Next Article in Journal
Revisiting the In-Between: Everyday Potentials of Intermediate Spaces for Social Sustainability in Ankara-Batıkent Housing Settlements
Previous Article in Journal
Investigating Spatial Distribution Characteristics and Influencing Factors of Traditional Villages in Southwest China from the Perspective of Human–Land Relationship
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Sustainability
Volume 18
Issue 4
10.3390/su18041893
sustainability-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Systematic Review

Soil Heavy Metals for Sustainable Risk Management: A Systematic Review and a Context-Aware Method Selection Framework

by
Leqi Yang
1,2,3,
Tianxiang Yue
3,* and
Maohua Ma
1,2
1
Chongqing Institute of Green and Intelligent Technology, Chinese Academy of Sciences, Chongqing 400714, China
2
Chongqing School, University of Chinese Academy of Sciences, Chongqing 400714, China
3
State Key Laboratory of Resources and Environmental Information System, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China
*
Author to whom correspondence should be addressed.
Sustainability 2026, 18(4), 1893; https://doi.org/10.3390/su18041893
Submission received: 23 December 2025 / Revised: 20 January 2026 / Accepted: 23 January 2026 / Published: 12 February 2026
Browse Figures
Versions Notes

Abstract

Sustainable land use requires precise monitoring of soil pollution, yet accurately predicting the spatial distribution of heavy metals often relies on post hoc accuracy comparisons with limited a priori diagnosis. To address the challenge of cost effective environmental monitoring, we conducted a PRISMA guided systematic review (2000–2024) and synthesized 135 studies to develop a mechanism-informed, context aware method selection framework. Evidence revealed three regularities: (i) element–driver coupling is structured (Pb/Cd/Zn predominantly anthropogenic; Cr/Ni geogenic; As/Hg mixed), with dominant influence scales from local to regional; (ii) model performance hinges on alignment between algorithmic assumptions, and context hybrid machine learning models integrating multi-source covariates tend to excel under strong, non-stationary anthropogenic heterogeneity, whereas kriging variants are more robust when geogenic continuity holds; and (iii) applicability is jointly constrained by environmental context, data foundations, and management objectives. Building on these insights, we propose a three-step decision workflow—goal definition, contextual diagnosis, and method matching. This framework serves as a decision support tool that shifts selection from trial and error to a priori alignment, optimizing resource allocation and enhancing the reliability of pollution assessments for sustainable soil remediation and policymaking.

1. Introduction

Soil supports plant growth and provides habitats for diverse organisms. In recent decades, rapid industrialization and intensive agriculture have accelerated pollutant inputs to soils. Wastewater and solid residues from mining, smelting, and industrial activities, as well as agricultural inputs, contribute metal-bearing wastes to soils [1,2,3]. Among the various pollutants, heavy metals such as Hg, Pb, Cd, Cu, and As are of particular concern due to their toxicity, persistence, and bioaccumulation potential [4,5,6,7]. Once entering the soil–plant–animal–human continuum, these elements can impair human health and ecosystem functions [8]. Therefore, reliable approaches to reduce the accumulation and bioavailability of heavy metals in soils are urgently needed.
Current strategies for heavy-metal-contaminated soils include physicochemical remediation, chemical remediation, phytoremediation, and bioremediation. These approaches adjust pH, redox status, and ion composition, enhance plant tolerance and uptake, or leverage microbial transformations to mitigate risks [9]. Electrokinetic processes have also shown promise for selected ions in specific matrices. In parallel with remediation technologies, environmental management increasingly requires credible spatial predictions of soil heavy metals to support risk assessment, zoning, and targeted interventions [10]. Spatial interpolation, defined as predicting continuous concentration fields from finite point observations using mathematical or statistical models, provides an essential bridge from soil-survey data to decision support [11].
Traditional interpolators such as Inverse Distance Weighting (IDW) and kriging rely primarily on spatial geometry under the assumption of spatial autocorrelation [12]. These methods offered early value for characterizing spatial patterns [4,13], yet their predictions can mirror sampling designs and provide limited insight into the environmental drivers of heavy-metal distributions [14,15]. With the growing understanding of soil processes, spatial prediction has evolved toward environmentally coupled models that explicitly incorporate terrain, soil properties, land use, and source proximity as covariates, thereby quantifying relationships between drivers and concentrations [16]. Practice has likewise shifted from single-model strategies to ensemble and hybrid approaches (e.g., regression kriging; random-forest regression kriging), combining nonlinear fitting strengths with explicit modeling of spatial structure [7,17].
Despite methodological progress, mainstream choices of interpolation methods still face three gaps that limit effective application. First, selection commonly relies on post hoc accuracy comparisons with limited prior mechanistic diagnosis, encouraging “trial-and-error” and case-dependent conclusions [18,19,20]. For example, Ordinary Kriging (OK) can be robust in settings with homogeneous parent materials but may underestimate localized enrichments where stationarity is violated [21,22,23,24]. Second, element-specific geochemical behaviors are rarely integrated into method choice, even though Cd/Pb often reflect anthropogenic inputs, Cr/Ni are frequently geology-dominated, and As/Hg commonly exhibit mixed controls [25,26,27,28]. Third, method applicability is constrained by environmental context and data foundations—sample size, sampling design, and covariate availability/quality—yet systematic guidance on matching strategies to context–data combinations remains scarce [29,30].
The rise of machine learning (ML) provides new opportunities for integrating multi-source covariates and capturing nonlinearities, supported by growing interpretability tools (e.g., SHAP values, partial-dependence plots) [31,32,33,34,35,36]. However, emphasis on accuracy gains often outweighs prior diagnosis and uncertainty quantification, while performance can be sensitive to sample size, data quality, and tuning, which complicates management use [37,38,39].
To address these gaps, we conducted a PRISMA-guided systematic review (2000–2024) to synthesize evidence on element–driver relationships, scale effects, and method performance across contexts [40]. We then developed a mechanism-informed, context-aware method-selection framework for spatial interpolation of soil heavy metals. Specifically, we: (i) clarify typical objectives and constraints; (ii) diagnose environmental mechanisms, spatial structure, and data conditions; and (iii) match method families to diagnostic profiles and data boundaries. Our goal is to shift method selection from post hoc accuracy ranking toward a priori, diagnosis-based decisions that balance predictive reliability with mechanistic plausibility, thereby supporting precise assessment and management of soil heavy-metal pollution.

2. Materials and Methods

2.1. Literature Search and Screening Strategy

This review followed PRISMA 2020 guidelines (see Supplementary Materials: Table S1—PRISMA checklist) [40], with an a priori protocol covering objectives, eligibility, data items, and synthesis plans (Appendix A). We searched the Web of Science Core Collection (Science Citation Index Expanded and Social Sciences Citation Index, Clarivate (Philadelphia, PA, USA)) for records published from 1 January 2000 to 31 December 2024, spanning the period from the mature adoption of geostatistics [41] to recent modeling advances [42]. The search strategy combined three concept groups—study object (e.g., “soil heavy metal(s)” or specific elements), core methods (e.g., “spatial interpolation”, “kriging”, “geostatistics”, “regression kriging”, “inverse distance weighting”, “random forest” AND “spatial prediction” [43,44]), and application goals (e.g., “spatial distribution”, “spatial pattern” [45,46])—linking terms within groups with OR and combining groups with AND. Full Boolean strings and field tags are provided in Appendix A to ensure reproducibility. Records were exported with full bibliographic fields and abstracts. Zotero 7 (Corporation for Digital Scholarship, Fairfax, VA, USA) was used for initial de-duplication and machine-assisted title–abstract triage based on keyword rules targeting non-soil media and non-interpolation studies; all automation decisions were subsequently verified by human reviewers. The PRISMA flow diagram is presented in Figure 1, and the completed checklist is provided in Appendix A [40].

2.2. Inclusion and Exclusion Criteria

Eligible studies investigated one or more heavy-metal elements in surface soils (typically 0–20 cm), applied at least one spatial interpolation method to predict concentrations, reported accuracy metrics based on cross-validation or equivalent procedures (e.g., RMSE, MAE, R2), were peer-reviewed original research, and were published in Chinese or English. We excluded reviews, commentaries, conference abstracts, and theses/dissertations; studies providing only descriptive analyses or visualization without interpolation; primarily methodological/theoretical papers without a soil case application; and records for which the full text was unavailable after reasonable efforts (institutional access, author contact, and interlibrary loan). We prespecified narrative/structured synthesis rather than quantitative meta-analysis because outcomes, validation protocols, environmental contexts, and reporting completeness were highly heterogeneous, making effect-size pooling statistically and substantively inappropriate.

2.3. Screening Process

The database search returned 2354 records. After removing 427 duplicates, the automation tool flagged 400 records as out of scope at the title–abstract stage; reviewer checks reinstated 18, resulting in 382 confirmed automation exclusions and 1545 records entering human screening. Two reviewers independently screened these records against prespecified criteria, excluding 646 as clearly irrelevant (e.g., non-soil media or no spatial interpolation). Discrepancies were resolved through discussion, with senior arbitration when necessary; inter-rater agreement (Cohen’s κ) was 0.81, indicating substantial agreement [47]. Full texts were sought for 899 records; 245 could not be retrieved despite institutional access, author contact, and interlibrary loan attempts. The remaining 654 articles underwent independent full-text assessment, of which 519 were excluded for mutually exclusive reasons: no spatial interpolation (n = 228), non-original research (n = 112), descriptive/visual only (n = 123), and methodological only with no soil case (n = 56). Ultimately, 135 studies met all eligibility criteria and were retained for data extraction and synthesis. Counts and reasons at each stage are summarized in Figure 1. The detailed list of the 135 reviewed studies is provided in the References section [1,2,3,4,5,6,7,8,9,10,11,12,13,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,30,31,32,34,35,36,37,38,39,41,42,43,44,45,46,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136,137,138,139], and in the Supplementary Materials: Table S2—Characteristics of the included studies.

2.4. Data Extraction and Synthesis

We used a standardized, piloted form for independent data extraction and cross-checking by two reviewers. The extracted data covered context and foundations (such as target elements, environmental setting, spatial scale, and sampling design), method characteristics (including interpolation family and parameterization), validation schemes, and performance metrics like RMSE and R2. Additionally, we assessed the risk of bias—focusing on selective reporting, data leakage, and sample stability—to inform our interpretation without excluding studies [14]. Due to significant heterogeneity, we conducted a structured narrative synthesis stratified by element, setting, and method family. Performance was summarized via directional vote counting and metric ranges, using cross-tabulations to link specific contexts with optimal methods. Finally, sensitivity analyses were performed by excluding high-risk studies, with no deviations from the original protocol.

3. Results

3.1. Overview of the Literature and Publication Characteristics

The 135 English-language studies were published across a wide range of journals, though distinct concentrations emerged. The distribution of the main journals publishing these studies is presented in Table 1. The International Journal of Environmental Research and Public Health was the leading outlet (25 papers; 18.52%), followed by Science of the Total Environment and Scientific Reports (8 each; 5.93%). Open-access journals were prominent, including Sustainability (6), Frontiers journals (5), Toxics (3), and Minerals (3). Specialized environmental venues also featured frequently, such as Journal of Hazardous Materials (4), Environmental Pollution (3), and Environmental Research (3). Discipline-specific contributions appeared in soil science (Geoderma, Journal of Soils and Sediments), remote sensing (Remote Sensing), and agriculture (Agronomy).
The remaining 57 studies were dispersed among over 50 journals covering geochemistry, atmospheric science, ecology, and hydrology, with journals containing only one or two papers representing 42.22% of the sample. Categorically, the literature was dominated by multidisciplinary environmental science journals (31.85%), open-access outlets (29.63%), and earth-science journals (18.52%), with a minor share (5.19%) in regional journals. This distribution reflects both a core set of recurring channels and broad disciplinary applicability (see Appendix A for grouping details). In addition to the journal distribution, the overall publication trends are illustrated in Figure 2.

3.2. Element-Specific Drivers and Patterns in Method Selection

Our analysis revealed two foundational regularities in the literature on spatial interpolation of soil heavy metals, providing an empirical basis for the framework developed later. Drawing on statistical synthesis of the 135 included studies, we identified core patterns from two perspectives: environmental driving mechanisms and methodological performance.
Table 2 consolidates, for major heavy metal elements, the reported methods, high-frequency drivers, and counts of relevant studies. Analysis of this table indicates several systematic correspondences between method choice and driver type.
On the methodological side, ordinary kriging and inverse distance weighting are widely used as baseline tools across study designs. However, in cases where dominant drivers are anthropogenic sources characterized by high intensity and spatial discreteness (e.g., specific industrial point sources, major traffic corridors), machine-learning models capable of integrating multi-source environmental covariates (e.g., land-use type, distance to pollution sources)—such as random forests and their kriging-based hybrids—appear with markedly higher frequency. In contrast, where the principal drivers are large-scale, spatially continuous natural background factors (e.g., specific parent materials or lithologic units), the chosen methods concentrate strongly within the kriging family, with co-kriging and regression kriging particularly prevalent.
With respect to driver composition, reporting patterns vary by element. For some elements (e.g., Pb, Cd, Cu, Zn), high-frequency drivers cluster around anthropogenic activities; others (e.g., Cr, Ni) show strong associations with specific geological backgrounds; still others (e.g., As, Hg) are frequently reported as influenced jointly by natural background and multiple anthropogenic inputs.
Taken together, there is a clear association between method selection and the dominant driver types emphasized in the literature. Studies focusing on areas primarily shaped by human activities—where spatial variability can be strong and locally anisotropic—tend to favor models that accommodate complex nonlinear relationships and spatial non-stationarity. In contrast, studies centered on settings governed by natural background—typically exhibiting strong spatial autocorrelation and broad regional trends—more commonly rely on geostatistics-centered interpolation approaches.
This association suggests that, in practice, prior analysis of the primary sources of spatial variation (i.e., putative dominant drivers) in the target area substantively guides method choice. The value of Table 2 lies in its systematic aggregation of numerous empirical cases, rendering this selection logic as quantifiable statistical regularities. This offers an evidence-based reference for future studies: once the dominant driver types in a given study area are diagnosed, one can consult the prevailing methodological tendencies under analogous contexts as empirical justification for designing the technical workflow. Building on this foundation, the subsequent sections formalize these empirical regularities into a systematic and operational method-selection framework.

3.3. Performance and Applicability Boundaries of Spatial Interpolation Methods

Analysis of method applications and head-to-head comparisons across the included studies reveals clear patterns in validation practices and contextual applicability. Table 3 synthesizes the occurrence frequencies of major method families alongside their deployment scales and target elements.
The evidence in Table 3 indicates that method selection is not random but systematically associated with research scale and element type. Specifically, the choice of interpolation strategy is jointly shaped by spatial scale (e.g., local vs. regional) and element-specific drivers (anthropogenic vs. natural).
Table 3 shows a divergence between classical geostatistical methods (e.g., ordinary kriging) and machine-learning models (e.g., random forests) across scales: the former are widely used at multiple scales, whereas the latter cluster in small- to medium-scale studies with complex anthropogenic influence. Furthermore, machine-learning approaches (e.g., RF, RFRK) are predominantly applied to elements such as Cd, Pb, and Zn, which are frequently reported as strongly associated with human activities, while classical geostatistics dominate studies of elements like Cr and Ni, where geological background is the principal driver.
This pattern suggests that the observed regularities in method selection reflect researchers’ empirical strategies for addressing spatial variability structures shaped by specific drivers at given scales. In other words, when contexts involve complex heterogeneity induced by intense human disturbance, models capable of integrating multi-source data and capturing nonlinear relationships are preferred; conversely, where the natural background predominates and spatial continuity is strong, classical methods grounded in spatial autocorrelation theory are more commonly employed.
Accordingly, we argue that evaluating the performance and suitability of spatial interpolation methods must be situated within a two-dimensional contextual space defined by “element–driver attributes” and “research scale”. The empirical associations summarized in Table 3 provide direct literature-based support for this contextualized perspective. This reframing shifts the focus from debating the superiority of individual models to systematically analyzing the alignment between study context and method characteristics, thereby laying the theoretical groundwork for a more explanatory and practically informative framework for method selection.

3.4. Construction of the Decision-Support Framework

Synthesizing the systematic evidence from Table 2 and Table 3, we constructed a mechanism-informed decision-support framework (Figure 3) that distills the implicit logic of successful studies into an explicit, step-by-step workflow. This framework is designed to bridge the gaps identified in the introduction by shifting method selection from post hoc accuracy comparison to a priori diagnosis and matching, fundamentally transforming how researchers approach spatial interpolation in soil heavy metal studies.
The framework directly addresses the lack of mechanistic diagnosis by departing from traditional “trial-and-error” approaches through a mandatory diagnostic phase at its foundation. Rather than immediately testing multiple algorithms, the process requires researchers to first refine their objectives and conduct attribute diagnosis, followed by a rigorous assessment of spatial scale and heterogeneity. This structure enforces a theoretical check before any model is run, ensuring that method selection is grounded in understanding the underlying soil processes. As demonstrated in the flowchart, the decision path diverges fundamentally based on whether the spatial process exhibits stationarity, which is typical of natural geological backgrounds, or non-stationarity, which characterizes anthropogenic disturbances. This diagnostic branching ensures that the chosen method is not merely statistically convenient but physically plausible for the specific contamination mechanism at work.
Building on this diagnostic foundation, the framework explicitly incorporates the element–driver relationships revealed in our review to address the insufficient integration of element-specific behaviors. The decision logic distinguishes between scenarios driven by strong non-stationarity, often associated with anthropogenic elements like Cd and Pb at urban and sub-regional scales, and those exhibiting relatively stationary patterns, typically linked to geological elements like Cr and Ni at broader regional scales. For pollution characterized by complex, localized sources along the left branch of the framework, users are guided toward data-driven priority methods such as Random Forest and Geographically Weighted Regression, or hybrid approaches, provided that high-quality covariates like points of interest and industrial proximity data are available to explain the observed non-stationarity. Conversely, for elements controlled by broad geological backgrounds along the right branch, the framework directs users toward kriging-based priority methods or multi-scale approaches that effectively leverage spatial autocorrelation, thereby aligning method strengths with the continuous nature of geogenic dispersion processes.
Crucially, the framework addresses the scarcity of guidance on data constraints by embedding specific decision nodes that prevent method misuse under data-limited conditions. At critical junctures, researchers must evaluate whether covariates are sufficient and of high quality, and whether sample size and spatial coverage are adequate for the contemplated approach. When covariates are insufficient or of poor quality, the framework explicitly advises against complex machine learning models to avoid overfitting, instead suggesting hybrid and semi-parametric methods or even simplified approaches like Inverse Distance Weighting if the sample size is too small to reliably capture non-stationary patterns. Similarly, when the spatial process exhibits anisotropy or multi-scale characteristics, which are common in large-scale regional studies, the framework suggests specific methodological refinements such as nested variograms rather than generic algorithms, ensuring that the selected approach matches both the spatial complexity and the available data infrastructure.

3.5. Retrospective Validation: A Case Study

To verify the practical applicability and robustness of our proposed decision-making framework, we conducted a retrospective validation using a study that was not included in our initial literature review pool. We selected a highly recent study by Han et al. [140], published in the Journal of Hazardous Materials, titled “Accurate prediction of spatial distribution of soil heavy metal in complex mining terrain using an improved machine learning method”. Since this paper was published after our data collection cutoff (December 2024), it serves as an independent “blind test” dataset. We applied our framework to the environmental context described in their study to generate a method recommendation, and then compared this recommendation against the empirical results reported by the authors.

3.5.1. Application of the Framework to the Case Study

Following the logical flow of our Principles (Figure 3), we first diagnosed the environmental characteristics of the Shimen realgar mining area described by Han et al. [140] The study area is defined by complex terrain with an elevation difference of approximately 200 m and a dominant anthropogenic point source (a smelter). Pollution dispersion in this context is heavily influenced by wind direction and topography. These factors indicate a highly non-stationary process characterized by complex, non-linear interactions between the pollutant (As) and environmental variables, suggesting that the assumption of stationarity required by pure geostatistics (e.g., Ordinary Kriging) is likely violated.
Regarding data availability, the study utilized high-resolution auxiliary data, including Digital Elevation Models (DEMs), distance to the smelter, and wind direction/speed data. These represent high-quality, spatially continuous covariates that are physically correlated with the pollutant. Consequently, based on the diagnosis of a “Non-stationary” process combined with the “Covariates Available” condition, our framework directs the user towards Hybrid Machine Learning methods (e.g., RFRK, XGBoost-RK). These methods are prioritized in our decision tree because they can model the non-linear trend using covariates to address the complex terrain and smelter influence while simultaneously using Kriging to handle the spatial autocorrelation of the residuals.

3.5.2. Verification Against Empirical Findings

This study was not included in the 135 studies synthesized under PRISMA; it was used solely for external validation of the framework. The empirical findings reported by Han et al. [140] strongly corroborate our framework’s recommendation. The authors compared four distinct approaches: Ordinary Kriging (OK), Geographically Weighted Regression Kriging (GWRK), Random Forest (RF), and Random Forest Regression Kriging (RFRK). As predicted by our framework, the RFRK model achieved the highest accuracy on the test set (R2 = 0.78, RMSE = 23.46 mg/kg) [140].
The study explicitly noted the limitations of traditional methods in this context; Ordinary Kriging suffered from the “bull’s-eye effect” and failed to capture the diffusion pattern caused by the smelter and wind, resulting in lower accuracy (R2 = 0.72) [140]. The authors concluded that incorporating environmental variables via a non-linear model combined with residual interpolation effectively addressed the spatial heterogeneity caused by the complex terrain. This retrospective test demonstrates that our framework correctly identified the optimal methodological class based solely on the initial environmental diagnosis. By following the framework, a researcher would have avoided the less effective traditional methods and selected the high-performing RFRK approach, thereby validating the framework’s utility as a pre-study decision support tool.

4. Discussion

4.1. Core Findings and Conceptual Approach

For decades, research on the spatial interpolation of soil heavy metals has followed a seemingly efficient yet intrinsically contradictory methodological paradigm: although models rest on fundamentally different mathematical assumptions (e.g., spatial stationarity, linearity), researchers commonly adopt a generic workflow of multi-model trial-and-error and cross-validated model selection [4,29,73]. Our systematic analysis of a large body of empirical studies reveals the fundamental limitations of this paradigm. The core problem is that it reduces method choice to a post hoc “goodness-of-fit competition” based solely on statistical indices (e.g., RMSE, R2) while neglecting that spatial interpolation is, at heart, “spatial modeling of specific environmental processes” [20,30]. This inversion of priorities makes conclusions highly contingent on the idiosyncrasies of sampling configurations, impedes the accumulation of generalizable methodological insights beyond individual case studies, and often leads to model failure when transferring from Area A (where a method performs well) to Area B with different driving mechanisms [39,93].
Our in-depth review provides clear empirical grounds for overcoming this impasse. We found that methodological effectiveness is not dictated by the intrinsic sophistication of a model, but by the degree of alignment between its underlying assumptions and the generative mechanisms of spatial variation in the study region [58]. This alignment is governed primarily by two contextual dimensions: (i) the type of dominant drivers—defined jointly by the geochemical properties of the target element and the intensity of human activities (e.g., anthropogenic point-source emissions versus geogenic background control) [16,34]; and (ii) the spatial scale of interest [141]. For example, in industrial areas characterized by strong anthropogenic disturbance, pronounced heterogeneity, and short-range dependence, machine-learning hybrids capable of integrating multi-source covariates and capturing nonlinearity and non-stationarity (e.g., random-forest regression kriging) tend to perform better [50]. Conversely, in natural settings governed by continuous geological background with gradual variation and long-range dependence, classical kriging methods—whose performance relies on stationarity—demonstrate inherent robustness and efficiency [4].
Establishing this empirical regularity marks a shift in perspective from “searching for a universally optimal tool” to “understanding context-specific requirements”. It clarifies that, before asking “which method should we choose”, one must first address a more fundamental question: “What are the principal characteristics of the spatial process we aim to represent?” [95]. This reframing provides a solid foundation for a more scientific and interpretable decision logic. It requires moving the focus of decision-making upstream—from post hoc accuracy comparisons to a priori, systematic diagnosis of the study context, including pollution origins, spatial structure, and data foundations [57].

4.2. Implications for Method Selection and Practice

Grounded in the core insight that “method effectiveness depends on its alignment with the underlying spatial process”, a sound decision logic must begin with understanding the study context rather than end with comparing model accuracies. Accordingly, we distilled from numerous successful cases a structured “diagnose–match” decision approach. This approach reframes method selection as a rational analysis guided by prior diagnosis and followed by validation, ensuring that the chosen technical pathway rests on explicit environmental mechanisms.
The starting point is to clarify application objectives and objective data constraints. Any technical choice serves specific management or scientific aims and is constrained by real-world conditions. It is therefore essential to define whether the final outputs are in-tended for risk tiering, source apportionment, remediation planning, or baseline mapping, as these goals differ fundamentally in emphasis (e.g., exceedance probabilities, boundary delineation, extremes) [9,13]. In parallel, one must realistically assess sample size and spatial representativeness, the availability and quality of key environmental covariates, and computational resources [12]. These considerations delimit the feasible decision space and prevent the pursuit of theoretically “optimal” yet practically infeasible solutions.
Once objectives and constraints are established, the core of the decision process shifts to diagnosing the dominant spatial processes. This step aims to identify the key drivers shaping the spatial patterns of the target metals and to characterize their essential spatial-structure features. Two tasks proceed in parallel: (i) driver-source analysis, which integrates multiple information sources—regional geology, historical land use, distributions of industrial and agricultural activities—together with elemental chemistry to infer primary contributors through logical reasoning and spatial overlays; and (ii) spatial-structure probing, which conducts exploratory spatial data analysis on samples, such as quantifying autocorrelation and characteristic scales via semi-variograms and identifying clustering patterns via spatial autocorrelation indices, thereby revealing continuity, stationarity, and operative scales.
With diagnosis in hand, method selection shifts from “blind choice” to “targeted matching”. Decision-makers then select from the method library those classes whose core assumptions are most compatible with the diagnosed process characteristics. For example, when diagnostics indicate intense anthropogenic point-source influence with strong, non-stationary heterogeneity, models that integrate multi-source covariates and capture nonlinearity and non-stationarity—such as geographically weighted regression or random forest regression kriging—should be prioritized. When diagnostics indicate geo-genic control with continuous, stationary variation, kriging-based methods—robust and efficient for stationary series—are preferred. The initially defined objectives and data constraints must also inform final selection. For instance, if the primary goal is probabilistic risk mapping, indicator kriging is functionally well-matched because it directly outputs exceedance probabilities; if sample sizes are very limited, even when diagnostics suggest complex processes, simpler, more parameter-stable models may be advisable to mitigate overfitting risk.

4.3. Challenges and Future Directions

Despite the proposed structured decision logic, limitations remain. First, reliance on the published literature introduces potential bias towards anthropogenic hotspots (e.g., mining and industrial areas) over natural backgrounds, necessitating validation in diverse geological contexts. Second, the “diagnose–match” approach requires systematic prospective testing to verify its operational feasibility in complex real-world settings.
Future research should focus on three directions:
  • Develop objective diagnostics for spatial processes.
Research should transition from qualitative expert judgment to quantitative “spatial process fingerprints”. This can be achieved through multi-parameter geostatistical analysis (e.g., semi-variograms) and machine learning-based feature extraction, standardizing the identification of dominant drivers.
2.
Strengthen multi-dimensional validation frameworks.
Beyond simple metrics like RMSE, validation systems must integrate spatial-structure fidelity, mechanistic plausibility, and uncertainty quantification. Aligning these metrics with risk assessment objectives will provide a more relevant benchmark for environmental management.
3.
Build interactive decision-support tools.
To lower barriers to application, the decision logic and element–driver associations identified here could be encoded into software tools or expert systems. Integrating these with background databases and exploratory analysis modules would facilitate the practical translation of this framework from research to practice.

5. Conclusions

Spatial interpolation of soil heavy metals has evolved into a multidisciplinary domain integrating geo-statistics, environmental processes, and machine learning; yet, selecting appropriate methods for specific conditions remains a critical bottleneck. Through a systematic analysis of 135 studies, this review draws the following core conclusions:
First, the effectiveness of spatial interpolation methods fundamentally depends on the alignment between their algorithmic assumptions and the generative mechanisms of spatial variability in the study area. This alignment is primarily governed by two contextual dimensions: the dominant drivers (defined by the interplay of elemental geochemistry and human activities) and the spatial scale of analysis. Specifically, in areas dominated by anthropogenic disturbance with pronounced variability, models that integrate multi-source covariates and capture nonlinearity and spatial non-stationarity are required. Conversely, in areas governed by natural backgrounds with continuous variation, classical geostatistical approaches prove more robust and efficient.
To address the challenge of method selection systematically, we proposed a three-step decision framework—goal definition, contextual diagnosis, and method matching—grounded in these mechanistic insights. While the rise of Machine Learning (ML) offers powerful tools for modeling complex, non-linear relationships, it is not a universal panacea. Our review highlights that the superiority of ML is strictly conditional on data quantity and quality. This framework shifts method selection from post hoc accuracy comparisons to a priori mechanistic diagnosis. The process begins by clarifying management objectives and data constraints, proceeds to diagnose dominant drivers and spatial structures using geological and land-use data, and culminates in selecting method families compatible with the diagnostic profile.
The principal contribution of this study is the consolidation of fragmented empirical knowledge into a structured decision logic. By offering a mechanism-informed and reproducible pathway, this framework reduces arbitrariness in technical decisions, thereby enhancing the scientific rigor and reliability of spatial interpolation for pollution assessment and risk management. Future research will focus on applying these models to in situ soil sampling to verify the theoretical predictions against the observed contamination levels.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/su18041893/s1, Table S1: PRISMA Checklist; Table S2: Characteristics of the included studies.

Author Contributions

Conceptualization, T.Y. and L.Y.; methodology, L.Y.; software, L.Y.; validation, L.Y., T.Y. and M.M.; formal analysis, L.Y.; investigation, L.Y.; resources, T.Y.; data curation, L.Y.; writing—original draft preparation, L.Y.; writing—review and editing, T.Y. and M.M.; visualization, L.Y.; supervision, T.Y.; project administration, T.Y.; funding acquisition, T.Y. All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported by the National Key Research & Development Program of China (2024YFD1700904), the National Natural Science Foundation of China (42330707), the Science Fund for Creative Research Groups of the National Natural Science Foundation of China (72221002), the Strategic Priority Research Program of Chinese Academy of Sciences (XDB0740100), and the National Key R&D Program of China (2021YFB3901300).

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The data from the literature search conducted for this review are available from the corresponding author upon reasonable request (email: yue@lreis.ac.cn).

Acknowledgments

We thank the editors and anonymous reviewers for their constructive comments.

Conflicts of Interest

The authors declare no conflicts of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

Abbreviations

The following abbreviations are used in this manuscript:
BMEBayesian Maximum Entropy
BPNNBackpropagation Neural Network
CNNConvolutional Neural Network
COKCo-Kriging
EBKEmpirical Bayesian Kriging
FKFixed Kriging
GBDTGradient Boosting Decision Trees
GEOHealthGlobal Environmental and Occupational Health
GTWRGeographically and Temporally Weighted Regression
GWRKGeographically Weighted Regression Kriging
HASMHigh-Accuracy Surface Modeling
IDWInverse Distance Weighting
IKIndicator Kriging
LSTMLong Short-Term Memory
MAEMean Absolute Error
MLMachine Learning
OKOrdinary Kriging
PMFPositive Matrix Factorization
PRISMAPreferred Reporting Items for Systematic Reviews and Meta-Analyses
PXRFPortable X-Ray Fluorescence
R2Coefficient of Determination
RBFRadial Basis Function
RFRandom Forest
RFRKRandom-Forest Regression Kriging
RKRegression Kriging
RMSERoot Mean Square Error
RSRemote Sensing
SHAPSHapley Additive exPlanations
SOMSoil Organic Matter
STCKSpatio-Temporal Co-Kriging
STKSpatio-Temporal Kriging
SVMSupport Vector Machine
TPSThin-Plate Spline
UKUniversal Kriging

Appendix A. Protocol, Search Strategy, Extraction Scheme, and Included Studies

A1. Protocol and Deviations
  • Objective: Synthesize application patterns and methodological performance of spatial interpolation for soil heavy metals; derive a mechanism informed, context aware method selection framework.
  • Registration: Protocol drafted a priori (date: [insert]) including objectives, eligibility, data items, and synthesis plan. No registry ID (discipline typical).
  • Deviations: None in eligibility. A post hoc addition of risk of bias annotations was made after piloting extraction; this did not affect inclusion.
A2. Search Strategy (Web of Science Core Collection)
  • Databases: Science Citation Index Expanded (SCIE) and Social Sciences Citation Index (SSCI).
  • Time window: 2000 01 01 to 2024 12 31.
  • Language: English or Chinese.
  • Document types: Article; Early Access included; Reviews excluded.
  • Query concepts and Boolean logic:
Group 1 (object): TS=(“soil heavy metal*” OR cadmium OR lead OR mercury OR arsenic OR copper OR zinc OR chromium OR nickel OR “potentially toxic element*” OR PTE*)
Group 2 (methods): TS=(“spatial interpolation” OR kriging OR “ordinary kriging” OR “universal kriging” OR cokriging OR “regression kriging” OR IDW OR “inverse distance weight*” OR “geostatistic*” OR “geograph* weight* regression” OR “random forest” OR “support vector*” OR “gradient boosting” OR XGBoost OR “neural network” OR “deep learning”) AND TS=(“spatial prediction” OR “spatial distribution” OR “spatial pattern”)
Group 3 (application scope): TS=(soil OR topsoil OR farmland OR agricultural OR urban OR mining OR industrial)
Final Boolean: (Group1) AND (Group2) AND (Group3)
  • Field tags: TS=Topic; Language filter: English, Chinese; Document Type filter: Article.
  • Export fields: Full record and abstract, cited references, author keywords, Keywords Plus, funding.
A3. Information Sources and Yields
  • Source: Web of Science (WoS) only; no trial registers or grey literature.
  • Records identified: 2354 (SCIE/SSCI combined).
  • De duplication: 427 records removed.
  • See Figure 1 for PRISMA flow.
A4. Automation and Human Screening
  • Tools: Zotero 7 (Corporation for Digital Scholarship) for de duplication; Rayyan v1.0 (Qatar Computing Research Institute) for machine assisted title–abstract triage.
  • Automation rules: keyword filters to flag non soil media and non-interpolation studies; conservative thresholds to minimize false negatives.
  • Human verification: All automation flagged records were checked by a reviewer; 18 reinstated.
  • Title–abstract screening: Two independent reviewers; Cohen’s κ = 0.81 (computed on all records at this stage).
  • Full text retrieval: Institutional access, author contact, interlibrary loan; 245 not retrievable.
A5. Eligibility Criteria (verbatim)
  • Inclusion: (i) surface soil (typically 0–20 cm); (ii) ≥1 spatial interpolation method predicting concentrations; (iii) accuracy via cross validation or equivalent (RMSE, MAE, R2); (iv) peer reviewed original research; (v) English or Chinese.
  • Exclusion: reviews/commentaries/abstracts/theses; descriptive only without interpolation; methodological/theoretical without soil case; full text unavailable after reasonable efforts.
A6. Data Extraction Form (Variables and Codebook)
  • Study identifiers: first author, year, journal, DOI, country/region.
  • Elements: single/multiple; list of metals(loid)s.
  • Environmental setting: anthropogenic dominated vs. geology dominated; land use (urban/agricultural/mining/industrial/mixed); terrain.
  • Scale: site/local (≤100 km2), district/city, regional/national, catchment.
  • Sampling: n (bins: <100, 100–300, 300–1000, >1000); design (grid/stratified/opportunistic/mixed); depth (cm).
  • Covariates: types (terrain/soil/land use/source proxies/remote sensing/atmospheric); richness (none/low/moderate/rich); quality notes.
  • Methods: family (IDW; OK/UK; Co-Kriging; Regression Kriging; ML regression; ML+RK; GWR/GTWR; spatiotemporal kriging; depth trend kriging); key parameters (variogram model, neighborhood, ML hyperparameters); uncertainty reporting (prediction intervals, error surfaces: yes/no).
  • Validation: scheme (LOOCV/k fold/hold out); metrics (RMSE/MAE/R2; others); within study rankings.
  • Risk of bias: selective reporting (Y/N); leakage risk (Y/N); variogram diagnostics adequate (Y/N/NA); uncertainty reported (Y/N); small sample risk (Y/N).
  • Notes: special modeling (e.g., depth function, diffusion informed, network constrained, two point ML).
A7. Synthesis Plan and Justification for No Meta Analysis
  • Primary synthesis: structured narrative, stratified by element category (Cd/Pb; Cr/Ni; As/Hg; mixed), environmental setting, data conditions (sample size/design/covariate richness), and method family.
  • Descriptive comparisons: directional vote counting within strata; ranges of RMSE/R2; uncertainty reporting frequency.
  • Cross tabulation: elements × settings × methods to identify regularities.
  • Sensitivity: exclude high risk of bias studies; assess stability of qualitative conclusions.
  • Rationale against meta-analysis: heterogeneity in outcomes and validation protocols, environmental contexts, and incomplete reporting makes effect size pooling inappropriate.

References

  1. Davis, H.T.; Marjorie Aelion, C.; McDermott, S.; Lawson, A.B. Identifying Natural and Anthropogenic Sources of Metals in Urban and Rural Soils Using GIS-Based Data, PCA, and Spatial Interpolation. Environ. Pollut. 2009, 157, 2378–2385. [Google Scholar] [CrossRef]
  2. Santos-Frances, F.; Martinez-Grana, A.; Avila Zarza, C.; Garcia Sanchez, A.; Alonso Rojo, P. Spatial Distribution of Heavy Metals and the Environmental Quality of Soil in the Northern Plateau of Spain by Geostatistical Methods. Int. J. Environ. Res. Public Health 2017, 14, 568. [Google Scholar] [CrossRef] [PubMed]
  3. Ballabio, C.; Panagos, P.; Lugato, E.; Huang, J.-H.; Orgiazzi, A.; Jones, A.; Fernández-Ugalde, O.; Borrelli, P.; Montanarella, L. Copper distribution in European topsoils: An assessment based on LUCAS soil survey. Sci. Total Environ. 2018, 636, 282–298. [Google Scholar] [CrossRef] [PubMed]
  4. Xie, Y.; Chen, T.; Lei, M.; Yang, J.; Guo, Q.; Song, B.; Zhou, X. Spatial Distribution of Soil Heavy Metal Pollution Estimated by Different Interpolation Methods: Accuracy and Uncertainty Analysis. Chemosphere 2011, 82, 468–476. [Google Scholar] [CrossRef] [PubMed]
  5. Liu, Y.; Ma, Z.; Lv, J.; Bi, J. Identifying Sources and Hazardous Risks of Heavy Metals in Topsoils of Rapidly Urbanizing East China. J. Geogr. Sci 2016, 26, 735–749. [Google Scholar] [CrossRef]
  6. Hu, B.; Zhao, R.; Chen, S.; Zhou, Y.; Jin, B.; Li, Y.; Shi, Z. Heavy Metal Pollution Delineation Based on Uncertainty in a Coastal Industrial City in the Yangtze River Delta, China. Int. J. Environ. Res. Public Health 2018, 15, 710. [Google Scholar] [CrossRef]
  7. Xu, X.; Qian, J.; Xie, E.; Shi, X.; Zhao, Y. Spatio-Temporal Change and Pollution Risk of Agricultural Soil Cadmium in a Rapidly Industrializing Area in the Yangtze Delta Region of China. Int. J. Environ. Res. Public Health 2018, 15, 2743. [Google Scholar] [CrossRef]
  8. Hoque, M.M.; Islam, A.; Islam, A.R.M.T.; Pal, S.C.; Mahammad, S.; Alam, E. Assessment of Soil Heavy Metal Pollution and Associated Ecological Risk of Agriculture Dominated Mid-Channel Bars in a Subtropical River Basin. Sci. Rep. 2023, 13, 11104. [Google Scholar] [CrossRef]
  9. Mohamed, E.S.; Jalhoum, M.E.M.; Hendawy, E.; El-Adly, A.M.; Nawar, S.; Rebouh, N.Y.; Saleh, A.; Shokr, M.S. Geospatial Evaluation and Bio-Remediation of Heavy Metal-Contaminated Soils in Arid Zones. Front. Environ. Sci. 2024, 12, 1381409. [Google Scholar] [CrossRef]
  10. Saha, A.; Sen Gupta, B.; Patidar, S.; Hernández-Martínez, J.L.; Martín-Romero, F.; Meza-Figueroa, D.; Martínez-Villegas, N. A Comprehensive Study of Source Apportionment, Spatial Distribution, and Health Risks Assessment of Heavy Metal(Loid)s in the Surface Soils of a Semi-Arid Mining Region in Matehuala, Mexico. Environ. Res. 2024, 260, 119619. [Google Scholar] [CrossRef]
  11. Santos-Frances, F.; Martinez-Grana, A.; Alonso Rojo, P.; Garcia Sanchez, A. Geochemical Background and Baseline Values Determination and Spatial Distribution of Heavy Metal Pollution in Soils of the Andes Mountain Range (Cajamarca-Huancavelica, Peru). Int. J. Environ. Res. Public Health 2017, 14, 859. [Google Scholar] [CrossRef]
  12. Cao, J.; Li, C.; Zhang, L.; Wu, Q.; Lv, J. Source Apportionment of Potentially Toxic Elements in Soils Using APCS/MLR, PMF and Geostatistics in a Typical Industrial and Mining City in Eastern China. PLoS ONE 2020, 15, e0238513. [Google Scholar] [CrossRef]
  13. Duan, Y.; Zhang, Y.; Li, S.; Fang, Q.; Miao, F.; Lin, Q. An integrated method of health risk assessment based on spatial interpolation and source apportionment. J. Clean. Prod. 2020, 276, 123218. [Google Scholar] [CrossRef]
  14. Higgins, J.P.T.; Thomas, J.; Chandler, J.; Cumpston, M.; Li, T.; Page, M.J.; Welch, V.A. Cochrane Handbook for Systematic Reviews of Interventions, 2nd ed.; Wiley: Chichester, UK, 2019. [Google Scholar]
  15. Wang, G.; Liu, H.-Q.; Gong, Y.; Wei, Y.; Miao, A.-J.; Yang, L.-Y.; Zhong, H. Risk Assessment of Metals in Urban Soils from a Typical Industrial City, Suzhou, Eastern China. Int. J. Environ. Res. Public Health 2017, 14, 1025. [Google Scholar] [CrossRef]
  16. Shi, T.; Zhang, J.; Shen, W.; Wang, J.; Li, X. Machine Learning Can Identify the Sources of Heavy Metals in Agricultural Soil: A Case Study in Northern Guangdong Province, China. Ecotoxicol. Environ. Saf. 2022, 245, 114107. [Google Scholar] [CrossRef] [PubMed]
  17. Correcting Correlation Quality of Portable X-Ray Fluorescence to Better Map Heavy Metal Contamination by Spatial Co-Kriging Interpolation. Ecotoxicol. Environ. Saf. 2024, 271, 115962. [CrossRef]
  18. Yang, W.; Deng, M.; Yang, X.; Wei, D. Predictive Soil Pollution Mapping: A Hybrid Approach for a Dataset with Outliers. IEEE Access 2019, 7, 46668–46676. [Google Scholar] [CrossRef]
  19. Castro-Hinostroza, N.; Mestana Novoa, C.J.; Vizarreta-luna, G.A.; Gutierrez Vilchez, P.P.; Guerrero Barrantes, J.A.; Nazario Rios, J.C.; Tello Peramas, L.D. Spatial Distribution of Lead in the Urban Soil of Callao, Peru. Rev. Int. Contam. Ambient. 2024, 40, 423–436. [Google Scholar] [CrossRef]
  20. Li, Y.; Yang, D.; Zhang, P.; Geng, W.; Liu, Y.; Chang, H.; Qin, M.; Rong, T.; Shao, J. Optimal Interpolation and Ecological Risk Assessment of Spatial Distribution of Heavy Metals in Soil Based on GIS and Geostatistics—A Case Study of Low-Lying Land of the Lower Yellow River, China. Pol. J. Environ. Stud. 2021, 30, 2631–2644. [Google Scholar] [CrossRef] [PubMed]
  21. Chen, D.; Wang, X.; Luo, X.; Huang, G.; Tian, Z.; Li, W.; Liu, F. Delineating and Identifying Risk Zones of Soil Heavy Metal Pollution in an Industrialized Region Using Machine Learning. Environ. Pollut. 2023, 318, 120932. [Google Scholar] [CrossRef]
  22. Ma, Q.; Han, L.; Zhang, J.; Zhang, Y.; Lang, Q.; Li, F.; Han, A.; Bao, Y.; Li, K.; Alu, S. Environmental Risk Assessment of Metals in the Volcanic Soil of Changbai Mountain. Int. J. Environ. Res. Public Health 2019, 16, 2047. [Google Scholar] [CrossRef]
  23. Kabir, M.S.; Salam, M.A.; Paul, D.N.R.; Hussain, M.I.; Rahman, N.M.F.; Latif, M.A. Geo-Statistical Models for Determining Spatial Variation and Spatial Dependency of Soil Arsenic in Bangladesh. J. Natl. Sci. Found. Sri Lanka 2017, 45, 179–191. [Google Scholar] [CrossRef]
  24. Zhao, D.; Xie, D.; Yin, F.; Liu, L.; Feng, J.; Ashraf, T. Estimation of Pb Content Using Reflectance Spectroscopy in Farmland Soil near Metal Mines, Central China. Remote Sens. 2022, 14, 2420. [Google Scholar] [CrossRef]
  25. Ahado, S.K.; Nwaogu, C.; Sarkodie, V.Y.O.; Boruvka, L. Modeling and Assessing the Spatial and Vertical Distributions of Potentially Toxic Elements in Soil and How the Concentrations Differ. Toxics 2021, 9, 181. [Google Scholar] [CrossRef] [PubMed]
  26. Cai, Z.; Lei, S.; Zhao, Y.; Gong, C.; Wang, W.; Du, C. Spatial Distribution and Migration Characteristics of Heavy Metals in Grassland Open-Pit Coal Mine Dump Soil Interface. Int. J. Environ. Res. Public Health 2022, 19, 4441. [Google Scholar] [CrossRef] [PubMed]
  27. Zhang, S.; Liu, H.; Luo, M.; Zhou, X.; Lei, M.; Huang, Y.; Zhou, Y.; Ge, C. Digital Mapping and Spatial Characteristics Analyses of Heavy Metal Content in Reclaimed Soil of Industrial and Mining Abandoned Land. Sci. Rep. 2018, 8, 17150. [Google Scholar] [CrossRef]
  28. Javaid, A.; Ahmad, S.R.; Mahmood, R.; Mahmood, K.; Qadir, A.; Batool, R. Depletion Risks of Essential Trace Metals in a Transboundary Drain and Its Surrounding Soil Using GIS Techniques. Pak. J. Agric. Sci. 2021, 58, 1215–1221. [Google Scholar] [CrossRef]
  29. Radočaj, D.; Jurišić, M.; Župan, R.; Antonić, O. Spatial Prediction of Heavy Metal Soil Contents in Continental Croatia Comparing Machine Learning and Spatial Interpolation Methods. Geod. List 2020, 74, 357–372. [Google Scholar]
  30. Yang, Y.; Jia, M. 3D Spatial Interpolation of Soil Heavy Metals by Combining Kriging with Depth Function Trend Model. J. Hazard. Mater. 2024, 461, 132571. [Google Scholar] [CrossRef]
  31. Jia, Z.; Zhou, S.; Su, Q.; Yi, H.; Wang, J. Comparison Study on the Estimation of the Spatial Distribution of Regional Soil Metal(Loid)s Pollution Based on Kriging Interpolation and BP Neural Network. Int. J. Environ. Res. Public Health 2018, 15, 34. [Google Scholar] [CrossRef]
  32. Wang, H.; Zhao, M.; Huang, X.; Song, X.; Cai, B.; Tang, R.; Sun, J.; Han, Z.; Yang, J.; Liu, Y.; et al. Improving Prediction of Soil Heavy Metal(Loid) Concentration by Developing a Combined Co-Kriging and Geographically and Temporally Weighted Regression (GTWR) mode. J. Hazard. Mater. 2024, 468, 133745. [Google Scholar] [CrossRef]
  33. Li, K.; Liu, J.; Zheng, H.; Hou, Y.; Fu, Q.; Chen, C.; Yuan, L.; Lv, B.; Liu, S.; Huang, Y.; et al. Quantifying Realized Ecosystem Services from the Supply-Flow-Demand Perspective and Understanding the Management Implications: A Case Study of the Chengdu Prefecture, China. Ecol. Indic. 2025, 181, 114337. [Google Scholar] [CrossRef]
  34. Chen, T.; Chang, Q.; Liu, J.; Clevers, J.G.P.W.; Kooistra, L. Identification of Soil Heavy Metal Sources and Improvement in Spatial Mapping Based on Soil Spectral Information: A Case Study in Northwest China. Sci. Total Environ. 2016, 565, 155–164. [Google Scholar] [CrossRef]
  35. Cortes, J.L.; Bautista, F.; Delgado, C.; Quintana, P.; Aguilar, D.; Garcia, A.; Figueroa, C.; Gogichaishvili, A. Spatial Distribution of Heavy Metals in Urban Dust from Ensenada, Baja California, Mexico. Rev. Chapingo Ser. Cienc. For. Ambiente 2017, 23, 47–60. [Google Scholar] [CrossRef]
  36. Chen, H.; Wang, J. Active Learning for Efficient Soil Monitoring in Large Terrain with Heterogeneous Sensor Network. Sensors 2023, 23, 2365. [Google Scholar] [CrossRef] [PubMed]
  37. Zeng, Y.; Shi, T.; Liu, Q.; Yang, C.; Zhang, Z.; Wang, R. A geographically weighted neural network model for digital soil mapping of heavy metal copper in coastal citie. J. Hazard. Mater. 2024, 480, 136285. [Google Scholar] [CrossRef] [PubMed]
  38. Zhang, H.; Lu, R.; Shi, W.; Shi, R. Application of RBF Neural Network in Determining Soil Heavy Metal Spatial Variability: Application of RBF Neural Network in Determining Soil Heavy Metal Spatial Variability. Chin. J. Eco-Agric. 2012, 20, 474–479. [Google Scholar] [CrossRef]
  39. Li, C.; Jiang, X.; Jiang, H.; Sha, Q.; Li, X.; Jia, G.; Cheng, J.; Zheng, J. Environmental Controls to Soil Heavy Metal Pollution Vary at Multiple Scales in a Highly Urbanizing Region in Southern China. Sensors 2022, 22, 4496. [Google Scholar] [CrossRef]
  40. Page, M.J.; McKenzie, J.E.; Bossuyt, P.M.; Boutron, I.; Hoffmann, T.C.; Mulrow, C.D.; Shamseer, L.; Tetzlaff, J.M.; Akl, E.A.; Brennan, S.E.; et al. The PRISMA 2020 statement: An updated guideline for reporting systematic reviews. BMJ 2021, 372, n71. [Google Scholar] [CrossRef]
  41. Hu, K.-L.; Zhang, F.-R.; Li, H.; Huang, F.; Li, B.-G. Spatial Patterns of Soil Heavy Metals in Urban-Rural Transition Zone of Beijing. Pedosphere 2006, 16, 690–698. [Google Scholar] [CrossRef]
  42. Han, H.; Suh, J. Spatial Prediction of Soil Contaminants Using a Hybrid Random Forest-Ordinary Kriging Model. Appl. Sci. 2024, 14, 1666. [Google Scholar] [CrossRef]
  43. Johnbull, O.; Abbassi, B.; Zytner, R.G. Risk Assessment of Heavy Metals in Soil Based on the Geographic Information System-Kriging Technique in Anka, Nigeria. Environ. Eng. Res. 2019, 24, 150–158. [Google Scholar] [CrossRef]
  44. Wang, H.; Yilihamu, Q.; Yuan, M.; Bai, H.; Xu, H.; Wu, J. Prediction models of soil heavy metal(loid)s concentration for agricultural land in Dongli: A comparison of regression and random forest. Ecol. Indic. 2020, 119, 106801. [Google Scholar] [CrossRef]
  45. Li, Y.; Xue, J.; Cai, J.; Zhang, Y.; Li, F.; Zha, X.; Fang, G. The Spatial Distribution and Influencing Factors of Heavy Metals in Soil in Xinjiang, China. Sustainability 2023, 15, 16379. [Google Scholar] [CrossRef]
  46. Krami, L.K.; Amiri, F.; Sefiyanian, A.; Shariff, A.R.B.M.; Tabatabaie, T.; Pradhan, B. Spatial Patterns of Heavy Metals in Soil under Different Geological Structures and Land Uses for Assessing Metal Enrichments. Environ. Monit. Assess. 2013, 185, 9871–9888. [Google Scholar] [CrossRef]
  47. Cohen, J. A Coefficient of Agreement for Nominal Scales. Educ. Psychol. Meas. 1960, 20, 37–46. [Google Scholar] [CrossRef]
  48. Cheng, W.; Wang, K.; Zhang, X. A Com-Gis Based Decision Tree Model Inagricultural Application. In Computer and Computing Technologies in Agriculture II, Volume 1; Li, D., Zhao, C., Eds.; IFIP Advances in Information and Communication Technology; Springer: Boston, MA, USA, 2009; Volume 293, pp. 455–464. [Google Scholar]
  49. Agyeman, P.C.; John, K.; Kebonye, N.M.; Boruvka, L.; Vasat, R.; Drabek, O. A Geostatistical Approach to Estimating Source Apportionment in Urban and Peri-Urban Soils Using the Czech Republic as an Example. Sci. Rep. 2021, 11, 23615. [Google Scholar] [CrossRef] [PubMed]
  50. Huang, G.; Wang, X.; Chen, D.; Wang, Y.; Zhu, S.; Zhang, T.; Liao, L.; Tian, Z.; Wei, N. A hybrid data-driven framework for diagnosing contributing factors for soil heavy metal contaminations using machine learning and spatial clustering analysis. J. Hazard. Mater. 2022, 437, 129324. [Google Scholar] [CrossRef]
  51. Molla, A.; Zhang, W.; Zuo, S.; Ren, Y.; Han, J. A Machine Learning and Geostatistical Hybrid Method to Improve Spatial Prediction Accuracy of Soil Potentially Toxic Elements. Stochastic Environ. Res. Risk Assess. 2022, 37, 681–696. [Google Scholar] [CrossRef]
  52. Mohamed, E.S.; Jalhoum, M.E.M.; Belal, A.A.; Hendawy, E.; Azab, Y.F.A.; Kucher, D.E.; Shokr, M.S.; El Behairy, R.A.; El Arwash, H.M. A Novel Approach for Predicting Heavy Metal Contamination Based on Adaptive Neuro-Fuzzy Inference System and GIS in an Arid Ecosystem. Agronomy 2023, 13, 1873. [Google Scholar] [CrossRef]
  53. Yin, G.; Chen, X.; Zhu, H.; Chen, Z.; Su, C.; He, Z.; Qiu, J.; Wang, T. A Novel Interpolation Method to Predict Soil Heavy Metals Based on a Genetic Algorithm and Neural Network Model. Sci. Total Environ. 2022, 825, 153948. [Google Scholar] [CrossRef]
  54. Abuzaid, A.S.; Jahin, H.S.; Shokr, M.S.; El Baroudy, A.A.; Mohamed, E.S.; Rebouh, N.Y.; Bassouny, M.A. A Novel Regional-Scale Assessment of Soil Metal Pollution in Arid Agroecosystems. Agronomy 2023, 13, 161. [Google Scholar] [CrossRef]
  55. Suh, J.; Lee, H.; Choi, Y. A Rapid, Accurate, and Efficient Method to Map Heavy Metal-Contaminated Soils of Abandoned Mine Sites Using Converted Portable XRF Data and GIS. Int. J. Environ. Res. Public Health 2016, 13, 1191. [Google Scholar] [CrossRef]
  56. Gao, B.; Stein, A.; Wang, J. A Two-Point Machine Learning Method for the Spatial Prediction of Soil Pollution. Int. J. Appl. Earth Obs. Geoinf. 2022, 108, 102742. [Google Scholar] [CrossRef]
  57. Ma, T.; Zhang, Y.; Hu, Q.; Han, M.; Li, X.; Zhang, Y.; Li, Z.; Shi, R. Accumulation Characteristics and Pollution Evaluation of Soil Heavy Metals in Different Land Use Types: Study on the Whole Region of Tianjin. Int. J. Environ. Res. Public Health 2022, 19, 10013. [Google Scholar] [CrossRef]
  58. Zhao, W.; Ma, J.; Liu, Q.; Dou, L.; Qu, Y.; Shi, H.; Sun, Y.; Chen, H.; Tian, Y.; Wu, F. Accurate Prediction of Soil Heavy Metal Pollution Using an Improved Machine Learning Method: A Case Study in the Pearl River Delta, China. Environ. Sci. Technol. 2023, 57, 17751–17761. [Google Scholar] [CrossRef]
  59. Zeng, W.; Wan, X.; Gu, G.; Lei, M.; Yang, J.; Chen, T. An interpolation method incorporating the pollution diffusion characteristics for soil heavy metals - taking a coke plant as an example. Sci. Total Environ. 2023, 857, 159698. [Google Scholar] [CrossRef]
  60. Yang, P.; Mao, R.; Shao, H.; Gao, Y. An Investigation on the Distribution of Eight Hazardous Heavy Metals in the Suburban Farmland of China. J. Hazard. Mater. 2009, 167, 1246–1251. [Google Scholar] [CrossRef]
  61. Alijagic, J.; Sajn, R. Application and Improvement of Soil Spatial Distribution Mapping Using Advanced Modeling Techniques. Geol. Croat. 2020, 73, 69–84. [Google Scholar] [CrossRef]
  62. Khalil, A.; Taha, Y.; Benzaazoua, M.; Hakkou, R. Applied Methodological Approach for the Assessment of Soil Contamination by Trace Elements around Abandoned Coal Mines-a Case Study of the Jerada Coal Mine, Morocco. Minerals 2023, 13, 181. [Google Scholar] [CrossRef]
  63. Manjarrez-Dominguez, C.B.; Prieto-Amparan, J.A.; Cecilia Valles-Aragon, M.; Del Rosario Delgado-Caballero, M.; Teresa Alarcon-Herrera, M.; Nevarez-Rodriguez, M.C.; Vazquez-Quintero, G.; Berzoza-Gaytan, C.A. Arsenic Distribution Assessment in a Residential Area Polluted with Mining Residues. Int. J. Environ. Res. Public Health 2019, 16, 375. [Google Scholar] [CrossRef]
  64. Khalil, A.; Hanich, L.; Bannari, A.; Zouhri, L.; Pourret, O.; Hakkou, R. Assessment of Soil Contamination around an Abandoned Mine in a Semi-Arid Environment Using Geochemistry and Geostatistics: Pre-Work of Geochemical Process Modeling with Numerical Models. J. Geochem. Explor. 2013, 125, 117–129. [Google Scholar] [CrossRef]
  65. Hammam, A.A.; Mohamed, W.S.; Sayed, S.E.-E.; Kucher, D.E.; Mohamed, E.S. Assessment of Soil Contamination Using GIS and Multi-Variate Analysis: A Case Study in El-Minia Governorate, Egypt. Agronomy 2022, 12, 1197. [Google Scholar] [CrossRef]
  66. Pecina, V.; Brtnicky, M.; Balkova, M.; Hegrova, J.; Buckova, M.; Baltazar, T.; Licbinsky, R.; Radziemska, M. Assessment of Soil Contamination with Potentially Toxic Elements and Soil Ecotoxicity of Botanical Garden in Brno, Czech Republic: Are Urban Botanical Gardens More Polluted than Urban Parks? Int. J. Environ. Res. Public Health 2021, 18, 7622. [Google Scholar] [CrossRef]
  67. He, J.; Yang, Y.; Christakos, G.; Liu, Y.; Yang, X. Assessment of Soil Heavy Metal Pollution Using Stochastic Site Indicators. Geoderma 2019, 337, 359–367. [Google Scholar] [CrossRef]
  68. Jaworska, H.; Klimek, J. Assessment of the Impact of a Motorway on Content andSpatial Distribution of Mercury in Adjacent Agricultural Soils. Minerals 2021, 11, 1221. [Google Scholar] [CrossRef]
  69. Dragović, R.; Gajić, B.; Dragović, S.; Đorđević, M.; Đorđević, M.; Mihailović, N.; Onjia, A. Assessment of the Impact of Geographical Factors on the Spatial Distribution of Heavy Metals in Soils around the Steel Production Facility in Smederevo (Serbia). J. Clean. Prod. 2014, 84, 551–562. [Google Scholar] [CrossRef]
  70. Cojocariu, L.; Bordean, D.M.; Borozan, A.B.; Nita, S.; Horablaga, A. Carduus Nutans L and the Effect on the Heavy Metals and Microenvironment Biota. Rev. Chim. 2017, 68, 1524–1527. [Google Scholar] [CrossRef]
  71. Wang, J.; Dong, C.; Sun, S.; Mu, L.; Zhang, N.; Bao, L. Characteristics and Correlation Analysis of the Spatial Distribution of Heavy Metals in Arable Soils with Different Soil-Forming Matrices. Sustainability 2024, 16, 10338. [Google Scholar] [CrossRef]
  72. Li, D.; Lu, Q.; Cai, L.; Chen, L.; Wang, H. Characteristics of Soil Heavy Metal Pollution and Health Risk Assessment in Urban Parks at a Megacity of Central China. Toxics 2023, 11, 257. [Google Scholar] [CrossRef]
  73. Zhang, J.; Peng, J.; Chen, X.; Shi, X.; Feng, Z.; Meng, Y.; Chen, W.; Liu, Y. Comparative Study on Different Interpolation Methods and Source Analysis of Soil Toxic Element Pollution in Cangxi County, Guangyuan City, China. Sustainability 2024, 16, 3545. [Google Scholar] [CrossRef]
  74. Qiao, P.; Lei, M.; Yang, S.; Yang, J.; Guo, G.; Zhou, X. Comparing Ordinary Kriging and Inverse Distance Weighting for Soil as Pollution in Beijing. Environ. Sci. Pollut. Res. 2018, 25, 15597–15608. [Google Scholar] [CrossRef]
  75. Ding, Q.; Wang, Y.; Zhuang, D. Comparison of the common spatial interpolation methods used to analyze potentially toxic elements surrounding mining regions. J. Environ. Manag. 2018, 212, 23–31. [Google Scholar] [CrossRef] [PubMed]
  76. Tian, H.; Zhang, C.; Qi, S.; Kong, X.; Yue, X. Concentration and Spatial Distribution of Potentially Toxic Elements in Surface Soil of a Peak-Cluster Depression, Babao Town, Yunnan Province, China. Int. J. Environ. Res. Public Health 2021, 18, 3122. [Google Scholar] [CrossRef]
  77. Du, H.; Wang, J.; Wang, Y.; Yao, Y.; Liu, X.; Zhou, Y. Contamination Characteristics, Source Analysis, and Spatial Prediction of Soil Heavy Metal Concentrations on the Qinghai-Tibet Plateau. J. Soils Sediments 2023, 23, 2202–2215. [Google Scholar] [CrossRef]
  78. Pan, C.; Yu, F.; Tao, X.; Guo, J.; Yu, Y. Contamination, Spatial Distribution and Source Analysis of Heavy Metals in Surface Soil of Anhui Chaohu Economic Development Zone, China. Sustainability 2020, 12, 8117. [Google Scholar] [CrossRef]
  79. Xu, S.; Tao, S. Coregionalization Analysis of Heavy Metals in the Surface Soil of Inner Mongolia. Sci. Total Environ. 2004, 320, 73–87. [Google Scholar] [CrossRef]
  80. Tan, M.; Zhang, X.; Luo, W.; Hao, M. Deep Learning Based Spatial Distribution Estimation of Soil Pb Using Multi-Phase Multispectral Remote Sensing Images in a Mining Area. Land 2023, 12, 1789. [Google Scholar] [CrossRef]
  81. Safarov, R.; Shomanova, Z.; Nossenko, Y.; Mussayev, Z.; Shomanova, A. Digital Visualization of Environmental Risk Indicators in the Territory of the Urban Industrial Zone. Sustainability 2024, 16, 5190. [Google Scholar] [CrossRef]
  82. Pan, Y.; Peng, H.; Xie, S.; Zeng, M.; Huang, C. Eight Elements in Soils from a Typical Light Industrial City, China: Spatial Distribution, Ecological Assessment, and the Source Apportionment. Int. J. Environ. Res. Public Health 2019, 16, 2591. [Google Scholar] [CrossRef] [PubMed]
  83. Hao, H.; Li, P.; Jiao, W.; Ge, D.; Hu, C.; Li, J.; Lv, Y.; Chen, W. Ensemble learning-based applied research on heavy metals prediction in a soil-rice system. Sci. Total Environ. 2023, 898, 165456. [Google Scholar] [CrossRef]
  84. Song, Y.; Kang, L.; Lin, F.; Sun, N.; Aizezi, A.; Yang, Z.; Wu, X. Estimating the Spatial Distribution of Soil Heavy Metals in Oil Mining Area Using Air Quality Data. Atmos. Environ. 2022, 287, 119274. [Google Scholar] [CrossRef]
  85. Shi, S.; Hou, M.; Gu, Z.; Jiang, C.; Zhang, W.; Hou, M.; Li, C.; Xi, Z. Estimation of Heavy Metal Content in Soil Based on Machine Learning Models. Land 2022, 11, 1037. [Google Scholar] [CrossRef]
  86. Khodoli Zangeneh, D.; Amanipoor, H.; Battaleb-Looie, S. Evaluation of Heavy Metal Contamination Using Cokriging Geostatistical Method (Case Study of Abteymour Oilfield in Southern Iran). Appl. Water Sci. 2023, 13, 200. [Google Scholar] [CrossRef]
  87. CoŞKun, M.; Steinnes, E.; Frontasyeva, M.V.; Sjobakk, T.E.; Demkina, S. Heavy Metal Pollution of Surface Soil in the Thrace Region, Turkey. Environ. Monit. Assess. 2006, 119, 545–556. [Google Scholar] [CrossRef]
  88. Delgado, A.; Bautista, F.; Gogichaishvili, A.; Luis Cortes, J.; Quintana, P.; Aguilar, D.; Cejudo, R. Identification of Areas Contaminated with Heavy Metals in the Urban Dust of the City of Mexico. Rev. Int. Contam. Ambient. 2019, 35, 81–100. [Google Scholar] [CrossRef]
  89. Golden, N.; Zhang, C.; Potito, A.P.; Gibson, P.J.; Bargary, N.; Morrison, L. Impact of Grass Cover on the Magnetic Susceptibility Measurements for Assessing Metal Contamination in Urban Topsoil. Environ. Res. 2017, 155, 294–306. [Google Scholar] [CrossRef] [PubMed]
  90. Li, F.; Li, X.; Hou, L.; Shao, A. Impact of the Coal Mining on the Spatial Distribution of Potentially Toxic Metals in Farmland Tillage Soil. Sci. Rep. 2018, 8, 14925. [Google Scholar] [CrossRef] [PubMed]
  91. Cheng, W.; Wang, K.; Zhang, X. Implementation of a COM-Based Decision-Tree Model with VBA in ArcGIS. Expert Syst. Appl. 2010, 37, 12–17. [Google Scholar] [CrossRef]
  92. Cao, J.; Li, C.; Wu, Q.; Qiao, J. Improved Mapping of Soil Heavy Metals Using a Vis-NIR Spectroscopy Index in an Agricultural Area of Eastern China. IEEE Access 2020, 8, 42584–42594. [Google Scholar] [CrossRef]
  93. Ju, L.; Guo, S.; Ruan, X.; Wang, Y. Improving the Mapping Accuracy of Soil Heavy Metals through an Adaptive Multi-Fidelity Interpolation Method. Environ. Pollut. 2023, 330, 121827. [Google Scholar] [CrossRef]
  94. Zha, Y.; Yang, Y. Innovative Graph Neural Network Approach for Predicting Soil Heavy Metal Pollution in the Pearl River Basin, China. Sci. Rep. 2024, 14, 16505. [Google Scholar] [CrossRef]
  95. Mirzaei, F.; Abbasi, Y.; Sohrabi, T.; Mirhashemi, S.H. Investigation of Heavy Metalloid Pollutants in the South of Tehran Using Kriging Method and HYDRUS Model. Geosci. Lett. 2022, 9, 27. [Google Scholar] [CrossRef]
  96. Gok, G.; Tulun, S.; Celebi, H. Mapping of Heavy Metal Pollution Density and Source Distribution of Campus Soil Using Geographical Information System. Sci. Rep. 2024, 14, 29918. [Google Scholar] [CrossRef] [PubMed]
  97. Wang, Z.; Nie, K. Measuring Spatial Distribution Characteristics of Heavy Metal Contaminations in a Network-Constrained Environment: A Case Study in River Network of Daye, China. Sustainability 2017, 9, 986. [Google Scholar] [CrossRef]
  98. Omondi, E. Modeling the Spatial Distribution of Soil Heavy Metals Using Random Forest Model—A Case Study of Nairobi and Thirirka Rivers’ Confluence. J. Geogr. Inf. Syst. 2020, 12, 597–619. [Google Scholar] [CrossRef]
  99. Nickel, S.; Hertel, A.; Pesch, R.; Schroeder, W.; Steinnes, E.; Uggerud, H.T. Modelling and Mapping Spatio-Temporal Trends of Heavy Metal Accumulation in Moss and Natural Surface Soil Monitored 1990-2010 throughout Norway by Multivariate Generalized Linear Models and Geostatistics. Atmos. Environ. 2014, 99, 85–93. [Google Scholar] [CrossRef]
  100. Li, Z.; Lu, W.; Huang, J. Monitoring, Diffusion and Source Speculation Model of Urban Soil Pollution. Processes 2020, 8, 1339. [Google Scholar] [CrossRef]
  101. Yotova, G.; Hristova, M.; Padareva, M.; Simeonov, V.; Dinev, N.; Tsakovski, S. Multivariate Exploratory Analysis of the Bulgarian Soil Quality Monitoring Network. Molecules 2023, 28, 6091. [Google Scholar] [CrossRef]
  102. Yang, P.; Yang, M.; Mao, R.; Shao, H. Multivariate-Statistical Assessment of Heavy Metals for Agricultural Soils in Northern China. Sci. World J. 2014, 2014, 517020. [Google Scholar] [CrossRef]
  103. Zhou, J.; Feng, K.; Pei, Z.; Lu, M. Pollution Assessment and Spatial Variation of Soil Heavy Metals in Lixia River Region of Eastern China. J. Soils Sediments 2016, 16, 748–755. [Google Scholar] [CrossRef]
  104. Xia, Q.; Zhang, J.; Chen, Y.; Ma, Q.; Peng, J.; Rong, G.; Tong, Z.; Liu, X. Pollution, Sources and Human Health Risk Assessment of Potentially Toxic Elements in Different Land Use Types under the Background of Industrial Cities. Sustainability 2020, 12, 2121. [Google Scholar] [CrossRef]
  105. Imo, D.; Dressel, H.; Byber, K.; Hitzke, C.; Bopp, M.; Maggi, M.; Bose-O’Reilly, S.; Held, L.; Muff, S. Predicted Mercury Soil Concentrations from a Kriging Approach for Improved Human Health Risk Assessment. Int. J. Environ. Res. Public Health 2018, 15, 1326. [Google Scholar] [CrossRef]
  106. Alijagic, J.; Sajn, R. Predicting the Spatial Distributions of Elements in Former Military Operation Area Using Linear and Nonlinear Methods across the Stavnja Valley, Bosnia and Herzegovina. Minerals 2020, 10, 120. [Google Scholar] [CrossRef]
  107. Keçeci, M.; Gökmen, F.; Usul, M.; Koca, C.; Uygur, V. Prediction of Cadmium Content Using Machine Learning Methods. Environ. Earth Sci. 2024, 83, 362. [Google Scholar] [CrossRef]
  108. Li, Q.; Wang, C.; Dai, T.; Shi, W.; Zhang, X.; Xiao, Y.; Song, W.; Li, B.; Wang, Y. Prediction of Soil Cadmium Distribution across a Typical Area of Chengdu Plain, China. Sci. Rep. 2017, 7, 7115. [Google Scholar] [CrossRef]
  109. Hou, Y.; Ding, W.; Xie, T.; Chen, W. Prediction of Soil Heavy Metal Contents in Urban Residential Areas and the Strength of Deep Learning: A Case Study of Beijing. Sci. Total Environ. 2024, 950, 175133. [Google Scholar] [CrossRef] [PubMed]
  110. Fu, P.; Yang, Y.; Zou, Y. Prediction of Soil Heavy Metal Distribution Using Geographically Weighted Regression Kriging. Bull. Environ. Contam. Toxicol. 2022, 108, 344–350. [Google Scholar] [CrossRef]
  111. He, P.; Li, Y.; Huo, T.; Meng, F.; Peng, C.; Bai, M. Priority Planting Area Planning for Cash Crops under Heavy Metal Pollution and Climate Change: A Case Study of Ligusticum Chuanxiong Hort. Front. Plant Sci. 2023, 14, 1080881. [Google Scholar] [CrossRef]
  112. Senoro, D.B.; Monjardin, C.E.F.; Fetalvero, E.G.; Benjamin, Z.E.C.; Gorospe, A.F.B.; de Jesus, K.L.M.; Ical, M.L.G.; Wong, J.P. Quantitative Assessment and Spatial Analysis of Metals and Metalloids in Soil Using the Geo-Accumulation Index in the Capital Town of Romblon Province, Philippines. Toxics 2022, 10, 633. [Google Scholar] [CrossRef]
  113. Zhao, K.; Zhang, L.; Dong, J.; Wu, J.; Ye, Z.; Zhao, W.; Ding, L.; Fu, W. Risk Assessment, Spatial Patterns and Source Apportionment of Soil Heavy Metals in a Typical Chinese Hickory Plantation Region of Southeastern China. Geoderma 2020, 360, 114011. [Google Scholar] [CrossRef]
  114. Chen, H.; Wu, W.; Cao, L.; Zhou, X.; Guo, R.; Nie, L.; Shang, W. Source Analysis and Contamination Assessment of Potentially Toxic Element in Soil of Small Watershed in Mountainous Area of Southern Henan, China. Int. J. Environ. Res. Public Health 2022, 19, 13324. [Google Scholar] [CrossRef]
  115. Su, H.; Hu, Y.; Wang, L.; Yu, H.; Li, B.; Liu, J. Source Apportionment and Geographic Distribution of Heavy Metals and as in Soils and Vegetables Using Kriging Interpolation and Positive Matrix Factorization Analysis. Int. J. Environ. Res. Public Health 2022, 19, 485. [Google Scholar] [CrossRef]
  116. Gan, T.; Zhao, H.; Ai, Y.; Zhang, S.; Wen, Y.; Tian, L.; Mipam, T.D. Spatial Distribution and Ecological Risk Assessment of Heavy Metals in Alpine Grasslands of the Zoige Basin, China. Front. Ecol. Evol. 2023, 11, 1093823. [Google Scholar] [CrossRef]
  117. Lv, Y.; Kabanda, G.; Chen, Y.; Wu, C.; Li, W. Spatial Distribution and Ecological Risk Assessment of Heavy Metals in Manganese (Mn) Contaminated Site. Front. Environ. Sci. 2022, 10, 942544. [Google Scholar] [CrossRef]
  118. Shaheen, A.; Iqbal, J. Spatial Distribution and Mobility Assessment of Carcinogenic Heavy Metals in Soil Profiles Using Geostatistics and Random Forest, Boruta Algorithm. Sustainability 2018, 10, 799. [Google Scholar] [CrossRef]
  119. Zhang, Q.; Yuan, L.; Lu, J.; Shi, Z.; Shen, F.; Pei, G.; Zeng, Y. Spatial Distribution and Pollution Assessment of Heavy Metals in Soils in the Jiefangzha Irrigation Area of the Hetao Irrigation District, China. Environ. Forensics 2024, 25, 431–441. [Google Scholar] [CrossRef]
  120. Xiao, T.; Muhammad, A.M.; Pan, X.; Wang, Y.; Guangming, H.; Tang, Z. Spatial Distribution and Risk Identification of Arsenic Contamination in Water and Soil through GIS-Based Interpolation Techniques in Jianghan Plain, Central China. Front. Environ. Sci. 2022, 10, 1001862. [Google Scholar] [CrossRef]
  121. Cheng, Y.; Li, P.; Xu, G.; Lu, K.; Wang, F.; Zhang, T.; Feng, Z. Spatial Distribution Characteristics and Pollution Evaluation of Soil Iron in the Middle Hanjiang River. Int. J. Environ. Res. Public Health 2019, 16, 4075. [Google Scholar] [CrossRef] [PubMed]
  122. Chen, G.; Yang, Y.; Liu, X.; Wang, M. Spatial Distribution Characteristics of Heavy Metals in Surface Soil of Xilinguole Coal Mining Area Based on Semivariogram. ISPRS Int. J. Geo-Inf. 2021, 10, 290. [Google Scholar] [CrossRef]
  123. Tedoldi, D.; Chebbo, G.; Pierlot, D.; Branchu, P.; Kovacs, Y.; Gromaire, M.-C. Spatial Distribution of Heavy Metals in the Surface Soil of Source-Control Stormwater Infiltration Devices-Inter-Site Comparison. Sci. Total Environ. 2017, 579, 881–892. [Google Scholar] [CrossRef]
  124. Solt, M.J.; Deocampo, D.M.; Norris, M. Spatial Distribution of Lead in Sacramento, California, USA. Int. J. Environ. Res. Public Health 2015, 12, 3174–3187. [Google Scholar] [CrossRef]
  125. Gan, W.; Zhang, Y.; Xu, J.; Yang, R.; Xiao, A.; Hu, X. Spatial Distribution of Soil Heavy Metal Concentrations in Road-Neighboring Areas Using UAV-Based Hyperspectral Remote Sensing and GIS Technology. Sustainability 2023, 15, 10043. [Google Scholar] [CrossRef]
  126. Nie, S.; Chen, H.; Sun, X.; An, Y. Spatial Distribution Prediction of Soil Heavy Metals Based on Random Forest Model. Sustainability 2024, 16, 4358. [Google Scholar] [CrossRef]
  127. Sun, Y.; Lei, S.; Zhao, Y.; Wei, C.; Yang, X.; Han, X.; Li, Y.; Xia, J.; Cai, Z. Spatial Distribution Prediction of Soil Heavy Metals Based on Sparse Sampling and Multi-Source Environmental Data. J. Hazard. Mater. 2024, 465, 133114. [Google Scholar] [CrossRef]
  128. Chen, W.; Cai, Y.; Zhu, K.; Wei, J.; Lu, Y. Spatial Heterogeneity Analysis and Source Identification of Heavy Metals in Soil: A Case Study of Chongqing, Southwest China. Chem. Biol. Technol. Agric. 2022, 9, 50. [Google Scholar] [CrossRef]
  129. Yang, H.; Hou, J.W.; Fu, J.M. Spatial Identification of Heavy Metal Pollution Sources in the Zhuxianzhuang Mining Region, China. Appl. Ecol. Environ. Res. 2023, 21, 5247–5264. [Google Scholar] [CrossRef]
  130. Chen, H.; An, J.; Wei, S.; Gu, J. Spatial Patterns and Risk Assessment of Heavy Metals in Soils in a Resource-Exhausted City, Northeast China. PLoS ONE 2015, 10, e0137694. [Google Scholar] [CrossRef] [PubMed]
  131. Wu, D.; Liu, H.; Gao, X.; Li, B.; Nyasha, N.K. Spatial Variability and Hotspot Prediction of Heavy Metal Pollution in Soil-Rice System: Based on a New Geostatistical Method. Int. J. Environ. Anal. Chem. 2024, 104, 8293–8315. [Google Scholar] [CrossRef]
  132. Sheng, J.; Yu, P.; Zhang, H.; Wang, Z. Spatial Variability of Soil Cd Content Based on IDW and RBF in Fujiang River, Mianyang, China. J. Soils Sediments 2021, 21, 419–429. [Google Scholar] [CrossRef]
  133. Wu, Z.; Xia, T.; Jia, X. Spatial Variation and Pollution Risk Assessment of Heavy Metals in Industrial Soil Based on Geochemical Data and GIS-a Case of an Iron and Steel Plant in Beijing? China. Int. J. Environ. Anal. Chem. 2024, 104, 6826–6847. [Google Scholar] [CrossRef]
  134. Spatio-Temporal Analysis of Heavy Metals in Arid Soils at the Catchment Scale Using Digital Soil Assessment and a Random Forest Model. Remote Sens. 2021, 13, 1698. [CrossRef]
  135. Zhang, B.; Yang, Y. Spatiotemporal Modeling and Prediction of Soil Heavy Metals Based on Spatiotemporal Cokriging. Sci. Rep. 2017, 7, 16750. [Google Scholar] [CrossRef]
  136. Barkat, A.; Bouaicha, F.; Ziad, S.; Mester, T.; Sajtos, Z.; Balla, D.; Makhloufi, I.; Szabó, G. The Integrated Use of Heavy-Metal Pollution Indices and the Assessment of Metallic Health Risks in the Phreatic Groundwater Aquifer—The Case of the Oued Souf Valley in Algeria. Hydrology 2023, 10, 201. [Google Scholar] [CrossRef]
  137. Zheng, X.; Yin, G.; Li, D. The Spatial Distribution Of Cu in Soil Around Copper Tailings Based On Kriging Interpolation. Res. J. Chem. Environ. 2011, 15, 462–465. [Google Scholar]
  138. Li, H.; Fu, P.; Yang, Y.; Yang, X.; Gao, H.; Li, K. Exploring Spatial Distributions of Increments in Soil Heavy Metals and Their Relationships with Environmental Factors Using GWR. Stochastic Environ. Res. Risk Assess. 2021, 35, 2173–2186. [Google Scholar] [CrossRef]
  139. Peng, Y.; Zhao, Y.; Chen, J.; Xie, E.; Yan, G.; Zou, T.; Xu, X. Simultaneously Mapping the 3D Distributions of Multiple Heavy Metals in an Industrial Site Using Deep Learning and Multisource Auxiliary Data. J. Hazard. Mater. 2024, 480, 136000. [Google Scholar] [CrossRef]
  140. Han, Z.; Wang, J.; Liao, X.; Yang, J. Accurate prediction of spatial distribution of soil heavy metal in complex mining terrain using an improved machine learning method. J. Hazard. Mater. 2025, 491, 137994. [Google Scholar] [CrossRef]
  141. Zhu, H.; Xu, Q.; Zheng, Y.; Cui, J.; Meng, Q. Ecological Health Assessment of the Middle Route of the South-to-North Water Diversion Project Using an Enhanced VORS Model. Ecol. Indic. 2025, 172, 113281. [Google Scholar] [CrossRef]
Sustainability 18 01893 g001
Figure 1. PRISMA 2020 flow diagram for study selection. Duplicates were removed; machine-assisted prescreening was human-verified; reasons for full-text exclusion are listed in the Eligibility box.
Figure 1. PRISMA 2020 flow diagram for study selection. Duplicates were removed; machine-assisted prescreening was human-verified; reasons for full-text exclusion are listed in the Eligibility box.
Sustainability 18 01893 g001
Sustainability 18 01893 g002
Figure 2. Spatial Interpolation of Soil Heavy Metals: Publication Trends. (a) Annual publications on soil heavy metals from 2000 to 2024 (bars; left y-axis) with cumulative publications (line; right y-axis). (b) Publication counts by heavy metal element (Pb, Zn, Cu, Cd, Cr, Ni, As, Hg, Mn, Co, Fe, Ba, V, Sr, Sb, Tl, Se, Mo).
Figure 2. Spatial Interpolation of Soil Heavy Metals: Publication Trends. (a) Annual publications on soil heavy metals from 2000 to 2024 (bars; left y-axis) with cumulative publications (line; right y-axis). (b) Publication counts by heavy metal element (Pb, Zn, Cu, Cd, Cr, Ni, As, Hg, Mn, Co, Fe, Ba, V, Sr, Sb, Tl, Se, Mo).
Sustainability 18 01893 g002
Sustainability 18 01893 g003
Figure 3. Selection Principles of Spatial Interpolation Methods for Soil Heavy Metal Analysis.
Figure 3. Selection Principles of Spatial Interpolation Methods for Soil Heavy Metal Analysis.
Sustainability 18 01893 g003
Table 1. Distribution of Main Journals.
Table 1. Distribution of Main Journals.
No.JournalArticles (n)Share (%)
1International Journal of Environmental Research and Public Health2518.52
2Science of the Total Environment85.93
3Scientific Reports85.93
4Sustainability64.44
5Frontiers (series)53.70
6Journal of Hazardous Materials42.96
7Agronomy32.22
8Environmental Monitoring and Assessment32.22
9Environmental Pollution32.22
10Environmental Research32.22
11Minerals32.22
12Toxics32.22
13Geoderma21.48
14IEEE Access21.48
15Journal of Soils and Sediments21.48
16Land21.48
17Remote Sensing21.48
18Other journals (50+ titles)5742.22
Total135100
Note: Data are based on 135 English-language studies; percentages are shown to two decimal places. The Frontiers series includes Frontiers in Environmental Science, Frontiers in Ecology and Evolution, Frontiers in Plant Science, etc.
Table 2. Literature Counts, Top-3 Methods, and Top-3 Dominant Drivers by Heavy Metal.
Table 2. Literature Counts, Top-3 Methods, and Top-3 Dominant Drivers by Heavy Metal.
Heavy MetalCountMethods Dominant Drivers
Pb105OK; RF; IDWTraffic emissions; Industrial/smelting; Atmospheric deposition
Cd103OK; RF; IDWAgricultural inputs; Industrial emissions; Topography/hydrography
Cu100OK/RK; RF/RFOK; IDW/RBFIndustrial processing; Traffic; Agricultural Cu agents
Zn101OK; RF; IDWIndustrial electroplating/galvanizing; Traffic; Soil properties pH/SOM
As92OK/COK/EBK; RF; IDWMining/industry and deposition; Parent material/lithology; Agricultural inputs/irrigation
Cr99OK; RF; RK/RFOKParent material/geologic background; Industrial electroplating/stainless steel; Traffic dust
Ni90OK/FK; RF; COKParent material/lithology; Industrial fallout; Precipitation/topography
Hg76OK; COK/EBK; RF/GTWRAtmospheric deposition; Industrial point sources; Soil pH/SOM
Mn50OK; IDW; COKParent material and topographic eluviation; Mining/tailings aeolian transport; Traffic abrasion
Co24OK; PCA+Kriging; IDWParent material/lithology; Organic fertilizers/agricultural inputs; Urban–industrial mixed sources
V22OK/COK; RF; IDWPetroleum/fuel combustion and industry; Parent material volcanic–mafic; Irrigation salinity/agriculture
Fe35OK; IDW; COKParent material/depositional topography; Irrigation water quality; Mining/industry
Ba16OK; IK; PCA+KrigingTraffic/industrial dust; Mining/processing; Parent material
Mo14OK/IDW; PCA/PMF + mapping; COKMining/smelting; Coal-combustion deposition; Parent material/tectonic belts
Sb12OK/IDW; IK/probability; RFIndustry/incineration/traffic; Parent material; Urbanization intensity
Se10OK/IDW; COK; SVM/RFParent material; Agricultural management; Topography/drainage
Sr12OK/IDW; COK; PCA + mappingParent material/carbonates; Drilling fluids/irrigation water; Livestock manure
Ti9COK; OK/IDW; RFPetroleum activities/drilling additives; Parent material sandstone; Traffic dust
Li9COK; OK/IDW; PCA + mappingFertilizers/livestock wastewater; Petroleum-associated; Parent material
S9COK; OK/IDW; RFFertilizers/drilling wastewater; Crude oil sulfur content; Agriculture–industry superposition
Ag11OK/IDW; PMF + mapping; RFMining/beneficiation; Traffic/industrial dust; Parent material
Table 3. Methods for Heavy-Metal Spatial Prediction.
Table 3. Methods for Heavy-Metal Spatial Prediction.
MethodCountScales (C/R/M)Common Elements
Ordinary Kriging (OK)85C 32R 41M 12PbCuZn
Universal Kriging (UK)12C 4R 6M 2PbCdZn
Co-Kriging (CK)18C 7R 9M 2CdZnPb
Indicator Kriging (IK)10C 4R 5M 1PbCdCr
Disjunctive Kriging (DK)6C 2R 3M 1CuPbCr
Empirical Bayesian Kriging (EBK)8C 2R 5M 1CdZnPb
Inverse Distance Weighting (IDW)65C 27R 33M 5CuPbZn
Radial Basis Function (RBF)15C 6R 9M 0CuZnPb
Spline/Thin-Plate Spline (TPS)10C 4R 6M 0CuPbCr
Random Forest (RF)28C 9R 17M 2CdZnPb
Support Vector Machine (SVM; incl. SVMOK)12C 4R 7M 1CdPbZn
Gradient Boosting (GBDT/XGBoost/LightGBM)15C 5R 9M 1CdZnPb
Neural Networks (ANN/BPNN/CNN/LSTM, etc.)20C 6R 12M 2CdPbZn
Regression Kriging (RK)22C 6R 14M 2CdZnPb
Random-Forest Regression Kriging (RFRK)10C 3R 6M 1CdZnPb
Geographically Weighted Regression Kriging (GWRK)8C 2R 6M 0CdPbZn
Spatio-Temporal Kriging/Co-Kriging (STK/STCK)7C 0R 7M 0CdZnPb
Kriging Neural Networks (e.g., Kriging-BPNN)6C 2R 4M 0CuPbZn
Bayesian Maximum Entropy (BME)4C 1R 3M 0CdPbZn
Ensemble Learning (e.g., Stacking, Bagging)5C 1R 3M 1CdZnPb
Deep Learning Geostatistics (e.g., CNN-Kriging, LSTM-RK)8C 2R 5M 1CdZnPb
High-Accuracy Surface Modeling (HASM)3C 1R 2M 0CuPbCr
Multi-Method Comparison Studies (separately counted)30C 13R 15M 2CuPbZn
Scale Legend: C = County scale (Local); R = Regional scale (Provincial/Basin); M = Macro scale (National/Continental). Note: Counts reflect the number of distinct studies utilizing each method. Common elements list the top 3 most frequently modeled heavy metals for that method.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Yang, L.; Yue, T.; Ma, M. Soil Heavy Metals for Sustainable Risk Management: A Systematic Review and a Context-Aware Method Selection Framework. Sustainability 2026, 18, 1893. https://doi.org/10.3390/su18041893

AMA Style

Yang L, Yue T, Ma M. Soil Heavy Metals for Sustainable Risk Management: A Systematic Review and a Context-Aware Method Selection Framework. Sustainability. 2026; 18(4):1893. https://doi.org/10.3390/su18041893

Chicago/Turabian Style

Yang, Leqi, Tianxiang Yue, and Maohua Ma. 2026. "Soil Heavy Metals for Sustainable Risk Management: A Systematic Review and a Context-Aware Method Selection Framework" Sustainability 18, no. 4: 1893. https://doi.org/10.3390/su18041893

APA Style

Yang, L., Yue, T., & Ma, M. (2026). Soil Heavy Metals for Sustainable Risk Management: A Systematic Review and a Context-Aware Method Selection Framework. Sustainability, 18(4), 1893. https://doi.org/10.3390/su18041893

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

Article Metrics

Back to TopTop