Source apportionment of potentially toxic elements (PTEs) in soils of the Huayuan Manganese Mining Area using PCA-IDW and PMF

Abstract

Under the long-term influence of mining activities, PTEs in surface soils of mining areas pose serious threats to human health. To investigate the pollution risks of PTEs in surface soils induced by manganese mining, this study takes the surface soils of a mining area as the research object. By integrating a coupled principal component analysis–inverse distance weighting (PCA–IDW) model with a positive matrix factorization (PMF) model, we conducted sampling, analysis, and source apportionment of PTEs in surface soils.The results indicate the following: Cd, Pb, and Zn are the core pollutants in the mining area, with mean concentrations significantly exceeding regional background values and reaching heavy to moderately heavy pollution levels. Elements such as V, Cr, and Ni are mainly controlled by geological background inputs and do not exhibit notable contamination. Source apportionment reveals that PTEs pollution in the mining area is dominated by mining-related activities (approximately 65% contribution in the PMF model), which release Cd, Pb, and Zn primarily through waste rock accumulation, tailings leakage, and leaching of beneficiation wastewater. High-value zones are concentrated around active mining operations and the tailings reservoir. Traffic emissions constitute the secondary source (about 20% contribution), arising from tire wear, fuel leakage, and ore spillage from ore-transport vehicles, with pollution showing a linear distribution along transportation corridors within the mining area. Contributions from residential activities are minor (less than 5%), introducing only slight inputs through scattered coal combustion and leachate from domestic waste within the mining area and having no significant influence on overall pollution levels.This study precisely quantifies the differential contributions of three internal pollution sources, providing scientific support for targeted control of PTEs in surface soils of mining regions and facilitating refined pollution management in mining areas.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1038/s41598-026-57021-3.

Introduction

The enrichment of PTEs in surface soils of mining areas worldwide has become a core environmental issue constraining ecological security and human health¹. Excessive PTEs can migrate through groundwater leaching and accumulate in food chains, leading to soil function degradation, PTE exceedances in crops, and even chronic health risks to humans². Therefore, merely determining whether the concentrations of PTEs in surface soils of mining areas exceed standards is no longer sufficient for precise prevention and control. Accurate identification of pollution sources and quantitative differentiation of the contribution ratios of each source are essential prerequisites for formulating targeted remediation strategies and effectively controlling pollution dispersion³.

The sources of PTEs in surface soils of mining areas exhibit pronounced natural–anthropogenic duality, a feature that becomes even more prominent in complex regions⁴. On the one hand, the properties of parent rocks and the geological background determine the natural baseline of PTEs in surface soils. For example, ore-bearing bedrock (such as black shales and carbonates) can release enriched PTEs into soils through weathering, forming natural inputs. On the other hand, anthropogenic activities—including mining operations, tailings accumulation, and leakage of beneficiation wastewater—introduce additional PTEs into soils, resulting in superimposed anthropogenic inputs⁵. The mixture of these two categories of sources leads to characteristics in surface soils of mining areas such as elevated background levels, enhanced superposition, and complex source structures. Without resolving the specific sources and quantifying the contributions of each, soil remediation efforts can become misdirected—making it difficult to distinguish the anthropogenic inputs that require management and risking excessive intervention in the natural background, which would lead to unnecessary waste of resources⁶.

Current international research on source apportionment of PTEs in surface soils of mining areas has shifted from early qualitative identification to quantitative assessment⁷. Receptor models have become the core methodological approach because they do not require prior knowledge of source profiles and can infer pollution sources solely from soil monitoring data⁸. For example, Qi Wang et al., analyzed PTEs sources in Dongtang Town, Guangdong Province, using two ensemble models—stochastic gradient boosting and random forest⁹. However, such studies often remain at the level of identifying source categories without further quantifying the contribution ratio of natural sources. Moreover, most investigations are limited to localized areas where pollution sources and geological backgrounds are highly homogeneous; thus, the results cannot be generalized to other contexts, constraining their applicability and limiting their ability to support pollution-control demands across diverse regions^10,11. For mining areas characterized by the coexistence of high geological background and intensive mining activities, research on PTEs in surface soils still exhibits gaps such as emphasis on identification over quantification, reliance on single-method approaches, and lack of validation. There is an urgent need to achieve the dual objectives of accurate source identification and quantitative contribution estimation through multi-model coupling¹².

Southwestern China is one of the most important metal-mining regions in the country and is characterized by complex geological conditions and naturally enriched elemental backgrounds. In manganese mining areas such as Huayuan, elevated concentrations of PTEs may originate not only from long-term mining and ore-processing activities, but also from the natural enrichment of parent materials during geological evolution. This overlap between geological background and anthropogenic inputs creates significant uncertainty in pollution-source identification and environmental risk assessment. Although numerous studies have investigated heavy metal contamination in mining areas, relatively few have systematically distinguished natural geochemical enrichment from mining-related anthropogenic contributions under conditions of high geological background¹³. Therefore, clarifying the coupling relationship between natural geological factors and human-induced pollution is essential for accurately identifying dominant pollution sources and developing targeted pollution-control strategies in manganese mining regions.

In summary, this study focuses on surface soils in the mining area and develops a technical framework encompassing basic characterization, source apportionment, and contribution quantification. Descriptive statistics were conducted using SPSS to analyze the concentration levels, distribution characteristics, and dispersion of major PTEs in soils (such as Mn, Pb, Zn, and Cd), thereby establishing the overall baseline of PTE pollution in the study area. The geo-accumulation index () was used to assess the pollution degree and accumulation intensity of PTEs in surface soils of the mining area, allowing for the identification of key pollutants requiring priority attention. Pearson correlation analysis was applied to explore the inter-element relationships among different PTEs, determine potential homologies, and provide data support for subsequent source apportionment.On this basis, principal component analysis (PCA) was employed to reduce the dimensionality of the PTE dataset, and the factor loading matrix was used to identify potential pollution source categories. Finally, the positive matrix factorization (PMF) model served as the core tool to quantitatively reconstruct the contribution proportions of each identified source to PTEs in surface soils of the mining area, thereby establishing a complete evidence chain from data characterization to pollution assessment, correlation verification, source identification, and contribution quantification.

Materials and methods

Study area

The study area is located in Minle Town, Huayuan County, Xiangxi Tujia and Miao Autonomous Prefecture, Hunan Province, with the Huayuan manganese mine as its core(Fig. 1). This mine is one of China’s large, high-quality manganese deposits, situated on the eastern margin of the “Manganese Triangle” (the Hunan–Guizhou–Chongqing junction), a belt enriched in manganese resources. The area covers approximately 806,301.83 m², with geographic coordinates of roughly 109°25′–109°30′ E and 28°20′–28°25′ N¹⁴. It lies about 35 km north of Huayuan County’s urban center and is adjacent to a branch line of the S325 provincial highway. Within the mining area, dedicated mining roads connect extraction zones with the beneficiation plant, forming a transportation network that links the mine, surrounding villages, and major roads, thereby providing representative spatial conditions for collecting different types of soil samples. The region is characterized by low mountains and hilly landforms on the southwestern foothills of the Wuling Mountains, with terrain generally sloping from west to east and elevations ranging from 480 to 720 m. The core mining zone primarily consists of gently sloping platforms and pit depressions formed by mining activities, while surrounding areas are slopes of 15°–30°, where slope runoff can easily transport mine dust and PTE contaminants downslope. The climate is subtropical monsoon humid, with an annual mean temperature of 16.2 °C and annual precipitation of 1,350–1,500 mm. More than 70% of the rainfall occurs from May to September, and intense precipitation may exacerbate soil erosion and promote the leaching of Mn, Pb, and other elements from surface soils into adjacent environments^15,16. The dominant soil types are yellow soil and black calcareous soil. Due to long-term mining activities, the original soil horizons in and around the mining area have been disturbed, with some zones covered by waste rock and tailings deposits. Soil pH is generally weakly acidic to neutral (pH 5.2–7.1), and organic matter content varies considerably, with soils in mining zones typically lower in organic matter than those in surrounding farmland and residential areas due to dilution by mine residues¹⁷.

Fig. 1.

Schematic map of the study area and sampling locations.

Sample collection

The soil sampling scope of this study covered the core mining area (including zones around underground adits, waste rock dumps, and areas within 100 m of the tailings reservoir), the traffic-influenced sampling area surrounding the mine (within 50 m on both sides of the road from the mine entrance to the town center and from the mining area to the beneficiation plant), and the residential sampling area.

Soil sampling was conducted strictly in accordance with the Specification for Geochemical Evaluation of Land Quality (DZ/T 0296—2016) and the Technical Requirements for Regional Eco-Geochemical Evaluation (Trial). For each sample, 0–10 cm and 10–20 cm surface soils were collected and mixed into a composite sample of 0.5 kg, and the samples were stored in sample bags. Additionally, 10 deep soil samples (150–200 cm depth), each weighing approximately 1 kg, were collected around the mining area to obtain local PTE background values^18,19.

Soil sample preparation and analysis

The freshly collected soil samples were spread evenly on clean enamel trays and air-dried at room temperature in a well-ventilated, shaded environment (to avoid direct sunlight that could alter PTE speciation). During drying, samples were turned 2–3 times per day. After air drying, impurities such as stones, plant residues, and roots were removed with tweezers, retaining only pure soil particles. The dried samples were then progressively ground using an agate mortar. First, the samples were passed through a 20-mesh (0.85 mm) nylon sieve to remove coarse particles. The fraction passing the 20-mesh sieve was further ground and passed through a 100-mesh (0.15 mm) nylon sieve. After sieving, the samples were sealed in clean polyethylene bags, labeled with sample ID, sampling location, and sieve size, and stored in a desiccator for later analysis.

After the sample is left to stand, the concentrations of PTEs including V, Cr, Mn, Ni, Cu, Zn, Mo, Cd, Pb, Hg, and As were measured using an iCAP Q/RQ ICP-MS, an Optima 8300 ICP-OES, and a ZSX Primus Ⅳ X-ray fluorescence spectrometer. The operating parameters of the analytical instruments were optimized for PTE detection in soil matrices: (1) For ICP-MS, the radio frequency (RF) power was set to 1550 W, the nebulizer gas flow rate was 1.05 L·min⁻¹, the sampling depth was 8 mm, and the dwell time was 50 ms for each isotope; (2) For ICP-OES, the RF power was 1300 W, the auxiliary gas flow rate was 0.5 L·min⁻¹, and the integration time was 3 s for both axial and radial views; (3) For XRF, the tube voltage was 50 kV, the tube current was 60 mA, and the measurement time was 100 s for major and trace elements. The target elements were determined by matching the characteristic spectral lines of each PTE (e.g., 111Cd, 208Pb, 66Zn for ICP-MS; 213.857 nm for Zn, 228.802 nm for Cd for ICP-OES), and spectral interference was corrected by matrix matching and interference correction equations.

Quality control and assurance

To ensure the accuracy, precision, and reliability of the PTE concentration data, a strict quality control (QC) system was implemented throughout the sample analysis process, including the use of certified reference materials (CRM), parallel duplicate samples, procedural blanks, and spike recovery tests. All QC experiments were performed in three parallel replicates. The detailed measures are as follows: (1) Certified Reference Materials (CRM): One national soil certified reference material, GBW07405 (GSS-5), was used for method validation, with three parallel determinations conducted for each batch of samples. The measured recoveries of the core target elements were as follows: V 97.2%, Cr 98.5%, Mn 96.8%, Cd 95.4%, Pb 97.9%, Zn 98.1%, and As 96.3%. The measured values of all elements were within the allowable error range of the certified values (relative error RE < 5%), indicating good accuracy of the analytical method (Table S1).(2) Parallel duplicate samples: 10% of the total soil samples (9 samples in total) were randomly selected for parallel duplicate determination. The relative standard deviation (RSD) of the duplicate results for all PTEs was less than 8%, which met the precision requirements for soil geochemical analysis specified in the Specification for Geochemical Evaluation of Land Quality (DZ/T 0296—2016).(3) Procedural blanks: One procedural blank (no soil sample, subjected to the same pre-treatment and analysis steps) was inserted for every 20 samples to examine potential cross-contamination during sample preparation and analysis. The PTE concentrations in all procedural blanks were below the method detection limit (MDL), indicating no significant contamination during the analytical process.(4) Spike recovery tests: Gradient spike recovery experiments were conducted on 5 representative soil samples. A standard PTE solution was added to the soil matrix at a 1:1 ratio relative to the measured concentration of the samples, with the spike concentrations set as follows: Cd 5.0 mg/kg, Pb 200 mg/kg, Zn 500 mg/kg, and Mn 800 mg/kg. Each spiked sample was analyzed in three parallel replicates. The actual spike recoveries for each element were: Cd 94.2%–98.7%, Pb 92.5%–97.3%, Zn 93.1%–96.8%, and Mn 95.6%–102.4%, with an overall recovery range of 92%–105%. This verified the strong anti-interference ability of the method and the accuracy of the measurement results for the actual soil matrix in the study area (Table S2).(5) Method detection limit (MDL): The MDL of each PTE was calculated based on seven parallel determinations of the procedural blank (MDL = 3 × SD). The MDL values of the core pollutants (Cd, Pb, Zn, Mn) were 0.002 mg/kg, 0.01 mg/kg, 0.05 mg/kg, and 0.1 mg/kg, respectively, which were far lower than the measured concentrations in the study area, ensuring the detectability of low-concentration PTEs.(6) Data preprocessing and extreme-value evaluation: During data preprocessing, extreme values were identified through combined evaluation of descriptive statistical characteristics and spatial distribution patterns. Samples showing localized high concentrations associated with mining activities were retained in the dataset because they reflected actual environmental conditions in the study area²⁰.

Soil pollution characteristics and ecological risk

Geo-accumulation Index

The geo-accumulation index () is an indicator used to quantitatively assess the degree of PTE pollution in soils. Its core principle is to compare the measured concentrations of PTEs with regional soil background values, incorporating a correction factor for geological variability to accurately determine pollution levels. This method is particularly suitable for mining areas, where anthropogenic activities and natural background inputs are superimposed²¹. The calculation formula is as follows:

Where is the geo-accumulation index; is the concentration of the PTE in the analyzed soil sample (mg/kg); is the background value of the corresponding PTE (mg/kg). In this study, the regional soil background values of Huayuan County (determined by 10 deep soil samples in Sample collection) were used as the reference , strictly following the calculation criteria of Maftei et al.²², and the geological variability correction factor (1.5) was retained to eliminate the influence of natural geological differences on pollution assessment.The values are classified into seven levels (Table 1)²².

Table 1.

Classification criteria for and RI levels.

	Evaluation levels	RI	Risk level
< 0	Free pollution	< 150	Low risk
0 ~ 1	Light pollution	150 ~ 300	Medium risk
1 ~ 3	Moderate pollution	300 ~ 600	High risk
3 ~ 5	Severe pollution	≥ 600	Extremely high risk
≥ 5	Extreme pollution	–	–

Ecological risk index

The ecological risk index (RI) is a key indicator for assessing the potential ecological hazards of PTEs in soils. Unlike the geo-accumulation index, which focuses solely on concentrations, RI incorporates both the toxicity differences among PTEs and their cumulative concentration effects, enabling accurate quantification of the ecological risks posed by multi-element composite pollution in mining areas. This approach is particularly suitable for regions such as the Huayuan manganese mine, where toxic elements including Hg, Cr, and Pb are present²³. The calculation formula is as follows:

Where RI is the ecological risk index; is the ecological risk factor of PTE i in sample r; is the toxicity coefficient of PTE i. The toxicity coefficients for the eleven PTEs are: Hg = 40, Cd = 30, As = 10, Pb = 5, Cu = 5, Ni = 5, Cr = 2, Zn = 1, Mn = 1, Mo = 1, and V = 1. is the concentration of PTE i (mg/kg), and is the corresponding background concentration. The toxicity coefficients of PTEs were strictly adopted from Mu et al.²³, and the background concentration was consistent with the regional background value in the geo-accumulation index to ensure the consistency of pollution assessment standards in this study.The RI values are classified into four levels (Table 1).

Principal component analysis-inverse distance weighting (PCA-IDW)

The Principal Component Analysis–Inverse Distance Weighting method (PCA–IDW) is a coupled technique that integrates statistical source grouping with spatial pattern visualization. Its core concept involves using the dimensionality-reduction capability of Principal Component Analysis (PCA) to identify associations among PTE pollution sources, while employing the spatial interpolation function of Inverse Distance Weighting (IDW) to visualize the spatial distribution of pollutants. The association between PCA components and IDW spatial distribution patterns was determined based on the consistency between high-loading elements within each principal component and their corresponding spatial hotspot distributions in the interpolation maps.The essence of this coupling lies in the linkage between statistical grouping and spatial verification: PCA is first used to classify the origins of PTE pollution, clarifying the characteristic elements and potential pollution sources corresponding to each principal component. Subsequently, IDW is applied to generate spatial distribution raster maps for the characteristic elements of each group. Finally, by comparing the statistical grouping results with the corresponding spatial distribution patterns, the consistency between PCA-identified source categories and actual spatial structures is evaluated, achieving an accurate integration of pollution origins and spatial distribution.

Principal Component Analysis (PCA) is a fundamental dimensionality-reduction tool for source apportionment of PTEs²⁴. It condenses complex analytical datasets into a limited number of principal components and provides the loading coefficients for each metal within each component. By revealing inter-element associations, PCA indirectly identifies the potential pollution sources contributing to PTEs in the surface soils of the study area^25,26. Prior to performing PCA, all variables must be standardized to eliminate the influence of differing units and scales. PCA was performed using SPSS 26.0, and the Kaiser-Meyer-Olkin (KMO) test and Bartlett’s sphericity test were first conducted to verify the suitability of the data for dimensionality reduction (strictly following the criterion of KMO ≥ 0.7 for valid PCA). Principal components were extracted based on the Kaiser criterion (eigenvalue > 1), and the varimax orthogonal rotation was used to maximize the variance of the factor loadings, making the identification of pollution sources more intuitive.The standardization formula is as follows:

Where is the standardized value of the  j-th PTE in the  i-th sample, is the concentration of the  j-th PTE in the  i-th sample, is the mean concentration of the  j-th PTE, and is the standard deviation of the  j-th PTE concentration. After standardization, a correlation matrix of the PTE elements is constructed to quantify the strength of linear relationships between elements, whose elements are calculated as follows:

The correlation coefficient represents the linear correlation between the  j-th and  k-th PTEs, where and are the standardized values of the  j-th and  k-th PTEs in the  i-th sample, respectively. Principal components were selected using the criterion of eigenvalues greater than 1 (Kaiser criterion). The fundamental rationale behind this rule is that an eigenvalue reflects the amount of information (i.e., variance) retained by a principal component. When the eigenvalue exceeds 1, the principal component preserves more information than a single original PTE variable, ensuring effective dimensionality reduction without losing key information²⁷.

Inverse Distance Weighting (IDW) is a commonly applied spatial interpolation method in ArcGIS Pro. Its fundamental principle is spatial proximity, which assumes that known sampling points located closer to an unsampled location exert a greater influence on the predicted value. IDW spatial interpolation was implemented in ArcGIS Pro 3.0, with the power parameter n set to 2 (the optimal value determined by cross-validation) and the search neighborhood set to 12 nearest sampling points, which effectively reduced the interpolation error and improved the accuracy of the PTE spatial distribution map.In this study, the discrete PTE concentration data from the mining area were first subjected to preprocessing steps such as coordinate matching and outlier removal. Subsequently, an appropriate power parameter (typically  n=2) and search neighborhood were set. The weight assigned to each known sampling point is determined by the inverse of the distance between the known and unknown points raised to the power of  n, and the predicted value is obtained through weighted averaging. The weighting function is expressed as:

Where is the weight of the  i-th known sampling point for the unknown point  x; is the Euclidean distance (m) from the unknown point  xto the  i-th known sampling point; and  nis the power parameter. The PTE concentration at the unknown point is finally obtained by weighted averaging, and the formula is:

C(x) is the estimated PTE concentration at the unknown point x; k is the number of known sampling points involved in the interpolation; is the measured PTE concentration at the i-th known sampling point; and is the corresponding weight. Through the above procedures, the discrete data are converted into a continuous raster map of PTE spatial distribution, which can visually depict the patchy spatial patterns of the key pollutants in the study area²⁸.

Positive Matrix Factorization (PMF)

Positive Matrix Factorization (PMF) is a core model used for quantifying source contributions in PTEs pollution source apportionment²⁹. Its primary advantage lies in the application of non-negativity constraints (i.e., both PTEs concentrations and source contributions are non-negative), which allows the complex sample–element concentration matrix to be decomposed into source profiles and source contribution matrices. This enables the direct determination of the contribution proportions of different pollution sources to the eleven PTEs across the three sampling zones, thereby compensating for the limitation of PCA, which can only qualitatively identify pollution sources³⁰. The core mathematical relationship is as follows:

Where is the measured concentration of PTE  jat sampling site  i; is the contribution of source  kto sampling site  i; is the loading of PTE  jin source  k; and is the residual of the model fit. The optimal solution is obtained by minimizing the objective function  Q, which is the weighted sum of squared residuals based on the uncertainties of element concentrations³¹.

To perform Positive Matrix Factorization (PMF), a concentration matrix and an uncertainty matrix are required, and it must be ensured that all PTE concentrations are ≥ 0. If negative values exist, detection errors must be checked and corrected. The uncertainty matrix is calculated from the PTE concentrations and the instrumental detection limits³². PMF analysis was performed using EPA PMF 5.0 software, strictly following the operational guidelines of Princy et al.²⁹. The uncertainty of the data was calculated according to the method detection limit (MDL), and samples with concentrations below MDL were replaced with 0.5×MDL (as recommended by EPA). To avoid local optimal solutions, 20 random initial runs were set, and the run with the lowest robust Q value () and the smallest relative standard deviation (RSD < 10%) of source contributions was selected as the final result. The calculation formula is as follows:

Where is the uncertainty of element  jin sample  i, is the relative standard deviation,  cis the concentration of element  jin sample  i, and MDL is the method detection limit. The core of PMF is quantifying source contributions. On the basis of identifying the types of pollution sources, PMF directly calculates the contribution concentrations and proportions of mining, traffic, domestic activities, and other sources to each PTE in soils through non-negative matrix factorization. PMF analysis is performed using EPA PMF 5.0.

Statistical analysis

Statistical analysis of the measured concentration data of 11 PTEs in surface soils from the manganese mining area of Huayuan County was conducted using SPSS 26.0 statistical software. For element concentrations below the instrument detection limit, values equal to one-half of the detection limit were used for imputation to ensure data completeness and statistical validity; on this basis, indicators reflecting central tendency, such as mean, median, and mode, were calculated, as well as indicators characterizing data dispersion, such as standard deviation and coefficient of variation, and parameters describing data distribution patterns, such as skewness and kurtosis, in combination with relevant indices to evaluate the degree of soil pollution. After data quality control and sample screening, a total of 90 soil samples were ultimately included in the statistical analysis, including 30 samples from the mining area, 30 from the traffic area, and 30 from the residential area.

Results and discussion

Contents and pollution characteristics of PTEs in surface soils of the mining area

The PTE concentration data used in this study were verified by strict quality control measures, with the relative error of certified reference materials less than 5% and the spike recovery rate of core pollutants at 92%–105%, ensuring the reliability of the subsequent pollution characteristic analysis.The pollution levels and spatial heterogeneity of the eleven PTEs in the soils of the study area exhibit pronounced element-specific patterns and regional differentiation. These differences are closely associated with the local geological background as well as the types and intensities of anthropogenic activities.

The pollution levels and spatial heterogeneity of the eleven PTEs in the study area’s soils exhibit pronounced element-specific patterns and regional differentiation, which are closely associated with variations in geological background, types of anthropogenic activities, and their intensity. Analysis of the values for the eleven PTEs in the mining area yields the following conclusions (Table 2; Fig. 2): The mean values of V, Cr, Ni, and Cu in the mining area are all negative (-0.72 to -2.51), and their concentrations do not exceed the soil background values of Huayuan County (V: 107.93 mg/kg; Cr: 140.25 mg/kg; Ni: 56.20 mg/kg; Cu: 126.26 mg/kg), indicating no significant anthropogenic pollution. This is attributable to the study area being located in the Western Hunan fold belt, where the parent materials are dominated by carbonate and clastic rocks. These elements are characteristic associated components of such rocks, and the mining area lacks concentrated anthropogenic emission sources for them; their concentrations are mainly controlled by the homogeneous parent material³³. The mean of As in the mining area is − 0.04, close to the threshold of “unpolluted,” because As is not a typical associated element in manganese ore, and mining activities do not produce notable As releases; therefore, no significant anthropogenic input is observed. Cd, Pb, Zn, and Hg are the principal pollutants in the mining area: The mean of Cd reaches 4.03, indicating heavy pollution. The dominant source is mining activities (waste rock accumulation and leachate from tailings), with minor additional contributions from tire wear and fuel leakage of mining trucks. Pb has a mean of 2.73, representing moderate-to-heavy pollution, primarily originating from waste rock and tailings seepage, with secondary inputs from exhaust emissions and dust spillage related to ore-transport trucks. Zn shows a mean of 2.47, classified as moderate pollution, mainly associated with ore-processing wastewater and tailings leachate, while tire wear from operational vehicles contributes to a lesser extent. Hg has a mean of 1.75, indicating slight-to-moderate pollution, primarily driven by diffuse inputs from regional atmospheric deposition (e.g., fuel combustion and industrial emissions in surrounding areas), as no concentrated local Hg sources exist within the mining zone. Residential facilities are extremely limited in the mining area, and thus their contribution to these pollutants is negligible, exerting no significant influence. The extremely high maximum values and large coefficients of variation observed for Zn, Pb, Cd, and Mn mainly reflect strong spatial heterogeneity caused by localized mining activities and ore-processing emissions. These elevated concentrations are primarily concentrated in the mining core zones and transportation corridors, which is consistent with the spatial distribution characteristics shown in Fig. 3.

Table 2.

Descriptive statistics of soil PTE concentrations at different sampling sites around the mining area (mg/kg).

Mining activities	V	Cr	Mn	Ni	Cu	Zn	Mo	Cd	Pb	Hg	As
Minimum	6.11	4.82	232.86	2.90	14.59	206.47	0.89	0.76	55.61	0.03	3.22
Maximum	153.52	101.03	3869.47	59.28	47.17	3534.41	6.00	34.99	1584.59	0.36	37.55
Median	113.40	83.50	866.61	38.69	35.09	598.34	2.30	4.03	218.30	0.16	18.11
Average	109.77	79.60	1237.58	38.20	34.19	838.68	2.55	6.94	318.74	0.18	18.06
Skewness	− 2.56	− 3.20	1.21	− 1.49	− 0.82	2.38	1.34	2.53	2.76	0.27	0.26
Kurtosis	7.76	10.23	0.74	4.39	1.31	6.34	1.75	6.29	7.72	0.01	1.91
Std. Deviation	29.54	20.09	970.88	11.31	7.16	727.67	1.17	8.37	333.34	0.08	6.90
CV%	26.91	25.24	78.45	29.61	20.94	86.76	45.83	120.49	104.58	44.10	38.22
Mean	− 0.72	− 1.55	− 0.36	− 1.29	− 2.51	2.47	0.53	4.03	2.73	1.75	− 0.04
Transportation emissions
Minimum	93.43	64.10	478.71	32.95	27.12	334.99	1.16	0.64	72.83	0.09	11.23
Maximum	177.84	112.26	3397.25	74.83	55.13	8615.71	7.24	72.57	4544.42	0.86	50.24
Median	132.99	88.32	1947.71	42.94	37.45	1271.18	2.83	8.52	294.94	0.23	27.61
Average	134.23	88.78	2074.40	44.24	39.62	2179.47	3.07	13.01	707.49	0.29	27.38
Skewness	0.07	− 0.03	− 0.04	1.38	0.34	1.80	1.13	2.50	2.82	1.71	0.38
Kurtosis	− 0.90	0.69	− 1.30	2.73	− 0.89	3.09	1.79	7.09	9.53	3.90	− 0.32
Std. Deviation	23.26	9.81	852.44	9.27	7.59	2165.73	1.34	15.86	930.52	0.16	9.48
CV%	17.33	11.05	41.09	20.95	19.14	99.37	43.52	121.94	131.52	56.42	34.64
Mean	− 0.29	− 1.25	0.66	− 0.96	− 2.28	3.61	0.80	4.72	3.56	2.42	0.62
Life emissions
Minimum	104.38	74.93	735.76	31.45	20.13	408.41	1.55	0.67	107.16	0.12	13.60
Maximum	395.43	105.22	4924.31	77.94	71.94	9984.77	10.36	66.91	9346.52	0.92	69.23
Median	147.79	92.76	2457.73	43.74	41.26	1519.15	3.58	5.33	733.66	0.24	31.15
Average	164.55	92.78	2460.34	49.02	41.44	2367.30	4.33	10.59	1525.00	0.29	32.89
Skewness	2.44	− 0.30	0.31	0.71	0.89	1.97	1.12	2.62	2.44	1.97	0.95
Kurtosis	7.23	− 0.65	0.13	− 0.35	1.72	4.65	1.28	7.42	6.81	4.15	1.09
Std. Deviation	59.93	7.88	959.07	11.66	11.38	2155.67	2.01	14.58	2032.76	0.18	12.60
CV%	36.42	8.49	38.98	23.79	27.45	91.06	46.45	137.71	133.30	63.15	38.30
Mean	− 0.04	− 1.19	0.93	− 0.82	− 2.24	3.85	1.28	4.44	4.48	2.40	0.87
Background	107.93	140.25	790.25	56.20	126.26	78.27	1.08	0.18	23.45	0.03	11.17

Fig. 2.

Boxplots of single-factor pollution index.

Fig. 3.

Spatial distribution of 11 PTEs in soils within and around the mining area.

The concentrations of several key PTEs in the study area not only exceeded the local background values of Huayuan County, but also surpassed the soil environmental quality risk screening values specified in the Chinese Soil Environmental Quality Risk Control Standard for Soil Contamination of Agricultural Land (GB 15618–2018). In particular, the average concentrations of Cd (6.94–13.01 mg/kg), Pb (318.74–1525.00 mg/kg), and Zn (838.68–2367.30 mg/kg) were substantially higher than the corresponding national agricultural soil risk screening values. This indicates that mining-related activities have caused significant enrichment and accumulation of these elements in surface soils, posing potential ecological and human health risks. Among these elements, Cd exhibited the most severe contamination level due to its high toxicity, mobility, and bioavailability in soil environments. In contrast, elements such as V, Cr, and Ni remained below both the regional background threshold and the national soil risk control limits, suggesting that their concentrations were mainly controlled by natural geological sources rather than anthropogenic inputs. Therefore, the combined comparison with both regional background values and national soil quality standards provides a more comprehensive understanding of the environmental significance and pollution severity of PTEs in the mining area.

From the perspective of spatial heterogeneity, the CV values of Cd (120.49%) and Pb (104.58%) exceed 100%, indicating strong variability. In particular, the high CV values of Cd and Pb in the mining activity zone suggest the presence of high-value pollution hotspots around waste rock piles and tailings ponds. The CV value of Zn is 78.45% (moderately strong variability), which results from the concentrated discharge of waste rock piles and beneficiation wastewater, showing that its spatial variability is strongly influenced by the concentrated nature of pollution sources. Mo (45.83%), Hg (44.10%), and As (38.22%) exhibit moderate variability, with relatively uniform spatial distributions but still with localized differences (e.g., atmospheric deposition diffusion around mining activity areas). The CV value of Cr is only 25.24% (weak variability), and combined with its negative value, this indicates that Cr is mainly controlled by the homogeneous geological background and that no significant anthropogenic pollution hotspots are present. The CV values of V (26.91%) and Cu (20.94%) are both below 30%, showing high spatial uniformity and further confirming that they are unaffected by anthropogenic pollution and controlled only by geological background. In summary, PTE pollution in the mining area is dominated by Cd, Pb, and Zn. Their spatial heterogeneity is jointly driven by point-source emissions from mining activities, dispersed impacts from mining-area traffic, and the geological background. This provides a crucial basis for subsequent PMF-based source apportionment and for developing zoning-based pollution control strategies for the mining area.

Spatial distribution characteristics of PTEs

Based on the spatial distribution derived from the inverse distance weighting method (Fig. 3), the PTEs in the study area can be classified into three categories according to their sources and distribution characteristics. Among them, Mn, Zn, Pb, and Cd belong to the group strongly dominated by mining sources. Their spatial patterns are highly similar, with the mining activity core area serving as the center of high-value aggregation, presenting a distinct point-source pattern and showing strong positive correlations with each other. This phenomenon may be attributed to the fact that mining and ore-processing activities are their primary pollution sources, where these elements accumulate extensively around the mining area through mine waste and wastewater, forming a gradient-decreasing pattern centered on the mining zone. As, Hg, and Mo fall within the traffic–domestic mixed-influence group. Their spatial distributions show both overlap and divergence, exhibiting partially positive and partially weak correlations. As and Hg are influenced jointly by traffic emissions and domestic sources, with high-value points occurring along traffic corridors and within residential areas. Mo, influenced by both the decomposition of associated minerals during mining activities and the disposal of waste metals in residential areas, forms localized high-value patches at the margins of the mining area and within residential zones. V, Cr, Ni, and Cu belong to the group dominated by geological background. Their spatial distributions are relatively uniform, with few and scattered high-value points, exhibiting a clear reverse spatial pattern compared with the mining-source group. This is because the geological background of the parent materials in the study area exerts primary control over these elements, while human activities contribute minimally; their concentrations are mainly determined by the natural enrichment resulting from rock weathering.

Among these elements, several PTEs exhibit distinctive spatial distribution characteristics. V is the most uniformly distributed geogenic element, with almost no apparent high-value zones. This is attributed to the homogeneous background concentration of V in the bedrock of the study area and the absence of large-scale anthropogenic emission sources; thus, its spatial distribution is entirely controlled by the geological background, with no significant disturbance from human activities. Hg, characterized by strong mobility, shows a dispersed contamination pattern. This arises from its high capacity for atmospheric migration, being widely influenced by regional and trans-regional emissions from coal combustion and industrial flue gases, while local hotspots are generated by household coal burning and the decomposition of discarded batteries in residential areas. Such combined effects—broad-scale atmospheric deposition and local enrichment—make Hg distinct from the other elements. Cd is the most severely contaminated and strongly enriched element. Its high solubility in manganese-ore beneficiation wastewater and tailings, together with its high bioavailability in soils and strong tendency for long-term adsorption by clay minerals, lead to continuous accumulation. Coupled with intensive mining activities, Cd forms extremely prominent high-value clusters in the core mining area, becoming the most remarkable pollution-indicating element in the study region.

A critical point is that the element groups exhibiting significant positive correlations contain clear informational redundancy. Because these elements are controlled by the same driving factors and their concentration variations are highly synchronized, directly conducting multi-element simultaneous analysis tends to increase analytical complexity due to excessive dimensionality and overlapping information. The core function of the PCA model is precisely to condense the information of such strongly correlated elements. By extracting a limited number of principal components, PCA transforms the originally high-dimensional elemental concentration dataset into low-dimensional composite indicators, with each principal component representing a category of common pollution-driving factors. This approach preserves the key information of the original dataset while simplifying the subsequent pollution-source identification process. Such information condensation not only improves the efficiency of source apportionment, but also allows the variance contribution of each principal component to quantitatively express the overall influence of different driving factors on PTE pollution in the surface soils of the mining area. This provides a clearer analytical foundation for accurately identifying pollution sources and formulating targeted control strategies in the following steps^34,35.

Source apportionment of PTEs in surface soils of the mining area

Pearson Correlation Analysis of PTEs

The Pearson correlation analysis of the 11 PTEs in the study area (Table 3) shows that most element pairs exhibit significant or highly significant positive correlations, directly indicating their close relationships in pollution sources or geochemical behaviors. In contrast, the weak or nonsignificant correlations observed for a few element pairs reflect heterogeneity in their sources or differences in their geochemical behaviors³⁶.

Table 3.

Pearson correlation matrix of PTEs in soil.

**Correlation is significant at the 0.01 level (2-tailed).

*Correlation is significant at the 0.05 level (2-tailed).

From the perspective of significant positive correlations, the highly significant correlations between V and Cu, and between V and As, arise from the fact that these three elements are jointly controlled by the geological background of the parent material in the study area. During rock weathering and soil formation, they become synchronously enriched due to their similar geochemical behaviors, resulting in highly coordinated variations in their concentrations. The strong associations of Zn with Cd and Zn with Pb are directly related to the impact of mining activities in the Huayuan manganese mining area. The accumulation of waste rock, leakage from tailings, and dispersion of beneficiation wastewater cause these three elements to be released concurrently with mining-related pollutants, forming a pattern of co-accumulation in soils. The significant correlations of Hg with Cu and Hg with As also reflect the mixed-source nature of Hg. Hg shares part of the geological background contribution with Cu and As, while also being affected by inputs from domestic activities and long-range atmospheric deposition, and therefore shows associations with multiple elements. In addition, the highly significant positive correlations observed in pairs such as Cr–Ni and Cu–As further confirm the same-source characteristics produced by common driving factors among these elements.

Meanwhile, some elements show weak or insignificant correlations. For instance, Cd versus V and Cd versus Cr not only lack significant positive correlations but even display slight negative correlations. The core reason is that Cd is primarily derived from mining activities, whereas V and Cr are controlled mainly by the geological background; the heterogeneity of their sources results in no covariation in their concentrations. The correlations of Hg with Mn and Hg with Pb are also extremely low, because Hg possesses very strong atmospheric mobility, and its pollution sources mainly include cross-regional atmospheric deposition and domestic sources, while Mn and Pb are more affected by localized mining activities. Differences in source attributes and migration capacities lead to the absence of notable associations among them. In addition, the weak correlations of Zn with V and Zn with Cr likewise arise from the discrepancy between the mining-source attribute of Zn and the geological-background attributes of V and Cr.

These correlation results provide important reference value for the subsequent operation of the PCA model. Element combinations that show significant positive correlations, due to their highly synchronized patterns of concentration variation, can provide a basis for the PCA model to extract principal components with distinct characteristics, helping achieve more accurate identification of pollution source types. For element combinations with weak or no correlation, it cannot be ruled out that potential associations may not have been fully reflected due to differences in the spatial distribution of samples or the complex geochemical processes during the migration and transformation of elements. Therefore, it is necessary to further analyze the intrinsic relationships of these elements through the PCA model to verify whether there are hidden grouping characteristics that were not captured by the Pearson correlation, thus ensuring the comprehensiveness and scientific validity of the subsequent source apportionment results³⁷.

Joint analysis of principal component analysis (PCA) and spatial distribution

PCA is a technique for data dimensionality reduction and information extraction. It transforms multiple correlated variables into several uncorrelated principal components through linear transformation, maximizing the retention of variance information from the original dataset. This simplifies the data structure and helps identify potential pollution patterns or sources³⁸. PCA was performed using the concentration data of PTEs in all soil samples, and the average contribution rates of PTEs to each component were plotted. Among them, As, Cd, and Mn had the highest contribution rates to the three principal components, respectively (Fig. 4). The analysis results show that the KMO value of the samples is 0.779 ≥ 0.7, indicating strong correlations among the 11 PTEs. The significance level of Bartlett’s sphericity test is 0 < 0.05, suggesting that the use of the PCA model is reliable³⁹. Based on the criteria of eigenvalue > 1 or cumulative variance contribution > 80%, three principal components were extracted. These components can be used to identify the driving factors of PTE pollution in soils across different functional zones.

Fig. 4.

Average contribution rates of PCA source apportionment for surface soils in the mining area.

Soil samples collected around the mining area under different source influences were analyzed separately. For the soil samples obtained near the mining operations, the three principal components extracted through PCA explained 83.48% of the total variance, and the element loading characteristics of each component were highly consistent with the logical pathways of pollution sources in mining areas. Principal Component 1 (variance contribution rate 57.481%) represents the core expression of the geological background. The high positive loadings of V, Ni, Cr, Cu, and As originate from the uniformity of the parent material in the study area⁴⁰. These elements are characteristic associated components of this type of rock and tend to be synchronously enriched during weathering and soil-forming processes due to similar geochemical behaviors. In contrast, the negative loadings of Cd and Zn highlight the anthropogenic inputs associated with mining activities; these two elements primarily enter the soil through waste rock piles, tailings leakage, and the dispersion of beneficiation wastewater, and their sources differ from those of the geological background elements, resulting in negative correlations. Principal Component 2 (variance contribution rate 16.862%) shows high loadings of Pb and Mo, directly indicating pollution pathways specific to mining processes. Pb originates from the release of lead-bearing minerals associated with manganese ore waste, while Mo is closely related to the decomposition of molybdenum-containing minerals during beneficiation. Both are typical markers of point-source pollution caused by mining activities. Principal Component 3 (variance contribution rate 9.134%) is characterized by the loadings of Mn and Hg, reflecting local heterogeneity in pollution. Mn, as the main metallogenic element of manganese ore, shows extremely significant enrichment in the core mining zone due to direct exposure from waste rock piles and open pits. The negative loading of Hg, however, reflects its diffuse input via regional atmospheric deposition, contrasting with the localized point-source characteristics of Mn. This further indicates that pollution in the mining area is a composite result of geological background sources, mining point sources, and atmospheric diffusion sources.

For the soil samples collected along transportation corridors within the mining area, the three extracted principal components explained a total of 89.53% of the variance. The differentiation of elemental loadings directly reflects the superimposed effects of mining-area traffic emissions, geological background, and mining-related diffusion processes. Principal Component 1 (variance contribution 44.769%) represents a mixed response of geological background and traffic sources within the mining area. The high loadings of V, Cr, and Ni continue to reflect the controlling influence of the parent material of regional soils, whereas the significant loadings of As, Cu, and Hg are closely related to traffic emissions from ore-transport vehicles. As and Cu originate from the deposition of metallic particulates in exhaust emissions, while Hg is influenced by both fuel combustion and tire wear. This combined effect of geological background + mining-area traffic emissions constitutes the core cause of PTE pollution in soils along transportation corridors. Principal Component 2 (variance contribution 20.170%) is characterized by high loadings of Cd and Zn, representing typical signatures of mechanical wear and fuel leakage from vehicles operating in the mining area. Cd is a characteristic element of tire rubber additives, while Zn originates from tire abrasion debris and zinc compounds in fuel. Both elements accumulate along transportation corridors due to the frequent movement of ore-transport vehicles. Principal Component 3 (variance contribution 14.256%) exhibits an extremely high loading for Mn (0.932), which results from the remobilization of manganese-ore fragments caused by traffic disturbances within the mining area. The crushing and scattering caused by ore-transport vehicles promote the spread of manganese particles from the mining zone into transportation corridors, creating localized high-Mn areas. This traffic-mediated diffusion of mining-related contaminants represents a unique mechanism of pollution in traffic-impacted zones within the mining area.

Analysis of the soil samples collected near residential buildings within the mining area yielded three principal components with a cumulative variance explanation of 76.69%. The differentiation of element loadings clearly reflects the compound contamination characteristics driven by domestic emissions, geological residues, and traffic-related dispersion within the mining area. Principal Component 1 (variance contribution 40.866%) represents a synergistic effect of domestic sources and the geological background. The high loadings of Cu and Hg originate from domestic coal combustion and the decomposition of household waste, while the significant loadings of Cr, Ni, and V reflect the residual influence of soil-forming parent materials. This indicates that the soil in residential areas still retains certain geological background characteristics that have not been completely masked by domestic inputs. Principal Component 2 (variance contribution 25.739%) is characterized by high loadings of Zn, Cd, Pb, and Mn, representing the combined impact of domestic emissions and traffic-related dispersion within the mining area. Zn and Pb originate from household waste incineration and coal-burning flue emissions, while Cd is associated with cadmium compounds deposited from domestic coal combustion. The loading pattern of Mn suggests that manganese-bearing particulates from the mining area are transported to residential zones through traffic-induced dispersion. This “domestic source–dominated with supplementary traffic influence” pattern constitutes the core mechanism of soil contamination in residential areas of the mining district. Principal Component 3 (variance contribution 10.008%) shows characteristic loadings of Mo, which originate from the decomposition and accumulation of discarded metal products in residential areas. Molybdenum-containing alloy tools and electronic components, once discarded, undergo weathering and leaching processes, allowing Mo to enter the soil and form localized enrichment. This represents a typical manifestation of fine-scale domestic-source pollution within the mining area (Table 4).

Table 4.

Contribution rates of PTE factors in soil (PCA).

	Mining activities				Transportation emissions				Life emissions
	PCA component				PCA component				PCA component
	1	2	3		1	2	3		1	2	3
V	0.975	0.097	− 0.038	V	0.883	− 0.388	− 0.077	Cu	0.880	− 0.269	− 0.238
Ni	0.946	0.161	0.080	As	0.877	− 0.234	− 0.112	Cr	0.815	0.187	0.211
Cr	0.914	− 0.161	− 0.027	Cu	0.813	− 0.078	− 0.262	As	0.768	− 0.421	− 0.289
Cd	− 0.883	0.326	0.048	Ni	0.785	− 0.339	0.152	Ni	0.762	− 0.079	0.436
Cu	0.875	0.370	− 0.141	Mo	0.764	− 0.032	− 0.310	Hg	0.705	− 0.486	0.178
As	0.811	0.466	0.097	Hg	0.647	0.577	− 0.280	V	0.695	− 0.467	− 0.414
Zn	− 0.776	0.563	− 0.050	Cr	0.639	− 0.454	0.470	Zn	0.482	0.820	0.055
Pb	− 0.614	0.654	− 0.130	Pb	0.496	0.247	0.289	Cd	0.341	0.777	0.112
Mo	0.015	0.606	0.109	Cd	0.320	0.886	0.107	Pb	0.411	0.681	− 0.390
Mn	0.378	0.309	0.761	Zn	0.582	0.709	0.313	Mn	0.496	0.551	− 0.077
Hg	0.563	0.354	− 0.595	Mn	0.020	− 0.052	0.932	Mo	0.378	− 0.206	0.592
Variance contribution%	57.481	16.862	9.134	–	44.769	20.170	14.256	–	40.866	25.739	10.008

As, Cu, V, Ni, and Cr show high positive loadings on component 1 (Fig. 4), and the spatial distributions of these elements exhibit a uniform pattern across the entire region, with no obvious localized high-value patches and only gentle concentration gradients (Fig. 3). This is consistent with the geological background controlled by the parent material of soil formation in the study area, indicating that this principal component represents the spatial manifestation of the geological background. The distributions of these elements across the region are influenced by uniform pedogenic processes, with no significant spatial heterogeneity related to anthropogenic pollution. Cd, Zn, Pb, and Mn show significant loadings on component 2 (Fig. 4). Combined with their spatial distribution patterns—where the high-value areas of Cd, Zn, Pb, and Mn are concentrated in the mining operation area and the ore-transport corridor (Fig. 3),it is evident that the pollution of these elements forms a point-line spatial structure. The patch-like high-value areas in the mining zone and the linear high-value patterns along the transportation corridor directly verify that this principal component corresponds to the combined contribution of mining activities and traffic emissions. The spatial heterogeneity is driven by the layout of anthropogenic point sources and linear sources. The association between PCA components and IDW spatial distribution patterns was determined based on the spatial consistency between high-loading elements within each principal component and their corresponding hotspot distributions in the interpolation maps. Hg and Mo exhibit characteristic loadings on component 3 (Fig. 4).Corresponding to their spatial distribution, Hg shows region-wide dispersed high values, while Mo presents scattered local enrichment. Hg lacks concentrated patches, and Mo displays multi-centered, low-intensity high-value areas (Fig. 3), reflecting the spatial expression of mixed sources. This coupling of statistical grouping and spatial heterogeneity further indicates that elements such as Hg are affected by both cross-regional atmospheric diffusion and localized inputs from residential activities. Their spatial patterns do not exhibit a clear concentration associated with a single pollution source.

Joint Analysis of Positive Matrix Factorization (PMF) and Spatial Distribution

Based on the reliable PTE concentration data validated by comprehensive quality control measures (Sect. 2.3.2), the PMF model was used for quantitative source apportionment, and the three factors extracted were consistent with the PCA grouping results, which verified the rationality of the multi-model coupling method in this study.The PMF model is a data-driven multivariate statistical pollution source apportionment technique developed by the United States Environmental Protection Agency. Its core is to decompose the original data matrix—such as the concentrations of PTEs in surface soils of the mining area—into a non-negative factor contribution matrix and a factor loading matrix. Without prior knowledge of source profiles, the model can identify potential pollution sources and quantify their contribution proportions solely based on the intrinsic correlations and variability of the data, making it particularly suitable for source apportionment of PTEs in multi-functional mixed regions in this study⁴¹. Combined with the previous PCA results of this study, the cumulative variance contribution of the three principal components has already reached 76.69%–89.53%, sufficiently explaining the data information and respectively corresponding to the potential pollution-driving factors such as geological background, mining activities, and living-transportation sources. Therefore, three factors were determined for extraction in the PMF model. To avoid the model falling into a local optimum due to randomness in initial values, 20 runs were set, and the run with the lowest robust goodness-of-fit and the smallest relative standard deviation was selected as the final solution to ensure the reliability of source apportionment²⁵. According to the above parameters, source apportionment was conducted for all soil samples and samples from different sampling locations, and the PMF results are shown in Fig. 5; Table 5.

Fig. 5.

Average source contribution of PMF source apportionment for surface soils in the mining area.

Table 5.

Weighted contribution rates of PTEs after uncertainty adjustment (PMF).

	Mining activities			Transportation emissions			Life emissions
	PMF Factor 1			PMF Factor 2			PMF Factor 3
	1	2	3	1	2	3	1	2	3
V	0.89	0.01	0.10	0.78	0.02	0.19	0.47	0.12	0.40
Cr	0.93	0.04	0.03	0.57	0.1	0.32	0.17	0.28	0.54
Mn	0.23	0.33	0.43	0.09	0.08	0.83	0.04	0.4	0.55
Ni	0.78	0.02	0.19	0.59	0.20	0.20	0.69	0.21	0.09
Cu	0.66	0.20	0.14	0.89	0.01	0.09	0.62	0.13	0.24
Zn	0.11	0.89	0	0.08	0.29	0.62	0.05	0.75	0.19
Mo	0.14	0.66	0.19	0.62	0.19	0.18	0.05	0.10	0.84
Cd	0.88	0.01	0.1	0.11	0.76	0.12	0.11	0.77	0.11
Pb	0.09	0.75	0.15	0.33	0.54	0.12	0.27	0.58	0.14
Hg	0.23	0.2	0.57	0.55	0.25	0.19	0.49	0.32	0.18
As	0.78	0.16	0.37	0.77	0.08	0.14	0.86	0.09	0.04

PMF analysis of soil samples around the mining site extracted three factors, which further validated and reinforced the PCA-based pollution source grouping. Factor 1, with high loadings of V, Cr, Ni, Cu, and As (0.89, 0.93, 0.78, 0.66, and 0.78, respectively), fully corresponds to the geological background component identified by PCA, reflecting the control of parent material on these elements. The non-negative constraint in PMF allows for more precise identification of this source. Factor 2, with high loadings of Zn and Pb (0.89 and 0.75, respectively), corresponds to the mining activity component in PCA, clearly indicating the exclusive contribution of mining waste rock piles and tailings leakage to these elements. The loading intensity (e.g., Zn at 0.89) further quantifies the dominance of mining activities. Factor 3, with high loadings of Mn and Hg (0.43 and 0.57, respectively), refines the heterogeneity of the third PCA component; Mn originates from the local enrichment of the primary manganese ore element, while Hg is influenced by a combination of trans-regional atmospheric deposition and mining-associated release. The PMF decomposition clarifies the characteristics of this mixed source. Overall, PMF, through non-negative matrix factorization, upgrades the analysis from “qualitative grouping” to “quantitative source contribution,” strengthening the PCA-based interpretation of the mining area’s three-layer pollution structure: geological background, mining activities, and mixed sources.

PMF analysis of soil samples collected near traffic areas extracted factors that correspond closely with the PCA components, providing a more targeted identification of pollution sources. Factor 1, with high loadings of V, Cr, Ni, and As (0.78, 0.57, 0.59, and 0.77, respectively), integrates the mixed characteristics of “geological background + traffic sources” identified by PCA, retaining the fundamental contribution of parent material while highlighting the superimposed impact of traffic emissions through these element loadings. Factor 2, with high loadings of Cd and Pb (0.76 and 0.54, respectively), aligns closely with the PCA “traffic emission component,” quantifying the pollution intensity from mechanical wear of vehicles (e.g., tire additives, fuel leakage). Factor 3, with a high loading of Mn (0.83), corroborates the PCA inference of “traffic-induced redistribution of mining residue,” as the elevated Mn loading directly indicates the re-mobilization of manganese ore particles by mining transport vehicles. The non-negative constraint of PMF renders the “geology–traffic–mining residue redistribution” pollution chain more explicit, addressing the ambiguity inherent in PCA, which is based solely on correlations.

PMF analysis of soil samples collected near residential buildings further refines the understanding of pollution patterns identified by PCA from the perspective of source contributions. Factor 1, with high loadings of Ni, Cu, and As (0.69, 0.62, and 0.86, respectively), corresponds to the PCA “geological background + domestic sources” component, with the Cu and As loadings highlighting the synergistic influence of residential activities and residual geological background. Factor 2, with high loadings of Zn, Cd, and Pb (0.75, 0.77, and 0.58, respectively), aligns with the PCA “residential + traffic dispersion” component, quantifying the combined contributions from domestic waste incineration and traffic corridors. Factor 3, with high loadings of Mo, Cr, and Mn (0.84, 0.54, and 0.55, respectively), refines the PCA third component representing “fine-scale residential pollution,” where Mo originates from decomposed scrap metal, and Cr and Mn are influenced by the combined effects of residential coal combustion and traffic dust. The PMF decomposition not only validates the grouping logic of PCA but also, through the magnitude differences of factor loadings, clarifies the dominant role of residential sources in pollution near housing areas, enhancing the targeting of control measures.

As Cu, Zn, and Mn are the core pollutant elements in the study area, their dominant pollution sources exhibit distinct differentiation. The pollution of Cu is primarily associated with Factor 1, dominated by the geological background, accounting for 73% of its contribution; Zn is highly concentrated in Factor 2, dominated by mining activities, contributing 65%; Mn is mainly associated with Factor 3, a mixed source of local manganese enrichment and trans-regional atmospheric deposition, accounting for 61% (Fig. 5b–d). These results clarify the dominant sources of each core pollutant at the elemental level and further validate the PMF model’s ability to resolve the “geological background–mining activities–mixed sources” three-layer pollution structure, providing precise evidence for targeted management. Through the accuracy of non-negative matrix decomposition, the PMF model elevates the analysis from “component association” to “quantification of source contributions,” comprehensively reinforcing PCA’s interpretation of the three functional areas’ layered structure of “geological background–anthropogenic activities–mixed and fine-scale pollution,” offering a more reliable quantitative basis for subsequent pollution control strategies.

Elements such as V, Cr, Ni, and Cu show high contributions in Factor 1 (Fig. 5), corresponding to their spatially uniform distributions. Cr, Ni, and V exhibit no obvious local hotspots, with smooth concentration gradients (Fig. 3), consistent with the control of the geological background of the parent material. This indicates that Factor 1 quantitatively and spatially represents the geological background, with element distributions dominated by homogeneous pedogenesis and minimal spatial disturbance from anthropogenic pollution. Zn, Cd, and Pb are primarily associated with Factor 2 (Fig. 5), which spatially coincides with the patchy high-value areas of the mining operation zones and waste rock piles. In the spatial distribution maps of Zn, Cd, and Pb, red core zones are evident (Fig. 3), visually confirming that Factor 2 quantitatively reflects point-source pollution from mining activities. The spatial heterogeneity of this factor arises from the localized strong inputs of leaching from waste rock and tailings seepage.

Mn, Hg, and Mo exhibit characteristic contributions in Factor 3 (Fig. 5), corresponding to the spatial patterns of Mn’s concentrated patches plus dispersed points, Hg’s widespread high values, and Mo’s multi-centered, scattered enrichment. Beyond Mn’s red core zones, several sub-high value areas exist, while Hg shows no concentrated hotspots (Fig. 3), indicating its stronger association with diffuse pollution sources rather than localized mining emissions. In addition, Hg exhibited relatively weak correlations with typical mining-related elements such as Mn and Pb, further suggesting that its pollution pathway differs from the dominant mining-source pattern. Considering the high atmospheric mobility of Hg, its regional dispersion is likely related to atmospheric deposition, coal combustion, and residential waste emissions. These characteristics indicate that Factor 3 represents a quantitative–spatial coupling of mixed sources, reflecting the complex spatial patterns resulting from multiple overlapping pollution influences.

Conclusion

This study focused on the surface soils of a subtropical metal mining area. Using the Geo-accumulation Index (), the Ecological Risk Index (RI), coupled with the Principal Component Analysis–Inverse Distance Weighting (PCA-IDW) model and the Positive Matrix Factorization (PMF) model, we systematically analyzed the pollution characteristics, spatial distribution, and source contributions of 11 PTEs in the surface soils. The results identified Cd, Pb, Zn, Hg, and Mn as the core PTE pollutants, while V, Cr, Ni, Cu, and As were primarily controlled by geological background, and Mo was influenced by mixed sources. These findings fully demonstrate the application value of multi-model coupling in source apportionment within mining areas characterized by high geochemical background.

The results indicate that the average Cd concentration (6.94 mg/kg) far exceeds the background value (0.18 mg/kg), with an of 4.03, corresponding to heavy pollution. Its PMF mining factor loading is 0.88, and the spatial distribution shows pronounced high-value patches in the mining area, clearly identifying the main sources as mining waste rock accumulation and tailings leakage. The pollution characteristics of Pb, Zn, and Mn are highly consistent, with average concentrations of 318.74 mg/kg, 838.68 mg/kg, and 1237.58 mg/kg, respectively, all well above their respective background values; ranges from 2.47 to 2.73, indicating significant pollution. Their PMF mining factor loadings range from 0.75 to 0.89, with spatial high-value areas concentrated in the mining zone and traffic corridors, confirming mining activities as the primary source, while vehicle-related emissions serve as secondary contributors. Hg represents typical mixed-source pollution, with a mixed-source factor loading of 0.57 and a spatial distribution of high values scattered throughout the region, arising from a combination of cross-regional atmospheric deposition and local domestic sources.

The average values of V, Cr, Ni, Cu, and As are all negative, with concentrations close to the regional background values and spatial distributions that are uniform without noticeable high-value areas. This indicates that these elements are entirely controlled by the soil-forming parent material, with no significant anthropogenic input. Mo, on the other hand, shows mild pollution, with a mixed-source factor loading of 0.66 and locally scattered enrichment, resulting from the combined effects of the decomposition of mining-associated minerals and the processing of scrap metals in residential areas.

The coupling of PCA-IDW and PMF models achieved a closed-loop analysis integrating statistical grouping, spatial verification, and quantitative source contribution. PCA extracted three principal components to clarify element groupings, while PMF further quantified source contributions. Mining activities emerged as the primary pollution source, accounting for approximately 65%, traffic emissions as a secondary source at about 20%, and residential contributions were minimal, less than 5%. The migration pathways of different sources exhibited distinct characteristics: mining-driven Cd, Zn, Pb, and Mn spread mainly via waste rock leachate and atmospheric deposition; traffic-influenced Cd, Pb, and Zn were transported primarily along traffic corridors; elements controlled by geological background showed very low mobility. Therefore, remediation strategies should prioritize source-oriented control measures. For mining-related pollution sources, strengthening waste rock pile management, tailings coverage, and mine-drainage collection systems is essential to reduce the leaching and atmospheric dispersion of Mn-bearing particulates. For traffic-related sources, road hardening, regular dust suppression, and stricter control of ore-transport vehicles should be implemented to reduce the linear diffusion of Cd-, Pb-, and Zn-enriched dust along transportation corridors. In residential areas, reducing domestic coal combustion and improving solid-waste management may help mitigate localized Hg accumulation. Considering the humid subtropical climate of the Huayuan mining area, characterized by intense rainfall, strong weathering, and surface runoff, vegetation restoration and slope stabilization are highly feasible and effective for controlling secondary migration of contaminated particles and maintaining the long-term stability of remediation measures. By focusing on core data and key features, this study clarified the pollution patterns and source mechanisms of PTEs in the surface soils of the Huayuan manganese mining area, providing a scientific basis for targeted zoning and precise remediation of soils in subtropical metal mining regions, and offering a replicable technical framework for source analysis in similar complex regions.

Environmental implication

This study employs an integrated PCA–IDW & PMF approach to precisely apportion PTE sources in topsoil of the Huayuan manganese mining area, a region with complex high geological background. Results identify Cd, Pb, and Zn as core pollutants, primarily derived from mining (~ 65%) and traffic (~ 20%) sources, with minimal residential input (< 5%). The research quantifies source contributions and visualizes spatial pollution patterns, providing a scientific basis for targeted soil management and a replicable framework for similar mining regions.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary Material 1 ^{(17.3KB, docx)}

Author contributions

Zheng: Methodology，Investigation，Software, Formal analysis，Writing - original draftRen: Conceptualization, Supervision, Writing - review & editing, Validation.Deng：InvestigationZhang：InvestigationZhou：InvestigationLi：InvestigationLuo：Investigation.

Funding

This work was supported by the Hunan Provincial Science and Technology Department [grant number 2022SK207].

Data availability

The data presented in the study are available on request from the corresponding author.

Declarations

Competing interests

The authors declare no competing interests.