人工智能在土壤表征与农业中的应用:技术、模型与应用系统综述(Agronomy 16(13) 1241)

Logical Operator Operator

Search Text

Search Type

add_circle_outline

remove_circle_outline

Article Menu

Font Type:

Arial Georgia Verdana

Font Size:

Aa Aa Aa

Line Spacing:

  

Column Width:

  

Background:

Open Access Systematic Review

by
Cesar Augusto Navarro Rubio

,
Hugo Martínez Ángeles

*,
Mario Trejo Perea

*,
José Luis Reyes Araiza

,
Guillermo Ronquillo-Lomeli

,
Ivan Gonzalez-Garcia

,
Eusebio Ventura Ramos

and
José Gabriel Ríos Moreno

Facultad de Ingeniería, Universidad Autónoma de Querétaro, Santiago de Querétaro 76010, Mexico

Authors to whom correspondence should be addressed.

Submission received: 22 May 2026 / Revised: 19 June 2026 / Accepted: 24 June 2026 / Published: 26 June 2026

Abstract

Artificial Intelligence (AI) has become a key enabler in soil science and agriculture, supporting advanced modeling, monitoring, and decision-making processes. This systematic review synthesizes recent developments in AI-based soil characterization and agricultural applications, with emphasis on soil physicochemical properties, digital soil mapping, irrigation management, and crop yield prediction. Following the PRISMA 2020 framework, a structured search of the Scopus database identified 196 eligible studies published between 2018 and 2026. The reviewed literature reveals a clear transition toward data-driven approaches, with machine learning and deep learning models dominating recent research. Random Forest, Support Vector Machines, gradient boosting methods, artificial neural networks, Convolutional Neural Networks, and Long Short-Term Memory architectures were the most frequently reported techniques. The primary data sources included in situ sensors, laboratory measurements, remote sensing imagery, and environmental covariates, often integrated through multi-source data fusion frameworks. The results indicate that tree-based ensemble models provide robust performance across diverse soil properties, whereas deep learning models are particularly effective for spatiotemporal prediction and remote sensing applications. AI-driven systems are increasingly used to support precision agriculture through irrigation optimization, crop yield forecasting, digital soil mapping, and soil health monitoring. However, challenges remain regarding data quality and availability, model transferability across regions, and the limited interpretability of complex models. The findings highlight current research trends, methodological challenges, and future opportunities for the development of reliable and scalable AI-driven soil and agricultural systems.

1. Introduction

Soil is a fundamental component of terrestrial ecosystems and plays a critical role in agricultural productivity, environmental sustainability, and land management. The accurate characterization of soil properties such as moisture content, texture, salinity, temperature, and nutrient availability is essential for optimizing agricultural practices and ensuring efficient resource utilization. These properties are inherently variable in both space and time, making soil systems highly complex and challenging to model using traditional analytical approaches.

Among the various soil parameters investigated in the literature, soil electrical resistivity (SER) has gained increasing attention due to its sensitivity to multiple physicochemical properties and its suitability for data-driven analysis. Electrical conduction in soils occurs primarily through the pore fluid and dissolved ions, making resistivity strongly dependent on moisture content, salinity, temperature, and soil structure. As a result, SER is widely used as an indirect proxy for assessing soil conditions in applications ranging from geophysical exploration to precision agriculture and environmental monitoring [1,2,3].

The relationship between soil electrical resistivity and environmental variables is highly nonlinear and influenced by complex interactions among multiple parameters. Moisture content has been consistently identified as the dominant factor controlling resistivity, exhibiting a strong inverse correlation (r ≈

$- 0.88$
), while temperature and soil suction generally show weaker positive relationships [1,2,3]. In addition, soil properties such as clay content, plasticity, and dry unit weight significantly affect electrical conduction mechanisms, highlighting the need for multivariable modeling approaches capable of capturing these interactions [2,4]. Furthermore, electrical conductivity, which represents the inverse of resistivity, is closely linked to soil salinity and chemical properties such as pH and nutrient concentration, enabling its use in irrigation management and soil fertility assessment [3,4,5].

Traditional methods for soil characterization and resistivity measurement, including laboratory testing and field-based techniques such as the Wenner four-electrode method, provide reliable results but are often labor-intensive, time-consuming, and limited in spatial resolution. These limitations become particularly significant in agricultural contexts, where large-scale monitoring and real-time decision-making are required. Consequently, there has been a growing interest in developing data-driven approaches capable of estimating soil properties based on easily measurable variables.

In this context, Artificial Intelligence (AI) has emerged as a powerful tool for modeling complex soil systems. Machine learning (ML) and deep learning (DL) algorithms have demonstrated strong capabilities in capturing nonlinear relationships between soil variables and predicting key properties with high accuracy. Models such as Artificial Neural Networks (ANNs), Support Vector Machines (SVMs), Random Forests (RF), and Extreme Gradient Boosting (XGBoost) have been widely applied for soil resistivity and property prediction. Some studies have reported coefficients of determination (R 2) values approaching 0.99 under specific experimental settings, controlled conditions, or dataset characteristics; however, model performance varies substantially depending on soil type, environmental conditions, data quality, and validation procedures [1,2,6]. In addition, advanced deep learning architectures, including Deep Neural Networks (DNNs), Long Short-Term Memory (LSTM) networks, and Gated Recurrent Units (GRU), have shown strong performance in modeling temporal and spatial variability in soil systems, particularly for moisture prediction and hydraulic property estimation [7,8,9].

The integration of AI with sensing technologies has further enhanced soil characterization capabilities. Real-time monitoring systems based on electrical resistivity measurements, such as robotic Wenner arrays, enable rapid spatial mapping of soil properties across agricultural fields [3]. Similarly, Electrical Resistivity Tomography (ERT) combined with soil moisture measurements has been successfully applied to monitor irrigation processes and analyze wetting–drying cycles in crops such as cotton and tomatoes [10,11]. Moreover, multi-sensor platforms incorporating electrical, dielectric, optical (Vis–NIR), and acoustic measurements, coupled with AI models, have enabled the development of high-resolution digital representations of soil systems, often referred to as “digital soil twins” [12].

Beyond resistivity-focused applications, AI techniques have been widely applied in broader agricultural contexts, including soil fertility mapping, crop yield prediction, crop recommendation systems, and smart irrigation management. These applications leverage large-scale datasets combining soil, climate, and remote sensing information to support data-driven decision-making in precision agriculture [4,13,14,15,16,17].

Despite these advances, several challenges remain in the application of AI to soil characterization and agricultural systems. Soil heterogeneity, environmental variability, and the dependence of models on site-specific conditions limit the generalizability of AI approaches [1,2]. Additionally, the availability of high-quality labeled data remains a major constraint, particularly when integrating field measurements, laboratory analyses, and remote sensing data at different spatial and temporal scales [12,14]. To address these issues, recent studies have explored synthetic data generation techniques, such as variational autoencoders, and multi-sensor data fusion strategies to improve model robustness and predictive performance [1,12,16]. Furthermore, there is a growing emphasis on the development of hybrid models that combine physical principles with AI techniques, as well as interpretable machine learning approaches, to enhance both predictive accuracy and model transparency [2,4,5].

Despite the rapidly expanding body of research in this field, the existing literature remains fragmented due to variations in datasets, modeling approaches, soil types, environmental conditions, and application domains. Many studies focus on specific regions or individual soil properties, while others emphasize particular AI techniques, making it difficult to establish a unified understanding of current advancements and limitations.

This fragmentation highlights the need for a systematic and comprehensive synthesis of the existing literature on AI applications in soil characterization and agriculture. A structured review is essential to identify dominant research trends, evaluate the performance of different AI models, and determine the most relevant soil parameters for predictive modeling.

Although several review studies have examined specific aspects of AI in agriculture, such as crop yield prediction, remote sensing applications, digital soil mapping, or precision agriculture technologies, a comprehensive synthesis integrating AI methodologies, soil properties, sensing technologies, and agricultural applications within a unified framework remains limited. Addressing these interconnected dimensions is essential to better understand how AI supports soil characterization and agricultural decision-making across multiple scales and application domains.

Accordingly, this study conducts a systematic review of the applications of Artificial Intelligence in soil characterization and agricultural systems, with particular emphasis on soil electrical resistivity, multivariable modeling approaches, and precision agriculture applications.

To structure this systematic analysis, the following research questions are defined. Addressing these questions provides insights into the current state of knowledge on AI-driven soil analysis and agricultural applications, while acknowledging that the findings are conditioned by the scope and characteristics of the selected studies.

• What types of AI models are most commonly used for soil characterization and agricultural applications?

• Which soil properties are most frequently modeled using AI techniques?

• What sensing technologies and data sources are employed in AI-driven soil studies?

• What are the main agricultural applications of AI in soil analysis, including precision agriculture and irrigation management?

• What limitations, challenges, and research gaps exist in current AI-based soil modeling approaches?

This systematic review is not hypothesis-driven; rather, it is structured around predefined research questions that provide a framework for synthesizing and analyzing the existing literature. The objective is to analyze and compare AI-based methodologies, identify key soil variables influencing predictive performance, and evaluate their applicability in agricultural systems.

This paper is structured as follows. Section 1 presents the Introduction, describing the importance of soil characterization, the role of electrical resistivity, and the emergence of AI techniques in agricultural applications. Section 2 describes the Materials and Methods, including the systematic review protocol, search strategy, and selection criteria. Section 3 presents the Results, summarizing the main AI techniques, soil properties, and application domains identified in the literature. Section 4 provides the Discussion, analyzing methodological trends, limitations, and future research directions. Finally, Section 5 presents the Conclusions, highlighting the main findings and contributions of this study.

2. Materials and Methods

2.1. Systematic Review Protocol

This systematic review was conducted following the principles established in the PRISMA 2020 framework [18], which provides a standardized methodology for identifying, screening, and selecting scientific literature. The application of this protocol ensures a transparent, structured, and reproducible workflow throughout the review process.

The methodological procedure was organized into four main phases: (i) identification of relevant studies through database searching, (ii) screening of records based on titles and abstracts, (iii) eligibility assessment through full-text evaluation, and (iv) final inclusion of studies that met the predefined criteria. This sequential approach allows for a clear traceability of the selection process while minimizing potential biases associated with subjective study selection.

The adoption of the PRISMA methodology enables a systematic synthesis of the existing literature on Artificial Intelligence applications in soil characterization and agricultural systems, with particular attention to soil electrical resistivity, multivariable soil analysis, and precision agriculture technologies. The PRISMA 2020 checklist is provided as Supplementary Material to ensure compliance with reporting standards.

To further support transparency and reproducibility, the review protocol was registered in the Open Science Framework (OSF) platform (registration DOI: https://doi.org/10.17605/OSF.IO/6ACSK, accessed on 30 March 2026). The literature search, screening, and study selection processes were completed prior to the final preparation of the manuscript, and the year 2026 represents the final time boundary defined in the search strategy. The OSF access date corresponds solely to the consultation and verification of the registered review protocol.

2.2. Search Strategy and Database Selection

The literature search was performed using the Scopus database, which was selected due to its extensive coverage of peer-reviewed journals, robust indexing system, and advanced search capabilities. Scopus offers a comprehensive representation of research across multiple disciplines, including soil science, agriculture, environmental engineering, and artificial intelligence, making it particularly suitable for the multidisciplinary scope of this study.

Although other databases such as Web of Science and Google Scholar also provide valuable sources of scientific information, Scopus was chosen to ensure consistency in metadata, reproducibility of results, and reliable citation tracking. Nevertheless, it is acknowledged that relying on a single database may lead to the omission of some relevant studies, particularly those published in regional or non-indexed journals.

To mitigate this limitation, the search strategy was carefully designed to capture a broad range of relevant publications by combining keywords related to artificial intelligence, soil characterization, agricultural applications, and soil electrical resistivity.

The final search query used was as follows:

TITLE-ABS-KEY ((“machine learning” OR “deep learning” OR “artificial intelligence”) AND (“soil characte