Integrated assessment and forecasting of heavy metal contamination in soils affected by Pb–Zn mining in Guizhou, China

Abstract

Heavy metal contamination in soils surrounding Pb–Zn mining areas poses persistent ecological and public health risks in Guizhou Province, China. However, ecological risk transfer among ecosystem compartments and the future decline in soil environmental capacity remain poorly understood. This study investigated heavy metals in surface soil (0–20 cm) and subsoil (20–60 cm) around a Pb–Zn mining area in Shuicheng District, Liupanshui City, by integrating the geo-accumulation index (I_geo), pollution load index (PLI), potential ecological risk index (PERI), network environ analysis (NEA), Monte Carlo health-risk assessment, and dynamic environmental-capacity forecasting. This framework characterized contamination levels and human-health risks, traced the redistribution of ecological risk among trophic components, and evaluated the future decline of residual soil buffering capacity under continued pollution pressure. The results revealed severe contamination by Pb, Cd, and Zn in both surface and subsurface soils, with Pb showing the highest pollution intensity. PERI classified most sampling sites as moderate to high ecological risk. The NEA model further showed that Pb, Zn, and Cd generated the highest ecological pressures across trophic levels and that soil microorganisms were the major ecological risk sink, accounting for 66.4% of integrated ecosystem risk, followed by vegetation, herbivores, and carnivores. Health risk assessment indicated markedly higher non-carcinogenic and carcinogenic risks for children than adults, primarily due to Pb and As exposure. Dynamic capacity forecasting suggested a rapid reduction in the residual soil environmental capacity over the next 10 years, especially for Hg. These findings support remediation planning, land-use management, and exposure reduction in mining-affected areas.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-026-27490-2.

Introduction

Lead–zinc mining is one of the major anthropogenic sources of soil heavy metal contamination worldwide [40, 41]. In mining-affected landscapes, ore extraction, crushing, flotation, smelting, and tailings storage release Pb, Zn, Cd, As, and Hg to surrounding soils through atmospheric deposition, runoff, seepage, and physical dispersion of tailings and waste rock [50]. In karst regions such as Guizhou, carbonate-rich parent materials can partly buffer soil acidity, but they can also influence metal transport by controlling Cd mobility, sorption–desorption equilibria, and the redistribution of dissolved and particulate metals from tailings to nearby farmland and semi-natural land [12]. As a consequence, mining-derived metals may show contrasting behaviors across soil profiles: surface enrichment often reflects recent deposition, whereas elevated subsoil concentrations may reflect downward migration, historical accumulation, or lithogenic inheritance [5]. Because Pb and Cd are highly toxic and persistent, and Zn commonly co-occurs with them in Pb–Zn ore bodies, understanding their coupled distribution is essential for ecological risk control and human health protection [16].

Recent studies have widely used pollution indices, multivariate statistics, and health risk models to evaluate heavy metal contamination in mining areas. These approaches are valuable for identifying contamination levels and likely sources, but they are often applied as static screening tools [8]. In particular, conventional index-based assessments do not explicitly describe how risks are redistributed among ecosystem compartments, and most studies do not evaluate how the remaining soil environmental capacity may decline under continued pollution pressure [4]. Network environ analysis (NEA) can be used to trace direct and indirect ecological-risk transmission beyond concentration-based screening [25]. Dynamic environmental-capacity forecasting can further indicate whether remaining soil buffering potential may decline under continued pollution pressure [49]. This study combines established methods within a single Pb–Zn mining system to provide an integrated assessment of current contamination, ecological-risk transmission, health risk, and future buffering decline.

Shuicheng District contains an important Pb–Zn mining area in Guizhou Province and has experienced long-term mineral exploitation and associated waste accumulation [53]. However, the vertical distribution of heavy metals, the relative importance of ecological transfer pathways, and the future decline in soil environmental capacity have not been evaluated together for this area. To address these gaps, this study examined surface and subsoil samples collected along the production area-tailings storage facility gradient and tested three working hypotheses: (1) Pb, Zn, and Cd are the dominant anthropogenic contaminants and are enriched mainly in surface soils; (2) ecological risk is concentrated in lower trophic components and subsequently redistributed through the food web; and (3) soil environmental capacity for the main pollutants is already depleted or will decline rapidly under a screening-level forecast scenario. The specific objectives were to (1) combine Igeo, PLI, and PERI to evaluate pollution status; (2) use NEA to assess direct and indirect ecological risk transmission; (3) apply Monte Carlo simulation for probabilistic health risk assessment; and (4) forecast dynamic environmental capacity as a scenario-based indicator of future buffering decline.

Materials and methods

Study area and sample collection

The study area encompasses soils surrounding a Pb–Zn mining enterprise in Shuicheng District, Liupanshui City, Guizhou Province. Twenty sampling sites were distributed across the ore-processing zone, tailings-affected area, and surrounding impacted land to represent the principal exposure settings and land-use conditions within the mining system (Fig. 1; Table S1). According to the coordinates listed in Table S1, the sampling network spans approximately 1.76 km from the near-source zone to the most distant impacted site, with a mean nearest-neighbor spacing of about 73 m. The layout was designed to compare soil contamination patterns across the mining–agricultural interface under practical field conditions. Each site provided paired samples from two depths. Surface soil (0–20 cm) was used to characterize recent atmospheric deposition and shallow runoff accumulation, whereas subsoil (20–60 cm) was used to explore deeper storage that may reflect historical inputs, downward migration, or lithogenic background contributions. At each site, plant residues and dead branches were removed before sampling. Using stainless steel shovels and soil drills, approximately 2 kg of mixed soil was collected, placed in clean cloth bags, and transported to the laboratory. Samples were cleaned to remove impurities, then air-dried, ground, sieved (200 mesh), and stored for analysis.

Fig. 1.

Sampling location map

Sample testing and quality control

Soil samples were subjected to microwave digestion using ultra-pure HNO₃-HF-HClO₄ [30]. Concentrations of Cd, Cr, Cu, Ni, Pb, Zn, As, and Hg were determined using inductively coupled plasma mass spectrometry (ICP-MS, Thermo Fisher Scientific, USA) and atomic fluorescence spectrometry (AFS-8220, Beijing Jitian Instrument Co., China). Soil pH was measured using a calibrated pH meter (SevenExcellence S400, Mettler Toledo, Switzerland). Two procedural blanks and three national soil reference materials (GSS-3a, GSS-1a, and GSS-5a) were analyzed together with the samples. The recoveries of standard sample residues were 86.5–105%, and the relative standard deviation was less than 3%, indicating acceptable analytical precision. The detection limits and methods used for elements are listed in Table S2.

Evaluation methodology

Geo-accumulation index (I_geo) and pollution load index (PLI)

The geo-accumulation index (I_geo), proposed by Muller in 1969 [34], quantifies pollution by comparing measured concentrations (C_i) with background values (C_ib), adjusted by a factor (K = 1.5) [27] to account for natural variability:

1

Heavy metal background values were obtained from regional soil data for Liupanshui City [21], as detailed in Table 1. Pollution severity is classified into seven levels (Table S3).

Table 1.

Descriptive analysis of soil elements

Elements/Indicator		pH	Cd (mg/kg)	Cu (mg/kg)	Pb (mg/kg)	Cr (mg/kg)	Zn (mg/kg)	Ni (mg/kg)	As (mg/kg)	Hg (mg/kg)
Maximum	0–20 cm	7.94	7.74	459	1260	93.7	3490	44	140	1.89
	20–60 cm	7.98	9.11	159	1350	90.4	4740	38.5	136	2.21
Minimum	0–20 cm	5.02	0.827	11.6	68.2	34.9	177	7.1	5.28	0.0694
	20–60 cm	5.05	0.416	12.8	43.1	25.1	132	4.85	4.36	0.0516
Median	0–20 cm	7.26	1.49	27.8	241	73.7	450	22.7	11.6	0.110
	20–60 cm	7.21	1.17	31.1	164	79.1	279	22.5	11.6	0.127
Mean	0–20 cm	6.99	2.30	62.8	381	72.3	796	23.6	20.5	0.300
	20–60 cm	6.96	2.12	39.4	334	74.0	913	22.9	21.1	0.385
SD	0–20 cm	0.720	1.94	98.8	346	13.1	957	7.71	29.4	0.451
	20–60 cm	0.910	2.19	30.5	375	14.8	1367	7.01	30.3	0.658
CV	0–20 cm	0.103	0.844	1.57	0.908	0.183	1.20	0.327	1.43	1.50
	20–60 cm	0.131	1.034	0.774	1.12	0.201	1.50	0.307	1.44	1.71
Cib	-	-	0.398	48.41	17.67	112.2	119.7	40.38	16.3	0.109
Cic	-	-	0.3	100	120	200	250	100	25	0.6

To assess the overall contamination level of multiple metals, the pollution load index (PLI) was used [45].

2

3

where CF_i represents the pollution coefficient of heavy metal i, and n denotes the number of heavy metal elements (in this study, n = 8). PLI classification is detailed in Table S3.

Ecological risk

Potential ecological risk index (PERI)

The potential ecological risk index (PERI), introduced by Hakanson in 1980 [14], integrates metal concentration and toxicity to evaluate ecological hazards:

4

5

where E_i is the ecological risk for metal i, and T_i is its toxicity coefficient: Cd (30), Cu (5), Pb (5), Cr (2), Zn (1), Ni (5), As (10), and Hg (40) [6, 26]. Risk classification is provided in Table S3.

Network environmental analysis (NEA)

The NEA model was used to quantify both direct and indirect ecological risk transmission within the ecosystem, thereby extending the interpretation beyond concentration-based screening. In this study, the NEA was implemented as a deterministic framework. Initial risk inputs (R_i) were derived from risk intensity, occurrence probability, and receptor sensitivity [9]. Receptor sensitivity values and the energy-flow/control-allocation structure were obtained from field observations and literature sources. Therefore, the NEA results should be interpreted as inferred ecological-risk pathways rather than fully uncertainty-resolved estimates.

6

where RI_x denotes the risk intensity resulting from a specific change in risk factor x, P_x represents the probability of occurrence of such a change, and S_i indicates the sensitivity of community i. Specifically, these three parameters are calculated using the following formula.

7

8

9

In these equations, I_tx denotes the concentration of pollutant i in soil after pollution at time t, and I_ox represents the corresponding background reference value in an unaffected ecosystem with similar habitat conditions. P_ex is the probability of occurrence of pollution event x, calculated as the proportion of the probability of a single pollutant event relative to the total probability of all pollutant events. P_ox denotes the operational probability of pollutants under a given land-use condition, reflecting the actual presence of ecotoxicants. N_x is the number of susceptible components involved in the risk process, that is, the number of risk receptors in the ecosystem.

Network control analysis is the core component of the NEA model and differs fundamentally from conventional ecological risk assessment approaches [39]. It quantifies the influence of each ecological component by calculating control allocation (CA) among community components [37]. In this study, estimation of CA was based on food-web structure, field investigation, and literature evidence describing energy exchange among vegetation, herbivores, carnivores, and soil microorganisms [25]. The complete energy-flow matrix used in the analysis is shown in Fig. 7 and Table S4.

10

11

12

13

14

where f_ij denotes the energy information flow from community j to community i; T_i (or T_j) represents the total energy information inputs and outputs of communities i (or j); and ca_ij indicates the level of control exerted by community j over community i. The ecological risks induced by pollutants will propagate and accumulate in accordance with the outcomes dictated by CA.

Fig. 7.

Risk transport chain: (a) energy-flow network and flow matrix (kJ·ft⁻²·yr⁻¹); (b) control-allocation digraph and matrix

By combining the initial pollutant risk with the CA model, energy-information flows were further transformed into ecological risk flows. In this process, each community acts both as a receptor of risk from other components and as a transmitter of risk to the rest of the system. After pollutant disturbance, communities first respond to external risk and then propagate that risk cumulatively through the ecosystem. The risk-flow calculation is given as follows:

15

Here, rf_ji and rf_kj represent the risk allocation from i to j and from j to k, respectively, while ca_ji and ca_kj denote the control exerted by i on j and by j on k, respectively.

This study models internal risk flows as elements in the direct risk network (RF), whereas the integral risk network (RN) captures the complete risk process by incorporating all direct (path length = 1) and indirect (path length > 1) pathways throughout the ecosystem [25].

16

Probabilistic health risk assessment via Monte Carlo simulation

Health risks were evaluated using the US EPA exposure model, considering ingestion, inhalation, and dermal absorption. Separate assessments were performed for adults and children. Average daily doses (ADD) for each pathway are calculated as [31]:

17

18

19

In the above equations, ADD_ing, ADD_inh, and ADD_der represent the average daily exposure doses through ingestion, inhalation, and dermal contact, respectively. The remaining parameter values were derived from the Exposure Factors Handbook of Chinese Population and related studies [7, 17, 24]. Detailed parameter definitions are listed in Table S5.

Non-carcinogenic risk was assessed for eight heavy metals (Cd, Cu, Pb, Cr, Zn, Ni, As, and Hg), whereas carcinogenic risk was evaluated for five metals (Cd, Pb, Cr, Ni, and As). The total non-carcinogenic risk index (HI) was obtained by summing the hazard quotients (HQ_i) of individual metals, and the total carcinogenic risk index (CRI) was calculated from the corresponding carcinogenic risk values (CR_i) [19]. The formulas are shown below:

20

21

where ADD_ij represents the average daily exposure to heavy metal i through exposure pathway j. RfD_ij and SF_ij denote the reference dose and carcinogenic slope factor for heavy metal element i via exposure pathway j, respectively. The corresponding values are presented in Table S6 [57, 60]. Given the variability in quantitative Pb carcinogenic assessment across regulatory frameworks, Pb cancer-risk values in this study are presented as screening-level estimates.

An HQ_i or HI value of ≤ 1 indicates no significant non-carcinogenic risk, whereas a value > 1 suggests potential concern. For carcinogenic risk, CR_i or CRI values of ≤ 1 × 10⁻⁶ are considered negligible, values between 1 × 10⁻⁶ and 1 × 10⁻⁴ are considered acceptable, and values > 1 × 10⁻⁴ indicate unacceptable risk [11].

Due to the variability in sources of heavy metal pollution, contaminated areas, and exposure parameters, health risk assessments exhibit significant uncertainty and variability. To reduce uncertainty in health-risk evaluation, this study employed Monte Carlo simulation with 10,000 iterations [58, 59].

Soil environmental capacity

Static and residual capacity

The static environmental capacity (Q_s) represents the maximum allowable pollutant load without exceeding quality thresholds [36]. The residual soil environmental capacity (Q_i) reflects the remaining buffering potential [44], calculated as:

22

23

where M represents the weight of soil in the tillage layer, which is typically assumed to be 2.25 × 10⁶ kg/hm². This standard value is widely utilized in numerous studies for calculating environmental capacity [15, 18]. The variables C_ib and C_i maintain the same definitions as outlined in section " Geo-accumulation index (Igeo) and pollution load index (PLI)". C_ic represents the reference risk threshold for heavy metals in soil, with values derived from the Chinese Agricultural Land Soil Pollution Control Standards, as detailed in Table 1.

Dynamic forecast environmental capacity

The dynamic forecast environmental capacity (Q_n) represents the soil’s ability to retain and process pollutants over time. This concept incorporates pollutant migration and transformation, as well as the environmental capacity and self-purification properties of soil [28, 29]. The calculation formula is as follows:

24

where n represents the forecast period in years, and K denotes the annual average residual rate of pollutants. In this study, K was set to 0.9 as a screening-level scenario parameter following previous environmental-capacity studies [51]. The same K value was applied to all metals to facilitate comparison of relative capacity decline. However, this simplification does not capture metal-specific geochemical behavior.

Statistical analysis

All data processing was performed using SPSS 24.0. PCA and visualizations were generated using Origin 2019. Monte Carlo simulations were conducted in Oracle Crystal Ball 11.1. Geospatial analysis was carried out using ArcGIS 10.8.1 to map sampling sites.

Results and discussion

Distribution characteristics of soil elements

This study performed a statistical analysis of nine testing indicators for soil samples, with the results summarized in Table 1. The average pH values for the surface soil and subsoil were 6.99 and 6.96, respectively, indicating that the soil is generally neutral. In the surface soil, the average concentrations of Cd, Cu, Pb, Cr, Zn, Ni, As, and Hg were 2.30, 62.8, 381, 72.3, 796, 23.6, 20.5, and 0.300 mg/kg, respectively. Except for Cr and Ni, whose average concentrations were below the local soil background levels, the average concentrations of Cd, Cu, Pb, Zn, As, and Hg were 5.78, 1.31, 21.6, 6.65, 1.26, and 2.75 times higher than the local soil background values. According to China's "Soil Pollution Risk Control Standards for Agricultural Land," only the average concentrations of Cd, Pb, and Zn exceeded the pollution risk thresholds, which aligns with findings from a study in typical lead–zinc mining areas in Guangxi, China [35].

In the subsoil, the average concentrations of Cd, Cu, Pb, Cr, Zn, Ni, As, and Hg were 2.12, 39.4, 334, 74.0, 913, 22.9, 21.1, and 0.385 mg/kg, respectively. Except for Cu, Cr, and Ni, the average concentrations of Cd, Pb, Zn, As, and Hg were 5.33, 18.9, 7.63, 1.29, and 3.53 times higher than the local soil background values. As with the surface soil, only the average concentrations of Cd, Pb, and Zn exceeded the pollution risk thresholds. Compared with surface soils, the higher Zn and Hg concentrations observed in part of the subsoil profile may reflect the combined effects of downward migration, legacy contamination retained below the plow layer, and the contribution of local parent material. This pattern suggests possible vertical redistribution, but direct transport measurements were unavailable.

The coefficient of variation (CV) measures differences in the spatial distribution of elements, and it can also give an indirect sense of the level of human influence in a region [43]. In this study, the CVs for pH in the surface soil and subsoil were 0.103 and 0.131, respectively, indicating that the acidity levels are relatively stable. The CV values of heavy metal elements in the surface soil are ranked as follows: Cu > Hg > As > Zn > Pb > Cd > Ni > Cr. In contrast, the rankings for the subsoil are as follows: Hg > Zn > As > Pb > Cd > Cu > Ni > Cr.

The CV values for Cu, Hg, and As in the surface soil, and for Hg, Zn, and As in the subsoil, were relatively high, all exceeding 1.4, which suggests significant fluctuations in their concentrations that might be linked to human activities. On the other hand, the CV values for Cr and Ni were low, indicating a more uniform distribution, which is likely due to natural, uniform geological sources [28, 29], [55]. This pattern suggests that further source apportionment is needed to distinguish anthropogenic inputs from geological controls.

Source apportionment of soil heavy metals

Principal component analysis (PCA) was used to identify potential sources of heavy metals in the soil [1]. The Kaiser–Meyer–Olkin (KMO) value was 0.676, and Bartlett's test of sphericity was statistically significant (p < 0.001), confirming that the data were suitable for factor analysis [52]. Although the third component had an eigenvalue slightly greater than 1.0 (Table S7), only two components were retained because they captured the main geochemical structure of the dataset (67.8% cumulative variance), matched the scree-plot inflection, and provided interpretable source patterns. PC3 contributed limited additional explanatory power and did not yield a stable, independent source interpretation.

PC1, which accounted for 49.8% of the variance, showed strong associations with Pb, Zn, Cd, As, Hg, and Cu (Fig. 2). This component is considered the primary contamination source. Correlation analysis (Fig. 3) revealed a strong positive relationship between Pb and Zn (r = 0.91, p < 0.01), suggesting a common source related to Pb–Zn mining. Other significant correlations were found between As and Hg (r = 0.78), As and Cu (r = 0.76), Pb and Cd (r = 0.70), and Zn and Cd (r = 0.69), supporting co-release during ore extraction and smelting [33]. Sampling sites S1, S3, and S4 showed the highest PC1 scores, indicating the strongest source-related contamination signals in the dataset.

Fig. 2.

Loadings and scores from the PCA results

Fig. 3.

Correlation analysis of soil heavy metals

PC2, which explained 18.0% of the variance, was primarily associated with Cr and Ni. These two elements showed a significant positive correlation (r = 0.58, p < 0.01), suggesting a similar source. Since their concentrations were below screening thresholds, PC2 likely reflects natural sources such as parent material weathering [22, 23]. The minimal human influence on Cr and Ni is consistent with their low spatial variability. However, an anomaly was observed at site S7, which had an unusually low PC2 score (−3.65). This may reflect local heterogeneity, although the current sampling design was not intended for dense geostatistical interpolation.

Overall, PCA successfully distinguished between human and natural sources of heavy metals in the study area, with PC1 strongly pointing to mining activities as the primary cause of pollution.

Pollution level

I_geo

The I_geo values for the eight heavy metals are shown in Fig. 4. Across both soil layers, the average I_geo values were ranked as follows: Pb > Zn > Cd > Hg > As > Cu > Cr > Ni. In surface soil, Pb had a mean I_geo of 3.30, indicating moderate to heavy pollution. Subsoil Pb levels were also elevated (I_geo = 2.91), showing moderate pollution. Zn and Cd were classified as lightly to moderately polluted (1 < I_geo < 2), while Hg showed light contamination (0 < I_geo < 1). The remaining metals (As, Cu, Cr, Ni) had negative I_geo values, meaning there was no significant contamination. The higher I_geo values for Pb, Zn, and Cd suggest that pollution has come from several sources, including ore weathering, tailings runoff, and emissions from smelting. Pb showed the strongest contamination signal in both soil layers, indicating that it should be prioritized in subsequent remediation planning.

Fig. 4.

I_geo of surface soil (a) and subsoil (b)

PLI

The PLI results (Fig. 5) offer an overall measure of metal contamination. The PLI values showed significant variation across soil layers, averaging 2.11 (range: 0.80–7.99) in surface soils and 1.93 (range: 0.75–5.33) in subsoils. According to the classification criteria, most sites (65%) showed medium levels of pollution in both soil layers, whereas strong to extreme pollution was concentrated in a smaller subset of near-source sites. Only site S7 fell below the pollution threshold in surface soil, while S7 and S11 were below the threshold in the subsoil. The generally higher PLI values in surface soils suggest a stronger human impact at shallower depths, likely due to ongoing deposition and surface runoff [42].

Fig. 5.

PLI of soil heavy metals

Ecological risk assessment

PERI

The calculated results of the E_i and PERI using Eqs. (4) and (5) are presented in Fig. 6. For both soil layers, the average values of E_i for heavy metals consistently followed the same order: Cd > Hg > Pb > As > Zn > Cu > Ni > Cr. Cd had E_i values ranging from 160 to 320, placing it in the high ecological risk category. Hg and Pb had E_i values (mean) between 80 and 160, indicating considerable ecological risks. Other metals had low ecological risks (E_i < 40). Notably, although Hg concentrations were relatively low, its high toxicity coefficient (T_i = 40) and significant ecological effects warrant special attention [6].

Fig. 6.

E_i and PERI of soil heavy metals

The PERI values for surface soils ranged from 157 to 1570, with 65% of sites showing moderate risk, 15% showing considerable risk, and 20% showing high ecological risk. For subsoils, the PERI values ranged from 86.1 to 1570. Five sites (S11, S12, S14, S15, and S19) were classified as low risk, while the others fell into moderate- to high-risk categories: 30% moderate, 10% considerable, and 30% high.

These findings highlight that Cd, Pb, and Hg are the primary ecological stressors in the region. Despite their varying concentrations, their toxicity and persistence in the environment pose significant threats to soil health and biodiversity.

NEA

Initial risk intensities for Cd, Cu, Pb, Zn, As, and Hg were quantified for surface soil samples with exceedance rates over local background values (Liupanshui City, Guizhou Province) greater than 10% (Table S8). The assessment applied literature-based S_i values of 0.3 for vegetation and 0.5 for soil microorganisms [25], yielding a heavy metal risk ranking of Pb > Zn > Cd > Hg > Cu > As.

Based on field surveys and literature reviews [25, 26], we constructed energy flows between ecological communities. Figure 7(a) presents the numerical matrix and topological flow diagram of the energy network. Vegetation (V), herbivores (H), carnivores (C), and soil microorganisms (SM) serve as fundamental components of the ecosystem, representing producers, decomposers, and consumers that encompass core ecological functions. Vegetation and soil microorganisms directly absorb pollutants and bear the initial risk, while herbivores and carnivores acquire risk indirectly through energy transfer.

Using Eqs. (10, 11, 12, 13 and 14), a schematic diagram of the CA coefficient matrix between ecological communities was constructed. As illustrated in Fig. 7(b), the system's energy input is primarily derived from vegetation and is balanced by the total energy output from soil microorganisms, herbivores, and carnivores. The arrows in the diagram indicate the direction of energy flow and control, assuming that no risk transfer occurs in the absence of a CA relationship between components. A higher CA intensity signifies a greater inflow of risk.

To assess the influence of input parameters on the outputs of the NEA framework, we conducted a sensitivity analysis. In this analysis, only the input energy of vegetation was treated as a variable. Four alternative levels were examined: 1246 kJ·ft⁻²·yr⁻¹, 50% lower than the baseline; 1993.6 kJ·ft⁻²·yr⁻¹, 20% lower; 2990.4 kJ·ft⁻²·yr⁻¹, 20% higher; and 3738 kJ·ft⁻²·yr⁻¹, 50% higher. Fig. S1presents the CA values under these four levels and under the baseline input. The results show that variation in the input energy produced little change in control allocation from vegetation to herbivores, soil microorganisms, and carnivores. By contrast, only minor shifts were observed in the control allocation from herbivores to soil microorganisms and carnivores. When the input energy decreased by 50%, the CA from herbivores to soil microorganisms declined by 2.54%, whereas the CA from herbivores to carnivores increased by 2.82%. At a 20% decrease, the corresponding changes were − 0.61% and + 0.69%. When the input energy increased by 20%, the CA from herbivores to soil microorganisms rose by 0.38%, whereas the CA from herbivores to carnivores declined by 0.48%. At a 50% increase, these changes were + 0.76% and − 0.95%, respectively. Overall, variation in input energy had a limited effect on the evaluation results, indicating that the model is robust.

Figure 8 illustrates the risk flow among vegetation, herbivores, soil microorganisms, and carnivores in the study area, revealing distinct characteristics of heavy metal risk transmission and accumulation within the ecosystem. Although soil contamination does not have a direct link to herbivores or carnivores, pollutants can indirectly influence these organisms through energy transfer along the food chain. The direct risk from vegetation to herbivores ranges from 0.0583 for Pb (highest) to 0.0038 for As (lowest). Since herbivores are primarily regulated by vegetation, their direct risk also reflects the overall risk level. In contrast, carnivores are directly influenced by herbivores and indirectly by vegetation, with risk values ranging from 0.0404 for Pb (highest) to 0.0026 for As (lowest). The dominating control of vegetation over soil microorganisms markedly enhances risk transmission, exceeding the effects from herbivores and carnivores by factors of 3.69 and 17.0, respectively.

Fig. 8.

Analysis of ecological risk transfer of heavy metals in soil

An integrated risk assessment across trophic levels was conducted by evaluating cumulative risks from primary, direct, and indirect hazards (Fig. S2). The results indicate that Pb, Zn, and Cd pose the highest ecological risks to vegetation and soil microorganisms, with subsequent cascading effects on herbivores and carnivores. In contrast, Hg, Cu, and As primarily affect lower trophic organisms—such as microorganisms and plants—with minimal transfer to higher-level consumers. Analysis of risk contribution ratios further reveals that vegetation risk originates entirely from the external environment; herbivore risk is fully transmitted via vegetation; soil microbial risk derives from the external environment (62.5%), vegetation (28.2%), carnivores (5.30%), and herbivores (4.00%); and carnivore risk stems predominantly from herbivores (68.7%), with the remainder from vegetation (31.3%).

The overall integrated risk values for six heavy metals in vegetation, herbivores, soil microorganisms, and carnivores were 0.924, 0.188, 2.46, and 0.131, respectively. Soil microorganisms, influenced by the initial soil exposure risk and the other three components, accounted for the highest proportion of risk in the ecosystem (66.4%). As the medium directly exposed to soil pollution, soil microorganisms play a crucial mediating role in risk accumulation, bearing the majority of risks within the ecosystem. This finding aligns with previous research conclusions [48]. Vegetation, as the primary absorber of heavy metals (accounting for 25.0% of risk), absorbs these metals through its root system and transfers some elements to herbivores. Consequently, the elevated risk level of vegetation significantly impacts ecosystem nutrient cycling and food chain health. The risk multiplication relationship indicates that the threat level to soil microorganisms is 2.66 times that of vegetation, the vulnerability of vegetation is 4.91 times that of herbivores, and the potential risk to herbivores is 1.44 times that of carnivores. Under these conditions, long-term heavy metal pollution may severely threaten food chain balance and biodiversity [10]. A critical aspect of this dynamic is that not all initial risks are transmitted within the ecosystem: while vegetation exhibits multi-path risk transmission, the initial risk to soil microorganisms does not propagate further within the system.

Health risk assessment based on Monte Carlo simulation

A probabilistic health risk assessment was carried out for surface soils using Monte Carlo simulations, with the findings summarized in Figs. 9 and 10. For non-carcinogenic risks, the HQ values (Fig. 9) for the eight metals followed this order: Pb > As > Cr > Cd > Zn > Cu > Hg > Ni for both adults and children. All mean HQ values were below 1, suggesting there is little to no non-carcinogenic risk under normal exposure scenarios. However, at the 95th percentile, HQ values for Pb and As exceeded 1 in children, indicating potential health risks under high-exposure situations. For adults, the mean and 95th percentile HI values were 0.122 and 0.286, respectively, both of which are below the safety threshold. In contrast, for children, the mean HI was 1.22, and the 95th percentile was 2.74, well above the threshold of 1, indicating significant non-carcinogenic risk. These findings highlight that children are more vulnerable to heavy metal exposure due to their higher ingestion rates and lower body weight. Pb and As were found to be the primary contributors to this increased risk.

Fig. 9.

Probability distribution of HQ and HI

Fig. 10.

Probability distribution of CR and CRI

Carcinogenic risk showed a similar age-related contrast but with different dominant contributors. Fig. 10 presents the probability distribution of CR values for five metals (As, Cd, Cr, Ni, and Pb). For adults, the average and 95th percentile CR values for As, Cd, Cr, and Ni were within the acceptable range (1 × 10⁻⁶ to 1 × 10⁻⁴). Although the mean CR of Pb was below 1 × 10⁻⁶, the 95th percentile CR value reached the acceptable range for carcinogenic risk. In contrast, children showed higher carcinogenic risks for all the metals. While their average and 95th percentile CRI values were generally within the acceptable range, the 95th percentile CRI for children exceeded 1 × 10⁻⁴, indicating an unacceptable cancer risk under worst-case exposure scenarios. The findings indicate substantially higher health risks for children and support priority risk reduction in child-exposure settings.

Fig. S3 shows how different exposure factors and metal concentrations affect health-risk outcomes. In most exposure parameters, the ingestion rate (IR_ing) had the greatest sensitivity, while skin adherence (SL) had the strongest influence on carcinogenic risk in adults. Among the metals, Pb had the greatest influence on non-carcinogenic risk, whereas As contributed most strongly to carcinogenic risk. These findings suggest that contamination retained in soils and vegetation may increase human exposure through dust generation, incidental soil ingestion, and potential transfer into local food webs.

Soil environmental capacity

Static and dynamic environmental capacities were analyzed to assess the current and future ability of soils to buffer heavy metal inputs. The results are summarized in Table 2, Fig. S4, and Fig. S5.

Table 2.

Descriptive analysis of environmental capacity

Elements		Cd (kg/hm2)	Cu (kg/hm2)	Pb (kg/hm2)	Cr (kg/hm2)	Zn (kg/hm2)	Ni (kg/hm2)	As (kg/hm2)	Hg (kg/hm2)
Qs	-	*	116	230	198	293	134	19.6	1.10
Qi	0–20 cm	*	83.7	*	287	*	172	10.1	0.675
	20–60 cm	*	136	*	284	*	173	8.78	0.484
Qn (0–20 cm)	10 years	*	30.0	*	67.1	*	35.2	6.85	0.190
	30 years	*	25.4	*	51.4	*	25.8	6.30	0.153
	50 years	*	25.0	*	50.2	*	25.1	6.26	0.150
Qn (20–60 cm)	10 years	*	33.1	*	66.9	*	35.3	6.77	0.179
	30 years	*	25.7	*	51.4	*	25.9	6.29	0.152
	50 years	*	25.1	*	50.2	*	25.1	6.26	0.150

^*Denotes no environmental capacity

The static environmental capacity (Q_s) represents the maximum amount of pollution a soil system can handle without exceeding environmental safety thresholds. Of the eight metals analyzed, Cd has already exceeded its static capacity, meaning the soil is saturated and can no longer absorb additional Cd. The metals were ranked in terms of static capacity as follows: Zn > Pb > Cr > Ni > Cu > As > Hg. Zn had the highest buffering potential (293 kg/hm²), while Hg had the lowest (1.10 kg/hm²), highlighting its susceptibility to reaching saturation and posing greater environmental risks.

The residual soil environmental capacity (Q_i) for Cd, Pb, and Zn is completely depleted in both soil layers, indicating an overload of pollution. For the other metals, the residual capacities were ranked as Cr > Ni > Cu > As > Hg. Notably, the residual capacities of As and Hg have sharply declined to 44.0%–61.4% of their static capacities, indicating increasing pollution pressure. Interestingly, Cr and Ni show residual capacities above their static values, possibly due to conservative risk thresholds or natural processes like passivation or low mobility [2, 56].

The dynamic environmental capacity projections showed that Q_n for Cr, Ni, Cu, As, and Hg declined continuously with increasing prediction time under all decay-coefficient scenarios. A common pattern was observed for all elements. The decline was more pronounced during the first 10 years, then slowed after 30 years and gradually became stable. This result indicates a generally consistent trend of weakening soil buffering capacity in the future. Under the screening-level scenario of K = 0.9, the environmental capacities of Cr, Ni, Cu, As, and Hg were projected to decrease by 66.0%, 73.0%, 71.0%, 65.0%, and 82.0%, respectively, relative to their static capacities by year 10. Hg showed the largest decrease, indicating the strongest reduction in its buffering space. Comparison among different K values (0.7, 0.8, 0.9, and 0.95) further showed that Q_n was sensitive to parameter settings. In general, a higher K value led to faster Q_n depletion and lower long-term residual environmental capacity, whereas a lower K value resulted in a slower decline. Consideration of Q_i further indicated that different K scenarios affected not only the rate of Q_n decline but also its position relative to the initial residual environmental capacity baseline. For example, under lower K values, the Q_n curve of As remained above Q_i for a relatively long period. As K increased, the Q_n curve gradually approached Q_i and then fell below it, indicating a continued increase in the magnitude of long-term capacity contraction. Overall, the direction of change was consistent among elements under different K scenarios, but the extent of the predicted decline differed clearly. These results indicate that K is an important parameter in long-term environmental capacity prediction.

The depletion of residual capacity for Cd, Pb, and Zn shows that the soil is already under significant heavy metal stress and cannot handle additional inputs without risking further environmental degradation. The dynamic projections also suggest that, without intervention, the soil’s ability to buffer other metals—especially Hg—will quickly diminish within a decade. These findings support the prioritization of pollution control measures.

Discussion

The combined results indicate that mining-derived Pb, Zn, and Cd dominate current contamination, that ecological risk is concentrated in lower trophic components, and that residual soil environmental capacity is already depleted for several metals and may decline rapidly for others.

I_geo, PLI, and concentration data consistently indicate that Pb, Zn, and Cd contamination in the study area is primarily mining-derived, likely associated with tailings runoff and smelting emissions. Pb showed the highest I_geo values (> 3.0), indicating moderate to heavy contamination and clear enrichment near the ore-processing area. The generally stronger contamination signal in surface soils suggests an important contribution from recent atmospheric deposition and shallow runoff inputs.

However, elevated Zn and Hg concentrations in some subsoil samples indicate that vertical redistribution cannot be explained by surface deposition alone. In the karst setting of Guizhou, carbonate buffering, preferential flow, and heterogeneous sorption sites may facilitate selective downward migration of some elements, while historical deposition and lithogenic inputs may also contribute to deeper horizons [22, 23]. Together, these observations suggest partial connectivity between surface and subsoil layers in the redistribution of mining-derived metals.

The potential ecological risk assessment identified Cd as the dominant ecological stressor, followed by Hg and Pb. Cd deserves particular attention because of its high toxicity and relatively high mobility in neutral to slightly acidic soils, whereas Hg can exert substantial ecological pressure even at lower concentrations because of its high toxicity coefficient and bioaccumulation potential [20, 54].

The NEA results further showed that soil microorganisms accounted for 66.4% of the total ecosystem risk and acted as the major ecological risk sink. This finding indicates that ecological risk in the study area is shaped not only by contaminant concentration but also by the ecological position of receptors within the trophic network. Because the NEA framework relied on literature-based sensitivities and control-allocation parameters, these trophic pathways should be interpreted as inferred rather than directly measured ecological-risk routes.

The PCA results identified mining activities as the main contributor to heavy metal enrichment in the study area. PC1 explained 49.8% of the total variance and was strongly associated with Pb, Zn, Cd, As, Hg, and Cu, supporting a mixed mining-smelting signature. PC2 was dominated by Cr and Ni and was more consistent with geogenic control.

Two principal components were retained because they provided the most stable and interpretable source structure, even though PC3 had a marginal eigenvalue greater than 1.0. High PC1 scores at S1, S3, and S4 indicate strong source-related contamination signals. Combined with the elevated PLI and PERI values at these sites, this pattern supports their prioritization for monitoring and field verification.

Health-risk modeling based on Monte Carlo simulation showed that adults generally remained within acceptable screening-level risk ranges, whereas children experienced substantially higher risks, particularly under high-exposure conditions. The dominant contributions of soil ingestion, Pb, and As indicate that bare-soil contact, dust resuspension, and incidental intake are the major exposure pathways in the study area.

Because quantitative carcinogenic assessment for Pb differs across regulatory frameworks [3, 47], Pb cancer-risk values in this study are presented as screening-level indicators [46]. Even with this limitation, the results clearly support prioritizing exposure reduction in child-contact areas and controlling Pb- and As-rich soil sources.

The depletion of residual soil environmental capacity for Cd, Zn, and Pb indicates that these metals have already exceeded the soil’s natural buffering capacity. For Cr, Ni, Cu, As, and Hg, the scenario-based forecasts suggest that buffering capacity may decline rapidly within the next decade, especially for Hg. Because a uniform K value was applied to all elements, these forecasts should be interpreted as relative trends rather than metal-specific predictions.

Even with this simplification, the results indicate a limited intervention window before moderately accumulated contaminants become more difficult to manage. From a practical perspective, priority actions should include source control at tailings and ore-processing areas, dust suppression [32], soil cover or clean-soil replacement in high-exposure zones [13], phytostabilization using locally adapted tolerant plants [40, 41], and amendment-based immobilization using materials such as lime, biochar, phosphate, or clay minerals [38]. In residential or agricultural interface areas, land-use control and child-exposure reduction should be prioritized.

Limitations and future work

Several limitations should be acknowledged. First, although the 20-site sampling design was sufficient for a case-oriented assessment, it does not support high-resolution spatial interpolation or a rigorous evaluation of seasonal variability. As a result, the dataset may disproportionately reflect localized hotspots or transient conditions, potentially leading to conservative (i.e., higher) risk estimates at the regional scale. Second, vertical migration was inferred from surface–subsoil contrasts rather than direct process-based measurements, which may oversimplify transport dynamics and could overstate the extent of downward metal fluxes. Third, the NEA framework was implemented deterministically and relied on literature-derived receptor sensitivities and control-allocation parameters; such parameterization typically adopts precautionary assumptions that may bias ecological risk estimates toward the upper bound. Fourth, the dynamic environmental-capacity model applied a uniform decay coefficient (K = 0.9) and did not explicitly incorporate metal-specific attenuation processes, hydrological variability, climate change, or land-use transitions. These simplifications may underestimate the system’s intrinsic buffering and attenuation capacity, thereby overstating the persistence and magnitude of future risks. Fifth, the human health-risk assessment was based on total soil concentrations without accounting for bioaccessibility, biomonitoring data, or food-chain exposure pathways. Because total concentrations generally exceed the bioavailable fraction, the estimated exposure doses and associated risks are likely to be conservatively overestimated. Future work should therefore incorporate higher-resolution sampling, metal speciation and bioaccessibility measurements, process-based transport modeling, and scenario-driven forecasting to refine risk quantification.

Conclusion

1. Pb, Cd, and Zn were the primary pollutants in the study area, with average concentrations far above local background values and agricultural screening thresholds. Pb showed the strongest pollution signal, and PCA indicated that Pb–Zn mining and associated smelting were the dominant anthropogenic sources, whereas Cr and Ni were mainly controlled by natural geological factors.

2. PERI showed that most sampling sites were subject to moderate to high ecological risk. The NEA model further indicated that Pb, Zn, and Cd generated the strongest ecological pressures across trophic levels and that soil microorganisms were the main ecological risk sink, accounting for 66.4% of integrated ecosystem risk.

3. Children faced higher screening-level health risks than adults. The 95th percentile HQ values for Pb and As in children exceeded 1.0, and the cumulative carcinogenic risk under high-exposure conditions exceeded the commonly accepted threshold, highlighting the importance of controlling Pb and As exposure in child-contact areas.

4. Soil environmental capacity for Cd, Zn, and Pb was already exhausted, and scenario-based forecasting indicated substantial further decline in the environmental capacity for Cr, Ni, Cu, As, and Hg over the next 10 years, with Hg showing the fastest projected depletion. These results support early source control and targeted remediation before residual capacity declines further.

Abbreviations

I_geo

Geo-accumulation index

PLI

Pollution Load Index

PERI

Potential Ecological Risk Index

NEA

Network Environmental Analysis

HQ

Hazard Quotient

HI

Hazard Index

CR

Carcinogenic Risk

CRI

Carcinogenic Risk Index

Authors’ contributions

T.S. was responsible for research design, data analysis, and drafting the manuscript. J.H., S.C., and H.C. participated in data collection, experimental implementation, and result analysis. H.X. and J.G., as co-corresponding authors, supervised the research, validated the methodology, and finalized the manuscript. X.W., Q.L., and Q.T. assisted with sample collection, literature review, and partial graphic preparation. F.X. and G.S. contributed critical technical support and data validation. All authors have reviewed and approved the final version of the manuscript for submission.

Funding

This work was supported by Zunyi Outstanding Youth Science and technology innovation talents training project (ZunYouQingKe (2021) 7); Moutai institute Joint Science and Technology Research and Development Project (ZunShiJiaoHe HZ Zi (2021) 322); Modern Baijiu Brewing Technology Engineering Research Center of Guizhou Universities Qianjiaoji (2023) No. 028; Zunyi Science and Technology Bureau of Guizhou Province and Moutai Institute Joint Science and Technology Cooperation Fund Project, Zunyi Ke (2023) No. 114; Research Foundation for Scientific Scholars of Moutai Institute (Grant No. mygccrc[2022]007, mygccrc[2022]097, mygccrc[2022]022, mygccrc[2022]020, mygccrc[2022]056, mygccrc[2022]045 and mygccrc[2022]031); Guizhou Province Distillers Grain and Agricultural Waste Resource Utilization Engineering Research Center (Grant No. 04500052); A Project on Characteristic Key Laboratory of Guizhou Ordinary Colleges and Universities by Department of Education of Guizhou Province (grant No: Qian Jiao He KY Zi [2018] 003); Guizhou Engineering Research Center for Comprehensive Utilization of Distillers' Grains;Youth Guidance Project of Guizhou Province Basic Research Program (Natural Sciences) in 2024 (Qiankehe Jichu [2024] Qingnian 203).

Data availability

All data generated or analyzed during this study are included in this published article and its supplementary information files.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.