H.; So, J. An Ensemble of Simple Convolutional Neural Network Models for MNIST Digit Recognition. arXiv; 2020; arXiv: 2008.10400

4. Kamilaris, A.; Prenafeta-Boldú, F.X. Deep Learning in Agriculture: A Survey. Comput. Electron. Agric.; 2018; 147, pp. 70-90. [DOI: https://dx.doi.org/10.1016/j.compag.2018.02.016]

5. García-Vera, Y.E.; Polochè-Arango, A.; Mendivelso-Fajardo, C.A.; Gutiérrez-Bernal, F.J. Hyperspectral Image Analysis and Machine Learning Techniques for Crop Disease Detection and Identification: A Review. Sustainability; 2024; 16, 6064. [DOI: https://dx.doi.org/10.3390/su16146064]

6. Mohanty, S.P.; Hughes, D.P.; Salathé, M. Using Deep Learning for Image-Based Plant Disease Detection. Front. Plant Sci.; 2016; 7, 1419. [DOI: https://dx.doi.org/10.3389/fpls.2016.01419] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/27713752]

7. Hornero, A.; Hernández-Clemente, R.; North, P.R.J.; Beck, P.S.A.; Boscia, D.; Navas-Cortes, J.A.; Zarco-Tejada, P.J. Monitoring the Incidence of Xylella Fastidiosa Infection in Olive Orchards Using Ground-Based Evaluations, Airborne Imaging Spectroscopy and Sentinel-2 Time Series through 3-D Radiative Transfer Modelling. Remote Sens. Environ.; 2020; 236, 111480. [DOI: https://dx.doi.org/10.1016/j.rse.2019.111480]

8. Pantazi, X.E.; Moshou, D.; Alexandridis, T.; Whetton, R.L.; Mouazen, A.M. Wheat Yield Prediction Using Machine Learning and Advanced Sensing Techniques. Comput. Electron. Agric.; 2016; 121, pp. 57-65. [DOI: https://dx.doi.org/10.1016/j.compag.2015.11.018]

9. Mammone, A.; Turchi, M.; Cristianini, N. Support Vector Machines. Wiley Interdiscip. Rev. Comput. Stat.; 2009; 1, pp. 283-289. [DOI: https://dx.doi.org/10.1002/wics.49]

10. Cubero, S.; Lee, W.S.; Aleixos, N.; Albert, F.; Blasco, J. Automated Systems Based on Machine Vision for Inspecting Citrus Fruits from the Field to Postharvest—A Review. Food Bioprocess Technol.; 2016; 9, pp. 1623-1639. [DOI: https://dx.doi.org/10.1007/s11947-016-1767-1]

11. Liakos, K.G.; Busato, P.; Moshou, D.; Pearson, S.; Bochtis, D. Machine Learning in Agriculture: A Review. Sensors; 2018; 18, 2674. [DOI: https://dx.doi.org/10.3390/s18082674]

12. Wang, S.; Li, G.; Ji, X.; Wang, Y.; Xu, B.; Tang, J.; Guo, C. Machine Learning-Driven Assessment of Heavy Metal Contamination in the Impounded Lakes of China’s South-to-North Water Diversion Project: Identifying Spatiotemporal Patterns and Ecological Risks. J. Hazard. Mater.; 2024; 480, 135983. [DOI: https://dx.doi.org/10.1016/j.jhazmat.2024.135983] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/39348756]

13. Suarez Baron, M.J.; Gomez, A.L.; Diaz, J.E.E. Supervised Learning-Based Image Classification for the Detection of Late Blight in Potato Crops. Appl. Sci.; 2022; 12, 9371. [DOI: https://dx.doi.org/10.3390/app12189371]

14. Hastie, T.; Tibshirani, R.; Friedman, J.H. The Elements of Statistical Learning: Data Mining, Inference, and Prediction; Springer Series in Statistics; Springer: Berlin/Heidelberg, Germany, 2001; ISBN 9780387952840

15. Zhu, W.; Miao, J.; Qing, L.; Chen, X. Deep Trans-Layer Unsupervised Networks for Representation Learning. arXiv; 2015; arXiv: 1509.08038

16. Gómez-Candón, D.; De Castro, A.I.; López-Granados, F. Assessing the Accuracy of Mosaics from Unmanned Aerial Vehicle (UAV) Imagery for Precision Agriculture Purposes in Wheat. Precis. Agric.; 2014; 15, pp. 44-56. [DOI: https://dx.doi.org/10.1007/s11119-013-9335-4]

17. Lin, S.; Qi, Z.; Li, X.; Zhang, H.; Lv, Q.; Huang, D. A Phenological-Knowledge-Independent Method for Automatic Paddy Rice Mapping with Time Series of Polarimetric SAR Images. ISPRS J. Photogramm. Remote Sens.; 2024; 218, pp. 628-644. [DOI: https://dx.doi.org/10.1016/j.isprsjprs.2024.09.035]

18. Chai, A.Y.H.; Lee, S.H.; Tay, F.S.; Bonnet, P.; Joly, A. Beyond Supervision: Harnessing Self-Supervised Learning in Unseen Plant Disease Recognition. Neurocomputing; 2024; 610, 128608. [DOI: https://dx.doi.org/10.1016/j.neucom.2024.128608]

19. Moussaid, A.; El Fkihi, S.; Zennayi, Y. Tree Crowns Segmentation and Classification in Overlapping Orchards Based on Satellite Images and Unsupervised Learning Algorithms. J. Imaging; 2021; 7, 241. [DOI: https://dx.doi.org/10.3390/jimaging7110241] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34821872]

20. Cristianini, N.; Shawe-Taylor, J. An Introduction to Support Vector Machines and Other Kernel-Based Learning Methods; Cambridge University Press: Cambridge, UK, 2000; ISBN 9780521780193

21. Camps-Valls, G.; Bruzzone, L. Kernel-Based Methods for Hyperspectral Image Classification. IEEE Trans. Geosci. Remote Sens.; 2005; 43, pp. 1351-1362. [DOI: https://dx.doi.org/10.1109/TGRS.2005.846154]

22. Pal, M.; Mather, P.M. Support Vector Machines for Classification in Remote Sensing. Int. J. Remote Sens.; 2005; 26, pp. 1007-1011. [DOI: https://dx.doi.org/10.1080/01431160512331314083]

23. Liu, S.; Whitty, M. Automatic Grape Bunch Detection in Vineyards with an SVM Classifier. J. Appl. Log.; 2015; 13, pp. 643-653. [DOI: https://dx.doi.org/10.1016/j.jal.2015.06.001]

24. Jamaoui, I.E.; Sánchez, M.J.M.; Sirvent, C.P.; Mana, A.A.; López, S.M. Machine Learning-Driven Modeling for Soil Organic Carbon Estimation from Multispectral Drone Imaging: A Case Study in Corvera, Murcia (Spain). Model. Earth Syst. Environ.; 2024; 10, pp. 3473-3494. [DOI: https://dx.doi.org/10.1007/s40808-024-01963-y]

25. Islam, S.; Samsuzzaman,; Reza, M.N.; Lee, K.-H.; Ahmed, S.; Cho, Y.J.; Noh, D.H.; Chung, S.-O. Image Processing and Support Vector Machine (SVM) for Classifying Environmental Stress Symptoms of Pepper Seedlings Grown in a Plant Factory. Agronomy; 2024; 14, 2043. [DOI: https://dx.doi.org/10.3390/agronomy14092043]

26. Dhal, S.; Braga-Neto, U.; Alvarado, J.; Wherley, B. Machine Learning-Based Smart Irrigation Controller for Runoff Minimization in Turfgrass Irrigation. Smart Agric. Technol.; 2024; 9, 100569. [DOI: https://dx.doi.org/10.1016/j.atech.2024.100569]

27. Xiao, T.; Yang, L.; Zhang, D.; Cui, T.; Wang, L.; Du, Z.; Xie, C.; Li, Z.; Gong, C.; Li, H. et al. Early Prediction of Maize Resistance to Nicosulfuron Using Hyperspectral Imaging and Deep Learning: Method and Mechanism. Comput. Electron. Agric.; 2024; 227, 109511. [DOI: https://dx.doi.org/10.1016/j.compag.2024.109511]

28. Goldberg, D.E. Genetic Algorithms in Search, Optimization, and Machine Learning; Addison Wesley series in artificial intelligence; Addison-Wesley: Boston, MA, USA, 1989; ISBN 9780201157673

29. Liu, Y.C.; Sun, S.H.; Yang, S.M.; Chuang, C.Y. Application of Genetic Algorithm in Production Scheduling: A Case Study on the Food Processing Business. Information; 2012; 15, pp. 6063-6075.

30. Alonso Campos, J.C.; Jiménez-Bello, M.A.; Martínez Alzamora, F. Real-Time Energy Optimization of Irrigation Scheduling by Parallel Multi-Objective Genetic Algorithms. Agric. Water Manag.; 2020; 227, 105857. [DOI: https://dx.doi.org/10.1016/j.agwat.2019.105857]

31. Mahmud, T.; Datta, N.; Chakma, R.; Das, U.K.; Aziz, M.T.; Islam, M.; Salimullah, A.H.M.; Hossain, M.S.; Andersson, K. An Approach for Crop Prediction in Agriculture: Integrating Genetic Algorithms and Machine Learning. IEEE Access; 2024; 12, pp. 173583-173598. [DOI: https://dx.doi.org/10.1109/ACCESS.2024.3478739]

32. Perea, R.G.; Moreno, M.Á.; da Silva Baptista, V.B.; Córcoles, J.I. Decision Support System Based on Genetic Algorithms to Optimize the Daily Management of Water Abstraction from Multiple Groundwater Supply Sources. Water Resour. Manag.; 2020; 34, pp. 4739-4755. [DOI: https://dx.doi.org/10.1007/s11269-020-02687-1]

33. Ghosh, A.; Sufian, A.; Sultana, F.; Chakrabarti, A.; De, D. Fundamental Concepts of Convolutional Neural Network. Intell. Syst. Ref. Libr.; 2019; 172, pp. 519-567. [DOI: https://dx.doi.org/10.1007/978-3-030-32644-9_36]

34. Li, Y.; Zhang, X. Lightweight Deep Learning Model for Underwater Waste Segmentation Based on Sonar Images. Waste Manag.; 2024; 190, pp. 63-73. [DOI: https://dx.doi.org/10.1016/j.wasman.2024.09.008]

35. Thakur, P.S.; Khanna, P.; Sheorey, T.; Ojha, A. Explainable Vision Transformer Enabled Convolutional Neural Network for Plant Disease Identification: PlantXViT. arXiv; 2022; arXiv: 2207.07919

36. García-Duarte, L.; Cifuentes, J.; Marulanda, G. Short-Term Spatio-Temporal Forecasting of Air Temperatures Using Deep Graph Convolutional Neural Networks. Stoch. Environ. Res. Risk Assess.; 2023; 37, pp. 1649-1667. [DOI: https://dx.doi.org/10.1007/s00477-022-02358-0]

37. Kerkech, M.; Hafiane, A.; Canals, R. Deep Leaning Approach with Colorimetric Spaces and Vegetation Indices for Vine Diseases Detection in UAV Images. Comput. Electron. Agric.; 2018; 155, pp. 237-243. [DOI: https://dx.doi.org/10.1016/j.compag.2018.10.006]

38. Chang, W.; Wen, W.; Gu, S.; Li, Y.; Fan, J.; Lu, X.; Chen, B.; Xu, T.; Wang, R.; Guo, X. et al. Canopy Height Uniformity: A New 3D Phenotypic Indicator Linking Individual Plant to Canopy. Comput. Electron. Agric.; 2024; 227, 109491. [DOI: https://dx.doi.org/10.1016/j.compag.2024.109491]

39. Hassan Hayatu, I.; Mohammed, A.; Ahmad Isma’eel, B.; Yusuf Ali, S. K-MEANS CLUSTERING ALGORITHM BASED CLASSIFICATION OF SOIL FERTILITY IN NORTH WEST NIGERIA. Fudma J. Sci.; 2020; 4, pp. 780-787. [DOI: https://dx.doi.org/10.33003/fjs-2020-0402-363]

40. Mazzia, V.; Comba, L.; Khaliq, A.; Chiaberge, M.; Gay, P. UAV and Machine Learning Based Refinement of a Satellite-Driven Vegetation Index for Precision Agriculture. Sensors; 2020; 20, 2530. [DOI: https://dx.doi.org/10.3390/s20092530] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/32365636]

41. Belgiu, M.; Drăgu, L. Random Forest in Remote Sensing: A Review of Applications and Future Directions. ISPRS J. Photogramm. Remote Sens.; 2016; 114, pp. 24-31. [DOI: https://dx.doi.org/10.1016/j.isprsjprs.2016.01.011]

42. Rodríguez-Díaz, F.; Chacón-Maldonado, A.M.; Troncoso-García, A.R.; Asencio-Cortés, G. Explainable Olive Grove and Grapevine Pest Forecasting through Machine Learning-Based Classification and Regression. Results Eng.; 2024; 24, 103058. [DOI: https://dx.doi.org/10.1016/j.rineng.2024.103058]

43. Zhou, T.; Ruud, N.; Šimůnek, J.; Brunetti, G.; Levintal, E.; García, C.P.; Dahlke, H.E. The Impact of Managed Aquifer Recharge on the Fate and Transport of Pesticides in Agricultural Soils. Water Res.; 2024; 267, 122442. [DOI: https://dx.doi.org/10.1016/j.watres.2024.122442] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/39305528]

44. Sharma, P.; Dadheech, P.; Aneja, N.; Aneja, S. Predicting Agriculture Yields Based on Machine Learning Using Regression and Deep Learning. IEEE Access; 2023; 11, pp. 111255-111264. [DOI: https://dx.doi.org/10.1109/ACCESS.2023.3321861]

45. Tanase, M.A.; Mihai, M.C.; Miguel, S.; Cantero, A.; Tijerin, J.; Ruiz-Benito, P.; Domingo, D.; Garcia-Martin, A.; Aponte, C.; Lamelas, M.T. Long-Term Annual Estimation of Forest above Ground Biomass, Canopy Cover, and Height from Airborne and Spaceborne Sensors Synergies in the Iberian Peninsula. Environ. Res.; 2024; 259, 119432. [DOI: https://dx.doi.org/10.1016/j.envres.2024.119432] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/38944104]

46. Sutton, R.S.; Barto, A.G. Reinforcement Learning: An Introduction; Adaptive Computation and Machine Learning Series; MIT Press: Cambridge, MA, USA, 1998; ISBN 9780262303842

47. Elavarasan, D.; Durairaj Vincent, P.M. Crop Yield Prediction Using Deep Reinforcement Learning Model for Sustainable Agrarian Applications. IEEE Access; 2020; 8, pp. 86886-86901. [DOI: https://dx.doi.org/10.1109/ACCESS.2020.2992480]

48. Chen, H.; Yuan, X.; Kang, J.; Yang, D.; Yang, T.; Ao, X.; Li, S. Reinforcement Learning-Based Assimilation of the WOFOST Crop Model. Smart Agric. Technol.; 2024; 9, 100604. [DOI: https://dx.doi.org/10.1016/j.atech.2024.100604]

49. Yang, Y.; Hu, J.; Porter, D.; Marek, T.; Heflin, K.; Kong, H.; Sun, L. Deep Reinforcement Learning-Based Irrigation Scheduling. Trans. ASABE; 2020; 63, pp. 549-556. [DOI: https://dx.doi.org/10.13031/trans.13633]

50. Gautron, R.; Maillard, O.A.; Preux, P.; Corbeels, M.; Sabbadin, R. Reinforcement Learning for Crop Management Support: Review, Prospects and Challenges. Comput. Electron. Agric.; 2022; 200, 107182. [DOI: https://dx.doi.org/10.1016/j.compag.2022.107182]

51. Nguyen, T.T.; Brownstein, K.J. Utilization of a Natural Language Processing-Based Approach to Determine the Composition of Artifact Residues. BMC Bioinform.; 2024; 25, 311. [DOI: https://dx.doi.org/10.1186/s12859-024-05888-2]

52. Hernández-Romero, I.M.; Niño-Caballero, J.C.; González, L.T.; Pérez-Rodríguez, M.; Flores-Tlacuahuac, A.; Montesinos-Castellanos, A. Waste Management Optimization with NLP Modeling and Waste-to-Energy in a Circular Economy. Sci. Rep.; 2024; 14, 19859. [DOI: https://dx.doi.org/10.1038/s41598-024-69321-7] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/39191830]

53. Domínguez, A.G.; Roig-Tierno, N.; Chaparro-Banegas, N.; García-Álvarez-Coque, J.M. Natural Language Processing of Social Network Data for the Evaluation of Agricultural and Rural Policies. J. Rural. Stud.; 2024; 109, 103341. [DOI: https://dx.doi.org/10.1016/j.jrurstud.2024.103341]

54. Ortego-Antón, M.T.; Ortego-Antón, M.T. Dried-Meats Descriptive Cards in Spanish and English: A Corpus Based Study of the Rhetoric Structure. Rev. Signos; 2020; 53, pp. 170-194. [DOI: https://dx.doi.org/10.4067/S0718-09342020000100170]

55. Ross, T.J. Fuzzy Logic with Engineering Applications; Wiley: Chichester, UK, 2016; ISBN 9781119235866

56. Ben Ali, R.; Bouadila, S.; Mami, A. Development of a Fuzzy Logic Controller Applied to an Agricultural Greenhouse Experimentally Validated. Appl. Therm. Eng.; 2018; 141, pp. 798-810. [DOI: https://dx.doi.org/10.1016/j.applthermaleng.2018.06.014]

57. Jamroen, C.; Jirayus, P.; Saneseepon, P.; Suttikul, T. Toward Sustainable Greenhouses: Self-Tuning Adaptive Temperature Control and Its Renewable Energy Potential. Energy Rep.; 2024; 12, pp. 1608-1622. [DOI: https://dx.doi.org/10.1016/j.egyr.2024.07.043]

58. Keswani, B.; Mohapatra, A.G.; Keswani, P.; Khanna, A.; Gupta, D.; Rodrigues, J. Improving Weather Dependent Zone Specific Irrigation Control Scheme in IoT and Big Data Enabled Self Driven Precision Agriculture Mechanism. Enterp. Inf. Syst.; 2020; 14, pp. 1494-1515. [DOI: https://dx.doi.org/10.1080/17517575.2020.1713406]

59. Stitou, H.; Atillah, M.A.; Boudaoud, A.; Aqil, M. Load Forecasting Using Fuzzy Logic, Artificial Neural Network, and Adaptive Neuro-Fuzzy Inference System Approaches: Application to South-Western Morocco. Int. J. Electr. Comput. Eng. IJECE; 2024; 14, pp. 7067-7079. [DOI: https://dx.doi.org/10.11591/ijece.v14i6.pp7067-7079]

60. Kim, T.H.; Shahroz, M.; Alabdullah, B.; Innab, N.; Baili, J.; Umer, M.; Majeed, F.; Ashraf, I. ANFIS Fuzzy Convolutional Neural Network Model for Leaf Disease Detection. Front. Plant Sci.; 2024; 15, 1465960. [DOI: https://dx.doi.org/10.3389/fpls.2024.1465960]

61. Kennedy, J.; Eberhart, R.C.; Shi, Y. Swarm Intelligence; Evolutionary Computation Series; Elsevier Science: Amsterdam, The Netherlands, 2001; ISBN 9781558605954

62. Boudjerda, M.; Touaibia, B.; Mihoubi, M.K.; Kisi, O.; Ehteram, M.; El-Shafie, A. Optimization of Reservoir Operation by Sine Cosine Algorithm: A Case of Study in Algeria. Sustain. Comput. Inform. Syst.; 2024; 44, 101035. [DOI: https://dx.doi.org/10.1016/j.suscom.2024.101035]

63. Singh, N.; Tewari, V.K.; Biswas, P.K.; Dhruw, L.K.; Hota, S.; Mahore, V. In-Field Performance Evaluation of Robotic Arm Developed for Harvesting Cotton Bolls. Comput. Electron. Agric.; 2024; 227, 109517. [DOI: https://dx.doi.org/10.1016/j.compag.2024.109517]

64. Jawad, H.M.; Jawad, A.M.; Nordin, R.; Gharghan, S.K.; Abdullah, N.F.; Ismail, M.; Abu-Alshaeer, M.J. Accurate Empirical Path-Loss Model Based on Particle Swarm Optimization for Wireless Sensor Networks in Smart Agriculture. IEEE Sens. J.; 2020; 20, pp. 552-561. [DOI: https://dx.doi.org/10.1109/JSEN.2019.2940186]

65. Granata, F.; Di Nunno, F.; de Marinis, G. Advanced Evapotranspiration Forecasting in Central Italy: Stacked MLP-RF Algorithm and Correlated Nystrom Views with Feature Selection Strategies. Comput. Electron. Agric.; 2024; 220, 108887. [DOI: https://dx.doi.org/10.1016/j.compag.2024.108887]

66. Box, G.E.P.; Jenkins, G.M.; Reinsel, G.C.; Ljung, G.M. Time Series Analysis: Forecasting and Control; Wiley Series in Probability and Statistics; Wiley: Hoboken, USA, 2015; ISBN 9781118674918

67. Dhal, S.B.; Kalafatis, S.; Braga-Neto, U.; Gadepally, K.C.; Landivar-Scott, J.L.; Zhao, L.; Nowka, K.; Landivar, J.; Pal, P.; Bhandari, M. Testing the Performance of LSTM and ARIMA Models for In-Season Forecasting of Canopy Cover (CC) in Cotton Crops. Remote Sens.; 2024; 16, 1906. [DOI: https://dx.doi.org/10.3390/rs16111906]

68. Jariwala, K.A.; Agnihotri, P.G. Time Series Modelling of Drought in a Semi-Arid Region of South Gujarat, India. J. Earth Syst. Sci.; 2024; 133, 154. [DOI: https://dx.doi.org/10.1007/s12040-024-02357-5]

69. Zahoor, A.; Zhang, J.; Wu, D.; Lin Chen, J.; Nihed, B.; Sen, T.; Yu, Y.; Mao, G.; Yang, P. A Systematic Study Involving Patent Analysis and Theoretical Modeling of Eco-Friendly Technologies for Electric Vehicles and Power Batteries to Ease Carbon Emission from the Transportation Industry. Energy Convers. Manag.; 2024; 321, 118996. [DOI: https://dx.doi.org/10.1016/j.enconman.2024.118996]

70. Hachimi, C.E.; Belaqziz, S.; Khabba, S.; Hssaine, B.A.; Kharrou, M.H.; Chehbouni, A. Advancements in Weather Forecasting for Precision Agriculture: From Statistical Modeling to Transformer-Based Architectures. Stoch. Environ. Res. Risk Assess.; 2024; 38, pp. 3695-3717. [DOI: https://dx.doi.org/10.1007/s00477-024-02778-0]

71. Han, J.; Kamber, M.; Pei, J. Data Mining: Concepts and Techniques; The Morgan Kaufmann Series in Data Management Systems; Morgan Kaufmann: Waltham, MA, USA, 2011; ISBN 9780123814807

72. Imran, S.; Sarker, P.; Mahamud, M.A.; Paul, N.C.; Chakrobortty, J.; Harine, I.J.; Rahman, M.A.; Rahimi, M. Copper Mitigates Salinity Stress by Regulating Water Status, Photosynthetic Pigments and Ion Homeostasis and Increases the Yield of Eggplant (Solanum Melongena). BMC Plant Biol.; 2024; 24, 927. [DOI: https://dx.doi.org/10.1186/s12870-024-05625-0] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/39367326]

73. Cheng, Z.; Liu, X.; Zhao, B.; Liu, X.; Yang, X.; Zhang, X.; Feng, X. A Portable Europium Complex-Loaded Fluorescent Test Paper Combined with Smartphone Analysis for the on-Site and Visual Detection of Mancozeb in Food Samples. Food Chem.; 2024; 458, 140311. [DOI: https://dx.doi.org/10.1016/j.foodchem.2024.140311] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/38968718]

74. Schnack, A.; Bartsch, F.; Osburg, V.S.; Errmann, A. Sustainable Agricultural Technologies of the Future: Determination of Adoption Readiness for Different Consumer Groups. Technol. Forecast. Soc. Change; 2024; 208, 123697. [DOI: https://dx.doi.org/10.1016/j.techfore.2024.123697]

75. Barriere, V.; Claverie, M.; Schneider, M.; Lemoine, G.; d’Andrimont, R. Boosting Crop Classification by Hierarchically Fusing Satellite, Rotational, and Contextual Data. Remote Sens. Environ.; 2023; 305, 114110. [DOI: https://dx.doi.org/10.1016/j.rse.2024.114110]

76. Bregaglio, S.; Ginaldi, F.; Raparelli, E.; Fila, G.; Bajocco, S. Improving Crop Yield Prediction Accuracy by Embedding Phenological Heterogeneity into Model Parameter Sets. Agric. Syst.; 2023; 209, 103666. [DOI: https://dx.doi.org/10.1016/j.agsy.2023.103666]

77. Goodfellow, I.; Bengio, Y.; Courville, A. Deep Learning; Adaptive Computation and Machine Learning Series; MIT Press: Cambridge, MA, USA, 2016; ISBN 9780262035613

78. Hu, H.; Mei, Y.; Zhou, Y.; Zhao, Y.; Fu, L.; Xu, H.; Mao, X.; Huang, L. Optimizing Starch Content Prediction in Kudzu: Integrating Hyperspectral Imaging and Deep Learning with WGAN-GP. Food Control; 2024; 166, 110762. [DOI: https://dx.doi.org/10.1016/j.foodcont.2024.110762]

79. Fawakherji, M.; Suriani, V.; Nardi, D.; Bloisi, D.D. Shape and Style GAN-Based Multispectral Data Augmentation for Crop/Weed Segmentation in Precision Farming. Crop Prot.; 2024; 184, 106848. [DOI: https://dx.doi.org/10.1016/j.cropro.2024.106848]

80. Khare, O.; Mane, S.; Kulkarni, H.; Barve, N. LeafNST: An Improved Data Augmentation Method for Classification of Plant Disease Using Object-Based Neural Style Transfer. Discov. Artif. Intell.; 2024; 4, 50. [DOI: https://dx.doi.org/10.1007/s44163-024-00150-3]

81. Morales-García, J.; Bueno-Crespo, A.; Terroso-Sáenz, F.; Arcas-Túnez, F.; Martínez-España, R.; Cecilia, J.M. Evaluation of Synthetic Data Generation for Intelligent Climate Control in Greenhouses. Appl. Intell.; 2023; 53, pp. 24765-24781. [DOI: https://dx.doi.org/10.1007/s10489-023-04783-2]

82. Alonso, R.S.; Sittón-Candanedo, I.; García, Ó.; Prieto, J.; Rodríguez-González, S. An Intelligent Edge-IoT Platform for Monitoring Livestock and Crops in a Dairy Farming Scenario. Ad. Hoc Netw.; 2020; 98, 102047. [DOI: https://dx.doi.org/10.1016/j.adhoc.2019.102047]

83. Belletti, B.; Garcia de Leaniz, C.; Jones, J.; Bizzi, S.; Börger, L.; Segura, G.; Castelletti, A.; van de Bund, W.; Aarestrup, K.; Barry, J. et al. More than One Million Barriers Fragment Europe’s Rivers. Nature; 2020; 588, pp. 436-441. [DOI: https://dx.doi.org/10.1038/s41586-020-3005-2] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33328667]

84. Heintz, F.; Belbachir, N.; Curry, E. Strategic Research, Innovation, and Deployment Agenda 2025–2027. Available online: https://adr-association.eu/news/adra-releases-strategic-research-innovation-and-deployment-agenda (accessed on 12 December 2024).

85. de Castro, A.I.; Peña, J.M.; Torres-Sánchez, J.; Jiménez-Brenes, F.M.; Valencia-Gredilla, F.; Recasens, J.; López-Granados, F. Mapping Cynodon Dactylon Infesting Cover Crops with an Automatic Decision Tree-OBIA Procedure and UAV Imagery for Precision Viticulture. Remote Sens.; 2019; 12, 56. [DOI: https://dx.doi.org/10.3390/rs12010056]

86. Peña, J.M.; Torres-Sánchez, J.; Serrano-Pérez, A.; López-Granados, F. Detección de Malas Hierbas En Girasol En Fase Temprana Mediante Imágenes Tomadas Con Un Vehículo Aéreo No Tripulado (UAV). Rev. De Teledetección; 2017; 42, pp. 39-48. [DOI: https://dx.doi.org/10.4995/raet.2014.3148]