Introduction
Potentially toxic elements (PTEs) in atmospheric dust pose significant ecological and health risks (Sobhanardakani Reference Sobhanardakani2017, Khalijian et al. Reference Khalijian, Lorestani, Sobhanardakani, Cheraghi and Tayebi2022, Selahvarzi & Sobhanardakani Reference Selahvarzi and Sobhanardakani2022), driving extensive research into their concentrations, distribution, pollution levels, sources and associated ecological risks (Hao et al. Reference Hao, Lu, Yu, Yang, Lei, Pan and Wang2023). Accurate pollution assessment and source identification are fundamental for effective pollution control and resource prioritization (Zhao et al. Reference Zhao, Li and Li2024).
Current methods for assessing pollution levels and ecological risks primarily rely on element concentrations and toxicity coefficients, implemented through a suite of indices such as the modified contamination degree (mCd), geo-accumulation index (Igeo), pollution load index (PLI), toxic risk index (TRI) and potential ecological risk index (PERI; Wang et al. Reference Wang, Huo, Wang, Lin, Liu, Huang and Lin2023b). Among these, the Igeo, PERI, mCd and PLI were recognized as superior to the contamination factor (Cf) for their enhanced accuracy and objectivity in evaluating PTE status (Li et al. Reference Li, Qian, Xu, Zhang, Chen, Hou and Chen2022b). Due to the limitations of individual methods, recent studies have highlighted that multivariate statistical methods, which incorporate multiple indicators, offer a more rational and efficient framework for assessing PTE pollution risks in environmental substrates, as they account for the complex interplay of factors that single-index approaches may overlook (Jiang et al. Reference Jiang, Pan, Lin, Wang, Wang, Liu and Lin2022).
Source identification is the primary step in the prevention and control of PTE pollution (Zhao et al. Reference Zhao, Li and Li2024) because not all pollution sources can be managed promptly due to limitations in resources and costs. Conventional statistical techniques such as correlation, principal component analysis and cluster analysis can determine potential sources qualitatively (Lei et al. Reference Lei, Li, Guo and Ju2022), yet they are not able to accurately evaluate how much each source contributes to a specific rate (Yang et al. Reference Yang, Lu, Fan, Yu, Wang, Lei and Zuo2023). Positive matrix factorization (PMF) is a receptor model that is widely used to overcome this limitation (Zhao et al. Reference Zhao, Li and Li2024). Nonetheless, its reliance on the input concentration dataset (i.e., the measured levels of PTEs) may not accurately represent the actual source contributions due to the complex solubility of trace elements (Galvão et al. Reference Galvão, de Cassia Feroni and Orlando2021). Regional pollution is the result of the combined effects of various land-use types, including industrial, mining, agricultural and transportation activities (Li et al. Reference Li, Dong and Ren2015). The correlation between the land-use areas surrounding the sampling sites and PTE concentrations can corroborate the accuracy of source results derived from methods such as PMF; however, such validation techniques have apparently not been developed.
Guiyang is a major central city in south-west China; it is a hub for regional innovation and a prominent destination for ecological leisure tourism within the country. Its distinctive industrial framework places considerable pressure on air quality. Tree leaves, including both evergreen and deciduous varieties, can act as effective passive pollution collectors by providing a large surface area for particles to settle upon (Norouzi et al. Reference Norouzi, Khademi, Cano and Acosta2016). The tree Osmanthus fragrans (sweet osmanthus) accounts for c. 10% of the tree species utilized in urban greening projects in Guiyang, is widely found in urban settings such as streets, residential areas and parks and is thus a highly suitable indicator of spatiotemporal variations in atmospheric particulate matter and its associated PTEs. Here, we use the leaf dust of O. fragrans to: (1) explore the spatial distribution and levels of the atmospheric PTEs; (2) identify potential sources of these PTEs based on the PMF model; and (3) corroborate the source apportionment results of the PMF by correlating PTE concentrations with land-use areas around sampling sites.
Methods
Sampling and analysis
Leaf samples were collected from O. fragrans trees in the greenbelts situated on both sides of urban traffic thoroughfares at a minimum distance of 50 m from the roadway. Due to the geographical distribution of the greenbelts and O. fragrans trees, the distances between sampling sites varied (Appendix S1 & Fig. S1), with the range being c. 2–5 km. Within each 100-m radius site, we gathered five individual samples, each consisting of c. 30 leaves from a single tree. These five samples were then combined into one composite sample per site, providing a representative measure of site-specific dust deposition. Overall, 105 leaf dust samples (21 composite samples) were collected. Samples were sealed in plastic bags and transported to the laboratory for analysis. The atmospheric dust on the leaf surface was then washed away using ultrasonic oscillation for 20 min to ensure that the particles were completely removed from the surface. The cleaning solution was filtered using 0.2 μm quartz filter paper. Afterwards, the filter paper was dried, and the dust was brushed off and weighed. Concentrations of the metals Cr, Ni, Cu, Zn, As, Cd and Pb in the dust were measured at the Institute of Geochemistry, Chinese Academy of Sciences (IGCAS, China; test conditions, methods and instrument details are given in Appendix S1).
To evaluate the influence of different anthropogenic emission sources on atmospheric PTE pollution, we captured the spatial distribution map of land-use types in urban Guiyang using Baidu Maps (https://lbsyun.baidu.com/) during the sampling period. Utilizing the Create Fishnet tool in ArcGIS, we quantified the areas of the four most predominant land-use types, including industrial area (IA), governmental area (GA), commercial area (CA) and agricultural area (AA), within a 2-km radius of each sampling site.
Evaluation of pollution status and source identification
We assessed the pollution levels of PTEs in leaf dust using the single-factor pollution index (Pi), Igeo, enrichment factor (EF) and Nemerow pollution index (NPI; explained in Appendix S1), including the grading criteria for each (Appendix S1 & Table S1). The PMF model was then applied to identify PTE sources, following the methodology of Su et al. (Reference Su, Yang, Han, Bai, Luo and Wang2024). Pearson’s correlation was utilized to quantify relationships between land-use areas surrounding the sample sites and PTE concentrations in leaf dust in order to compare this information with the PMF source apportionment results.
Results
Descriptive statistics and distribution of PTEs
As illustrated in Table 1, the concentration ranges of PTEs were 56.4–405 mg/kg for Cr, 12.2–50.9 mg/kg for Ni, 46.2–404.0 mg/kg for Cu, 277.8–1280 mg/kg for Zn, 7.20–20.2 mg/kg for As, 0.91–2.86 mg/kg for Cd and 48.2–375.5 mg/kg for Pb. The mean concentrations of Cr, Cu, Zn, Cd and Pb were, respectively, 1.3, 4.8, 5.2, 4.4 and 3.9 times higher than their respective background values (China National Environmental Monitoring Center 1990). Conversely, the average concentrations of Ni and As were below their corresponding background levels, suggesting a predominantly natural origin for these elements. With the exception of As, the coefficient of variation of all PTEs exceeded 30%, indicating substantial spatial variation.
Concentrations (mg/kg) of potentially toxic elements from leaf dust in urban Guiyang.
BV = soil background value of Guizhou Province; CV = coefficient of variation; SD = standard deviation.
Nearly all PTEs exhibited higher concentrations in the north-western (sample site 11) and southern (sample site 14) regions of the city (Appendix S1 & Fig. S1). Significantly, elevated levels of Cr, Ni and Cu were also observed in the central area (sample site 3). The highest concentration of Zn occurred in the north-eastern part (sample site 8). High-concentration areas were primarily located at intersections of major urban roads or stations. The spatial distribution patterns of the seven PTEs were broadly similar, indicating that PTE accumulation may be influenced by a common dominant factor. However, site-specific differences suggested variations in pollution sources and intensities, as well as differences in local environmental conditions (Li et al. Reference Li, Dong and Ren2015).
Assessment of PTE pollution
The mean Pi of PTEs in leaf dust decreased in the following order: Zn (5.23) > Cu (4.85) > Cd (4.41) > Pb (3.94) > Cr (1.30) > As (0.67) > Ni (0.60) (Fig. 1). Based on Appendix S1 & Table S1, Ni and As were classified as entirely uncontaminated to slightly contaminated (Pi ≤ 2), as were 95% of the sites for Cr. In contrast, Cu, Zn, Cd and Pb exhibited notable contamination: 48%, 52%, 62% and 33% of the sites were moderately contaminated (3 < Pi ≤ 5), whereas 33%, 43%, 24% and 24% of the sites were strongly contaminated (Pi > 5), respectively (Appendix S1 & Table S2).
Pollution degree and enrichment levels of potentially toxic element based on the pollution index (Pi), geo-accumulation index (Igeo), enrichment factor (EF) and Nemerow pollution index (NPI).
Based on mean Igeo values (Fig. 1), PTEs were ordered by pollution level as follows: Zn (2.88) > Cu (2.64) > Cd (2.63) > Pb (2.36) > Cr (0.82) > As (−0.03) > Ni (−0.28). Ni and As were assessed as uncontaminated, Cr as slightly contaminated and Cu, Zn, Cd and Pb as moderately contaminated. Zn showed the most severe accumulation, with 29% of sites being moderately to strongly contaminated. The Igeo evaluations indicated low Ni and As pollution levels, followed by Cr, whereas Cu, Zn, Cd and Pb were identified as the primary pollutants in leaf dust.
The mean EF for PTEs in leaf dust decreased in the following order: Zn (7.13) > Cu (6.61) > Cd (6.01) > Pb (5.37) > Cr (1.79) > As (0.92) > Ni (0.81) (Fig. 1). According to Appendix S1 & Table S1, Ni and As exhibited negligible enrichment (EF < 1), whereas the majority of sampling sites indicated that Cr was slightly enriched. In contrast, Zn, Cu, Cd and Pb demonstrated significant enrichment (EF ≥ 5), indicating that these were strongly linked to anthropogenic activities (Su et al. Reference Su, Shen, Yang, Liu, Yang and He2025).
The NPI – an integrated assessment of the overall pollution level that considers all monitored PTEs – showed mean values in leaf dust decreasing in the following order: Zn (9.78) > Cu (9.56) > Pb (8.08) > Cd (6.87) > Cr (3.12) > As (1.01) > Ni (0.86). Ni and As fell within the ‘slight contamination’ category (NPI ≤ 2), Cr showed ‘moderate contamination’ and Zn, Cu, Cd and Pb all demonstrated ‘strong contamination’ levels (NPI > 5), indicating that they were the primary contributors to the overall pollution.
PMF source profiles and corroboration
The PMF model elucidated three discrete source factors contributing to the observed PTEs. As illustrated in Fig. 2, the quantified variance contributions of these factors were 39.41% for Factor 1, 20.17% for Factor 2 and 40.42% for Factor 3. Each factor presented a distinct geochemical profile: Factor 1 exhibited dominant loadings of As (55.27%) and Cr (52.25%); Factor 2 was principally defined by a high loading of Cu (64.57%); and Factor 3 was markedly associated with Pb (66.90%) and Cd (56.57%). Factor 3 emerged as the predominant contributor, accounting for the largest share of the explained variance (40.42%), thereby underscoring its relative significance within the identified source apportionment.
Profiles of the three sources (Factors 1–3) identified by the positive matrix factorization model.
Pearson’s correlation coefficients between potentially toxic element concentrations of leaf dust and the areas of four land-use types surrounding the sampling sites (*p ≤ 0.05, **p ≤ 0.01). AA = agricultural area; CA = commercial area; GA = governmental area; IA = industrial area.
As demonstrated in Fig. 3, the concentrations of Pb and Cu were positively associated with both GA and CA (p ≤ 0.05), with the strongest correlation observed between Pb and GA (r = 0.63). Conversely, almost all metals (except Zn) showed negative or non-significant correlations with IA or AA (p > 0.05). These results align with the PMF-derived source profiles and reinforce the link between certain PTE concentrations and specific anthropogenic activities.
Discussion and conclusion
The limited anthropogenic impacts on As and Cr, as indicated by the Pi, Igeo, EF and NPI indices (Fig. 1), highlight their geogenic origins (Li et al. Reference Li, Sun, Yang, Liang, Ma, Huang and Luo2019). Multiple lines of evidence substantiate the geogenic origins of As and Cr. For As, its presence in the Yangtze Estuary sediments was predominantly ascribed to natural geological processes (Han et al. Reference Han, Cheng, Hu, Jiang, Mo, Xu and Wang2017). Similarly, surface erosion and weathering of terrigenous crustal materials were primary contributors to Cr concentrations in marine sediments (Proshad et al. Reference Proshad, Kormoker, Al, Islam, Khadka and Idris2022). Collectively, these studies support the interpretation that Factor 1 represents natural sources.
Cu was one of the four elements with the highest levels of enrichment (Fig. 1), denoting that its presence was closely associated with anthropogenic activities. Coal-fired power generation, printing and dyeing operations and electroplating industries have all been identified as significant contributors to elevated Cu levels (Wang et al. Reference Wang, Zheng, Li and Wang2023a). Additionally, industrial effluents discharged into rivers released Cu at concentrations far exceeding those of other PTEs (Pratap et al. Reference Pratap, Mani and Prasad2020, Niu et al. Reference Niu, Li, Luo, Fu, Chen and Yang2021). Thus, it can be concluded that Factor 2 represents the source contribution of industrial processes.
The assessments for PTEs indicated moderate contamination levels for Pb and Cd (Fig. 1), suggesting that their presence was primarily attributed to anthropogenic sources. Pb accumulation largely stemmed from transportation activities, which are major contributors to PTE pollution (Singh et al. Reference Singh, Bhat, Bhat, Dervash and Ganei2018). Environmental Pb maintains strong vehicular associations, originating primarily from three persistent vectors: (1) historical leaded petrol combustion residues; (2) brake lining abrasion particulates; and (3) mechanical detachment of lead wheel weights (Wang et al. Reference Wang, Zheng, Li and Wang2023a). Cd served as an indicator of agricultural activity (Fei et al. Reference Fei, Lou, Xiao, Ren and Lv2022). The application of phosphate fertilizers, manure and pesticides constitutes the primary source of Cd (Li et al. Reference Li, Yu, Chen, Leng, Zhao, Jia and An2022a). Therefore, Factor 3 represents a blend of traffic and agricultural sources. Given the limited distribution of agriculture plots within urban Guiyang, traffic emissions emerged as the dominant contributor to this factor. The aforementioned characteristics arose from the extensive use of motor vehicles associated with GA and CA, making traffic emissions a significant contributor to atmospheric PTEs. In contrast, due to the relocation of traditional polluting industries from urban Guiyang, IA exerted relatively minor impacts on PTEs. Additionally, the limited AA in urban Guiyang resulted in lower proportional contributions of PTEs from agricultural sources.
Unlike most prior studies that relied solely on their PMF models’ statistical fit, our correlation analysis with land-use areas revealed that PTE concentrations (particularly Pb and Cu) were more strongly associated with GA and CA than with IA or AA (Fig. 3). This provides crucial external validation, significantly strengthening our PMF conclusion that traffic sources are the dominant contributor to atmospheric PTEs in urban Guiyang. Specifically, our results robustly identified traffic emissions (particularly from stop-and-go conditions at major intersections) as the principal sources of atmospheric PTEs (Cu, Zn, Cd, Pb). This mechanistic link aligns with global research indicating that non-exhaust emissions (stemming from brake, tyre and road wear) are becoming major concerns in urban areas (Sagnotti et al. Reference Sagnotti, Taddeucci, Winkler and Cavallo2009, Limo et al. Reference Limo, Paturi and Mäkinen2018). Further supporting this interpretation, the spatial distribution of high-concentration sites was clustered around congested traffic nodes in the north-western and southern parts of urban Guiyang (Appendix S1 & Fig. S1). Thus, our research not only corroborates the findings of Su et al. (Reference Su, Yang, Han, Bai, Luo and Wang2024) through a comprehensive land-use dataset but also provides a spatially explicit evidence base. Accordingly, mitigation strategies for traffic emissions need to evolve to include measures such as optimizing traffic light sequences to minimize engine idling and promoting infrastructure improvements, such as overpasses, at critical congested intersections.
The PMF model identified a mixed source from traffic and agriculture (Factor 3), but future work should use specific markers (e.g., hopanes, Sb or rare earth elements) for clearer sourcing of the PTEs. Our analysis was also temporally limited to a single period of time. Seasonal variations in traffic, agriculture and weather probably influence PTE levels and sources, underscoring the need for long-term monitoring to better understand emission dynamics and inform mitigation. By providing a better-validated and more holistic model for PTE source apportionment, we highlight a reproducible framework that environmental managers can use to identify and prioritize targeted pollution control measures.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S0376892925100295.
Acknowledgements
None.
Financial support
This work was funded by the National Natural Science Foundation of China (No. 41964005) and the Guizhou Provincial Science and Technology Program (zk [2022]023, YWZ [2023]006).
Competing interests
The authors declare none.
Ethical standards
Not applicable.
