Source Apportionment of Soil Heavy Metals in Urban Agglomerations Based on the APCS-MLR Model

Abstract

In order to study the differential characteristics of heavy metal contamination levels and their sources in soils under various land use types and anthropogenic activities at a regional scale, this study focused on the Beijing–Tianjin–Hebei (BTH) urban agglomeration in North China. We analyzed heavy metal content in three land use types (urban green spaces, croplands, and vegetable fields/orchards) through field sampling and laboratory analysis, with content determined by inductively coupled plasma mass spectrometry (ICP-MS). The sources of heavy metals were quantitatively apportioned their sources using the absolute principal component score–multiple linear regression (APCS-MLR) method. Results of this study are as follows: (1) Heavy metal content varied among different soil types, with vegetable fields/orchards soils showing relatively higher content. Urban green spaces and cropland soils exhibited comparable heavy metal levels, though urban green spaces displayed higher spatial heterogeneity, while cropland soils showed more homogeneous distributions. (2) The APCS-MLR model identified five pollution sources: mixed traffic–coal combustion sources, industrial sources, agricultural sources, natural sources, and unknown sources. Natural sources were consistently the dominant contributors of arsenic (As), chromium (Cr), and nickel (Ni) across all three land use types, with contribution rates of 32.62–70.26%. Traffic and coal combustion emissions were the primary sources of lead (Pb) and copper (Cu) in urban green spaces, accounting for 40.28–66.26%, while industrial activities showed the highest contributions to zinc (Zn) and cadmium (Cd) in urban green spaces, at 45.88–65.25%. Agricultural activities contributed similarly to Cd accumulation in both cropland and vegetable fields/orchards soils (41.68–51.32%), but their contributions to Cu and Zn in vegetable fields/orchards soils (46.62–55.58%) were significantly higher than those in cropland (9.21–13.40%). Notably, unexplained sources accounted for 18.64–42.59% of heavy metals in vegetable fields/orchards soils, suggesting particularly complex sources in these systems. This study provides a scientific basis for sustainable soil management strategies and promoting coordinated pollution control in urban agglomeration regions.

1. Introduction

Soil heavy metals have drawn widespread attention due to their characteristics of persistence, high toxicity, and bioaccumulation potential through the food chain [1]. In addition to being influenced by parent materials, the occurrence of heavy metals in soil is substantially affected by anthropogenic activities such as industrial production, agricultural practices, transportation emissions, and others. Pollutants from these sources enter surrounding or downstream soils via atmospheric deposition, surface runoff, and waste disposal [2], further exacerbating regional heavy metal accumulation risks. In many cases, anthropogenic contributions even exceed natural sources of soil heavy metals. Thus, identifying and quantifying major pollution sources and their contributions are crucial for developing targeted control and mitigation measures to combat soil heavy metal pollution [3].

In current studies, methods for source apportionment of soil heavy metals can be categorized into qualitative source identification and quantitative source apportionment. The former typically employs correlation analysis, principal component analysis (PCA), and cluster analysis to qualitatively determine major pollution sources [4]. The latter primarily relies on receptor models, including the chemical mass balance (CMB) model, positive matrix factorization (PMF), absolute principal component score–multiple linear regression (APCS-MLR), the UNMIX model, and isotopic ratio analysis [5,6], which quantitatively estimates the contributions of different sources. The APCS-MLR model combines principal component analysis/factor analysis with multiple linear regression and does not require predefined source profiles, offering higher efficiency than the CMB method for identifying complex pollution sources. It also exhibits lower sensitivity to outliers and stronger robustness against outliers compared to the PMF model, demonstrating high applicability in soil heavy metal source apportionment [5]. This method is now widely used to quantify pollution sources in various land use types, such as urban green spaces [7], vegetable fields [8], and agricultural soils [9].

Although significant progress has been made in source apportionment of soil heavy metals, current research perspectives remain limited: (1) Most studies focus predominantly on a single administrative scale, such as municipal, county, or smaller levels. For instance, Shetaia et al. analyzed the distribution characteristics and sources of heavy metals in road dust from the Greater Cairo metropolitan area [10]; Sun et al. identified the sources of heavy metals in soils from the peri-urban areas of Ningbo in east China [11]; and Zhao et al. investigated the pollution characteristics and source apportionment of soil heavy metals on a county scale [12]. Given the pronounced transboundary transport characteristics of heavy metal pollution, a single and relatively small study scale may limit the understanding of anthropogenic versus natural contributions to soil heavy metal contamination, thereby posing governance challenges in guiding systematic regional prevention and control efforts. (2) Most studies focus on single land use types, such as cropland or parks, or treat mixed land use types as an integrated unit for source apportionment, making cross-type comparisons difficult. For example, although Peng et al. [13] analyzed anthropogenic contributions across urban, suburban, and rural environmental gradients in Chengdu, China, all study objects were croplands. Zhou et al. [14] conducted source apportionment in three major urban agglomerations of China using data without distinguishing land use types and found that industrial activities contributed over 60% to the content of Cd, Hg, and Pb in soils across these regions. In fact, the accumulation of heavy metals in the soil environment is closely related to the land use type [15]. Spatial material flows and metabolic processes vary significantly across different land use types, leading to distinct patterns in the types, sources, and accumulation characteristics of pollutants. For example, due to the application of organic fertilizers (such as livestock manure), the input fluxes of heavy metals like Cd, Cu, and Zn in vegetable soils are often significantly higher than those in conventional croplands [16]. Urban green spaces adjacent to major traffic arteries are more likely to accumulate elements such as Pb and Zn from vehicle emissions [17]. The separate analysis of soil heavy metal sources for each type allows for the clear identification of the “primary pollution source” for each category, rather than producing a vague conclusion such as “the region is collectively affected by both agricultural and traffic activities”. Thus, conducting source identification and apportionment based on single or aggregated land use types may obscure the specific differences in pollutant sources across different functional zones (e.g., industrial, agricultural, and residential areas), thereby hindering accurate diagnosis of heavy metal pollution causes in specific regions [18].

Therefore, this study takes the BTH urban agglomeration in North China as a case study. This region serves as a representative area where intensive coal consumption, concentrated industrial activities, dense transportation networks, and intensive agricultural development have all been identified as major anthropogenic drivers of regional soil contamination by heavy metals, such as cadmium (Cd) and lead (Pb) [14]. This study analyzed and determined the content of seven heavy metal elements (As, Cr, Cd, Cu, Ni, Zn, and Pb) in soils from three typical land use types: urban green spaces, croplands, and vegetable fields/orchards. The study sets two main objectives: (1) to analyze the differences in heavy metal content among different land use types at the large scale of an urban agglomeration and (2) to apply principal component analysis/absolute principal component scores (PCA/APCSs) to quantitatively identify and compare the contributions of different sources to soil heavy metals across these land use types at the urban agglomeration scale, thereby providing an in-depth exploration of the relationship between human activities and heavy metal accumulation under different land uses. This study aims to provide scientific foundations and technical support for developing sustainable soil management strategies and targeted heavy metal pollution mitigation measures, thereby facilitating the healthy and sustainable utilization of soil resources in rapidly urbanizing areas.

2. Materials and Methods

2.1. Study Area

The study area encompasses six major cities within the BTH urban agglomeration: Beijing (BJ), Tianjin (TJ), Baoding (BD), Shijiazhuang (SJZ), Tangshan (TS), and Langfang (LF). Geographically, the northwestern part of the region is dominated by the Yanshan and Taihang Mountain ranges, while the southeastern portion extends into the North China Plain. Climatically, the area experiences a temperate continental monsoon climate, characterized by distinct seasonal variations and pronounced dry–wet cycles. The mean annual temperature ranges from 8.8 to 11.7 °C, with an annual precipitation of 346–789 mm. The predominant soil types include fluvisols, luvisols, and calcic luvisols. Major land use types in the study area comprise croplands, built-up areas, forests, grasslands, and water bodies. Over recent decades, the region has undergone rapid urbanization and industrialization, leading to dramatic land use changes.

2.2. Sample Collection and Processing

Based on urban–rural patterns and land use characteristics within the BTH region, 182 sampling sites were established along urban–rural environmental gradients across six cities (Figure 1). The specific land use type for each site was determined by dominant land use, resulting in the collection of 92 urban green space samples (parks, residential areas, and roadside greenbelts), 71 cropland samples (land used for cultivating staple cereals), and 19 vegetable/orchard samples (land used for cultivating vegetables or fruits). Surface soil samples (0–10 cm depth) were collected at each site using a stainless-steel auger, homogenized through quartering, and approximately 1 kg of soil was sealed in polyethylene bags for laboratory analysis. In the laboratory, samples were air-dried at room temperature and then sieved using a nylon sieve with a 2.0 mm screen to eliminate any debris, gravel, and foreign substances. Subsequently, 100 g of the samples was crushed with a wooden roller and sieved through a 0.15 mm nylon mesh for subsequent analysis. Soil digestion followed the aqua regia microwave digestion method (US EPA Method 3051A) using a microwave digestion system (CEM Corporation, Matthews, NC, USA), and the content of As, Cd, Cr, Cu, Ni, Pb, and Zn was determined using inductively coupled plasma mass spectrometry (ICP-MS, NexION 350, PerkinElmer, Springfield, IL, USA). The standard reference material (GBW07408a/GSS-8a [19] and GBW07427/GSS-13 [20]) obtained from the Center of National Standard Reference Material of China was used in the digestion and determination as part of the quality assurance (QA) protocol. Reagent blanks and analytical duplicates were included to ensure the accuracy and precision of analysis.

2.3. Source Apportionment Method (APCS-MLR Model)

The APCS-MLR model, proposed by Thurston and Spengler in 1985, is a receptor model that combines multiple linear regression for the quantitative identification and apportionment of pollution sources, in which the measured pollutant contents are used as the dependent variable, and the results of principal component analysis (PCA) serve as independent variables [21]. The computational procedure begins with Z-score standardization of heavy metal content data using Equation (1):

$$X_{ij}={\displaystyle \frac{C_{ij}-\overline{C}}{{\sigma }_i}}$$

(1)

where X_ij represents the standardized value of heavy metal content (dimensionless), C_ij denotes the content of heavy metal i in sample j (mg·kg⁻¹), $\overline{C}$ signifies the mean content of heavy metal i (mg·kg⁻¹), and σi indicates the standard deviation of the distribution for heavy metal i.

Subsequently, artificial samples with zero content are introduced to calculate the factor scores for these zero-content samples, computed according to the following formula:

$$X_{0i}={\displaystyle \frac{0-\overline{C_i}}{{\sigma }_i}}=-{\displaystyle \frac{\overline{C_i}}{{\sigma }_i}}$$

(2)

The absolute principal component scores (APCS_k) for each sample are then obtained by subtracting the factor scores of the zero-content artificial samples from those of the actual samples (Equation (3)):

$$APC{S}_k=X_{0i}-X_{i\mathrm{j}}$$

(3)

where k represents the number of factors derived from principal component analysis, with each factor corresponding to a distinct pollution source, and X_0i represents the zero-content sample of heavy metal element i.

Finally, multiple linear regression analysis is conducted with the absolute principal component scores (APCS_k) as independent variables and individual heavy metal content as dependent variables, enabling the calculation of contribution rates for each pollution source according to Equation (4):

$$C_i=b_{j0}+{\displaystyle \sum _{k=1}^n(b_{jk}\times }APC{S}_k)$$

(4)

where b_j0 represents the constant term in the multiple linear regression for heavy metal j, indicating contributions from unexplained sources; b_jk denotes the regression coefficient of pollution source k for heavy metal j; and the product b_jk × APCS_k quantifies the contribution rate of pollution source k to the content of element i.

It is noteworthy that prior to conducting principal component analysis (PCA), the suitability of the dataset should be verified using the Kaiser–Meyer–Olkin (KMO) test and Bartlett’s test of sphericity. Only when the overall KMO value exceeds 0.6 and Bartlett’s test of sphericity reaches a significance level (p < 0.05) can the dataset be considered suitable for PCA.

2.4. Statistical Analyses

The principal component analysis and linear regressions were conducted using SPSS 22 software. Maps of the sampling locations and spatial distribution of heavy metals were generated using ArcGIS 10.7, while other figures were produced using Excel and Origin 2022 software.

3. Results

3.1. Characteristics of Soil Heavy Metal Contents

Statistical results of soil heavy metal contents in the study area are presented in

Table 1. Descriptive statistics of soil heavy metals in the BTH region.

Category		As	Cr	Cd	Cu	Ni	Pb	Zn
All Samples (n = 182)	min	5.42	39.8	0.06	14.01	17.99	12.85	45.54
	max	17.11	300.53	4.47	201.83	53.11	173.81	552.35
	mean	9.65	63.43	0.21	29.86	30.56	30.31	97.08
	CV	0.22	0.36	1.55	0.59	0.24	0.57	0.57
Urban Green Spaces (n = 92)	min	0.22	0.36	1.55	0.59	17.99	0.57	0.57
	max	14.39	142.66	0.55	106.38	44.61	123.34	324.54
	mean	9.17	59.49	0.17	28.16	28.58	30.23	87.72
	CV	0.17	0.23	0.42	0.44	0.18	0.58	0.43
Croplands (n = 71)	min	6.46	39.8	0.08	15.53	18.31	14.13	49.77
	max	17.11	105.09	0.36	50.67	51.58	55.63	198.45
	mean	9.99	63.25	0.2	26.73	31.61	27.92	89.28
	CV	0.26	0.23	0.29	0.27	0.26	0.26	0.31
Vegetable Field/Orchard (n = 19)	min	7.85	51.35	0.15	28.73	24.89	20.02	85.01
	max	16.11	300.53	4.47	201.83	53.11	173.81	552.35
	mean	10.63	83.22	0.47	49.74	36.28	39.61	171.58
	CV	0.23	0.65	2.08	0.82	0.25	0.84	0.68
Background Value [22]		10.50	71.20	0.10	23.90	29.90	21.90	81.90

. The mean contents (mg/kg) of seven heavy metals in soils were as follows: As (9.65), Cr (63.43), Cd (0.21), Cu (29.86), Ni (30.56), Pb (30.31), and Zn (97.08). Compared with the regional background values in the BTH area [22], the mean contents of As, Cr, and Ni were either lower than or comparable to background levels, with most coefficients of variation (CVs) ≤ 0.3, indicating minimal anthropogenic influence. In contrast, the mean contents of Cd, Cu, Pb, and Zn were 2.1, 1.25, 1.38, and 1.19 times the background values, respectively, with coefficients of variation (CVs) ranging from 0.57 to 1.55, indicating moderate to high variability [23]. This suggests obvious exogenous inputs for these four elements.

3.2. Characteristics of Soil Heavy Metal Contents in Different Land Use Types

Significant variations in heavy metal contents were observed among different land use types (Table 1, Figure 2). Vegetable fields/orchards soils exhibited the highest mean contents (mg/kg) for all measured elements—As (10.63), Cr (83.22), Cd (0.47), Cu (49.74), Ni (36.28), Pb (39.61), and Zn (171.58)—surpassing levels found in both urban green spaces and croplands. Notably, with the exception of As and Ni, they demonstrated substantial spatial heterogeneity, as indicated by variation coefficients ranging from 0.65 to 2.08. This pronounced variability may be attributed to intensive agricultural practices characterized by high multiple cropping indices, frequent tillage operations, and substantial agrochemical inputs, all of which elevate heavy metal accumulation risks [24]. Furthermore, the predominant location of vegetable plots within peri-urban areas renders them susceptible to mixed pollution sources from industrial, agricultural, and residential activities [25].

Results showed that there were also significant differences in the accumulation and distribution of heavy metals between urban green space soils and cropland soils. The average contents of Cu and Pb in urban green space soils were 1.05 and 1.08 times those in cropland soils, respectively, while the average contents of As, Cr, Cd, Ni, and Zn in cropland soils were 1.09, 1.06, 1.18, 1.11, and 1.02 times those in urban green space soils, respectively. Although the average contents of most heavy metals in the two land use types were relatively close, the coefficients of variation (CVs) of Cd, Cu, Ni, Pb, and Zn in urban green space soils ranged from 0.36 to 0.58, indicating a moderate variation level. In contrast, the variation levels of all soil heavy metals in croplands were relatively low (CV < 0.3). This suggests that the spatial variability of soil heavy metal contents in urban green spaces in the study area is relatively large, especially for elements such as Cu, Pb, and Zn, which show significantly more abnormal high values in urban green space soils, while their distributions in croplands are relatively homogeneous. This may be attributed to the impacts of various urbanization and industrialization processes: urban soils are affected by complex and diverse sources of anthropogenic disturbances, and the use of exogenous imported soils during urban construction has further exacerbated the spatial heterogeneity of soil heavy metals [26]. In contrast, cropland soils are used for growing field crops with similar agronomic management practices, which have consistent impacts on the soil environment.

3.3. Source Apportionment of Soil Heavy Metals Under Different Land Use Types

3.3.1. Source Identification

Following data standardization, the suitability of principal component analysis (PCA) was evaluated using the Kaiser–Meyer–Olkin (KMO) measure and Bartlett’s test of sphericity. All three land use types exhibited KMO values exceeding 0.6 with significance levels (p < 0.001), confirming the appropriateness of PCA for each dataset. To minimize information loss during factor extraction, components were retained based on a cumulative variance explanation threshold of >85%, ensuring comprehensive representation of source characteristics (

Table 2. Principal component extraction for soil heavy metals in three land use types.

Category	Urban Green Spaces			Croplands			Vegetable Fields/Orchards
	PCA1	PCA2	PCA3	PCA1	PCA2	PCA3	PCA1	PCA2
As	0.52	0.11	0.72	0.68	−0.02	0.67	0.90	0.03
Cr	0.09	0.86	0.38	0.91	0.18	0.12	0.92	0.12
Cd	0.38	0.84	0.08	0.16	0.59	0.75	0.21	0.87
Cu	0.88	0.31	0.22	0.79	0.44	0.26	0.21	0.94
Ni	0.01	0.25	0.91	0.91	0.22	0.22	0.94	0.01
Pb	0.91	0.11	0.09	0.65	0.40	0.36	0.84	0.28
Zn	0.70	0.56	0.09	0.27	0.91	0.15	−0.10	0.88

). Factor loadings were classified into three significance levels for interpretation: strong (>0.75), moderate (0.5–0.75), and weak (0.3–0.5) [27].

(1)

Urban Green Space soils

Three principal factors were extracted from urban green space soils, collectively explaining 85.66% of the total variance (Table 2). The first component (PC1), accounting for 35.74% of variance, showed strong loadings (>0.75) for Cu and Pb, moderate loadings (0.5–0.75) for Zn and As, and weak loadings (0.3–0.5) for Cd. Numerous studies have confirmed traffic emissions as dominant sources of urban Cu, Pb, and Zn [28], with specific mechanisms including tire wear (Cu, Zn, and Cd), brake lining abrasion (Cu and Pb), lubricant leakage (Zn and Cd), and automotive component corrosion (Cu and Zn) [29]. Although leaded gasoline was banned in China post-2000, Pb persists as a legacy pollutant [30,31,32], while coal combustion from power plants and heavy industries (e.g., steel, chemicals) emits substantial Pb, Zn, As, and Cd [33], particularly in the coal-dependent BTH region. The spatial distribution patterns (Figure 2d,f,g) showed that elevated Cu, Pb, and Zn contents in older urban parks, historic residential areas, and roadside green spaces strongly correlated with traffic density and industrial coal use. Therefore, PC1 is interpreted as representing a mixed source of traffic emissions and coal combustion byproducts.

The second principal component (PC2) accounted for 26.69% of the total variance, with strong loadings of Cd and Cr, moderate loading of Zn, and weak loading of Cu. Cadmium is widely recognized as a tracer element for ore smelting and refining processes [34], while steel production—particularly electric arc furnace operations using scrap steel as raw material—represents a major emission source of both Cd and Cr [35]. Additional industrial activities, including stainless steel/alloy manufacturing, electroplating, chemical production, textile dyeing, energy generation, and waste incineration, contribute significantly to Cr-, Cd-, Zn-, and Cu-bearing waste streams [36]. Spatial distribution patterns (Figure 2c,b) revealed elevated Cd and Cr contents clustered near the Tanggu industrial district of Tianjin, a major northern China industrial hub hosting metallurgical, petrochemical, coal-fired power, machinery manufacturing, and residential heating operations. The substantial impacts of industrial waste gas, wastewater, and solid waste on soil heavy metal accumulation align with previous findings from Tanggu [37], confirming PC2 as an industrial source.

The third principal component (PC3) explained 22.23% of variance, characterized by strong loadings of As and Ni with weak Cr contributions. The observed low mean contents and variation coefficients of As, Cr, and Ni in urban green spaces, coupled with their minimal enrichment relative to background levels, suggest stable accumulation processes dominated by natural factors. Existing research confirms that these elements primarily derive from pedogenic parent materials and natural weathering processes, with limited anthropogenic influence [26]. For this reason, we attribute PC3 to natural lithogenic sources.

(2)

Cropland Soils

Three principal factors were extracted from cropland soils, collectively explaining 87.85% of the total variance. PC1 contributed 46.53% of the total variance, with strong loadings of Cr, Ni, and Cu, and moderate loadings of As and Pb. Similar to PC3 in urban soils, the mean contents of As, Cr, and Ni in cropland soils were generally below or comparable to the background levels, with low coefficients of variation (<0.3), suggesting PC1 predominantly derives from natural pedogenic processes.

PC2 contributed 22.83%, exhibiting strong Zn loading, moderate Cd loading, and weak Cu and Pb loadings. Unlike in urban areas with dense road networks, traffic emissions generally have a limited contribution to heavy metals in cropland soils. The spatial distribution of soil Zn (Figure 2g) showed that high-content areas of Zn in cropland soils were mainly distributed in Tianjin, particularly along the Haihe River and the banks of the Long River from Beijing to Tianjin, with relatively higher levels also observed near the Dashi River in Beijing. Tianjin has a dense river network and is located in the lower reaches of the study area, where river water is the main irrigation source. Croplands in this area often rely on upstream urban drainage for irrigation [30]. Meanwhile, these river areas are also home to many industrial enterprises, such as metal smelting and machinery manufacturing, which may discharge pollutants containing large amounts of Zn, Cd, Cu, Pb, and other elements [31]. These pollutants may enter cropland through atmospheric deposition, surface runoff, and irrigation, which is likely the main cause of heavy metal accumulation in soils. Therefore, it is inferred that PC2 mainly originates from industrial activities.

PC3 contributed 18.49%, exhibiting strong Cd loading, moderate As loading, and weak Pb loading. Agricultural Cd and As accumulation primarily correlates with the application of fertilizers and pesticides [38,39]. A study from Yu et al. also revealed that the average Cd and As contents in phosphate fertilizers in China can reach 0.91 mg/kg and 19.83 mg/kg, respectively, constituting a significant source of heavy metal accumulation in agricultural soils [39]. The BTH region has a long history of agriculture, characterized by extensive use of phosphate fertilizers and pesticides. Unlike other heavy metals, cadmium (Cd) levels were commonly found to exceed background value, even at many sampling sites with few surrounding industrial enterprises that relied on well water irrigation, indicating relatively single pollution sources. This consistent pattern of Cd enrichment across diverse sites suggests that PC3 likely represents a source from agricultural activities.

(3)

Vegetable fields/orchards soils

In contrast to urban green spaces and croplands, two principal components explained 83.60% of the total variance in vegetable fields/orchards soils. PC1 contributed 47.90% of the total variance with strong loadings of As, Cr, Ni, and Pb. Similar to PC1 in cropland soils, mean contents of As, Cr, and Ni were comparable to or slightly higher than background values, and after excluding one extreme outlier, all four elements exhibited coefficients of variation below 0.3, indicating PC1 primarily originates from natural lithogenic sources.

PC2 explained 35.71%, with strong loadings of Cd, Cu, and Zn. The elevated mean contents and high coefficients of spatial variation for these three heavy metals in soil collectively indicate significant exogenous inputs. This may be related to the extensive application of agricultural inputs such as chemical fertilizers, pesticides, organic fertilizers, and soil conditioners [40]. For example, phosphate fertilizers such as calcium superphosphate and ammonium dihydrogen phosphate contain Cd, Zn, Cu, and Cr [41], while Cu/Zn-based fungicides, including Bordeaux mixture, oxine–copper, and zinc thiazole, are routinely applied in vegetable cultivation [42]. Heavy metal-containing additives are also commonly added to livestock feed to enhance animal growth rates and immune responses [41]. Notably, most of the sampled vegetable plots in the study area employed greenhouse cultivation systems, which typically exhibit reduced atmospheric deposition inputs. However, these systems paradoxically demonstrated higher Cd, Cu, and Zn contents compared to open-field cultivation. This counterintuitive pattern strongly suggests that PC2 primarily represents agricultural chemical inputs.

3.3.2. Source Apportionment of Soil Heavy Metals Using the APCS-MLR Model

Multiple linear regression analysis based on the APCS-MLR model revealed that the goodness-of-fit (R²) for seven heavy metals (As, Cr, Cd, Cu, Ni, Pb, and Zn) exceeded 0.75 in all land use types, except for Pb in cropland soils, which had an R² of 0.69, and the ratio of predicted values to measured mean values was basically close to 1 (Figure 3). These results demonstrate that the APCS-MLR approach provides a credible and accurate estimation of source contributions.

The analytical results revealed distinct dominant sources across different land use types (Figure 4). Overall, natural sources consistently demonstrated the highest contribution rates (32.62–70.26%) for As, Cr, and Ni across all three land use types, suggesting these heavy metals primarily originate from weathering of natural parent materials in the BTH region. Notably, the contribution from industrial sources to Cr in urban green spaces was 37.12%, slightly exceeding that of natural sources (33.62%). Meanwhile, agricultural activities accounted for 33.17% of the total As in croplands. This result indicates that urban industrialization enhances Cr accumulation, whereas agricultural intensification promotes As enrichment.

The combined influence of traffic and coal combustion sources accounted for 40.28–66.26% of the total deposition of Pb and Cu in urban green spaces, markedly surpassing the contribution from other sources. Industrial activities showed the highest contribution rates to Zn (45.88%) and Cd (65.25%) in urban green spaces, while their contributions to cropland soils are 49.73% for Zn and 29.76% for Cd, respectively. These results demonstrate that industrial development has substantially enhanced regional Zn and Cd accumulation. Notably, the contributions of industrial sources to urban green spaces were generally higher than those to cropland soils, likely due to proximity effects and industrial structure variations, as most industrial activities are concentrated in or near urban areas, with their influence declining with increasing distance from industrial facilities [43].

Agricultural activities accounted for 41.68–51.32% of the total Cd in both cropland and vegetable field/orchard soils, consistent with findings reported by Peng et al. [35]. Cd is regarded as a characteristic element in agricultural fertilizers, and agricultural activities can significantly enhance its accumulation in soils. However, the contribution rates of agricultural activities to Cu and Zn in vegetable fields/orchards soils (46.62–55.58%) substantially exceeded those in cropland (9.21–13.40%), demonstrating distinct accumulation patterns under different agricultural intensities. The markedly higher fertilizer application rates in vegetable cultivation systems constitute the primary driver for heavy metal accumulation [24].

Notably, unexplained sources accounted for considerable proportions (18.64–42.59%) of all seven heavy metals in vegetable fields/orchards soils, while croplands and urban green spaces showed fewer unexplained elements with lower percentages. This pattern suggests high complexity in heavy metal sources of vegetable fields/orchards soils, potentially attributable to (1) variability in livestock manure composition and its pretreatment methods [44] and (2) compounded influences from the diverse peri-urban industrial activities that might further complicate heavy metal accumulation patterns in peri-urban vegetable fields/orchards systems.

4. Discussion

4.1. Relationship Between Soil Heavy Metal Sources and Land Use Types

Results of this study indicate significant differences in the accumulation levels and sources of heavy metals among different land use types (urban green spaces, cropland, and vegetable fields/orchards) in the BTH urban agglomeration. Although the contents of various heavy metals were comparable between urban green spaces and cropland soils, the spatial variability of Cu, Pb, Cd, and Zn was substantially higher in urban green spaces, suggesting more pronounced accumulation of these elements. The APCS-MLR model revealed that Cu and Pb in urban green spaces were primarily derived from a mixed source of traffic and coal combustion, Cd mainly originated from industrial emissions, and Zn was co-influenced by industrial activities and the mixed traffic/coal combustion source. In contrast, all heavy metals in cropland soils exhibited low coefficients of spatial variation, indicating limited influence from external point sources. The APCS-MLR results further identified natural pedogenic parent materials as the dominant source for most heavy metals in cropland soils. However, Cd was primarily attributed to agricultural activities, and Zn was mainly attributed to industrial sources; the agricultural origin of As also warranted attention. Vegetable fields/orchards, which had the highest heavy metal accumulation, showed significant enrichment of Cd, Cu, and Zn, largely due to agricultural practices. For the remaining elements, natural parent materials remained the primary source. Nevertheless, likely due to the complexity of pollution sources in these peri-urban systems, the contribution of unknown sources constituted a substantially larger proportion for multiple heavy metals in vegetable field and orchard soils.

Long-term emissions from activities such as industry, transportation, and coal burning have become the main sources of Pb, Cu, Zn, and other heavy metal accumulation in urban green space soils. With ongoing urbanization, soil disturbances, and the introduction of imported soil may further increase the spatial heterogeneity of urban green spaces. Cropland, focused on large-scale grain production with relatively low intensity of use, is mainly influenced by a combination of regional-scale factors, such as regional atmospheric deposition, chemical fertilizer application, and irrigation water quality. These factors generally result in lower levels of heavy metal accumulation and relatively homogeneous distribution. However, the historical use of Cd-containing phosphate fertilizers, As-based pesticides, and long-term wastewater irrigation continues to have persistent legacy effects, which may represent significant sources of heavy metals in cropland soils. To meet the increasing food demand of urban residents, peri-urban vegetable fields typically adopt highly intensive agricultural production, characterized by the extensive application of livestock manure, chemical fertilizers, and pesticides. At the same time, urban expansion has introduced transportation and industrial emissions, along with the diversification and fragmentation of landscape types, inevitably adding new exogenous pollution sources. The dual pressures of intensive agricultural inputs and urban development collectively exacerbate the composite risks of heavy metal pollution in vegetable field soils.

As is evident, different land use types reflect distinct socio-economic activities and support varied material flows and metabolic processes, consequently leading to spatial heterogeneity in the dominant pollution sources, types, and content of soil heavy metals. Liu et al.’s study in Datong found that heavy metals in urban park soils primarily originated from traffic and coal combustion activities, while Khan et al. identified industrial emissions as the main source of Cd pollution in urban park soils in Pakistan. These findings are consistent with the results of this study. However, Cai et al. [6] found that mixed traffic/agricultural sources were the primary contributors to heavy metal pollution in urban park soils. Jin et al. [8] found agricultural activities to be a significant source of Zn in soils, whereas this study indicates that Zn is minimally influenced by agricultural activities, creating a contrast that may be attributed to greater regional differences in pollution sources. Another study on agricultural land (primarily vegetable fields and orchards) in Chengdu found that livestock manure and agricultural activities significantly contributed to various heavy metals such as Ni, Cu, Zn, Pb, As, and Cd [31]. However, in peri-urban areas, which have high-intensity human disturbance, the proportion of unexplained source contributions significantly increased. These results indicate that, at the city or urban agglomeration, extensive application of livestock manure resulting from urbanization and agricultural intensification is a major source of heavy metals in vegetable field soils. Nevertheless, the complex landscape diversity in peri-urban areas makes the source apportionment of heavy metals more challenging, necessitating further refined research in the future, potentially incorporating methods such as isotope analysis.

4.2. Strengths and Limitations of the Study

This study employed the APCS-MLR model to reveal the sources and contribution variations in soil heavy metals across different land use types in an urban agglomeration area. The findings offer a robust scientific foundation and technical support for formulating differentiated and refined pollution control strategies for soil heavy metals. The established model demonstrated robust performance, with regression coefficients (R²) of over 0.75 for seven heavy metals (As, Cr, Cd, Cu, Ni, Pb, and Zn) across all land use types, except for Pb in cropland soil. The predicted contents showed strong agreement with measured values, confirming the validity and accuracy of the research results. However, the study still has some limitations. While classifying land use types enhances the understanding of spatial heterogeneity, it does so at the expense of depicting the overall regional pollution pattern and cross-boundary transport processes. Moreover, it cannot fully elucidate the complex interactions among multiple sources of specific heavy metals. A key challenge is to develop new-generation integrated models capable of simultaneously simulating local pollution and regional-scale multimedia transport (e.g., atmospheric deposition from industrial to agricultural areas). Integrating receptor models with atmospheric dispersion or hydrological models offers a promising avenue for future research.

5. Conclusions

The results of this study allowed the following conclusions to be drawn:

• The mean contents of As, Cr, and Ni in the study area were either below or close to their background values, with most coefficient of variation (CV) values ≤ 0.3, indicating minimal anthropogenic influence. In contrast, Cd, Cu, Pb, and Zn exhibited mean contents 1.19–2.1 times their background values, accompanied by moderate to high variability (CV: 0.57–1.55), demonstrating significant exogenous inputs.
• Distinct differences in heavy metal contents were observed among different land use types. Vegetable soils showed the highest average heavy metal contents, while urban green spaces and croplands exhibited comparable mean values. Urban green spaces displayed strong spatial heterogeneity in heavy metal distribution, reflecting pronounced urbanization impacts, whereas cropland soils maintained relatively homogeneous patterns.
• APCS-MLR analysis revealed that natural sources dominated contributions to As, Cr, and Ni (32.62–70.26%), except for Cr in urban green spaces. Combined traffic emissions and coal combustion constituted primary sources for Cu and Pb in urban green spaces (40.28–66.26%). Agricultural activities contributed similarly to Cd accumulation in cropland and vegetable fields/orchards soils (34.29–41.68%). Notably higher agricultural contributions to Cu and Zn were observed in vegetable fields/orchards soils (31.18–55.33%) versus conventional cropland soils (9.21–13.40%), reflecting differential fertilizer application intensities.
• At the urban agglomeration scale, significant disparities exist in predominant pollution sources and their respective contribution levels across different land use types. Urbanization and traffic emissions constitute the primary drivers of heavy metal accumulation in urban green spaces, whereas the degree of heavy metal accumulation in cropland soils demonstrates close correlation with cultivation types and intensity levels. The establishment of land use-specific heavy metal input inventories, coupled with the implementation of differentiated management strategies, offers a scientifically grounded pathway to address the complex challenges of coordinated soil pollution control in evolving urban landscapes.