Soil Nutrient Monitoring Technologies for Sustainable Agriculture: A Systematic Review

Abstract

Soil nutrient monitoring plays a vital role in advancing sustainable agriculture by maintaining soil health, optimizing crop productivity, and minimizing environmental impacts. This study addresses gaps in unified definitions and standard methodologies by systematically analyzing 93 articles using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) framework. The results highlight five major monitoring approaches: traditional methods, Remote Sensing (RS), Internet of Things (IoT) and smart systems, in situ sensors, and Artificial Intelligence (AI)-based models, each contributing uniquely to nutrient assessment. A noticeable trend toward integrating machine learning and deep learning with sensor technologies underscores the advancement toward real-time, data-driven precision agriculture. The study also explores spatial and temporal publication trends, criteria for site selection, and the validation techniques used to assess monitoring accuracy. A synthesized definition of soil nutrient monitoring is proposed to support future research and standardization. This review highlights the crucial role of soil nutrient monitoring technologies in sustainable agriculture, crop optimization, and environmental management. It provides a comprehensive overview of the techniques employed in monitoring soil nutrients for precision soil management.

1. Introduction

Soil nutrient monitoring is a cornerstone of sustainable agriculture, as it enables the assessment of soil health for long-term productivity [1], and to environmental protection through minimizing agricultural non-point source pollution [2]. Soil nutrients are fundamental in regulating ecosystem structure and function [3], as they influence critical processes such as water retention, carbon storage, soil stability, crop health, and the maintenance of biodiversity and ecological resilience [4,5]. However, increasing pressure on soil poses significant challenges as the global population is expected to approach nine billion by 2050, according to the United Nations estimates [6,7]. This rapid growth intensifies concerns over food security and underscores the urgent need to preserve soil, maintain nutrient balance, and prevent further degradation [8].

Soil health and fertility are determined by the balanced availability of macronutrients, such as nitrogen (N), phosphorus (P), and potassium (K), alongside micronutrients, including iron, zinc, and copper, which play crucial roles in plant metabolism and crop productivity [9]. For example, nitrogen contributes to protein synthesis and carbon sequestration, phosphorus is essential for plants’ root development and energy transfer, and potassium regulates key physiological processes and overall plant health, significantly contributing to drought tolerance and disease resistance [10,11,12,13]. The natural process of soil nutrient cycling was altered due to the production and widespread application of agrochemicals, including fertilizers, pesticides, and herbicides [7].

Imbalances of soil nutrients present serious challenges to sustainable agriculture. An insufficient nutrient application weakens plant development, reducing agricultural land coverage and accelerating soil erosion over time [14,15]. An excessive application of nutrients leads to nutrient leaching, groundwater contamination, and the eutrophication of freshwater and coastal systems, which not only impair water quality but also disrupt aquatic biodiversity and ecosystem stability [16]. This carries significant economic consequences, as a 30% reduction in nitrate loading and 36% reduction in phosphorus levels is estimated to cost US$1.4 billion annually [17]. Moreover, the overuse of nitrogen fertilizers is closely linked to increased emissions of nitrous oxide, a major greenhouse gas that contributes significantly to climate change [5,10,18].

Soil nutrient monitoring directly addresses these challenges, helps evaluate management practices, optimizes fertilizer utilization, and guides agricultural management [1]. Thus, it functions not only to optimize crop production but also as a vital measure for preserving environmental quality and sustaining soil resilience.

Nevertheless, only a limited number of review studies have addressed soil nutrient monitoring technologies, as summarized in Supplementary Table S1 and Figure 1. These reviews have a different formatting style and are a comprehensive, systematic, and critical review. Additionally, they vary in scope and monitoring technologies from sensor-based technology, which has a more focused approach among the reviewed articles due to the variety of sensor types. For example, Mohapatra et al. [19] present a review of sensor technology for measuring soil ground parameters, including various nutrients, while Burton et al. [20] discuss optical and electrochemical sensor technologies for real-time soil information. In Nadporozhskaya [21], recent advances in chemical sensors for soil pH, moisture, and nutrients were discussed. Yadav et al. [22] highlight nanotechnology’s role in agriculture, noting that nanomaterials like graphene help monitor and bind soil nutrients for better sensors, while Fan et al. [23] critically review the sensor technology for real-time continuous soil monitoring (RTCSM) to highlight the effect of the time gap and its role in data uncertainty. Abdulraheem et al. [24] reviewed Remote Sensing (RS) for soil, highlighting ground sensors and aerial hyperspectral imaging for detecting soil nitrogen. Chen et al. [25] reviewed the role of printed electrochemical sensors for real-time measurements of nutrients such as nitrogen and phosphorus. Also, the sensors’ technologies for nitrogen, phosphorus, and potassium (NPK) were broadly discussed by Dattatreya et al. [26]. While Ameer et al. [27] comprehensively reviewed sensor technologies for measuring soil NPK, highlighting the potential of optical and electrochemical methods for real-time analysis.

Artificial Intelligence (AI) technology for soil nutrients was represented by Avhad et al. [28]. This review discusses the uses of machine learning (ML) in fertilizer prediction, while Jain et al. [9] provide a systematic review of ML and deep learning (DL) applications for soil nutrient prediction. Mhoro et al. [29] further explored the NUTOM model, emphasizing combined biophysical and socio-economic factors for sustainable soil fertility. Internet of Things (IoT) technology and wireless sensors were reviewed by Kumar et al. [30] to enhance agricultural practices, while Ndjuluwa et al. [31] and Musa et al. [32] address the uses of the IoT and NPK wireless sensor to optimize fertilizer application.

Despite the valuable research offered by existing reviews, there remains a lack of a common definition, and few studies iteratively focus on nutrient monitoring across all sensing and data-driven techniques. This review seeks to address this gap by carrying out the following:

• Analyze the spatial and temporal patterns of relevant scientific publications, providing insights into geographic focus and the evolving research trends.
• Synthesize definitions from diverse studies to establish a clear and unified understanding of soil nutrient monitoring.
• Assess the available monitoring technologies, evaluating their strengths and limitations.
• Identify and classify different sensor technologies utilized in soil nutrient monitoring, detailing the specific types of nutrients each technology detects.
• Analyze the input and output data utilized by AI technology in soil nutrient analysis.
• Investigate validation and accuracy approaches of monitoring technologies.
• Describe soil nutrient sampling protocols, including sampling frequency and locations.
• Address the criteria used in selecting suitable monitoring sites.

By addressing these objectives, this study aims to provide a thorough, multi-dimensional understanding of soil nutrient monitoring practices, highlighting current challenges, knowledge gaps, and opportunities for future research and practical implementation.

2. Materials and Methods

2.1. Framework Adoption and Article Selection Process

This study employed a rigorous and structured methodology, guided by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA 2020) standards (see Supplementary Materials). This review has been conducted following the PRISMA 2020 guidelines and has been registered with the Open Science Framework (OSF). The registration details are publicly available at https://doi.org/10.17605/OSF.IO/Q7XAY (accessed on 10 September 2025). The adoption of the PRISMA framework enabled the systematic identification and extraction of key findings, thereby advancing understanding of soil nutrient monitoring. The PRISMA approach was selected for its transparent and comprehensive procedures, which ensure that each step of the review process is clearly documented. There are essential elements of systematic reviews that are outlined in the PRISMA checklist, which includes details about the databases and sources used, the search tactics employed, specified eligibility criteria for study inclusion or exclusion, and the assessment of study quality in terms of internal and external validity.

A central feature of the PRISMA framework, illustrated in Figure 2, is a flow diagram that visually represents the major phases of the review process: identification, screening, eligibility assessment, and final inclusion of studies. This diagram assists in selecting a set of high-quality, pertinent articles that accurately reflect the current landscape of research in soil nutrient monitoring. In this study, the Google Scholar database [33] was utilized for a systematic search, employing the precise keyword phrase “soil nutrient monitoring” with quotation marks to ensure accurate retrieval of relevant studies. An initial comprehensive search yielded 255 records on 11 August 2024 for these keywords. From these records, 191 records are available to be downloaded and evaluated, while 64 are not available to be downloaded.

2.2. Quality Assessment and Study Selection Process

Following PRISMA 2020 guidelines, the downloaded articles were evaluated for the systematic review. After applying the inclusion and exclusion criteria, conference abstracts, reports, book chapters, and full books were systematically eliminated. The review process incorporated multiple staged screenings and analyses. Initial screening was conducted through title and abstract examination, followed by full-text review. Multiple review cycles ensured strict alignment with study objectives and keyword relevance.

After applying inclusion and exclusion criteria (assessing, judging, and identifying potential bias risks, and appraising internal or external validity, 98 papers were excluded from the final analysis, and 93 studies were included in this study.

2.3. Data Extraction and Analysis

The Data from the eligible articles is extracted according to each objective of this paper and organized in an Excel sheet or a Word table to be able to gather and analyze the information. To understand the spatial and temporal distribution of the articles, the publication year and the place of the study are extracted in an Excel sheet, and the results are represented with different visualization methods, such as line and bar graphs, to understand the distribution of publications. Also, geographical maps were created using Map Chart [34] to represent the global distribution of the studies related to soil nutrient monitoring. The definition figure was created with the Icograms tool for clearer representation.

To come up with a common and general definition of soil nutrient monitoring, the descriptive and definitional data are extracted from the eligible articles. Key elements: soil nutrient types and monitoring practices were collected into a comparative table using the key words “soil nutrients” and “monitoring” separately, to identify recurring aspects and conceptual overlaps of existing descriptions. Additionally, the keywords “monitoring”, “measurement”, and “test” were used to extract various soil nutrient monitoring techniques and technologies. For validation, the terms “validation” and “accuracy” were used to extract relevant information, which was summarized in a table. Data on sampling techniques were also collected using the keywords “ sampled” and “sampling” to extract the different monitoring techniques used for soil nutrients. Selection criteria for nutrient monitoring sites were systematically gathered from the reviewed articles, together with GPS records and site-specific characteristics, to identify common factors influencing site selection.

3. Results

3.1. Temporal and Spatial Distribution of Studies

The temporal distribution of publications on soil nutrient monitoring between 2005 and 2024 shows a clear upward trajectory. Research activity remained relatively low and stable from 2005 to 2019, but a gradual increase began in 2020, leading to a sharp surge between 2022 and 2024. The number of publications reached its highest point in 2024 with 28 studies, marking the peak of the period analyzed, as shown in Figure 3. This rapid growth highlights the rising global interest in soil nutrient monitoring, driven in part by the growing emphasis on sustainable agriculture and the advancement of precision farming technologies.

The global distribution of soil nutrient monitoring research reveals significant international engagement with the topic. The bar chart in Figure 4 shows that India leads in publication output with 22 studies, followed by China (17) and the United States (14). Figure 5 shows a complementary world map that highlights the geographic spread of research activity, emphasizing the broad relevance of soil nutrient monitoring across diverse agricultural regions.

This widespread participation reflects the increasing global recognition of soil nutrient monitoring as a critical tool for promoting sustainable agriculture and informed environmental management. The growing interest from various countries signals a collective response to pressing issues such as food security and soil degradation.

3.2. Soil Nutrient Monitoring Definition

Among the 93 reviewed articles, no common definition of soil nutrient monitoring was recognized. Instead, 18 articles provided varying descriptions or interpretations of the term, as summarized in

Table 1. Variation in the definitions and descriptions of soil nutrient monitoring across studies.

S.N.	Definition or Description	Citation
1	The monitoring system provides farmers with valuable data about the soil’s condition, allowing for timely decisions regarding nutrient management and fertilization. Soil has three major nutrients: nitrogen (N), phosphorus (P), and potassium (K), which are represented as NPK.	[30]
2	Soil nutrient monitoring can provide valuable indicators for sustainable soil fertility management by linking nutrient balances and soil nutrient stocks.	[39]
3	Monitoring soil nutrient levels is essential for the effective utilization of fertilizers and the mitigation of the ecological footprint resulting from fertilization techniques.	[32]
4	The soil nutrient monitoring system is to master the nutrient status of the bare ground and quickly extract information on farmland nutrients which are categorized into macronutrients required for sustained plant health such as Nitrogen (N), Potassium (K), Phosphorus (P), Carbon (C), Hydrogen (H), Oxygen (O), Calcium (Ca), Magnesium (mg), and Sulfur (S), and micronutrients also essential to plant development and growth such as Chlorine (Cl), Iron (Fe), Boron (B), Manganese (Mn), Zinc (Zn), Copper (Cu), Molybdenum (Mo), Sodium (S), Silicon (Si) and Nickel (Ni).	[35]
5	The monitoring of levels of soil nutrients can be utilized by farmers, agriculturists, and soil enthusiasts, and this information is used in terms of future trends and application of the appropriate amount and type of fertilizers needed to ensure optimal plant growth and increased crop yield. Monitoring and detecting levels of soil macronutrients (i.e., nitrogen, phosphorus, and potassium) are vital in the practices and guidelines on sustainable farming and implementation.	[40]
6	Soil nutrient monitoring is explained as the Nutrient Expert tool (IoT) and is highlighted for its role in efficient crop nutrient management, leading to increased yields, farmer income, and reduced greenhouse gas emissions, thereby addressing the impacts of climate change.	[41]
7	Soil nutrient concentration can be monitored by IoT, as nutrients play a vital role in the growth and nourishment of the plant. Measurements of nutrients will allow us to know about the constituents of the nutrients present in the soil and the nutrients lacking in the soil.	[15]
8	Soil nutrient status can have a direct impact on the success and speed of rehabilitation of cut slopes. As one of the important soil nutrients, soil phosphorus (P) can potentially limit successful rehabilitation of cut slopes, play an important role in soil nutrient cycling, and is a potentially significant determinant of soil quality.	[37]
9	The soil nutrient monitoring system is used to master the nutrient status of the bare ground and quickly extract information on farmland nutrients. Because it has a significant impact on crops, soil nutrient monitoring is important.	[36]
10	Nutrient management, based on the best available information for soil test targets, a greater understanding of fluxes of nutrients on farms, and potential nutrient loss processes and pathways, will lead to improved nutrient efficiency on farms and hence the best return on fertilizer investment, as well as reduced risk of losses of nutrients to the environment. Assessment of N, P, K, and S fertility status by soil testing is now widely accepted and is a major tool in providing fertilizer advice for crops and pasture.	[38]
11	Soil nutrient monitoring is a critical aspect of modern agriculture, and biosensors offer a promising solution to this challenge, which have the capability to detect and quantify essential nutrients such as nitrogen, phosphorus, and potassium in the soil. By providing real-time data on nutrient levels, biosensors enable farmers to implement site-specific fertilizer application strategies.	[42]
12	Sensing the changes in the nutrient ion concentrations is vital for providing the nutrient-sufficient conditions for maximal plant growth and yield. Therefore, a soil nutrient sensor is important for optimizing nutrient management. The detected ions contain the most important elements for plant growth, such as Nitrogen (N), Phosphorus (P), and Sulfur (S).	[43]
13	This real-time monitoring by Hyperspectral remote sensing can quickly assess and monitor crop nutrient levels and soil nutrient content. Technology can provide an important basis for the rational application of nitrogen during fertilization.	[44]
14	Monitoring soil conditions (e.g., moisture, nutrients, and pollutants) over growing seasons enhances resource efficiency, ultimately leading to maximized agricultural yields while simultaneously minimizing environmental impacts.	[25]
15	Soil nutrients are an important factor in measuring soil fertility, and traditional farm management and agricultural systems have led to polarization of soil nutrients in farmlands. (Soil nutrients’ importance.)	[45]
16	Soil nutrient monitoring, especially for nitrogen, is essential for understanding nutrient dynamics and enabling timely management.	[46]
17	Soil nutrient monitoring involves resin-based measurements using Plant Root Simulator (PRS) probes to track nutrient availability in soil, particularly for nitrogen (N), calcium (Ca), and sulfate (SO4) levels. This approach allows for analyzing changes in nutrient levels resulting from different forest management treatments and understanding the nutrient dynamics.	[47]
18	Soil nutrient monitoring is essential for precision agriculture, aiming to optimize fertilization and crop yield.	[48]

. Most of these contributions were descriptive rather than definitional, with only two studies, Mamman &Lubega [35] and Zu et al. [36] providing definitions that shared similar aspects as “The soil nutrient monitoring system is to master the nutrient status of the bare ground and quickly extract information on farmland nutrients”. Ai et al. [37], and Gourley et al. [38], on the other hand, employed ‘nutrient management’ to describe approaches for determining nutrient availability in soil as alternatives to the term ‘soil nutrient monitoring’.

Table 1 shows that the interpretation of soil nutrients monitoring differs across studies, and the comparative table enables us to extract key definition components as illustrated in Figure 6 by highlighting recurring elements and conceptual overlaps found in existing descriptions. These components can be characterized across several dimensions: (A) Nutrient type; as monitoring may target either only an individual nutrient, such as nitrogen, as described in Fan et al. [46] or phosphorus as reported by Ai et al. [37] or focus on a group of nutrients, especially the macronutrients, Nitrogen (N), Phosphorus (P), and Potassium (K), which are required in large quantities, as described by Kumar et al. [30] and Montañez [40]. A smaller subset of studies addresses the measurement of secondary macronutrients, including sulfur (S), calcium (Ca), as reported by Xu et al. [43], and Johnson et al. [47]. Despite the recognized significance of micronutrients in soil fertility and plant growth, these elements are primarily mentioned in nutrient definitions but are rarely included in measurement or monitoring methodologies. The non-mineral elements, C, H, and O, are available to the plant either from the atmosphere or the soil [35]. These elements are involved in their definition as macronutrients because of their crucial role in the agricultural process.

(B) Monitoring methods; in some studies, the description of soil nutrient monitoring is centered primarily on the monitoring method or technology employed, as shown by Malde et al. [15] and Chojnacka et al. [41], which uses the IoT monitoring technology as the main component of the description, while sensor technology was represented by the description provided by Pratama et al. [42] and Xu et al. [43] and the remote sensing technology applied in the description by Zhang et al. [44].

(C) Participants; the descriptions by Kumar et al. [30] and Montañez [40] specify the actors involved in nutrient monitoring and indicate that nutrient monitoring is carried out by both practicing farmers and agricultural professionals who apply the information to improve management practices. (D) The Scale; as a process, this is carried out at different scales, both spatially, e.g., farmland and forest ecosystems, as shown by Zu et al. [36] and Johnson et al. [47], and temporally, to capture changes over time [25,30]. (E) Monitoring purpose; as the aim of soil nutrient monitoring varies from increasing the crop yield [40,41,48,49], supporting precision agriculture [48], guiding fertilizer recommendations [32,38,42,50], an indicator for sustainable use management [39], and reducing nutrient losses to the environment [38,49].

Nonetheless, according to Johnson et al. [51], the perfect definition should be logical, nontheoretically, conceptually straightforward, inclusive, and thorough. But it’s often challenging to come up with the perfect term. From the different definition components in Figure 6 and the comparative Table 1 of the various descriptions, the general definition for the soil nutrient monitoring was synthesized as follows: “It is the process of collecting data and information on the condition and status of soil nutrient stocks over time within the upper soil layer of a specific area using various techniques. This practice supports informed decision-making on soil fertility management, optimizes crop growth, and reduces the environmental impacts of fertilizers, thereby enhancing nutrient balance and promoting sustainable practices in precision agriculture.” (Figure 7).

3.3. Soil Nutrient Monitoring Technologies

Soil nutrient monitoring employs a wide range of technologies. As illustrated in Figure 8 and Supplementary Table S2, different approaches were identified in the 93 reviewed studies, which were categorized into five main groups: Traditional laboratory methods, remote sensing, Internet of Things (IoT) applications, smart systems and sensor technologies, and artificial intelligence (AI) approaches, including machine learning and deep learning. In many cases, studies employed more than one approach simultaneously, such as combining conventional soil sampling with remote sensing to improve spatial coverage and accuracy, as in Zhang et al. [50] and Cheng et al. [52], or integrating IoT with AI applications to enable real-time data collection and analysis [53,54].

Each method represents a distinct pathway for assessing soil nutrient availability and fertility. In the following sections, we provide a detailed overview of each approach, examining how they work, their potential applications, and their role in advancing precision soil management and sustainable agriculture.

3.3.1. Traditional Methods

Traditional techniques involve field sampling and laboratory experiments to measure soil nutrients. The samples are first dried, and all debris is removed in preparation for chemical extraction using liquid solvents [55], as the appropriate extractants can effectively release the maximum amount of available nutrients [49]. This laboratory methodology was represented in 26 studies in our analysis.

Total nitrogen (TN) is often determined using the Kjeldahl method as applied in many studies, for example, Li et al. [55], Mbibueh et al. [56], and Guzman et al. [57] to monitor soil erosion, examine the effect of the land use changes on soil NPK nutrient contents, and soil nutrient interaction with the groundwater watershed, respectively. The total phosphorus is measured by using the colorimetric technique by adding molybdenum blue [37,56,58,59,60].

Mbibueh et al. [59] included magnesium (Mg) and calcium (Ca) nutrients in their study using complexometric titration by 1 M KCl; measured soil organic carbon (SOC) concentrations were measured using the dichromate oxidation method, and a conversion factor of 1.724 was applied to calculate soil organic matter (SOM) from SOC, while total organic carbon (TOC) was measured using the K₂Cr₂O₇ oxidation method [60].

The traditional soil nutrient approach is considered the most accurate and reliable method for assessing soil fertility and measuring nutrient availability [50]. However, it has several limitations, as the nutrient values may change during sample collection and handling, the process is time consuming, and it can be particularly labor-intensive and costly when applied on a large scale [13]. Additionally, these methods are unable to provide real-time information on soil nutrient status, which can delay timely decision-making for crop management.

3.3.2. Remote Sensing

Remote sensing technology is a rapid, non-destructive method that captures characteristic feature information by measuring electromagnetic radiation, either reflected or emitted from objects [24,36]. Although this technology provides cost-effective, real-time data on soil properties across large areas, offering advantages over traditional field surveys [52], this method was represented by only six records in our analysis of the studies.

A number of RS-based methods have been developed to retrieve soil nutrient information, including a direct estimation model using soil spectroscopy and an empirical model based on crop growth [52]. For instance, Zhang et al. [44] applied spectral features and multiple hyperspectral indices, such as the Normalized Difference Index (NDVI), Ratio Index (RVI), and Difference Index (DVI), to explore the quantitative correlation between soil and plant nitrogen concentration, while ground sensors and aerial hyperspectral imaging for detecting soil nutrients were highlighted by Abdulraheem et al. [24].

In addition, Zhang et al. [50] combine remote sensing technology with the traditional sampling to investigate how the time difference between field sampling and satellite image acquisition influences soil potassium predictions.

The concept of the “3S” technology in soil monitoring was represented by Zu et al. [36] and He et al. [61], which is achieved by using the way RS and GPS are combined with the GIS functions. The kriging method of GIS was used to obtain information on organic matter pH, phosphorus, nitrogen, and potassium [61].

Remote sensing (RS) offers a significant potential technique; however, its use is limited by several factors. For smallholder farms, fragmented fields, diverse vegetation types, and atmospheric interference often diminish the precision and utility of RS data [62]. Additional challenges include the lack of sufficient ground truth information and the need for experts to interpret results. Despite these barriers, RS remains a valuable tool that can complement traditional data collection methods and significantly contribute to the advancement of sustainable agricultural practices.

3.3.3. Internet of Things (IoT) and Smart Systems

IoT refers to a network of interconnected devices that rely on sensors capable of collecting, transmitting, and analyzing data wirelessly without human intervention [31,63,64]. According to the reviewed literature, 24 studies applied IoT-based systems to soil nutrient monitoring and management, highlighting its growing importance in supporting sustainable agriculture.

In the context of agriculture, IoT has rapidly evolved into a key component of precision farming by enabling real-time monitoring of soil conditions, including nutrient content (N, P, K), pH, soil moisture, and temperature [42,65,66,67,68,69,70]. The combination, with machine learning, was shown to generate actionable recommendations for nutrient management in real time [54,67,71,72,73]. Wireless sensor networks (WSNs) are the main component of the IoT, which allow multimodal data collection, storage, and real-time display through mobile platforms [30,32,62,63].

Several smart systems have been developed to support real-time nutrient monitoring and management. For instance, the Real-Time Continuous Soil Monitoring (RTCSM) system integrates sensor networks and data analytics to provide continuous soil nutrient data, enhancing decision-making for precision agriculture as represented by Fan et al. [23], while Mesfin et al. [39] use the Monitoring for Quality Improvement (Mon-QI), a notable tool which adopts mathematical modeling to improve site-specific nutrient management practices and improve nutrient management practices. Other approaches include IoT-based fuzzy control systems for smart soil monitoring, which give the optimal solution according to the trained database [70]. Site-Specific Nutrient Management (SSNM) systems were discussed, which can enhance soil health management by enabling timely, site- and crop-specific fertilizer applications [68,74].

IoT offers an effective approach to sustainable agriculture by providing real-time soil data that enhances not only fertilization but also irrigation decisions to enhance crop yield [62,65].

3.3.4. Sensors

In situ sensor techniques provide immediate, on-field soil nutrient data, aiding real-time decision support for fertilization. These sensors include optical, electrochemical, conductive, and advanced sensing technologies, which enhance monitoring efficiency and precision. In situ sensing offers quick feedback with minimal sample preparation, making it ideal for precision agriculture. It was reported in 34 of the reviewed studies, either as a standalone technique [27,75], or integrated with IoT through wireless sensors [64] and with machine learning, as demonstrated by Dattatreya et al. [26].

Table 2. Types of sensor technology used in soil nutrient monitoring and the type of measurements.

No	Type of Sensor/Probe	Sensor Measurement	Citation
1	Electrochemical sensors (3Printed sensors)	NPK	[76]
2	Electrochemical Sensors	NPK, pH, Carbon, moisture	[71]
3	Biosensor	NPK, EC,	[27]
4	Optical sensors (color sensors)	NPK	[49]
5	Optical sensors (color sensor)	pH and NPK, Moisture, Temperature	[28]
6	Wireless Sensor	Nutrient and pH levels	[64]
7	Printed Potentiometric sensor	Nitrate	[77]
8	Wireless Sensor Network	NPK	[62]
9	Ion-selective sensor	NPK	[48]
10	- Optical sensor (Vis-IR, ATR, and Raman spectroscopy) - Electrochemical sensors (ISFET’s, ISE’s)	NPK, continue nutrient monitoring	[20]
11	Optical methods (colorimetric, spectroscopic) Electrochemical methods (Ion Selective Membrane (ISM), Ion-Selective Field Transistor (ISFET) Conductivity electrodes	NPK, pH, temperature, and moisture	[26]
12	PRS (Plant Root Simulator) TM probes	Nitrogen	[78]
13	The 3D Electrospray sensor (Potentiometric solid-state ion-selective membrane (ISM)	Soil Nitrogen	[46]
14	Electro-Chemical Sensing (ISE sensors)	Nitrate	[79]
15	ISFET electrochemical microsensors	Nitrate, ammonium, pH	[80]
16	Printed soil sensors using electrochemical sensing mechanisms (Potentiometric sensors, Voltametric Sensors, Amperometric Sensors)	NPK, Cd, Pb, Cu, Hg	[25]
17	Electrochemical sensing methods	NPK	[7]
18	Wireless sensor networks (WSN)	Soil moisture and nutrients	[30]
19	In situ Soil NPK sensor	NPK	[81]
20	ISFET (Ion Sensitive Field Effect Transistor)	The sensing elements are K+, NO3−, H2PO4− as well as pH	[82]
21	Optical sensors	Nitrogen	[83]
22	- Electrochemical sensors (ISE, ISFET) - Spectroscopic sensors (UV–Vis spectrophotometry - Raman, and infrared, and biosensors. - Electrochemical sensors exhibit	Nitrite and nitrate	[84]
23	Optical chemical sensor	Nutrients	[75]
24	1-Sensors using optical principles 2-Sensors using electrical conductivity	NPK, EC, moisture	[19]
25	Wireless Sensors	NPK	[32]
26	Chemical Sensors	Soil pollutants, nutrients, moisture, and temperature	[21]
27	Electrochemical sensors	NPK	[85]
28	Biosensor	Soil health, moisture, temperature, and metals	[42]
29	Color Sensor	NPK	[86]
30	Ion-selective field-effect transistors (ISFETs) offer potential as micro-sensors	NPK and soil moisture monitoring	[87]
31	Ion-selective electrode (ISE)	Phosphate	[88]
32	Electrical conductivity-based sensors	NPK	[89]
33	Potentiometric ion sensors	Potassium ion	[90]
34	Electrophoresis-based Microfluidic ion nutrient sensor	Chloride, nitrate, sulphate, and dihydrogen phosphate	[43]

represents the type of sensors and the type of measurements for 34 articles.

The categorization of soil sensors for monitoring nutrients can be grouped into five primary groups, as shown in Figure 9: electrochemical sensors, biosensors, optical sensors, and wireless/IoT sensors that reveal distinct technological trends and applications in soil nutrient monitoring, and other sensors that are not widely represented among the scanned articles.

(I)

Electrochemical sensors

Electrochemical sensors dominate the sensor technology in this review, with 16 articles, due to their wide applicability and cost-effectiveness, and since they can quickly and automatically detect a variety of soil nutrients, electrochemical sensors have drawn interest in soil nutrient detection. Ion-selective electrodes are used in electrochemical sensors for soil nutrient determination in order to produce a current or voltage output that represents the required ion concentration [7,20,76,84,85]. They are divided into several subcategories: Ion-Selective Membranes (ISM) and Ion-Selective Field-Effect Transistors (ISFETs) are particularly useful for detecting NPK, pH, and other essential soil parameters. These sensors offer high precision and real-time feedback, making them ideal for in-field monitoring [20,26,84,87,88]. Potentiometric sensors offer high sensitivity, ion selectivity, and portability, making them well suited for real-time soil nutrient monitoring, while conventional designs require frequent calibration and are often limited to laboratory use due to the need for inner filling solutions [20,46,90]. High sensitivity, quick detection speed, high accuracy, and cost-effectiveness are among the benefits of the voltametric approach. These benefits are essential for identifying nutrients in soil [20]. Also, Amperometric sensors, which measure faradaic current under a constant applied potential, offer high sensitivity and selectivity; however, they may suffer from electrode fouling over time due to enzyme activity [25].

(II)

Biosensors

Real-time soil biosensors can be used to enable farmers to detect the NPK nutrient status of soil and choose the appropriate fertilizer application rate accordingly. This type of sensor is represented by three articles, for example: [27,42,84].

(III)

Optical Sensors

Optical sensor technology is extremely promising and developing quickly. The optical approach’s basic idea is the relationship between incident light and soil surface qualities, wherein the physical and chemical makeup of the soil influences the reflected light’s characteristics [26]. This type is represented by seven articles, and they are categorized into two approaches: Colorimetric Sensors, which provide rapid detection of NPK through visual color changes. These sensors are often employed in laboratory settings for controlled experiments [49]. Spectroscopic sensors (including UV–Vis, Raman, and FTIR) enable the analysis of oil organic matter (SOM), available potassium (AK), and other soil components. These techniques are gaining popularity due to their non-invasive nature and ability to assess multiple parameters simultaneously [26,49,75,83].

(IV)

Wireless Sensors

Wireless Sensor Networks (WSNs) and IoT are rapidly emerging in smart agriculture. These multisensory systems integrate measurements of moisture, temperature, salinity, and NPK. These sensors are integrated with ML technology, offering real-time monitoring and predictive modeling capabilities. This type of sensor is found in 10 articles, e.g., [32,38,62,63,64].

3.3.5. Artificial Intelligence Applications

Artificial Intelligence (AI) technologies are increasingly being integrated into soil nutrient monitoring systems, offering advanced capabilities for data analysis, pattern recognition, and real-time decision-making to optimize soil health and agricultural productivity [9]. This approach is represented by 14 studies, as summarized in

Table 3. Identifying the AL Algorithms and the type of input and output data.

No	Type of Al (ML/DL)	Algorithms/Models Used	Input Data/Source	Output/Prediction Target	Citation
1	ML	RF, KNN, SVM, Naive Bayes	Soil samples (NPK, pH by colorimetry)	Nitrogen, Phosphorus, Potassium, and Soil pH Prediction	[71]
2	DL	CNN	Images of maize leaves (nutrient deficiencies)	Type of nutrient deficiency in maize leaves	[28]
3	ML	MLR	Soil nutrient results from the near-infrared spectroscopy	Building a soil nutrient information extraction model	[91]
4	ML	RF, SVM, MLP	Soil images (analyzed for NPK content)	Fertilizer recommendation, nutrient requirement prediction	[26]
5	ML	RF, Logistic Regression, SVM,	Field sensor data (NPK, pH, rainfall, temperature, crop)	Fertilizer recommendation	[67]
6	ML	RF, XGB, MLR, decision tree	NPK, PH, temperature, humidity	Crop recommendation	[53]
7	ML/DL	SVR, PLS-ANN, GBRT, Cubist PLSR, BPNN, GPR,	Spectral bands/imaging	Nutrient estimation	[9]
8	DL	ANN	Soil samples to represent the nitrogen application rate (F)	Prediction of soil urea conversion	[92]
9	ML/DL	PLSR, BPNN, SVM	Soil sampling + Landsat 5 TM multispectral images	Spatial distribution of soil total nitrogen	[45]
10	ML/DL	SVM, RF, CNN	Custom crop dataset (NPK, S, Fe, Zn, temp, rainfall, pH)	Crop recommendations	[93]
11	ML/DL	ADABOOST, MLR, CNN, SVR	Soil samples’ spectral data	Organic Matter, P, K prediction	[94]
12	ML	Gradient Boosting Classifier	Soil nutrition, pH, and weather	Fertilizer recommendation	[72]
13	ML	RF, SVM, XGB, GBDT (GBRT)	Meteorological, soil physical/chemical parameters	C: N P imbalance, soil net N mineralization rate	[10]
14	ML/DL	MLR, ANN	Soil samples + satellite imaging	Total potassium prediction	[50]

Where ML: Machine Learning, DL: Deep Learning, RF: Random Forest, KNN: K-Nearest Neighbor, SVM: Support Vector Machine, SVR: Support Vector Regression, MLR: Multiple Linear Regression, MLP: Multilayer Perceptron, GBRT: Gradient Boosted Regression Trees, XGB: Extreme Gradient Boosting, Naive Bayes: Naive Bayes Classifier, PLSR: Partial Least Squares Regression, PLS-ANN: Partial Least Squares-Artificial Neural Network, CNN: Convolutional Neural Network, BPNN: Backpropagation Neural Network, ANN: Artificial Neural Network.

.

According to Jain et al. [9] Machine Learning (ML) is a subset of Artificial Intelligence (AI) that involves training computers to recognize patterns and make decisions based on data, allowing them to improve their performance over time without the need for explicit programming, while Deep Learning (DL), considered as a branch of Machine Learning (ML), focuses on training artificial neural networks, particularly deep, multi-layered networks, to analyze and learn from large and complex datasets. ML and DL algorithms are capable of accurately analyzing complex data to estimate the soil’s nutritional composition.

Machine learning (ML) has been applied to soil nutrient monitoring in several ways. (A) Sensor calibration: as ML algorithms can be used to calibrate NPK sensors, improving their accuracy by training models on datasets that link sensor outputs with corresponding nutrient values. This allows the model to adjust for potential biases or errors in sensor readings [9]. (B) Predictive Modeling: ML approaches are also used to develop predictive models that estimate Soil NPK levels in the soil based on various environmental factors, such as temperature, humidity, and rainfall. These models support pre-season planning and optimization of fertilizer application to enhance crop yields [26,50,58,71,91].

(C) Image Recognition: in cases where NPK sensors utilize imaging technologies, ML algorithms can be trained to identify soil features and patterns in images that correspond to nutrient levels, thereby enabling more efficient and accurate analysis [26,28]. ML algorithms can be trained to recognize patterns and features in these images corresponding to NPK levels, allowing for a more accurate and efficient analysis. (D) Fertilizer recommendations: ML models have also been applied to predict soil nutrients (NPK) and irrigation requirements, providing site-specific fertilizer recommendations [26,67,72]. Moreover, deep learning (DL) models extend these applications by offering advanced guidance for optimizing fertilizer use [26].

The descriptions of widely applied AI algorithms in soil nutrient monitoring are outlined below as follows:

(I) Random Forest (RF) stands as one of supervised machine learning’s widely used algorithms, effectively handling both classification and regression tasks through its ensemble approach. Rather than creating a single decision tree, RF constructs multiple trees that collectively contribute to the final prediction, significantly enhancing model strength and accuracy [10]. From Table 3, this algorithm is widely used for nutrient estimation and fertilizer, and crop recommendations [10,53,67,71].

(II) Support Vector Machines (SVMs) are a type of supervised learning method in machine learning, commonly used to solve classification and regression tasks. They are particularly effective in binary classification problems, where the goal is to separate data points into two distinct categories [10]. They were used to estimate the spatial distribution of the total soil nitrogen from the satellite image of the soil, as described by Liu et al. [45], addition to the crop and fertilizer recommendations [67,93].

(III) Extreme Gradient Boosting (XGBoost) is a powerful machine learning algorithm that builds an ensemble of decision trees, combining their outputs to produce highly accurate predictions. This technique iteratively improves model performance by focusing on errors from the previous tree [9,10]. This algorithm was used by Xu et al. [10] to understand the imbalance in the soil stoichiometry and estimate the mineralization rate of the Nitrogen.

(IV) Multiple Linear Regression (MLR) was applied to predict soil nutrients, pH, and soil organic matter (SOM), and it is a statistical modeling technique that examines how multiple independent variables collectively influence a single continuous dependent outcome, assuming that these relationships are linear in nature [9,91,94].

(V) Partial Least Squares Regression (PLSR) is particularly effective for analyzing high-dimensional datasets, where the number of variables or features exceeds the number of observations. In terms of nutrient monitoring, this approach is commonly applied in spectroscopy and chemometrics to predict a target variable based on a large set of spectral data in [9]. This ML algorithm was used for soil nitrogen estimation, predicting soil organic matter (SOM), pH, and other macronutrient predictions [45].

(VI) Gaussian Process Regression (GPR) is a non-parametric machine learning technique used for modeling continuous variables that yield variables that provide both point predictions and estimates of associated uncertainty, and is used for nutrient estimation [9].

(VII) Cubist is a predictive modeling technique that combines decision trees with rule-based refinement to generate accurate results, particularly effective for datasets with complex relationships. It has been successfully applied to soil nutrient prediction [9].

(VIII) Naïve Bayes is an ML algorithm that employs the Bayesian algorithm to perform classification. It presumes that the occurrence of a certain attribute is independent of the occurrence of other attributes. That is why this algorithm is called “Naïve”. One of its key advantages is that it requires only a small amount of training data to effectively classify input parameters [95].

(IX) Artificial Neural Intelligence (ANN): Recent studies have highlighted the successful application of artificial neural networks (ANNs) across a range of soil-related tasks. Lei et al. [92] have demonstrated the effectiveness of artificial neural networks (ANNs) as a deep learning algorithm in various soil-related applications. For instance, ANN models have successfully simulated soil ammonia volatilization, soil total nitrogen content, nutrient runoff losses, and nitrogen source pollution in surface water. However, the application of ANN models in predicting urea conversion remains an area requiring further investigation.

From Table 3, we can illustrate the different types of AL algorithms and the types of inputs and outputs that can be used in nutrient monitoring studies, as shown in Figure 10. The inputs are categorized into traditional and advanced data sources. Traditional inputs include soil nutrient data obtained from laboratory or field sampling, plant leaf imagery used to detect nutrient deficiencies, and publicly available datasets from national or global repositories. Advanced inputs comprise real-time IoT sensor data, spectral imaging (e.g., multispectral bands), and satellite imagery such as Landsat 5 TM, which offer large-scale environmental insights. At the core of the framework, ML/DL algorithms integrate and process these heterogeneous datasets to identify complex patterns and relationships in soil nutrient dynamics.

The resulting outputs include (I) nutrient analysis, such as spatial distribution maps, nutrient content estimations, and stoichiometric assessments (e.g., C: N P ratios and nitrogen mineralization rates), and (II) practical recommendations, such as fertilizer type and application rates, and crop suitability suggestions tailored to soil nutrient status.

Although ML and DL technologies have improved by merging with computer-aided precision agriculture services to achieve powerful data mining skills, environmental and nutrient conditions are highly dynamic, often characterized by complex, high-dimensional, and nonlinear interactions. These complexities pose considerable challenges in constructing an accurate predictive model [92].

3.4. Validation and Accuracy Techniques

Table 4. Validation and accuracy techniques in Soil Nutrient Monitoring.

No	Validation Techniques	Citation
1	Machine learning Help assess the prediction accuracy of the models: - Root Mean Squared Error (RMSE): $RMSE=\sqrt{{\displaystyle \frac{1}{n}}{\displaystyle \sum _{i=1}^n}{\left(y_i-{\hat{\mathrm{y}}}_i\right)}^2}$ - Ratio of Performance to Deviation (RPD):   $RPD={\displaystyle \frac{S_o}{RMSE}}$ - Ratio of Performance to Inter Quartile Distance (RPIQ):   $RPIQ={\displaystyle \frac{IQ}{RMSE}}={\displaystyle \frac{Q3-Q1}{\sqrt{\frac{1}{n}{\sum }_{i=1}^n{\left(y_i-{\hat{\mathrm{y}}}_i\right)}^2}}}$ - Mean Absolute Error (MAE): ${\displaystyle \frac{1}{n}}{\displaystyle \sum _{i=1}^n}\left(y_i-{\hat{\mathrm{y}}}_i\right)$ - Mean Squared Error (MSE): ${\displaystyle \frac{1}{n}}{\displaystyle \sum _{i=1}^n}{\left(y_i-{\hat{\mathrm{y}}}_i\right)}^2$ - Coefficient of Determination (R2): $R^2=1-{\displaystyle \frac{{\sum }_{i=1}^n{\left(y_{i }-{\hat{\mathrm{y}}}_i\right)}^2}{{\sum }_{i=1}^n{\left(y_i-{\stackrel{-}{\mathrm{y}}}_i\right)}^2}}$	[9,10,11,45,50,52,58,65,94]
2	Cross Validation Dividing the dataset into training data and testing data. Divided into three parts: 70% for training, 15% for testing, and 15% validation	[11,28,48,50,71]
3	Ground Validation Ground validation via field sampling and laboratory analysis is often necessary to calibrate and validate remote sensing data by taking direct measurements at specific locations, ensuring its accuracy and speed, and double-checking everything to ensure the remote sensing technology is reliable	[24,61]
4	Standard soil chemical properties were provided by trusted centers or institutions	[96]

Where: $y_i$ is the actual value, ${\hat{\mathrm{y}}}_i$ is the predicted value, n is the number of observations, ${\stackrel{-}{\mathrm{y}}}_i$is the mean of observed data, $S_o$ represents the standard deviation of observed data, and $IQ$is the interquartile range.

presents key validation techniques used to assess the accuracy and reliability of soil nutrient monitoring models. Machine learning validation methods include performance metrics such as RMSE, RPD, RPIQ, MAE, MSE, and R², which are widely applied to evaluate the prediction accuracy of regression models. The most common of the validation techniques in ML is R2, which is mentioned six times as in [9,10,11,44,45,50], and the RMSE, which is mentioned five times in the reviewed articles [11,44,50,65,94].

Cross-validation serves as an essential method for improving model robustness by dividing datasets and iteratively training the model. In most cases, it has been applied to verify the accuracy of soil data collected through in situ techniques [48,50]. Ground validation ensures remote sensing data accuracy by comparing field-sampled values with model predictions. Additionally, standard soil chemical properties provided by trusted institutions like the China Agricultural University further support accurate calibration and assessment [96].

3.5. Soil Nutrient Sampling Practices

The scanned articles show that of the 93 papers, only 13 of them mentioned the sampling techniques, and despite that, there is a variety between the sampling techniques mentioned in

Table 5. Different sampling techniques, sizes, and locations were used for soil nutrients.

Sampling Methods	Related Monitoring Technique	Sampling Techniques, Size, and Location	Citation
Random sampling	Traditional	In total, 12 Random sampling: Soil samples were collected for the four different rehabilitation ages: 1 year (1a), 3 years (3a), 5 years (5a), and 11 years (11a). Sampling location 1: Log 32°46′41″ N latitude 103°38′27″ E. Altitude 3050.15 m. Location 2: Log 32°46′48″ N latitude 103°37′17″ E Altitude 3000.67 m. Location 3: Log 32°49′4″ N latitude 103°39′51″ Altitude E 3110.38 m. Location 4: 32°47′39″ N 103°35′7″ E 3030.20 m.	[37]
		Random 30 points were generated using ArcGIS, with a minimum for the traditional monitoring method.	[97]
		A total of 588 soil samples were collected at a depth of 0 to 22.5 cm from the sampling location: between 21.54296910° N and 74.44691462° E.	[98]
	(IoT)+Sensors	In total, 24 for analysis, a few samples: GPS coordinates of [−7.97918385, 110.946764], [−8.06634869, 110.879614], [−8.03665317, 110.879614], and [−8.0366279, 110.884316].	[65]
	Electrochemical sensors	Random selection	[85]
	Remote sensing	Three 1 m × 1 m plots were randomly selected to understand the long-term grazing impacts on the soil. The sampled soil from sampling sites was collected at different distances (at 0, 300, 600, 900, 1200, and 1500 m).	[58]
Transects	Traditional	Four altitudinal transects, four sampling locations along each transect. Half of the sampling sites were located in cultivated land, and half were in uncultivated land—either fallow land or grassland. Sampling Location: at 37°25′18.9″ East, 11°21′44.1″ North.	[57]
		Two 20 m transects, 12 soil samples were collected at each site.	[99]
	Machine learning and traditional	In total, 210 soil samples were collected along a 3500-km transect; the sites were carefully selected, considering topography, climate conditions, crop types, cultivation practices, and local agricultural management policies.	[10]
Quadrat	Traditional and remote sensing	In total, 27 sampling quadrats were deployed for field SAN (soil available nutrients). Each quadrat had an area of 30 m × 30 m, and three sampling points were selected along the diagonal of each quadrat.	[52]
Grid	Sensor	Grid sampling, with 2 km × 2 km grid size, was adopted for the collection of soil samples.	[50]
	ML and T	A total of 800 soil samples were collected in the experimental area, and a 5 km grid was set as the sampling unit. Sampling study area: (Site 1 is located between40°04′–39°36′ N and 78°38′–79°50′ E. Site 2 range between 44°25′–44°27′ N and 85°40′–86°10′ E. Site 3 extends from 46°31–46°37′ N and 83°37′–83°41′ E.).	[94]
	Real-time sensors and traditional	A total of 145 soil samples were collected at a 0–20 cm depth. Samples were collected at 50 m × 50 m grid sampling points.	[11]

. Generally, sampling techniques could vary from Random techniques, which are dominant with six results [37,58,65,85,97,98], while the grid sampling is mentioned three times, and then comes the transaction techniques with two records by Cheng et al. [52] and Chen et al. [99], while the Quadrated techniques are mentioned once [52].

In a few articles that placed importance on the documentation of the GPS records for the sampling sites, as in Ai et al. [37], the sampling locations were specified as four different cut slopes in the mountains of Southwest China, which were related to different rehabilitation ages in this study. Hartono et al. [65] used the exact studied site location. In this study, the sensor and the Internet of Things were used, while Chaudhari et al. [98] recorded the GPS ranges of the randomly collected sampling in the study area, but the GPS range of the 118 villages was not recorded for every village. Five random soil samples were collected within every village. Guzman et al. [57] compared cultivated and non-cultivated areas by transect soil sampling techniques, but the GPS record was taken in the studied village in Ethiopia, not the exact transect location. The grid sampling techniques to identify the GPS records ranged from the examined location [11,94].

Overall, from all the reviewed articles, only eight articles documented the GPS record for the studied sites, and it was different from the studies that mentioned the exact location of the study sites, e.g., Ai et al. [37] and Hartono et al. [65], while others record the GPS location of the city or village, not the exact sampling or study point.

In terms of sample size, a wide range is observed, Wang et al. [94] conducted an intensive sampling of 800 samples across different fields, showcasing how high sampling densities can capture local soil nutrient variability, while Guzman et al. [57] sampled only 16 different samples from different locations.

3.6. Criteria for Selecting Monitoring Sites

There is no universal criterion for selecting study sites, as the selection process varies depending on the specific objectives of each study. For example, some studies prioritize topographical and rehabilitation factors, as in Ai et al. [37] who selected cut slopes with different rehabilitation ages, while others focus on land use classification to compare cultivated and non-cultivated land [57]. Soil type also plays a significant role, as demonstrated by Lindsay & French [99], who selected sand dunes with varying soil textures to examine the nitrogen cycle after weed invasion. Climate and agricultural practices also influence site selection, incorporating topography, climate conditions, and management policies into their study [10]. Some studies emphasize methodological considerations, such as Zhang et al. [50] who restricted soil sampling to the dry season to minimize moisture-related variability, and Zhang et al. [58], who assessed the impact of grazing by sampling at different distances from animal pens. This variability in site selection criteria highlights the absence of a standardized approach and the variation between the study objectives. Site selection criteria were documented in just 10 studies and are presented in

Table 6. The selection criteria of sites for nutrient monitoring within the articles.

No	Site Selection	Citation
1	The sampling locations were specified as four different 4-cut slopes in the mountains of Southwest China, corresponding to different rehabilitation ages: 1 year (1a), 3 years (3a), 5 years (5a), and 11 years (11a).	[37]
2	Half of the study sampling sites were located on cultivated land, and half were on non-cultivated land.	[57]
3	Sites were located on sand dunes, with soil types ranging from loose, loamy, quartz sands to sandy clay loams.	[99]
4	The site selection considered topography, climate, and soil type. Clay loam soil is typical of rice agriculture areas.	[85]
5	This is a forest farm area used for various experiments.	[60]
6	Due to historical land use, semi-arid soils commonly experience degradation, which leads to low levels of soil organic carbon (SOC) and poor structure.	[70]
7	This area is characterized by relatively high terrain, and the land use is cropland with corn, buckwheat, and flue-cured tobacco as the main crops.	[11]
8	The study included two types of agricultural land use (rice paddy field and vegetable field). Sites were carefully selected, considering topography, climate conditions, crop types, cultivation practices, and local agricultural management policies.	[10]
9	Soil sampling was restricted to the dry season to limit the impact of soil moisture on the satellite remote-sensing spectrum.	[50]
10	Sampling soil from sampling sites at different distances from the pens, to gain information about the longer-term impact of grazing on soil.	[58]

.

4. Discussion

4.1. Definition Challenges

The analysis of 93 studies revealed an obvious absence of agreement on the definition of soil nutrient monitoring. The different interpretations of how the researchers define or describe “soil nutrient monitoring” were placed in a table format (Table 1), to understand the overlaps, similarities, and differences. Only a small fraction of the literature provided direct definitions [35,36], while the majority provided incomplete or indirect descriptions.

The findings show some difficulties that were found in the analyzed studies; there is a conceptual overlap, as many of the reviewed studies associate soil nutrient monitoring with related but not identical concepts, such as “soil nutrient status” or “nutrient management,” as with Ai et al. [37] and Gourley et al. [38] who substitute “nutrient management” for monitoring, which shifts the focus from data collection to decision-making. Some studies frame the monitoring description narrowly to macronutrients [30,40,43,46], while the attention to micronutrients is obviously limited, which shows that, despite their agronomic importance, they are rarely incorporated into monitoring frameworks. A number of contributions described soil nutrient monitoring primarily in terms of the technologies employed as IoT, remote sensing, and sensors [15,41,44], which reduces the concept to its methodological dimension. These variabilities and challenges make it difficult to compare findings across studies. To address this gap, our study synthesizes the main components identified in previous descriptions (Figure 6) and proposes a general definition of soil nutrient monitoring that clarifies the concept and emphasizes the importance of monitoring for sustainable soil and nutrient management.

4.2. Technological Evolution and Nutrient Management

The field of soil nutrient monitoring is experiencing a technological revolution, as our results show a clear transition from the traditional laboratory-based approaches that were represented in 26 of the 93 reviewed studies (Figure 8), toward more advanced, real-time, and data-driven approaches. Although this methodology remains the most accurate technique for nutrient assessment, a dependency on a wide range of chemical extractions, high labor demand, and a lack of real-time measurement limit its widespread adoption [50]. Their dominance in early studies (Supplementary Table S2) highlights the dependence of agricultural systems on this technique but also underscores their limitations in meeting the dynamic needs for more precise agriculture in recent years.

Sensor technology is central to this revolution, as it is represented by the largest number of studies in our review (34). In situ sensors offer significant advantages by combining low cost with the ability to deliver high-density measurements, making them well-suited for large-scale soil nutrient monitoring compared to laboratory-based approaches [7]. Different sensors were highlighted in our study of soil nutrients (Figure 9). Among the wide array of options, optical and electrochemical sensors dominate soil nutrient sensors due to their ability to provide immediate field-level data. Optical sensors provide a useful analytical approach, capable of assessing irregular surfaces without damage and requiring little to no preparation of samples [21]. Despite the widespread usage of the electrochemical sensors, they have some limitations, as on-selective electrodes include the need for frequent calibration, potential requirement for additional extraction solutions, and dependence on soil moisture to obtain accurate nutrient measurements [20].

Building on these advances, the emergence of Wireless Sensor Networks (WSNs) represented a revolution in agricultural monitoring. By connecting multiple sensors into a network, WSNs enabled the real-time collection and transmission of soil data [62]. This advancement played a central role in realizing the conceptual framework of precision agriculture, offering structured and efficient solutions for optimizing fertilizer use, irrigation, and crop management [32].

This development from simple in situ sensors to interconnected WSNs marks a significant step toward the emergence of the Internet of Things (IoT) in agriculture, which is defined in this review with 24 records, as IoT builds directly on WSN infrastructure by integrating wireless communication with advanced analytics, cloud storage, and artificial intelligence [100]. As a result, soil nutrient monitoring has moved beyond sensor hardware toward a comprehensive digital ecosystem that connects data collection, processing, and decision-making into a unified framework. This development signals the true shift from traditional monitoring practices to smart, data-driven nutrient soil management systems. A major challenge in the transition from traditional to smart agriculture lies in technological accessibility and associated costs [62]. Although large-scale commercial farms are progressively adopting precision agriculture technologies, smallholder farmers face significant barriers due to the high initial costs of wireless sensors, data storage, and digital platforms. Moreover, infrastructural challenges in rural areas, such as poor network connectivity and limited digital literacy, further restrict widespread adoption [100,101]. Despite these challenges, our findings and analysis of the different soil nutrient monitoring approaches, suggests that the ongoing technological revolution represents a significant step toward more efficient, precise, and sustainable soil fertility monitoring.

4.3. Soil Nutrients Monitoring and Sustainability

The findings underscore the rising significance of soil nutrient monitoring in advancing sustainable agriculture. The recent growth in publications (Figure 3) reflects its increasing acknowledgment as an essential approach for improving soil fertility management, optimizing fertilizer use, and mitigating environmental impacts [1,102]. While the spatial distribution of these studies remains uneven across regions (Figure 4), there is a high research concentration in countries like India, China, and the United States, where stronger technological capacity supports efforts to address food security and sustainability. By contrast, regions like sub-Saharan Africa and Latin America remain underrepresented in nutrient monitoring, despite facing greater vulnerability to soil degradation and food insecurity [103].

Both macro and micronutrient levels are crucial for maintaining terrestrial ecosystems and sustaining soil fertility [104]. However, our results show that most scientific efforts and reviewed technologies for nutrient monitoring focus mainly on the primary macronutrients, nitrogen (N), phosphorus (P), and potassium (K) particularly through sensors (Table 2) and AI-driven models (Table 3), due to their central role in crop productivity and soil management, making their effective monitoring and management essential for supporting progress toward multiple SDG targets [105,106]. The purpose of nutrient monitoring differs between soil nutrient descriptions (Table 1 and Figure 6), which aligns with Sustainable Development Goals, as improving crop yields directly contribute to SDG 2 (Zero Hunger), while monitoring soil fertility is essential for land restoration and biodiversity protection [37,47], targeting SDG 15 (Life on Land). Fertilizer recommendations and minimizing environmental impact are intended to reduce greenhouse gas emissions linked to nitrogen mismanagement [41], thus addressing SDG 13 (climate action) and reducing nutrient losses and runoff [49], which contributes to SDG 6 (clean water) and SDG 14 (life below water) by reducing eutrophication [38].

Additionally, as soil nutrient monitoring advances toward more sustainable practices, it is concerning that only 13 studies reviewed reported sampling techniques, and documentation of GPS coordinates was even rarer (Table 5). Establishing standardized protocols for sampling and site GPS recording is essential to ensure transparency, reproducibility, and comparability across time and space [107], ultimately strengthening the role of nutrient monitoring in sustainable agriculture.

4.4. Assumptions, Limitations, and Challenges

This study followed the PRISMA framework, chosen for its transparent, systematic approach. A targeted search was conducted in Google Scholar on 11 August 2024, using the exact keyword phrase “soil nutrient monitoring.” While this strategy captured a wide range of relevant publications, studies that did not directly include the phrase in their titles or abstracts may have been omitted. Moreover, dependence on a single database may have limited the scope, as works indexed in other platforms were not considered. Therefore, the findings may not encompass the entirety of the existing literature but provide a representative and reliable overview of the field. The study addressed heterogeneity among the 93 reviewed articles through the systematic analysis process due to the variation of the monitoring approaches used for soil nutrients. Categorization of this approach into five groups allows the authors to discuss the different methodologies as distinct groups and summarize their unique contributions and limitations. Despite the limitations and challenges, the results provide a valuable foundation for future research on monitoring the nutrients of the soil for more sustainable agricultural practices.

5. Conclusions

The critical role of soil nutrient monitoring in safeguarding agricultural productivity, environmental sustainability, and global food security underscores its position at the center of modern agricultural systems. This review systematically analyzed definitions, methods, technologies, and validation approaches across 93 studies; there is no shared or standardized definition. To address this gap, the study integrates recurring components from existing research and proposes a comprehensive, unified definition of soil nutrient monitoring. Publication trends show a sharp increase after 2020, reflecting growing scientific and policy recognition of its importance for food security and ecosystem health. Furthermore, the findings highlight a clear transition from traditional laboratory analyses to innovative, real-time, and data-driven approaches, including sensors, Internet of Things (IoT) applications, and artificial intelligence models. These technological advances not only enhance the accuracy and real-time nutrient assessments but also reduce nutrient losses, improve fertilizer efficiency, and mitigate environmental degradation. Key recommendations from this review include standardizing soil sampling protocols with GPS site recording and integrating both macro- and micronutrient monitoring. Additionally, the adoption of in situ sensors, IoT, and AI models is encouraged to support rapid decision-making, optimize fertilizer use, enhance sustainability, and strengthen global food security.