Emerging Trends in AI-Based Soil Contamination Monitoring and Prevention

Abstract

Soil health directly impacts food security, so investigating contaminants is a topic of interest for the anticipatory study of the action–effect correlation. This paper conducts a systematic literature review through seven analyses, identifying researchers’ interest in soil health using artificial intelligence tools. The first study examines the distribution of articles over the years to assess researchers’ interest in soil health, and subsequently, the same analysis is conducted regarding artificial intelligence (AI) methods. Additionally, the productivity of authors, the distribution of articles by country, relevant publications, and the frequency of keywords are analyzed to identify areas of interest associated with soil health. Subsequently, the branches of AI and examples of applications that have already been investigated in the specialized literature are identified, allowing areas that are currently underexplored to be pinpointed. This paper also proposes a specialized analysis using an algorithm specifically developed by the author for this investigation, which evaluates the interdisciplinary potential of the articles analyzed in the literature. In this way, the authors of the present research will propose new research directions that include machine learning, natural language processing, computer visualization, and other artificial intelligence techniques for monitoring soil contaminants. They will also suggest using these tools as preventive measures to minimize the negative impact of contaminants on the soil. The direct consequence is the protection of soil health and its effects on human health.

1. Introduction

The agricultural sector, alongside the medical industry, is one of humanity’s most important fields of activity. This is because the agricultural sector provides a vital component of human existence [1]. For these reasons, current technologies are being explored intensively to streamline crop production. One of the most intensively studied components in the literature is the machine learning (ML) component of the artificial intelligence (AI) field [2]. This field is actively integrated into many human activity sectors, and the agricultural industry has also been included in the list of AI priorities. The ML component contributes to the progress of the farm field by anticipating outcomes through the analysis of different scenarios. This behavior allows farmers to identify the best scenario to achieve the best harvest. One of the directions in which ML operates as a component of anticipating crop evolution is related to soil health, including fertility, erosion, and the parameters that influence agricultural productivity. Soil health is a topic intensely debated by researchers because it influences crop quality and consumer health.

This paper aims to be a review article, presenting a bibliometric and systematic analysis of the scientific literature regarding the use of AI technologies in the field of soil health, especially in the context of contamination and contamination prevention in agriculture. The purpose of this paper is to map the trends and research directions related to AI technologies used in soil health studies. In this way, gaps in the literature are identified and future directions are proposed. This paper contains thematic reviews and literature syntheses, as is typical of review-type works. Thus, concrete applications of AI in contamination monitoring, organic carbon prediction, etc., are analyzed. Additionally, this paper includes lists of references selected according to a WOS search protocol. The bibliometric nature is also supported by the inclusion of figures with visual representations of academic collaborations, co-keyword maps, annual publication distributions, etc. Additionally, the use of the VOSViewer 1.6.20 tool for mapping research trends demonstrates the bibliometric nature of the review.

The central objective of this study is to demonstrate the interdisciplinary nature of AI tools in the agricultural field by identifying how ML models contribute to the assessment, monitoring, prevention, and action regarding soil contamination. Soil health encompasses a multitude of factors, including erosion, the decline of organic carbon, the use of chemical fertilizers, and microbiological biodiversity. These are not the central factors of the present work. These factors constitute contextual elements for the specialized literature. This study focuses on the intersection between AI technologies and soil contamination, aiming to highlight that the field has not been sufficiently explored to date and that researchers should delve more into this dual approach of AI and soil health.

The authors emphasize, through the approaches in this work, the necessity of intensifying interdisciplinary research between AI and soil health. Although this paper analyzes all AI tools, the emphasis is on ML models, as they are primarily used in the assessment, monitoring, prevention, and intervention of soil contamination. Therefore, the main focus is on ML due to its predictive capabilities in agricultural and ecological fields. These explanations justify the central purpose of this study, which is to demonstrate the necessity of an interdisciplinary approach between ML and soil contamination.

This study will identify the factors influencing soil health and the ML elements that anticipate each type of scenario. This paper aims to identify future research directions using ML to find less-researched subdomains. This paper focuses particularly on soil pollution from plastic, which directly affects consumer health in the long term.

This research paper has the following contributions:

• This paper identifies the major factors influencing soil health. This research identifies the soil contaminants and discusses their health implications for consumers.
• This paper establishes correlations between the types of soil contamination that can be addressed proactively through the AI component. Additionally, this paper evaluates researchers’ interest in exploring AI technologies for preventive assessment and proactive action for soil health.
• This paper identifies the correlations between the problem typologies modeled through AI and the problem typologies with direct implications on soil health. This paper discusses the measures through which soil contamination level forecasts can generate proactive measures to reduce the effects of pollution.
• This paper outlines the less explored elements in the literature regarding soil health treated in advance through AI.

By achieving these four significant contributions, this paper will indicate future research directions that can, in the long term, improve the techniques used in soil health monitoring to increase quality and productivity.

The primary objective of this study is to analyze soil contamination, considering preventive measures informed by AI technologies. Since the number of scientific papers directly addressing this topic is limited, it has been observed that the current state of research in which AI tools contribute to monitoring soil contaminants is insufficiently explored at this time. The bibliographic selection includes studies directly related to soil contamination and remediation, as well as works that address soil health, fertility, the impact on the food chain, and preventive measures. In this way, this work provides the reader with the necessary foundations for the further development of research in the preventive direction. Identifying current gaps highlights the potential of underutilized AI technologies concerning soil contamination. This study encourages future interdisciplinary research on AI tools in predicting, monitoring, preventing, and intervening in soil contamination.

2. Literature Review

Soil health refers to how the ecosystem associated with the soil is managed. The composition and functionality of the microbial communities in the soil influence the soil’s health. The nutrient cycle, the decomposition of organic matter, and the overall fertility of the soil constitute the microbial diversity in soil ecosystems [3]. They directly influence healthy plant growth as well as soil productivity. Griffiths and Philippot [4] mention that microbial diversity leads to the increased resilience and resistance of soil systems to various disturbances [5]. In this way, healthy soil management is ensured. Specialized works mention the importance of agricultural practices in soil health. The research by Lense et al. [6] specifies crop rotation and intercropping practices to reduce soil erosion. In this way, microbial diversity and activity are stimulated.

On the other hand, Raliya and Cappellari et al. [7,8] mention the excessive use of chemical fertilizers, which tend to diminish microbial diversity. This leads to a decline in soil health in the long term. Organic fertilizers, such as urine, have demonstrated benefits for soil microbial health and promote a diverse microbial community to support nutrient cycling and soil fertility [9]. The soil’s physicochemical properties influence microbial communities’ structure and functionality. Imarhiagbe and Onwudiwe [10] present the variations in soil properties at different depths. These variations have directly impacted microbial diversity and activity through the interconnection between soil parameters and microbial health. Abu-Qaoud et al. [11] explore the integration of microorganisms into agricultural practices to improve soil quality by creating a favorable environment for microbial populations that enhance plant growth and yield under stress conditions. Urban soils are also explored in the specialized literature due to the unique challenges related to contamination and degradation that negatively affect the microbial community and soil health. The researches [12,13,14] study the efforts of urban greening through diverse vegetation on microbial health, with positive implications for ecological stability and human health. The incorporation of different types of vegetation in urban spaces to effectively restore soil microbiota promotes a healthy soil system in urban landscapes.

2.1. Soil Contaminants

Soil pollution is an extensively studied phenomenon in the specialized literature. It is interesting from the perspective of identifying sources, the direct impact on plants, the overall impact on agriculture, and human health. Factors that pollute the soil include heavy metals, pollution with polycyclic aromatic hydrocarbons (PAHs), pollution with phthalate esters (PAEs), pollution with microplastics (MPs) and macroplastics, and other less-studied factors.

Heavy metals are the most common cause of soil pollution. They affect ecosystems through accumulation and transfer in the food chain. The main identified toxic metals include the following:

• Cadmium (Cd), which appears in high concentrations in agricultural soils in China [15,16,17], Peru [18], and other regions;
• Lead (Pb), which is present in industrial and mining areas, such as Chile [19] and China [15];
• Arsenic (As), which is identified in contaminated soils in China [20,21] and Serbia [22];
• Chromium (Cr), Copper (Cu), Zinc (Zn), and Nickel (Ni), which represent metals associated with geogenic and farming chemical fertilizers [23], mining activities [16,19], and industrial processes [17];
• Mercury (Hg), which is present in mining areas [16,24] and intensive agriculture [25].

Health risks include carcinogenic and non-carcinogenic exposure. Children represent the most vulnerable group. The primary sources of contamination are anthropogenic activities, including intensive agriculture, mining, and traffic [26,27].

PAHs are present in agricultural and urban soils in Bangladesh [28] and China [29]. PAHs originate from pyrogenic sources (biomass burning, coal) and petrogenic sources (industrial activities). These persistent compounds accumulate in the soil and enter the food chain. The direct consequences are visible at the level of plant and human health. The carcinogenic risks are considered moderate, and higher concentrations appear in industrial and coastal areas.

PAE compounds, such as di(2-ethylhexyl) phthalate (DEHP) and di-n-butyl phthalate (DnBP), are common pollutants in agricultural soils in Xinjiang [30] and the coastal areas of South China [31]. These endocrine-disrupting compounds affect human health through bioaccumulation in plants and soil. Higher concentrations are attributed to rapid urbanization and intensive agricultural activities [32].

Microplastics affect agricultural areas through irrigation systems and organic manure. Sharmin et al. [33] highlights the presence of various types of MPs (HDPE, LDPE, PP, PET, PVC) and forms (fiber, film, fragment) in agricultural soils.

Mercury pollution is a specific topic analyzed regarding artisanal and small-scale gold mining (ASGM) operations in Sudan [34]. The waste from amalgamation leads to extreme pollution in mining areas and moderate levels in agricultural and residential areas. The inhalation of mercury vapors is the main route of exposure and poses risks for children [35].

Zoghlami et al. [36] use deinking sludge (DS) and combined deinking sludge (DSC) as a soil amendment and offers benefits such as improving soil fertility and waste management. However, these materials may contain litter contaminants and can reduce soil porosity, affecting soil health. The study by Moriarity et al. [37] evaluates the risks introduced by the use of pesticides to the health of Indigenous communities in Australia and Canada. Children are at increased risk of exposure to Pb and As. This paper mentions that pesticides affect the health of agricultural soils. Ma et al. [38] assess exposure to herbicides (HBCs) in black soil regions, with maximum detected values of 6288 ng/L. Acetochlor, fomesafen, bentazon, and atrazine (ATZ) exhibit high detection rates. Non-carcinogenic risks are below the acceptable limit (<1), but the carcinogenic risk associated with ATZ is 10⁻⁵ for children and 10⁻⁶ for adults. Monitoring organic contaminants in the soil is a measure to support soil health and human health [39].

Residual nanomaterials [40] and zeolite [41] are being investigated as solutions for remediating contaminated soil. Nanobiochar (nB) and nano-water treatment residues (nWTRs) reduce heavy metal concentrations and improve soil fertility. Zeolite immobilizes potential toxic elements (PTEs), reducing their availability for plants.

Microcystins (MCs) affect the health of the soil and fava bean plants by inhibiting plant growth and contaminating the food chain [42]. The excess of fertilizers with NO₃ increases the concentrations of nitrates and nitrites in soil, water, and food. Secondary compounds, such as nitrosamines, are toxic. Kundu et al. [43] analyze the sources of contamination that affect soil health. The soil contaminants, such as Pb, regulate their mobility and bioavailability. In urban agriculture, the association of Pb with Al oxide phases and phosphates suggests an increased immobility through natural “soil aging” processes. Entwistle et al. [44] show that urban gardens are safe even at Pb levels 10 times above the recommended limits. Furthermore, the issue of these contaminants in agricultural soils arises.

Pollution with hydrophobic organic compounds, such as petroleum hydrocarbons, affects soil, water, and air and has toxic and carcinogenic effects on biota. These compounds, resistant to degradation, persist in the environment, generating risks for human health and ecosystems and affecting soil health [45]. Bioremediation methods, including solubilizing agents and surfactants, offer ecological solutions for environmental decontamination that should be intensively investigated in the specialized literature.

Air, water, and soil pollution cause diseases and premature deaths. Heavy metals, pesticides, plastics, and over-fertilization affect the soil by reducing beneficial microorganisms and contaminating groundwater. Volatile organic compounds (VOCs) emit greenhouse gases and particles that exacerbate climate change and cause respiratory, cardiovascular, and cancer-related diseases [46].

Figure 1 illustrates the main soil contaminant types, including traditional and emerging pollutants. Heavy metals (Cd, Pb, As, Cr, Cu, Zn, Ni, and Hg) represent inorganic soil pollutants due to their high toxicity. PAHs, PAE, and VOCs represent organic contaminants impacting human health and the environment. Additionally, Figure 1 includes petroleum residues, pesticides, and herbicides, all common in agricultural and industrial areas.

MPs, nitrates, phosphates from fertilizers, and pharmaceutical residues including DS and DSC highlight the emerging contaminants in Figure 1. MCs are a general factor that designates biologically active substances in very low concentrations. Overall, Figure 1 brings together various sources of soil pollution, highlighting the complexity of this phenomenon.

2.2. Soil Health and AI

AI is a discipline subdivided into six categories, each dedicated to different aspects of imitating human cognitive functions. These categories are represented by ML, natural language processing (NLP), computer vision (CV), robotics, expert systems (ESs), and evolutionary computation (EC). Figure 2 presents a synthesis of the main component elements in AI classification.

ML is one of the most studied subfields of AI [47]. This involves a series of algorithms that allow machines to learn from large datasets and also enable them, through these algorithms, to develop performance over time without being explicitly programmed [48,49,50]. In the category of ML algorithms, there are supervised learning, unsupervised learning, reinforcement learning, and last but not least, semi-supervised learning [51,52].

NLP is an AI component that uses natural language to interact with computers and humans. It enables speech recognition, sentiment analysis, and machine translation [53]. This field also relies on ML techniques to interpret and generate human language.

In addition to ML and NLP, another subdomain of AI is CV. This refers to the ability of machines to interpret and make decisions based on visual data. This component overlaps with deep learning (DL) through convolutional neural networks (CNNs). These represent a specific type of DL model and are used in image classification, object detection, and facial recognition [54,55,56].

Robotics is a component of AI in which intelligent agents are designed to perform a series of physical tasks. Robots are used in sectors such as healthcare, manufacturing, and logistics, where they can assist in performing complex tasks [57].

Additionally, the field of AI includes more specialized areas, such as expert systems based on knowledge bases, inference engines, and rule-based systems [58]. There are also fields such as evolutionary computation, which include genetic algorithms, evolutionary strategies, and swarm intelligence [59,60,61].

The interdisciplinary approach between soil health and artificial intelligence techniques is expanding within agricultural sciences. The application of advanced computational techniques is still in its early stages when analyzing and simulating soil properties. ML evaluates soil properties and predicts future crops based on soil conditions. Imam et al. [62] demonstrate that ML models predict potato yields by analyzing the structures of microbial communities in the soil. The correlation between prediction and crop yields was demonstrated in this study. In addition, specialized ML algorithms such as the random forest (RF) algorithm have shown the ability to quantitatively evaluate the influence of certain soil indicators on agricultural productivity in the study by Su et al. [63]. Other algorithms, such as support vector machines (SVMs), have demonstrated soil fertility evaluation through ML. Thus, the paper by Shevchenko et al. [64] highlights complex relationships between soil characteristics and crop health. ML algorithms are integrated into real-time applications for predicting and monitoring soil fertility in the context of [65]. This predicts crops based on soil fertility and weather conditions. Rosca et al. [66] integrates the Internet of Things (IoT) and the RF algorithm to analyze environmental data in real time. The practical implications of ML techniques are demonstrated in the paper by Condran et al. [67], which monitors soil moisture and nutrient levels in irrigation and fertilization optimization strategies. Additionally, Hengl et al. [68] show the importance of RF algorithms in mapping soil properties with direct implications for agricultural practices. In this way, large volumes of data from soil-associated sensors, satellite images, and environmental variables are exploited, which help in making informed decisions based on real-time agricultural data. Moreover, the paper by Tripathi et al. [69] integrates DL methodologies in assessing soil health by exploring large volumes of data. In this way, researchers have discovered soil characteristics that contribute to plant health and productivity. The dynamic soil models presented by Yang et al. [70] ensure ecological and agricultural practices concerning soil health.

The use of NLP in soil health refers to agricultural productivity and ecosystem sustainability. Singh et al. [71] demonstrated that the unbalanced use of fertilizers directly affects soil health and, implicitly, human health. Textual data from farmers’ reports are analyzed by integrating NLP technology into agriculture [72]. The article by Metwally et al. [73] ecologically evaluates, through specific techniques, how to discern patterns related to soil management practices that positively influence soil health. For example, NLP groups soil property data to identify site-specific management practices. In this way, Peter-Jerome et al. [74] align agronomic interventions with the varied state of the soil in different regions.

The chemical, physical, and biological properties of soil are studied in relation to agricultural productivity and ecosystem health in [75,76]. CV studies soil images to extract, through remote sensing techniques, values associated with moisture properties and nutrient levels over extensive areas. This approach enables decision-making in farm management [77,78]. Using CV techniques can also measure soil organic matter through visual and spectral analyses. The works by Zhou et al. and Jordán Vidal [79,80] mention the implications of CV in improving soil structure through the analysis of moisture retention and nutrient availability. CV algorithms trained on large datasets target soil characteristics through automated analyses on specific soil health indicators. Ghazal et al. [81] describe the application of fertilizers, irrigation needs, and crop rotation strategies as elements studied through CV algorithms. The analysis of these algorithms can also extend to the analysis of the soil microbiome. Venturini et al. and Babin et al. [82,83] mention the study of soil health through these algorithms. Visualizing soil microorganism structures and distributions through CV allows for understanding the relationships between soil microbial communities and their health indicators. The growth and health of plants at the soil level are closely correlated with the nutrient cycle and the decomposition of organic matter. Understanding the complexity of these subsystems of soil ecology through CV algorithms requires the analysis of chemical fertilizers that contribute to improving environmental quality [84].

The distribution and sources of heavy metals in the soils of the European Union (EU) are investigated in many studies, which illustrate the concerns of researchers in the field. Based on the LUCAS survey conducted by Ballabio et al. [85], it is shown that 5.5% of soil samples exceed the critical threshold of 1 mg Cd/kg, with an average level of 0.20 mg/kg, which is influenced by pH, texture, and the use of phosphate fertilizers. The paper by Fendrich et al. [86] presents research on arsenic concentrations in different countries in the EU. This study identifies the RF-based model as the most accurate in mapping associated risks. In the paper by Ballabio et al. [87], the concentrations of Cu are noted with values of 49.3 mg/kg due to the use of Cu-based fungicides. Ballabio et al. [88] analyze Hg concentrations in the EU and observes correlations with emissions from coal-fired thermal power plants and mining areas.

Wang el al. [89] combine statistical methods and knowledge-based ML methods to identify the main factors of soil contamination with HMs and PAHs in agricultural areas of China. Additionally, for predicting the distribution of contaminants, the paper by Salgado et al. [90] reported an R² value ranging from 50.1% to 63.0% using the RF algorithm. These results encourage the potential of these technologies in mapping soil contamination.

Remote sensing is used in assessing the condition of oil-contaminated soils in [91,92]. They use images to investigate the increase in these values using a dataset of images from the center of the Tamsag–Bulag oil field in Mongolia [91]. These investigations show the use of the satellite monitoring of areas affected by extractive activities. Dean et al. [92] investigate the impact of oil refining and storage on soil. This is based on images analyzed using ML models. Monitoring soil quality in areas with limited data to manage soil pollution can be achieved using ML techniques, as demonstrated in [93,94].

3. Methodology

The methodology for evaluating the impact of research on soil health will include seven types of analysis, focusing on the papers published between 1 January 2020 and 30 April 2025. Documentation will be carried out through the Web of Science (WOS) platform. This research aims to highlight directions that have not yet been investigated or promise to significantly improve soil health but have not yet been extensively studied. Through this approach, experts in the field can identify the directions to investigate in future research.

The articles included in this analysis are cited in the reference section when they are examined in detail in the literature review or detailed analysis sections (Section 4.6 and Section 4.7). Additionally, synthesis articles were included in the research to highlight the researchers’ interest in the respective topic. The inclusion of review articles is justified by the fact that they provide an overview of the current state of research, the interest of researchers, research trends, existing gaps, as well as future directions of investigation that need to be addressed in subsequent contribution works.

The evaluation of researchers’ interest in soil health regarding contaminants will be studied through seven analyses. Each analysis will also conduct a comparative analysis of researchers’ interest in soil health regarding AI technologies. The seven analyses include the following:

• The distribution of articles by year is investigated. This analysis aims to identify whether interest has been ascending or descending in soil health related to the contaminants and, further, in the combination of soil health investigated through specific AI methods.
• Author productivity is investigated concerning soil health. This analysis is conducted using the VOSviewer 1.6.20 tool through the co-authorship map, for which the minimum number of documents per author will be set to 2 and the minimum number of citations per author to 1. This reflects the author’s interest in exploring the topic of soil health.
• The distribution of articles by country is investigated to determine if there is a correlation between countries with a high level of agriculture and the number of research articles conducted by those countries. Subsequently, research in the field is investigated using AI tools to determine these countries’ interest in the agricultural sector’s technological advancement while adhering to soil health standards. This analysis highlights whether countries with a high level of agriculture have invested in soil health research.
• Publications that include the most research in soil health are investigated to provide authors with an overview of publishers supporting efforts to improve soil health through AI technologies and other auxiliary technologies. Thus, all searches in WOS will be narrowed to the publishers that have published these studies.
• The fifth analysis targets the frequency of keywords in research addressing soil health. This search identifies the fields that relate to this type of issue. The investigation is conducted through the co-occurrence map of keywords associated with each article identified in soil health. This analysis highlights the co-domains of analysis, meaning those related fields that address the issue of soil health. This study is conducted through the clusters created by the VOSviewer 1.6.20 tool.

Figure 3 presents the methodology related to this first set of five analyses. This figure shows that all these analyses are based on a WOS search that includes soil health elements concerning contaminants and the agricultural field for the five years analyzed. The use of AI technologies in addressing soil contamination is still in development at the time of writing this material, as AI technologies themselves are rapidly advancing. In the WOS searches, the authors of this material opted for an extended formulation to identify all relevant reference works in the field, even if they do not use the exact terms “preventive measures” or “contaminants”. This approach avoids excluding important studies solely due to differences in terminology. This study is bibliometric in nature and maps the global research trends of the AI versus soil health relationship. In addition to the bibliometric approach, the authors contributed with a customized analysis of the current state of research in the field and added their personal expertise to the area of soil health.

6.

Analysis number six presents the WOS branches of AI and examples of applications identified in the literature that address soil health issues related to each branch. It will highlight the branches that have not yet been explored.

7.

The Cross-AI Components Innovation Potential (CAI-CIP) analysis is a customized analysis that has not been conducted in any other review-type article. The CAI-CIP analysis identifies the interdisciplinary, innovative potential of the evaluated articles. To implement this analysis, the authors developed a program in C# that assesses the potential for multidisciplinary innovation by semantically analyzing articles between soil health and AI components. Thus, the key elements of the analysis and an interdisciplinary semantic similarity are identified. This similarity is achieved using a predefined list of 200 AI techniques. This way, the relationships between transversal concepts are identified according to the logical scheme in Figure 4. Through this analysis, new research opportunities are discovered by highlighting unexplored connections and prioritizing those studies that impact soil quality.

The algorithm presented in Figure 4 analyzes the abstracts provided by the general search in WOS. It identifies the predefined AI technologies as components of all existing technologies within these abstracts. Subsequently, the number of identified technologies at each abstract level is calculated, and the correlation between the analyzed paper and the number of identified technologies is saved. This algorithm allows for the analysis of AI trends concerning soil health. The algorithm identifies AI technologies in research related to soil health, contaminants, and agriculture. Thus, current research trends are explored through AI technologies, allowing for further innovations in the field. Additionally, the algorithm presented in Figure 4 analyzes the impact of AI technologies by the frequency of their mention at the abstract level. In this way, the obtained results will allow for the prioritization of research in specific directions, including AI technologies not mentioned in the papers identified in the WOS search.

The initial search strategy was based on general terms, such as artificial intelligence and AI. This initial approach acknowledged that this study’s purpose was not to conduct a comprehensive analysis of every type of AI algorithm, but rather to identify general trends in the application of AI in the field of soil health and its contamination. Moreover, most scientific journals recommend using more general terms in the initial selection stage to avoid the excessive fragmentation of the literature. In this way, the entirety of reference works for the respective field is identified. The methodology initially aimed to highlight the global level of interest in the integration of AI in agriculture and soil monitoring, regardless of the specific type of technology used. Subsequently, the methodology included conducting a detailed search regarding the specific technologies presented in Figure 2. For this purpose, the WOS search expression is as follows:

TS=(

  (“soil health” OR “soil contamination”)

  AND

  (

    “artificial intelligence” OR “AI” OR “machine learning” OR “deep learning” OR “neural network” OR “convolutional neural network” OR “recurrent neural network” OR “support vector machine” OR “decision tree” OR “random forest” OR “gradient boosting” OR “natural language processing” OR “text preprocessing” OR “language models” OR

    “computer vision” OR “image processing” OR “object detection” OR “image classification” OR “facial recognition” OR

    “robotics” OR “motion planning” OR “sensor fusion” OR “autonomous navigation” OR “human-robot interaction” OR

    “expert systems” OR “knowledge base” OR “inference engine” OR “rule-based systems” OR

    “evolutionary computation” OR “genetic algorithms” OR “evolution strategies” OR “swarm intelligence”

  )

) AND PY=(2020–2025)

This expression refers to AI technologies in studies related to health and soil contamination. The inclusion of all subfields and the most important algorithms identified in the literature highlights the breadth of existing research and, furthermore, illustrates the variety of AI applications in this interdisciplinary field.

To demonstrate the degree of fragmentation in the research, the search was reperformed to evaluate the impact of each AI component from Figure 2 on soil health through searches in WOS. These searches were structured to include in the keywords ML, NLP, computer vision, robots with AI, expert systems, and evolutionary computation. All searches were limited to the period 2020–2025.

4. Results

Each type of analysis proposed in the methodology section will generate results whose comments will highlight future research directions that specialists in the field should address in an interdisciplinary manner. This interdisciplinarity refers to integrating AI techniques in the preventive study of soil contamination. Alongside this idea related to soil contamination, researchers should expand their searches for preventive simulations in the actions they propose or undertake in the agricultural field concerning soil health.

4.1. Distribution of Articles by Year

The first search, “TS=((“soil” AND “health” AND “contaminants” AND “agriculture”)) AND PY=(2020–2025)”, generated a result of 217 articles on the WOS platform. The results of the soil health query related to agricultural contaminants highlighted in Figure 5 show the increased interest of researchers in this field. Thus, in 2020, there were only 27 studies, while in 2024, the total reached 54 studies, with the trend continuing to rise.

The annual distribution of research primarily aims to illustrate the dynamics of researchers’ interest in soil health, both in the general context of contaminants and agriculture, as well as in the context of integrating AI technologies. Figure 5 compares the evolution of the number of published papers between 2020 and 2025 based on the two initially proposed distinctive search strategies. These analyses regarding the number of publications are standard for review articles and represent a way in which researchers are encouraged to explore certain technologies.

The second search, “TS=((“soil” AND “health” AND (“Artificial Intelligence” OR “AI”))) AND PY=(2020–2025)”, generated 222 articles in WOS. The second search targeting soil health concerning AI techniques showed upward results regarding researchers’ interest, increasing from seven articles in 2020 to ninety-two in 2024. The first quarter of 2025 has already recorded 30 articles in the field of artificial intelligence on soil health.

These results reflect the researchers’ interest in soil health over the past 5 years, but the results are low considering the volume of work in the specialized literature. Moreover, the search that includes soil health concerning contaminants in the agricultural field and AI “TS=((“soil” AND “health” AND “contaminants” AND “agriculture”) AND (“artificial intelligence” OR “AI”)) AND PY=(2020–2025)” yielded only one article in WOS. This result signifies the need to intensify research in the agricultural field regarding soil health and contaminants through specific AI techniques.

4.2. Author Productivity and Co-Authorship Analysis

The analysis of author productivity started from the constraint of a minimum threshold of two articles per author and at least one citation per author, so the work would be considered a reference in soil health. The two initial constraints were established to limit the authors’ scientific output and provide a clearer picture of the academic community’s interest in soil health. The co-authorship network analysis in soil health studies is presented in Figure 6. The VOSViewer 1.6.20 analysis divided the WOS results into two clusters.

The first cluster contains seven elements, while the second contains five. The first cluster indicates the existence of a scientific community that is constantly active in soil health. The second cluster indicates an emerging area of research where collaborations are rarer or more recent. The ideas that emerge from this division into the two clusters highlight the need to broaden the scientific community involved in soil health studies. Attracting new specialists from various interdisciplinary fields, such as artificial intelligence, soil biology, agricultural engineering, and meteorological engineering, along with fostering international collaborations, could significantly expand research in soil health.

The second approach interprets results regarding the productivity of authors in soil health investigated through the lens of AI. The results provided a single cluster of six authors, as seen in Figure 7.

Compared to the first approach, this limited result indicates a minimal base of researchers who systematically publish in soil health within an AI approach. The fact that there are no more clusters indicates no broad collaboration networks centered around AI concerning soil health. Comparing these results with the approach that included terms such as contaminants and agriculture, it can be observed that the community is much more restricted, leading to the observation that the topic related to AI is much less explored. This finding demonstrates that the potential for developing interdisciplinary research in soil health using AI technologies is currently unexplored.

4.3. Geographical Distribution of Research Output

The analysis of the distribution of articles by country provided 24 items grouped into four clusters. The analysis investigates the existence of a correlation between the level of agricultural activities in a given country and the number of studies published by that country in the field of soil health. The results are presented in Figure 8, where the countries China, India, the USA, Pakistan, and Australia can be identified with a dominant number of studies in the field.

Countries with high agricultural production levels essentially correspond with the results identified in this map. For example, China (the world’s largest agricultural country) and the USA (the third-largest agricultural country in the world) are in the same green cluster from Figure 8, with the highest number of research studies on soil health and agriculture. Another noteworthy aspect is that India (the second-largest agricultural country in the world) belongs to the second most important academic research cluster (yellow marked), along with Saudi Arabia and Malaysia. The third cluster, colored in red, includes countries such as Italy, Spain, England, South Korea, Denmark, Portugal, Chile, Cyprus, and Canada, whereas the fourth cluster (blue marked) comprises Brazil, France, Sweden, and Poland. Thus, the countries with high agricultural production are simultaneously investing in soil health research.

Additionally, the map shows many connections among countries, suggesting collaboration among authors in the global academic community. These results highlight the importance of international cooperation in addressing soil health issues. These observations serve as benchmarks for subsequent research that should explore the implications of modern technology in optimizing agricultural systems, especially at the soil health level.

India, China, and the USA are the only countries that have provided results regarding academic research related to the integration of AI techniques in soil health. These are the only results grouped into a single cluster regarding the second approach. These results reinforce the idea that other states should invest equally in soil health because this field is a fundamental source of human life.

4.4. Identification of Key Publishers

Regarding identifying the leading publishers disseminating research in soil health, it can be noted that in the first approach, where the theme is traditional, ecological, chemical, and agricultural, publishers dominate (e.g., Amer Chemical Soc). In the second approach, where the central element is AI, a series of differences emerge due to the transition to technical publishers, such as IEEE or Taylor & Francis. Figure 9 presents the top publishers in soil health research, while Figure 10 shows the prominent publishers in AI-augmented soil health investigations.

Beyond these delimitations, the top publishers remain the same in both cases: Elsevier, MDPI, and Springer Nature. These results show a noticeable increased interest from technical publishers in a multidisciplinary approach, the dominance of top publishers who stand out through the many articles disseminating soil health issues, and opportunities for new publishers who should accept more and more of such interdisciplinary research.

4.5. Keyword Frequency and Thematic Mapping

This analysis focuses on the frequency of authors’ keywords in research dedicated to soil health. This study is reported on the co-occurrence map of terms associated with each article. Figure 11 presents the results obtained from the WOS query, processed with the VosViewer 1.6.20 tool. In this figure, the 29 elements grouped into seven distinct clusters can be observed.

The central cluster is “agriculture” and it is marked in green. This includes approaches targeting microplastics, groundwater, environment, and degradation. Directly connected to this cluster is the blue cluster, with “heavy metals” as its central element. This is associated with the medical area, and in relation to this cluster is the purple one which targets the soil and its derivatives. Connected with the “soil” cluster are the red clusters associated with “soil health” and the yellow cluster, which has existing soil and water contamination as its central element. This interconnection of the clusters demonstrates the need for a simultaneous approach to agriculture in relation to soil health, contamination, water pollution, and heavy metals. The results presented in Figure 11 illustrate the diverse research themes related to soil health. These observations provide a global picture of the dominant themes in the current literature. The observations that emerge from this analysis focus on prioritizing interdisciplinary research, especially that which includes elements of AI.

In the second approach, the co-occurrence map of the authors’ keywords generated 22 elements grouped into five distinct clusters. These highlight the main directions of research that address soil health through the lens of AI technologies (Figure 12).

The yellow cluster focuses on AI, ML, and advanced algorithms. This cluster is associated with remote sensing, sensors, and IoT and demonstrates how digital technologies integrate data from various sources to capture an image of soil health. This approach enables informed decision-making. This cluster demonstrates how AI is a key tool in modern research on soil health management. The red cluster is associated with disease detection from the DL, precision, and sustainability systems perspective. The cluster highlights how AI combined with robotics reduces reliance on manual monitoring methods. The purple cluster corresponds to soil health, microbiome, and bioremediation, with academic relevance highlighting the importance of integrating AI technologies in researching these concepts that directly impact agricultural productivity. The blue cluster focuses on AI algorithms, especially on ML combined with robotics technologies. The green cluster is generated by the convergence of the advanced ML algorithms (Artificial Neural Network - ANN, SVM) with the IoT field, applied in agriculture monitoring applications. The results presented in Figure 12 highlight the trend of research in integrating AI at the level of soil health research. These results reflect the potential of AI transformations in soil health through the collaboration between engineering and ecology.

4.6. AI Branches and Applications in Soil Health

The authors analyzed the results of the WOS search for soil health related to AI and identified seven branches of AI applications. The seven branches are summarized in Table 1 and discussed in the following sections:

• The first branch focuses on monitoring and predicting soil health. In this category, AI applications use predictive models employing the biological and physical properties of the soil, as well as data from the microbiome, sensors, and satellite images. For example, Kalantzopoulos el al. [95] monitor the soil in real time, allowing for the automatic generation of recommendations regarding decisions aimed at soil actions. Andrade et al. [96] use intelligent indices for soil quality assessment from microbiological data trained with ML models. Papers by Novielli et al. [97,98] employ explainable AI (XAI) components to predict the sensitivity of microbial respiration to temperature. This is an indicator of soil quality in the context of climate change.
• The second branch focuses on analyzing the soil microbiome for fertility. Soil microorganisms are responsible for its health and fertility. The analysis of metagenomic data can be performed using AI tools, which build predictive models regarding microbial interactions. AI tools can also analyze the impact of these interactions on nutrient cycles. García el al. [99] study the use of AI to design probiotics and optimize microbial consortia. This research analyzes the impact on the fertility and remediation of degraded soils. Andrade el al. [96] argue for creating a universal soil quality index using the microbiome and ML techniques.
• Soil organic carbon is also predicted using specific AI techniques. The tools predict soil organic carbon levels with direct implications for resilience. Minasny and McBratney [100] use global-scale soil carbon dynamics forecasting. This forecast optimizes carbon sequestration strategies. The works [97,98,101] identify soils with a higher probability of CO₂ release. The analysis of this probability provides directions for sustainable carbon management.
• Soil contamination and remediation strategies are associated with the fourth branch of applications. A consistent number of papers in the specialized literature analyze the implications of AI tools in detecting and remedying soil contaminants. Algarni et al. [102] estimate the concentrations of heavy metals such as arsenic, copper, cadmium, and lead using soil reflectance spectra.
• AI techniques are also used to estimate soil quality from images. These techniques use multispectral images obtained from drones or satellites. The study by Negiş et al. [103] shows how AI tools interpret soil color, vegetation, and other visual indicators to estimate organic matter, moisture, or texture parameters. Lithuania’s Soil Data Cube system combines images with AI techniques to create thematic soil maps [104].
• Estimating nutrient quantities in the soil is also integrated through AI tools in real time. In this way, the growth level of the plants is directly controlled. Hossen el al. [105] propose a system equipped with multispectral sensors that includes an AI component trained with spectroscopy data. This system predicts the total nitrogen content in the soil to optimize the timing and amount of fertilizer application. The study by Kaur and Gupta [106] demonstrates the importance of these applications in increasing food resources.
• AI and IoT tools are used for soil moisture and water management. Most soil moisture management applications use monitoring and prediction tools. Gaitan et al. [107] integrate both types of tools in the real-time analysis of climatic parameters to generate reports and suggestions for farmers. The works [95,108,109] propose systems that correlate humidity with plant growth for irrigation management and to reduce water stress.

Table 1. Branches of AI applications related to soil health with corresponding references.

Type of AI Application	References
Soil health monitoring and prediction	[95,96,97,98,100,104,105,110,111,112]
Soil microbiome analysis for fertility and health	[96,99,110,113]
Soil organic carbon prediction and carbon management	[97,98,100,101]
Soil contamination and remediation (heavy metals, etc.)	[102,114]
Soil quality estimation from remote sensing (UAV/satellite)	[103,104,105,115]
Soil nutrient detection and management (N, P, K, etc.)	[105,106,112]
Soil moisture and water management (AI + IoT)	[95,107,108,109]

Note: UAV—unmanned aerial vehicle; N—nitrogen; P—phosphorus; K—potassium.

Table 1. Branches of AI applications related to soil health with corresponding references.

Type of AI Application	References
Soil health monitoring and prediction	[95,96,97,98,100,104,105,110,111,112]
Soil microbiome analysis for fertility and health	[96,99,110,113]
Soil organic carbon prediction and carbon management	[97,98,100,101]
Soil contamination and remediation (heavy metals, etc.)	[102,114]
Soil quality estimation from remote sensing (UAV/satellite)	[103,104,105,115]
Soil nutrient detection and management (N, P, K, etc.)	[105,106,112]
Soil moisture and water management (AI + IoT)	[95,107,108,109]

Note: UAV—unmanned aerial vehicle; N—nitrogen; P—phosphorus; K—potassium.

This analysis highlights the importance of AI tools in soil health monitoring applications. The results presented in Table 1 highlight the need for further investigations that bring together researchers from modern technology fields, such as AI and IoT, with engineers from the agricultural sector.

4.7. Interdisciplinary Innovation Potential in AI-Soil Studies

The AI technologies identified by the CAI-CIP method indicate their frequency of occurrence in research materials. The results are presented in

Table 2. AI Technologies identified in soil health research and their frequency of use.

AI Technology	Frequency of Use
ViT	79
NER	72
OWL	29
PCA	4
Selection	4
SHAP (Explainable AI)	3
ML	2
T5	2
Inception (CNN architecture)	1
NLP	1
BERT (Transformer model)	1
Transformers (general class)	1
Mutation (Genetic Algorithm)	1
RDF	1
Stemming	1
Clustering	1
GloVe (Word Embeddings)	1
GANs	1
LIME (Model interpretability)	1
SfM	1
Topic Modeling	1

Note: ViT—Vision Transformer; NER—Named Entity Recognition; OWL—Web Ontology Language; PCA—Principal Component Analysis; SHAP—SHapley Additive exPlanations; T5—Text-to-Text Transfer Transformer; BERT—Bidirectional encoder representations from transformers; RDF—Resource Description Framework; GloVe—Global Vectors; GAN—Generative Adversarial Network; LIME—Local Interpretable Model-agnostic Explanations; SfM—Structure from Motion.

.

ViT is a technique for assessing soil quality, having been identified in 79 scientific papers. This method analyzes satellite or drone images to monitor soil conditions visually.

NER and OWL appear as secondary methods, demonstrating the focus on natural language processing and structured knowledge representation. NER is identified 72 times and allows for the automatic extraction of information from textual descriptions. OWL appears 29 times and facilitates the modeling of knowledge and the relationships between concepts such as nutrients, microorganisms, and soil quality.

These results identify 20 AI technologies that are applied in soil quality research. The results demonstrate that the field of image analysis is the preferred one in soil quality analysis. The fact that AI technologies are applied in a fragmented manner, with a high degree of dispersion, but also with varied purposes of monitoring, prediction, remediation, and text analysis, suggests a lack of standardization between the two approaches, AI and soil health. This article highlights this gap through the conceptual map of AI directions in soil health from Table 2. The CAI-CIP results show that the technologies are underutilized. This underutilization does not stem from a lack of potential but from the need for future strategic research directions. The analysis integrated a predefined list of 200 AI technologies, which ensured a level of methodological originality not encountered in other works.

The detailed search for AI technologies yielded 497 papers. This result shows the scientific interest in integrating these modern tools into agriculture. These technologies are used in various applications such as monitoring, prediction, remediation, microbiological analysis, nutrient detection, etc. Although the value of 497 may seem small, the huge number of AI technologies included in the search actually demonstrates the fragmentation of research. Analyzing the value of 497 papers reveals a lack of standardization and fragmentation of approaches, which limits the replicability of the solutions. From these considerations, there is a need for research development that combines AI with holistic assessments of soil health. This result supports the need for future studies that incorporate interdisciplinary collaborations and projects demonstrating the impact of AI technologies in preventing soil contamination or remediating contaminated sites.

The results obtained from the WOS searches on components demonstrate the fragmentation of the research. Most of the research is conducted in relation to ML algorithms, yielding 378 articles. The NLP search yielded three results, while computer vision generated six articles. These modest values once again demonstrate the need to intensify research in the field of AI technologies. Additionally, the robotics sector in relation to AI and soil health or soil contaminants includes only three articles, highlighting the need for future directions in developing automated soil monitoring systems.

Regarding expert systems related to contaminants, only one article was obtained, and it demonstrates a gap that should be addressed by integrating specific knowledge into soil management decisions. The complete absence of results regarding evolutionary computation indicates that computational technologies have not been explored in the context of soil health.

5. Discussion

The results presented in this paper provide an overview of the application of AI-based technologies in determining and predicting soil health. The paper included seven analyses, each analyzing a different perspective.

• The temporal interest analysis through the distribution of articles over the years shows interest in soil health and contaminants, with the number of articles increasing from 27 in 2020 to 54 in 2024. This increase confirms researchers’ interest in studying soil health issues concerning the impact of climate change and intensive agriculture. The analysis dedicated to the AI component also shows an increase from seven articles in 2020 to ninety-two articles in 2024, with thirty articles already published in the first quarter of 2025. The evolution confirms the global trend of integrating modern technologies into agriculture. This paper identified a single article that explicitly combines the themes of soil health and contaminants with AI techniques. This gap outlines a new research direction that the scientific community should be interested in.
• The productivity of authors and collaboration analysis highlighted two main clusters in soil health research. These clusters show a growing interest, but when the AI component was integrated, the results identified a single cluster with only six primary authors. This identification suggests that the application of AI in soil health requires intense exploration in the coming years. This need should serve as a directive to stimulate interdisciplinary collaborations and attract new researchers to collaborate on soil health studies.
• The geographical distribution of research confirms that countries with intensive agriculture (China, India, the USA, Pakistan, and Australia) dominate the field. Integrating the AI component in WOS searches regarding soil health has suggested a strategic opportunity for other countries to develop technological projects concerning agricultural productivity.
• The publishers’ analysis showed differences between traditional research and AI research. Classic publishers, such as Elsevier, MDPI, and Springer Nature, have proven dominant in both cases. Regarding the less popular publishers, it was found that traditional research belongs to technical publishers, while the AI area reflected an openness to multidisciplinary approaches. These aspects are normal, considering the interdisciplinary nature of AI components, and they outline a future development direction, encouraging traditional publications to disseminate works that also contain AI components.
• The frequency of keywords and thematic directions were used to analyze the co-occurrence of keywords. This analysis identified seven clusters in general research and five clusters in AI research. In traditional research, the central element is represented by classic pollutants (heavy metals and pesticides), sustainable agriculture, water recycling, biosolids, and pharmaceutical pollutants. These themes show the researchers’ growing interest in ecological issues. In contrast, AI research has clearly focused on advanced technologies concerning bioremediation. This direction indicates that articles addressing the issue of soil health concerning AI simultaneously aim at the necessity of modernizing agriculture through the integration of modern technologies.
• The analysis of AI branches and their applications in soil health has led to seven main branches. Thus, branches targeting monitoring, the prediction of microbiological analysis, remediation, and water resource management, as well as other elements reflecting fragmented implementation, were identified. This finding suggests a need for methodological standardization to create specific tools for replicating systems in the agricultural field, such as soil health.
• The CAI-CIP analysis demonstrated the potential for interdisciplinary innovation. It highlighted the use of ViT as a technology frequently employed in soil health. This stems from the ability of these algorithms to analyze images and process subtle elements identified at the image level. The fact that many of the 200 technologies identified by the authors did not yield results highlights an unexplored area in the specialized literature. This lack of homogenization highlights the increased need for interdisciplinary collaborations in agricultural engineering concerning AI methods.

In this study, 217 original articles published between 1 January 2020 and 30 April 2025 were analyzed. The search conducted in WOS used specific terms related to soil health and contaminants. The second search involved integrating AI into the field of soil health. The result yielded 222 papers. From this collection of results, one article explicitly addresses the relationship between soil contamination, agriculture, and the use of AI technologies.

The results of detailed searches regarding the subfields of AI applied to soil health and contaminants show that ML stands out as the most active branch, with 378 studies published between 2020 and 2025. Although this result suggests that researchers are interested in using ML for prediction, monitoring, remediation, and intervention regarding soil contamination, the number is small compared to the number of existing ML algorithms. Subfields such as NLP, computer vision, robotics, and expert systems have generated lower results, and the interpretation of this low interest is correlated with the need to explore the potential of these technologies in the field. Regarding the evolutionary computation component, it provided zero articles, indicating a gap that could become a research opportunity, considering the potential of genetic algorithms or evolutionary strategies in optimizing soil remediation strategies and reducing associated costs.

This study’s main limitation is its exclusive focus on WOS data; however, integrating publications from other alternative databases could have created confusion in identifying future development directives. Additionally, the CAI-CIP analysis may not fully reflect the complexity of the analyzed articles. Another limitation is that AI research is continuously evolving, which means that the analysis window does not reflect the field’s maturity for the coming years.

As future development directions, the necessity of expanding international collaborations, diversifying the AI technologies used in soil health studies, standardizing AI analysis procedures in agriculture, and investigating data infrastructure to facilitate the integration of data from satellite sensors or other local sources into a single system, as well as large-scale pilot projects and case studies to demonstrate the impact of AI technologies in soil contamination remediation and prevention, are emerging. This discussion section confirms the progress in soil health and underscores the necessity of further integrating AI technologies to harness this field’s potential. The results of this research represent a starting point for a new generation of interdisciplinary studies that will incorporate specific elements of human health protection through soil health into a unique agricultural ecosystem.

6. Conclusions

This work investigated the current state of research dedicated to soil health. This paper emphasizes the integration of AI technologies in the context of soil health. Seven analyses were implemented to measure global scientific interest and identify the academic communities involved in this collaboration to identify the degree of cooperation between the two fields. Additionally, the paper mapped the geographical distribution and analyzed the applied thematic and technological directions.

This study provides an overview of the progress made and identifies the gaps that need to be addressed for a future where soil health is managed using AI techniques. The first conclusion from this research is that interest in soil health has steadily increased over the past five years. This increase reflects the recognition of the importance of protecting soil as a resource for the ecosystem and global food security. The increase from 27 articles in 2020 to 54 in 2024 confirms this trend. Analyses have also shown an increase in the use of AI in the agricultural field. In 2020, there were seven articles, which increased to ninety-two in 2024. This evolution underscores that the academic community perceives modern technologies as viable solutions for contemporary agricultural problems. However, the authors identified only one article that explicitly combines soil health themes with contaminants and AI technologies during the analyzed period. This finding highlights the acute lack of truly interdisciplinary studies.

The analysis of author communities highlighted two main clusters: the first was associated with traditional soil health research, and the second included a smaller circle of six authors who approached the issue of soil health through AI techniques. This finding reveals interest in the field but also highlights the need to strengthen this academic community by attracting a larger number of researchers from complementary domains.

From a geographical perspective, the research has shown that countries with intensive agriculture dominate the research area. This aspect raises a red flag for other countries that have opportunities to improve soil quality and thus intensify agriculture, including the European Union, which has not yet been examined in this field.

The analysis of publishers that have disseminated research in soil health showed that Elsevier, MDPI, and Springer Nature are at the top of researchers’ preferences. This research shows a predilection for disseminating works that include AI techniques in technical and multidisciplinary areas, which should encourage other publications to do the same.

The study on keyword frequency showed a thematic diversity regarding traditional research concerning AI techniques. This aspect reflects an awareness of the need for sustainable solutions, and in the area of AI, the confirmation of the potential of these technologies to revolutionize the way soil health can be managed and monitored has been noted.

Another strong point of this research was the analysis of AI branches applied to soil health. Here, seven branches were identified that provide a map of AI applications: monitoring and predicting soil health, microbiome analysis, organic carbon prediction, contaminant detection, soil image analysis, nutrient estimation, and water management. The fact that the applications are fragmented suggests the need to capitalize on these applications. The results of the CAI-CIP analysis demonstrated researchers’ interest in technologies such as Vision Transformers; however, this interest highlighted the underutilization of other technologies, indicating a significant opportunity for future research to expand the repertoire of AI technologies used.

Analyzing these results, the future research directions are as follows:

• Another directive for interdisciplinary collaboration should be to stimulate interdisciplinary collaborations to create partnerships between AI experts and soil science researchers and include agricultural practitioners in the development of integrated solutions applicable on a large scale.
• The AI technologies used should be diversified so that the entirety of them are inspected in relation to soil health. The authors recommend inspecting ML, neural networks, evolutionary algorithms, generative networks, or fuzzy systems in this category.
• Methodologies should be standardized regarding the integration of AI in soil monitoring and remediation so that results can be replicated and compared globally. This standardization would allow for a meaningful comparison between research conducted with the same purpose but in different environments.
• Investments in infrastructure for the development of data platforms allow researchers to study as many AI models as possible using information provided by sensors, satellite images, and local reports so that advancements in the academic environment support farmers.
• Education should be another directive in which new generations of specialists possess agricultural knowledge corroborated with digital skills, such as those specific to integrating AI tools in agriculture.

In conclusion, this research confirms an increased interest in soil health and highlights the necessity of integrating modern technologies through AI to study soil health. This paper aims to serve as a methodological and thematic benchmark for future research. Also, it seeks to strengthen the relationship between technology and environmental sciences, with the common goal of the benefits generated by advancements in agricultural ecosystems.