Quantitative source apportionment and driver identification of soil heavy metals using advanced machine learning techniques

A comprehensive study of source apportionment, spatial distribution, and health risks assessment of heavy metal(loid)s in the surface soils of a semi-arid mining region in Matehuala, Mexico

BACKGROUND

This study investigates the contamination level, spatial distribution, pollution sources, potential ecological risks, and human health risks associated with heavy metal(loid)s (i.e., arsenic (As), copper (Cu), iron (Fe), manganese (Mn), lead (Pb), and zinc (Zn)) in surface soils within the mining region of Matehuala, located in central Mexico.

OBJECTIVES

The primary objectives are to estimate the contamination level of heavy metal(loid)s, identify pollution sources, assess potential ecological risks, and evaluate human health risks associated with heavy metal(loid) contamination.

METHODS

Soil samples from the study area were analysed using various indices including Igeo, Cf, PLI, mCd, EF, and PERI to evaluate contamination levels. Source apportionment of heavy metal(loid)s was conducted using the APCS-MLR and PMF receptor models. Spatial distribution patterns were determined using the most efficient interpolation technique among five different approaches. The total carcinogenic risk index (TCR) and total non-carcinogenic index (THI) were used in this study to assess the potential carcinogenic and non-carcinogenic hazards posed by heavy metal(loid)s in surface soil to human health.

RESULTS

The study reveals a high contamination level of heavy metal(loid)s in the surface soil, posing considerable ecological risks. As was identified as a priority metal for regulatory control measures. Mining and smelting activities were identified as the primary factors influencing heavy metal(loid) distributions. Based on spatial distribution mapping, concentrations were higher in the northern, western, and central regions of the study area. As and Fe were found to pose considerable and moderate ecological risks, respectively. Health risk evaluation indicated significant levels of carcinogenic risks for both adults and children, with higher risks for children.

CONCLUSION

This study highlights the urgent need for monitoring heavy metal(loid) contamination in Matehuala's soils, particularly in regions experiencing strong economic growth, to mitigate potential human health and ecological risks associated with heavy metal(loid) pollution.

Characteristics and partitions of traditional and emerging organophosphate esters in soil and groundwater based on machine learning

Organophosphate esters (OPEs) pose hazards to both humans and the environment. This study applied target screening to analyze the concentrations and detection frequencies of OPEs in the soil and groundwater of representative contaminated sites in the Pearl River Delta. The clusters and correlation characteristics of OPEs in soil and groundwater were calculated by self-organizing map (SOM). The risk assessment and partitions of OPEs in industrial park soil and groundwater were conducted. The results revealed that 14 out of 23 types of OPEs were detected. The total concentrations (Σ23OPEs) ranged from 1.931 to 743.571 ng/L in the groundwater, and 0.218 to 79.578 ng/g in the soil, the former showed highly soluble OPEs with high detection frequencies and concentrations, whereas the latter exhibited the opposite trend. SOM analysis revealed that the distribution of OPEs in the soil differed significantly from that in the groundwater. In the industrial park, OPEs posed acceptable risks in both the soil and groundwater. The soil could be categorized into Zone I and II, and the groundwater into Zone I, II, and III, with corresponding management recommendations. Applying SOM to analyze the characteristics and partitions of OPEs may provide references for other new pollutants and contaminated sites.

Quantitative tracing the sources and human risk assessment of complex soil pollution in an industrial park

Pollution in industrial parks has long been characterized by complex pollution sources and difficulties in identifying pollutant origins. This study focuses on a typical industrial park consisting of 11 factories (F1-F11) including organic pigment, inorganic pigment, and chemical factories in Hunan Province, China, here, a total of 327 sample points were surveyed. Eight pollutants (Mn, Cd, As, Co, NH3-N, l, 1,2-Trichloroethane, chlorobenzene, and petroleum hydrocarbons) were classified as contaminants of concern (COCs). This study assessed the contributions of driving factors to the distribution of COCs in the soil. Pollutant source apportionment was conducted using positive matrix factorization (PMF) and random forest (RF). The results revealed that the main factors driving pollution are groundwater migration, non-compliant emissions, leaks during production, and interactions among pollutants. The primary pollution sources were four chemical factories and an inorganic pigment factory. Source 5 demonstrates significant correlations with TCA (29.6%), CB (30%), and As (31.6%). Two chemical factories (F7 and F10) are the most significant pollution source with a risk assessment contribution rate of more than 60%. The present study sheds some light on the contamination characteristics, source apportionment and source-health risk assessment of COCs in industrial park. By utilizing the proposed research framework, decision-makers can effectively prioritize and address identified pollution sources.

Pollution area identification, receptor model-oriented sources and probabilistic health hazards to prioritize control measures for heavy metal management in soil

Identifying the primary source of heavy metals (HMs) pollution and the key pollutants is crucial for safeguarding eco-health and managing risks in industrial vicinity. For this purpose, this investigation was carried out to investigate the pollution area identification with soil static environmental capacity (QI), receptor model-oriented critical sources, and Monte Carlo simulation (MCS) based probabilistic environmental and human health hazards associated with HMs in agricultural soils of Narayanganj, Bangladesh. The average concentration of Cr, Ni, Cu, Cd, Pb, Co, Zn, and Mn were 98.67, 63.41, 37.39, 1.28, 23.93, 14.48, 125.08, and 467.45 mg/kg, respectively. The geoaccumulation index identified Cd as the dominant metal, indicating heavy to extreme contamination in soils. The QI revealed that over 99% of the areas were polluted for Ni and Cd with less uncertain regions whereas Cr showed a significant portion of areas with uncertain pollution status. The positive matrix factorization (PMF) model identified three major sources: agricultural (29%), vehicular emissions (25%), and industrial (46%). The probabilistic assessment of health hazards indicated that both carcinogenic and non-carcinogenic risks for adult male, adult female, and children were deemed unacceptable. Moreover, children faced a higher health hazard compared to adults. For adult male, adult female, and children, industrial operations contributed 48.4%, 42.7%, and 71.2% of the carcinogenic risks, respectively and these risks were associated with Ni and Cr as the main pollutants of concern. The study emphasizes valuable scientific insights for environmental managers to tackle soil pollution from HMs by effectively managing anthropogenic sources. It could aid in devising strategies for environmental remediation engineering and refining industry standards.

The heterogeneous distribution of heavy metal(loid)s at a smelting site and its potential implication on groundwater

The downward migration of soil heavy metal(loid)s (HMs) at smelting sites poses a significant risk to groundwater. Therefore, it is requisite for pollution control to determine the pollution characteristics of soil HMs and their migration risks to groundwater. 198 soil samples collected from a Pb-Zn smelting site were classified into 6 clusters by self-organizing map (SOM) and K-means clustering. Cd, Zn, As, and Pb were identified as the characteristic contaminants of the site. The driving factors for the heterogeneous distribution of HMs have been validated through the implementation of K-means clustering and multiple-hits calculation. Using ultrafiltration extraction and microscopic analysis, the soil colloids were identified as crucial carriers facilitating the migration of HMs. Specifically, the colloidal fractions of Cd, Zn, and As, Pb in deep soil (3-4 m) accounted for 91 %, 78 %, 88 %, and 82 %, respectively, consistently surpassing those found in topsoil (0-0.5 m). It was primarily attributed to the strong affinity of HMs toward soil colloids (franklinite, PbS, and kaolinite) and dissolved organic matter (humic acids and protein). The research findings highlight the potential risk of colloidal HMs to groundwater contamination, providing valuable insights for the development of targeted management and remediation strategies.

Identifying influence factors and thresholds of the next day's pollen concentration in different seasons using interpretable machine learning

The prevalence of pollen allergies is a pressing global issue, with projections suggesting that half of the world's population will be affected by 2050 according to the estimation of the World Health Organization (WHO). Accurately forecasting pollen allergy risks requires identifying key factors and their thresholds for aerosol pollen. To address this, we developed a technical framework combining advanced machine learning and SHapley Additive exPlanations (SHAP) technology, focusing on Beijing. By analyzing meteorological data and vegetation phenology, we identified the factors influencing next-day's pollen concentration (NDP) in Beijing and their thresholds. Our results highlight vegetation phenology data from Synthetic Aperture Radar (SAR), temperature, wind speed, and atmospheric pressure as crucial factors in spring. In contrast, the Normalized Difference Vegetation Index (NDVI), air temperature, and wind speed are significant in autumn. Leveraging SHAP technology, we established season-specific thresholds for these factors. Our study not only confirms previous research but also unveils seasonal variations in the relationship between radar-derived vegetation phenology data and NDP. Additionally, we observe seasonal fluctuations in the influence patterns and threshold values of daily air temperatures on NDP. These insights are pivotal for improving pollen concentration prediction accuracy and managing allergic risks effectively.

Single particle mass spectral signatures from on-road and non-road vehicle exhaust particles and their application in refined source apportionment using deep learning

With advances in vehicle emission control technology, updating source profiles to meet the current requirements of source apportionment has become increasingly crucial. In this study, on-road and non-road vehicle particles were collected, and then the chemical compositions of individual particles were analyzed using single particle aerosol mass spectrometry. The data were grouped using an adaptive resonance theory neural network to identify signatures and establish a mass spectral database of mobile sources. In addition, a deep learning-based model (DeepAerosolClassifier) for classifying aerosol particles was established. The objective of this model was to accomplish source apportionment. During the training process, the model achieved an accuracy of 98.49 % for the validation set and an accuracy of 93.36 % for the testing set. Regarding the model interpretation, ideal spectra were generated using the model, verifying its accurate recognition of the characteristic patterns in the mass spectra. In a practical application, the model performed hourly source apportionment at three specific field monitoring sites. The effectiveness of the model in field measurement was validated by combining traffic flow and spatial information with the model results. Compared with other machine learning methods, our model achieved highly automated source apportionment while eliminating the need for feature selection, and it enables end-to-end operation. Thus, in the future, it can be applied in refined and online source apportionment of particulate matter.

Isotopic insights and integrated analysis for heavy metal levels, ecological risks, and source apportionment in river sediments of the Qinghai-Tibet Plateau

The research was carried out to examine the pollution characteristics, ecological risk, and origins of seven heavy metals (Hg, As, Pb, Cu, Cd, Zn, and Ni) in 51 sediment samples gathered from 8 rivers located on the Qinghai-Tibet Plateau (QTP) in China. The contents of Hg and Cd were 5.0 and 1.1 times higher than their background values, respectively. The mean levels of other measured heavy metals were below those found naturally in the local soil. The enrichment factor showed that the study area exhibited significantly enriched Hg with 70.6% sampling sites. The Cd contents at 19.6% of sampling sites were moderately enriched. The other sampling sites were at a less enriched level. The sediments of all the rivers had a medium level of potential ecological risk. Hg was the major ecological risk factor in all sampling sites, followed by Cd. The findings from the positive matrix factorization (PMF) analysis shown agricultural activities, industrial activities, traffic emissions, and parent material were the major sources. The upper, middle, and low reaches of the Quanji river had different Hg isotope compositions, while sediments near the middle reaches were similar to the δ202Hg of the industrial source. At the upstream sampling sites, the Hg isotope content was very close to the background level. The results of this research can establish a strong scientific sound to improve the safety of the natural circumstances of rivers on the QTP.

Evaluation of the prediction effectiveness for geochemical mapping using machine learning methods: A case study from northern Guangdong Province in China

This study compares seven machine learning models to investigate whether they improve the accuracy of geochemical mapping compared to ordinary kriging (OK). Arsenic is widely present in soil due to human activities and soil parent material, posing significant toxicity. Predicting the spatial distribution of elements in soil has become a current research hotspot. Lianzhou City in northern Guangdong Province, China, was chosen as the study area, collecting a total of 2908 surface soil samples from 0 to 20 cm depth. Seven machine learning models were chosen: Random Forest (RF), Support Vector Machine (SVM), Ridge Regression (Ridge), Gradient Boosting Decision Tree (GBDT), Artificial Neural Network (ANN), K-Nearest Neighbors (KNN), and Gaussian Process Regression (GPR). Exploring the advantages and disadvantages of machine learning and traditional geological statistical models in predicting the spatial distribution of heavy metal elements, this study also analyzes factors affecting the accuracy of element prediction. The two best-performing models in the original model, RF (R2 = 0.445) and GBDT (R2 = 0.414), did not outperform OK (R2 = 0.459) in terms of prediction accuracy. Ridge and GPR, the worst-performing methods, have R2 values of only 0.201 and 0.248, respectively. To improve the models' prediction accuracy, a spatial regionalized (SR) covariate index was added. Improvements varied among different methods, with RF and GBDT increasing their R2 values from 0.4 to 0.78 after enhancement. In contrast, the GPR model showed the least significant improvement, with its R2 value only reaching 0.25 in the improved method. This study concluded that choosing the right machine learning model and considering factors that influence prediction accuracy, such as regional variations, the number of sampling points, and their distribution, are crucial for ensuring the accuracy of predictions. This provides valuable insights for future research in this area.

Historical construction, quantitative source identification and risk assessment of heavy metals contamination in sediments from the Pearl River Estuary, South China

Historical reconstruction of heavy metals (HMs) contamination in sediments is a key for understanding the effects of anthropogenic stresses on water bodies and predicting the variation trends of environmental state. In this work, eighteen sediment cores from the Pearl River Estuary (PRE) were collected to determine concentrations and geochemical fractions of HMs. Then, their potential sources and the relative contributions during different time periods were quantitatively identified by integrating lead-210 (210Pb) radioisotope dating technique into positive matrix factorisation (PMF) method. Pollution levels and potential ecological risks (PERs) caused by HMs were accurately assessed by enrichment factors (EF) based on establishment of their geochemical baselines (GCBs) and multiparameter evaluation index (MPE). HMs concentrations generally showed a particle size- and organic matter-dependent distribution pattern. During the period of 1958-1978, HMs concentrations remained at low levels with agricultural activities and natural processes being identified as the predominant sources and averagely contributing >60%. Since the reform and opening-up in 1978, industrial and traffic factors become the primary anthropogenic sources of HMs (such as Cu, Zn, Cd, Pb, Cr, and Ni), averagely increasing from 22.1% to 28.1% and from 11.6% to 23.4%, respectively. Conversely, the contributions of agricultural and natural factors decreased from 37.0% to 28.5% and from 29.3% to 20.0%, respectively. Subsequently, implementation of environmental preservation policies was mainly responsible for the declining trend of HMs after 2010. Little enrichment of sediment Cu, Zn, Pb, Cr and Ni with EFs (0.15-1.43) was found in the PRE, whereas EFs of Cd (1.16-2.70) demonstrated a slight to moderate enrichment. MPE indices of Cu (50.7-252), Pb (52.0-147), Zn (35.5-130), Ni (19.6-71.5), Cr (14.2-68.8) and Cd (0-9.90) highlighted their potential ecological hazards due to their non-residual fractions and anthropogenic sources.

Analysis of heavy metal and polycyclic aromatic hydrocarbon pollution characteristics of a typical metal rolling industrial site based on data mining

With the transformation and upgrading of industries, the environmental problems caused by industrial residual contaminated sites are becoming increasingly prominent. Based on actual investigation cases, this study analyzed the soil pollution status of a remaining sites of the copper and zinc rolling industry, and found that the pollutants exceeding the screening values included Cu, Ni, Zn, Pb, total petroleum hydrocarbons and 6 polycyclic aromatic hydrocarbon monomers. Based on traditional analysis methods such as the correlation coefficient and spatial distribution, combined with machine learning methods such as SOM + K-means, it is inferred that the heavy metal Zn/Pb may be mainly related to the production history of zinc rolling. Cu/Ni may be mainly originated from the production history of copper rolling. PAHs are mainly due to the incomplete combustion of fossil fuels in the melting equipment. TPH pollution is speculated to be related to oil leakage during the industrial use period and later period of vehicle parking. The results showed that traditional analysis methods can quickly identify the correlation between site pollutants, while SOM + K-means machine learning methods can further effectively extract complex hidden relationships in data and achieve in-depth mining of site monitoring data.

Application of machine learning in atmospheric pollution research: A state-of-art review

Machine learning (ML) is an artificial intelligence technology that has been used in atmospheric pollution research due to their powerful fitting ability. In this review, 105 articles related to ML and the atmospheric pollution research are critically reviewed. Applications of ML in the prediction of atmospheric pollution (mainly particulate matters) are systematically described, including the principle of prediction, influencing factors and improvement measures. Researchers can improve the accuracy of the prediction model through three main aspects, namely considering the geographical features of the study area into the model, introducing the physical characteristics of pollutants, matching and optimizing ML models. And by using interpretable ML tools, researchers are able to understand the mechanism of the model and gain in-depth information. Then, the state-of-art applications of ML in the source apportionment of atmospheric particulate matter and the effect of atmospheric pollutants on human health are also described. In addition, the advantages and disadvantages of the current applications of ML in atmospheric pollution research are summarized, and the application perspective of ML in this field is elucidated. Given the scarcity of source apportionment applications and human health research, standardized research methods and specialized ML methods are required in atmospheric pollution research to connect these two disciplines.

Assessment and machine learning prediction of heavy metals fate in mining farmland assisted by Positive Matrix Factorization

The accurate pollutant prediction by Machine Learning (ML) is significant to efficient environmental monitoring and risk assessment. However, application of ML in soil is under studied. In this study, a Positive Matrix Factorization (PMF) assisted prediction method was developed with Support Vector Machine (SVM) and Random Forest (RF) for heavy metals (HMs) prediction in mining farmland. Principal Component Analysis (PCA) and Redundancy Analysis (RDA) were selected to pretreat data. Experiment results illustrated Cd was the main pollutant with heavy risks in the study area and Pb was easy to migrate. The method effects of HMs total concentration predicting were PMF > Simple > PCA > PCA - PMF, and RF predicted better than SVM. Data pretreatment by RDA prior inspection improved the model results. Characteristic HMs Tessier fractions prediction received good effects with average R value as 0.86. Risk classification prediction performed good in Cd, Cu, Ni and Zn, however, Pb showed weak effect by simple model. The best classifier method for Pb was PMF - RF method with relatively good effect (Area under ROC Curve = 0.896). Overall, our study suggested the combination between PMF and ML can assist the prediction of HMs in soil. Spatial weighted attribute of HMs can be provided by PMF.

Machine learning combined with Geodetector quantifies the synergistic effect of environmental factors on soil heavy metal pollution

The critical prerequisite for the prevention and control of soil heavy metal (HM) pollution is the identification of factors that influence soil HM accumulation. The dominant factors have been individually identified and apportioned in existing studies. However, the accumulation of soil HMs results from a combination of multiple factors, and the influence of a single factor is less than the interaction of multiple parameters on soil HM pollution. In this study, we employed Geodetector to delve into the interaction effect of the influencing factors on the variations of soil HMs. We performed partial dependence plot to depict how these factors interact with each other to affect the HM content. We found that both individually and interactively, pH and agricultural activities significantly impact soil HM content. Except for Hg and Cu, the pairs with the most significant interaction effects all involve pH. For Pb, As and Zn, interaction with pH has the most significant driving force compared to the other factors. For Cu, Hg, and Ni, all environmental factor interactions increased their explanatory power, while for Cr, the single most significant driver decreased its driving power when interacting with other factors. Additionally, the study area exhibited a widespread prevalence of changes in HM concentration being governed by the synergistic effect of two factors. For the response of HMs to the interaction of pH and fertilizer, soil HM concentration was sensitive to pH, while fertilizer had less effect. These results provide a dependable method of investigating the interaction of environmental factors on soil HM content and put forth efficacious and potent tactical measures for soil HM pollution prevention and control based on the interaction type.

Differentiating environmental scenarios to establish geochemical baseline values for heavy metals in soil: A case study of Hainan Island, China

Soil heavy metal distributions exhibit regional heterogeneity due to the complex characteristics of parent materials and soil formation processes, emphasizing the need for appropriate regional standards prior to assessing soil risks. This study focuses on Hainan Island and employs the Multi-purpose Regional Geochemical Survey dataset to establish heavy metal geochemical baseline and background values for soil using an iterative method. Geographical detector analysis reveals that parent materials are the primary factor influencing heavy metal distribution, followed by soil types and land use. Heavy metal geochemical baseline values are established for the island's three environments and administrative regions. Notably, a universal geochemical baseline value cannot adequately represent regional variations in heavy metal distribution, with parent materials playing a crucial role in various scenarios. Locally applicable values based on parent material are the most representative for Hainan Island. This study provides a reference framework for developing region-specific environmental baseline values for soil heavy metal assessments.

Heavy metals contamination, receptor model-based sources identification, sources-specific ecological and health risks in road dust of a highly developed city

The present study quantified Ni, Cu, Cr, Pb, Cd, As, Zn, and Fe levels in road dust collected from a variety of sites in Tangail, Bangladesh. The goal of this study was to use a matrix factorization model to identify the specific origin of these components and to evaluate the ecological and health hazards associated with each potential origin. The inductively coupled plasma mass spectrometry was used to determine the concentrations of Cu, Ni, Cr, Pb, As, Zn, Cd, and Fe. The average concentrations of these elements were found to be 30.77 ± 8.80, 25.17 ± 6.78, 39.49 ± 12.53, 28.74 ± 7.84, 1.90 ± 0.79, 158.30 ± 28.25, 2.42 ± 0.69, and 18,185.53 ± 4215.61 mg/kg, respectively. Compared to the top continental crust, the mean values of Cu, Pb, Zn, and Cd were 1.09, 1.69, 2.36, and 26.88 times higher, respectively. According to the Nemerow integrated pollution index (NIPI), pollution load index (PLI), Nemerow integrated risk index (NIRI), and potential ecological risk (PER), 84%, 42%, 30%, and 16% of sampling areas, respectively, which possessed severe contamination. PMF model revealed that Cu (43%), Fe (69.3%), and Cd (69.2%) were mainly released from mixed sources, natural sources, and traffic emission, respectively. Traffic emission posed high and moderate risks for modified NIRI and potential ecological risks. The calculated PMF model-based health hazards indicated that the cancer risk value for traffic emission, natural, and mixed sources had been greater than (1.0E-04), indicating probable cancer risks and that traffic emission posed 38% risk to adult males where 37% for both adult females and children.

Genetic programming-based predictive model for the Cr removal effect of in-situ electrokinetic remediation in contaminated soil

Soil electrokinetic remediation is an emerging and efficient in-situ remediation technology for reducing environmental risks. Promoting the dissolution and migration of Cr in soil under the electric field is crucial to decrease soil toxicity and ecological influences. However, it is difficult to establish strong relationships between soil treatment and impact factors and to quantify their contributions. Machine learning can help establish pollutant migration models, but it is challenging to derive predictive formulas to improve remediation efficiency, describe the predictive model construction process, and reflect the importance of the predictors for better regulation. Therefore, this paper established a predictive model for the electrokinetic remediation of Cr-contaminated soil using genetic programming (GP) after determining the characteristic parameters which influenced the remediation effect, described the model's adaptive optimization process through the algorithm's function, and identified the sensitivity factors affecting the Cr removal effect. Results showed that the Cr(VI) and total Cr concentrations predicted by GP were in satisfactory agreement with the experimental values, 92% of the training data and 90% of the validation data achieved errors within 1%, and could fully reflect the target ions' content variation in different soil layers. By substituting the above prediction formulas into Sobol sensitivity analysis, it was determined that conductivity, pH, current, and moisture content dramatically affected the Cr content variation in distinguished regions. For overall contaminated area, the system current and soil pH were the most sensitive factors for Cr(VI) and total Cr contents. Remediation efforts throughout the contaminated area should focus on the role of current versus soil pH. GP and sensitivity analysis can provide decision support and operational guidance for in-situ soil electrokinetic treatment by establishing a remediation effect prediction model, expediting the development and innovation of electrokinetic technology.

Contamination and source apportionment of metals in urban road dust (Jinan, China) integrating the enrichment factor, receptor models (FA-NNC and PMF), local Moran's index, Pb isotopes and source-oriented health risk

Contamination and source identifications of metals in urban road dust are critical for remediation and health protection. Receptor models are commonly used for metal source identification, whereas the results are usually subjective and not verified by other indicators. Here we present and discuss a comprehensive approach to study metal contamination and sources in urban road dust (Jinan) in spring and winter by integrating the enrichment factor (EF), receptor models (positive matrix factorization (PMF) and factor analysis with nonnegative constraints (FA-NNC)), local Moran's index, traffic factors and Pb isotopes. Cadmium, Cr, Cu, Pb, Sb, Sn and Zn were the main contaminants, with mean EFs of 2.0-7.1. The EFs were 1.0-1.6 times higher in winter than in spring but exhibited similar spatial trends. Chromium contamination hotspots occurred in the northern area, with other metal contamination hotspots in the central, southeastern and eastern areas. The FA-NNC results indicated Cr contamination primarily resulting from industrial sources and other metal contamination primarily originating from traffic emissions during the two seasons. Coal burning emissions also contributed to Cd, Pb and Zn contamination in winter. FA-NNC model-identified metal sources were verified via traffic factors, atmospheric monitoring and Pb isotopes. The PMF model failed to differentiate Cr contamination from other detrital metals and the above anthropogenic sources, largely due to the model grouping metals by emphasizing hotspots. Considering the FA-NNC results, industrial and traffic sources accounted for 28.5 % (23.3 %) and 44.7 % (28.4 %), respectively, of the metal concentrations in spring (winter), and coal burning emissions contributed 34.3 % in winter. Industrial emissions primarily contributed to the health risks of metals due to the high Cr loading factor, but traffic emissions dominated metal contamination. Through Monte Carlo simulations, Cr had 4.8 % and 0.4 % possibilities posing noncarcinogenic and 18.8 % and 8.2 % possibilities posing carcinogenic risks for children in spring and winter, respectively.