Method for analyzing heavy metal pollution source of underground water based on characteristic ion screening

Abstract

The application discloses a groundwater heavy metal pollution source analysis method based on characteristic ion screening, which comprises the following steps: through multi-medium collaborative sampling, the concentration of heavy metals in groundwater, the BCR form content of sediments and the isotope ratio of lead, zinc and copper are determined; the characteristic ion combination is refined through three levels of screening of enrichment factor, ion ratio variation coefficient and Pearson correlation; the three types of fingerprint vectors of water chemistry, solid phase form and isotope are normalized and fused, and the candidate pollution source is matched through cosine similarity; the mass conservation is taken as the main likelihood, the BCR form and the isotope mass balance are taken as additional constraints, the MCMC is adopted to output the posterior mean value and the 95% confidence interval of the contribution rate, and the spatial distribution field and the high-risk area are obtained through Kriging interpolation. The application can identify four types of pollution sources, the reconstruction error is reduced by 73% compared with the PMF method, the confidence interval width is compressed by 65% to 73%, the on-site coincidence degree reaches 92%, and quantitative basis is provided for pollution prevention and control and responsibility tracing.

Description

Technical Field

[0001] This invention belongs to the field of groundwater pollution source tracing and environmental geochemistry, specifically involving a method for analyzing groundwater heavy metal pollution sources based on characteristic ion screening. Background Technology

[0002] Shallow groundwater around mining areas, electroplating industrial parks, and wastewater-irrigated agricultural areas often faces a complex situation of multiple heavy metal pollution sources coexisting and overlapping in space. Taking the downstream area of ​​a lead-zinc mine as a typical scenario: acidic leaching water from mine tailings ponds, historical dust deposition from smelters, and agricultural wastewater irrigation coexist in the same aquifer, resulting in arsenic concentrations reaching as high as 186 μg / L and lead concentrations reaching as high as 62 μg / L, far exceeding the Class III water limit of the Groundwater Quality Standard (GB / T 14848—2017). Because the three pollution sources overlap along the groundwater migration path, it is difficult to accurately distinguish the contribution ratio of each source based on single-medium concentration data, which restricts the scientific nature of pollution liability determination and precise prevention and control decisions.

[0003] The traditional positive definite matrix factorization (PMF) method is currently the most widely used tool for apportioning heavy metal sources in groundwater, relying on the groundwater hydrochemical concentration matrix to decompose pollution source factors. However, in multi-source scenarios, the hydrochemical fingerprints of heavy metals from different parts of the same deposit (such as acid mine wastewater and tailings leaching sources) are highly similar. The PMF method can only identify about three types of pollution sources, with a root mean square reconstruction error of up to 0.78 mg / L, and the consistency between the pollution source type and the field investigation is only about 70%. In addition, the PMF method completely ignores the occurrence speciation information (BCR morphological characteristics) and stable isotope fingerprints of heavy metals in aquifer sediments, and cannot use "solid-liquid multi-media" synergistic evidence to constrain the analysis results. As a result, when the end-member concentrations of multiple candidate pollution sources are similar, the posterior distribution width of the contribution rate reaches more than 0.32, which is difficult to meet the accuracy requirements for legal liability tracing.

[0004] While single-mixture models based on stable isotopes (lead, zinc, and copper isotopes) such as IsoSource can distinguish a few endmembers using isotopic fingerprints, they are limited to processing only 2-3 endmember sources per analysis. When four or more pollution sources coexist, they cannot construct effective constraints, and the average width of the 95% confidence interval for contribution rate is as high as 0.41, indicating extremely high uncertainty. At the same time, existing methods lack a systematic characteristic ion screening procedure before conducting source tracing analysis. A large number of ions with weak indicative power and large spatial variability are introduced into the analytical model, which not only increases the computational load but also significantly reduces the stability of contribution rate estimation due to noise interference. In addition, existing methods do not have the ability to fuse water chemical fingerprints, solid phase morphological fingerprints, and isotopic fingerprints into a unified multidimensional feature matrix. They cannot perform cosine similarity pre-screening of candidate pollution sources, resulting in a large number of pollution sources unrelated to the target sampling point being included in the Bayesian framework for calculation, further amplifying the uncertainty of the analysis.

[0005] The above three issues together restrict the engineering applicability of existing methods in complex, multi-source pollution scenarios, and there is an urgent need for a groundwater heavy metal pollution source analysis method based on characteristic ion screening. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for analyzing heavy metal pollution sources in groundwater based on characteristic ion screening, in order to solve the following technical problems existing in the prior art: Traditional PMF methods rely solely on single-medium hydrochemical data and neglect sediment morphology and isotopic multi-medium constraints, resulting in insufficient pollution source identification capabilities and low contribution rate resolution accuracy in multi-source superposition scenarios. Single isotopic mixing models can only handle a limited number of endmembers and have high uncertainty. Furthermore, existing methods lack a systematic characteristic ion screening mechanism, and the introduction of a large number of weak indicative indicators into the model amplifies noise.

[0007] To achieve the above-mentioned objectives, the specific technical solution is as follows: A method for analyzing heavy metal pollution sources in groundwater based on characteristic ion screening is applied to the analysis of heavy metal pollution sources in shallow groundwater in mining areas, electroplating and metal processing industrial parks, or wastewater irrigation agricultural areas. The heavy metals include at least three of arsenic, lead, cadmium, chromium, copper, and zinc.

[0008] Groundwater samples, sediment column samples from aquifers corresponding to the well locations, and surface soil samples were collected from the monitoring wells in the target area. The concentrations of heavy metals, major anions and cations, pH, redox potential, and dissolved oxygen in the groundwater samples were measured. The total amount of heavy metals and the content of various forms of BCR in the sediment column samples were measured. The stable isotope ratios of lead, zinc, and copper in the surface soil and sediment samples were measured.

[0009] The triple screening of characteristic ions in groundwater involves evaluating each ion in the groundwater sample based on enrichment factor screening, ion ratio stability screening, and source region correlation screening, thereby obtaining a combination of characteristic ions that can indicate the heavy metal pollution source.

[0010] A multidimensional feature fingerprint matrix is ​​constructed by: constructing a groundwater chemical fingerprint vector based on feature ion combinations; constructing a sediment morphological fingerprint vector based on BCR speciation content; and constructing an isotopic fingerprint vector based on stable isotope ratios. The groundwater chemical fingerprint vector, the sediment morphological fingerprint vector, and the isotopic fingerprint vector are then Z-score normalized and concatenated column-wise to obtain a normalized multidimensional feature fingerprint matrix. The normalized multidimensional feature fingerprint matrix is ​​then matched with the end-member spectra of candidate pollution sources using cosine similarity to select a set of candidate pollution sources to be analyzed.

[0011] A multi-media coupled Bayesian hybrid analysis was employed, with the mass conservation equation as the principal term of the likelihood function, and the sediment BCR morphology constraint equation and the stable isotope mass balance equation as additional constraint terms embedded in the Bayesian framework. The Dirichlet distribution was used as the prior distribution of the contribution rates of each candidate pollution source, and the Markov chain Monte Carlo method was used to sample the posterior distribution of the contribution rates of each candidate pollution source. The mean, standard deviation, and 95% confidence interval of the contribution rates of each heavy metal candidate pollution source were output, and the spatial distribution map of the contribution rates of each pollution source was obtained by ordinary Kriging interpolation.

[0012] Furthermore, the specific method of multi-media collaborative sampling is as follows: monitoring wells are deployed according to the grid method or the direction of the pollution halo, and groundwater samples are collected from each monitoring well no less than three times to cover the wet season, normal season and dry season; columnar samples of aquifer sediments are collected simultaneously at the location of each monitoring well or within a 20-meter radius around it.

[0013] Furthermore, the groundwater sample analysis included the total concentrations of arsenic, lead, cadmium, chromium, copper, and zinc; the cation concentrations of sodium, potassium, calcium, magnesium, and ammonium; the anion concentrations of chloride, sulfate, nitrate, and bicarbonate; as well as pH, redox potential, dissolved oxygen, conductivity, and water temperature. Isotope analysis was performed using a multi-receiver inductively coupled plasma mass spectrometer. Three sets of lead isotope ratios, and zinc isotope values Copper isotope values .

[0014] Furthermore, the specific method for enrichment factor screening is as follows: using aluminum or lithium as a reference element, the enrichment factor of each heavy metal in the groundwater sample (arsenic, lead, cadmium, chromium, copper, and zinc) is calculated respectively. ;in, For the first Enrichment factors for heavy metals; The first in the groundwater sample The concentration of certain heavy metals; This refers to the concentration of the reference element in the same water sample; and These are the ratios of heavy metal concentrations to reference element concentrations in the water samples from the sampling points and the background values, respectively; the background value is the arithmetic median of the concentrations from at least ten unpolluted sampling points in the upstream unpolluted area of ​​the target region.

[0015] Furthermore, the pollution sources of each heavy metal are classified indicatively according to the following criteria: when It was determined to be of natural origin at that time; when It was determined to be a mild form of anthropogenic enrichment. when It was determined to be of significant anthropogenic origin and the heavy metal was included in the candidate set of characteristic ions.

[0016] Furthermore, the specific method for screening the stability of ion ratios and screening the correlation between the source region is as follows: for the candidate set of characteristic ions, at least four sets of indicative ion ratios are constructed, including the arsenic to fluorine ratio, the iron to manganese ratio, the sulfate to chloride ratio, and the nitrate to chloride ratio.

[0017] Furthermore, the coefficient of variation of each ratio across all sampling points is evaluated: ;in, For the first The coefficient of variation of the group ion ratio; and The first The sample standard deviation and arithmetic mean of the group ion ratios at all sampling points; retain The ion ratio is used as a stability characteristic indicator.

[0018] Pearson correlation analysis was performed on the concentrations of each heavy metal in the candidate set of characteristic ions and the stability characteristic indices, and only those meeting the criteria were retained. and The ion pairs are used as the final characteristic ion combinations, which are used to construct the groundwater chemical fingerprint vector: ;in, For the first The characteristic ion and the first Pearson correlation coefficients of characteristic ions in groundwater samples; This represents the total number of groundwater samples. , The first At the sampling point The, the The concentration or ratio of a characteristic ion; , This is the corresponding sample mean; This represents the significance probability of the corresponding hypothesis test.

[0019] Furthermore, the sediment morphological fingerprint vector is constructed as follows: For each heavy metal in the sediment column sample—arsenic, lead, cadmium, chromium, copper, and zinc—a four-step sequential extraction is performed, followed by a final step of complete digestion. The four-step sequential extraction includes: The first step involves extracting acid-soluble or exchangeable substances using acetic acid at a concentration of 0.11 mol / L. ; The second step involves extracting the reducible form of hydroxylamine hydrochloride at a concentration of 0.5 mol / L and a pH of 2. ; The third step involves using hydrogen peroxide at a concentration of 8.8 mol / L and ammonium acetate at a concentration of 1 mol / L in combination to extract the oxidizable form. ; The fourth step involves using aqua regia to microwave digest the residue to obtain the residue state. Finally, the total amount of heavy metals was determined by complete digestion with three acids. .

[0020] Furthermore, define the first Mobility index of heavy metals: ;in, , , , The first The content of each form of heavy metal in the sediment column sample; when At that time, the first These heavy metals were identified as highly mobile heavy metals and were given priority for inclusion in the pollution source analysis system.

[0021] For each sediment column corresponding to a sampling point, the heavy metals were... The morphological proportions of the sediments were assembled in order of heavy metal species to obtain the morphological fingerprint vector of the sediments. .

[0022] Furthermore, the construction of the isotopic fingerprint vector and its coupling with the sediment morphological fingerprint are as follows: Constructing the isotopic fingerprint vector of the sampling point: The lead isotope ratios are determined by a multi-receiver inductively coupled plasma mass spectrometer using the acid digestion solution of the groundwater sample and the corresponding sediment column sample.

[0023] Zinc isotope value Using IRMM-3702 as a reference standard, copper isotope values Using NIST-976 as a reference standard: ;in, For elements The ratio of the abundance of heavy isotopes to that of light isotopes; subscript and These represent the sample and its corresponding international reference standard, respectively.

[0024] Furthermore, the morphological fingerprint vector of the sediment... With the isotopic fingerprint vector The solid-liquid Spearman rank correlation coefficient is calculated as a coupling weight in the following manner, and the coupling weight is used as the weight coefficient of the solid-phase constraint equation in the Bayesian mixture analytical model: ;in, For the first Spearman rank correlation coefficients between groundwater concentrations of certain heavy metals and the content of BCR mobile states in corresponding sediments; For the first The rank difference between the two variables at each sampling point; only when And when the significance probability of the corresponding hypothesis test is less than 0.05, the first... The solid-state constraint equations for heavy metals are incorporated into a Bayesian framework.

[0025] Furthermore, the specific method for constructing the normalized multidimensional feature fingerprint matrix and matching it with the cosine similarity of the candidate pollution source endmember spectrum is as follows: the groundwater chemical fingerprint vector Sediment morphological fingerprint vector With isotopic fingerprint vector After Z-score normalization, the vectors are concatenated column-wise to obtain the normalized multidimensional feature fingerprint vector: ;in, , , These are the results of Z-score normalization of the corresponding fingerprint vectors; all sampling points The normalized multidimensional feature fingerprint matrix is ​​obtained by stacking the rows.

[0026] For the mining areas, electroplating and metal processing industrial parks, or wastewater irrigation agricultural areas, a candidate pollution source end-member spectrum library is established in advance. The candidate pollution source end-member spectrum library includes at least three categories from tailings leaching sources, smelting dust deposition sources, electroplating wastewater discharge sources, acidic mine wastewater sources, agricultural wastewater irrigation sources, and natural background sources.

[0027] Calculate the cosine similarity between the normalized multidimensional feature fingerprint vector of each sampling point and the candidate pollution source endmember spectrum: ;in, For the first One candidate pollution source end-member spectral vector; for Norm; reservation The pollution sources are used as a set of candidate pollution sources to be analyzed.

[0028] Furthermore, the specific method of the multi-media coupled Bayesian hybrid analysis is as follows: for each sampling point, using the mass conservation equation as the main term of the likelihood function, a hybrid relationship between the observed concentration and the contribution of each candidate pollution source is constructed for each heavy metal: ;in, For the sampling point The observed concentrations of several heavy metals; For the first The contribution rate of each candidate pollution source; For the first The first candidate pollution source end-member spectrum The concentration of the end-members of the heavy metals; To conform to a mean of 0 and a standard deviation of The observation error of the normal distribution.

[0029] The contribution rates of candidate pollution sources satisfy the non-negativity and normalization constraints: Using the Dirichlet distribution as the joint prior for the contribution rates of each candidate pollution source: ;in, The prior concentration parameter has a range of values. .

[0030] Construct the log-likelihood function: ;in, The total number of heavy metals involved in the analysis is given; the Metropolis-Hastings algorithm is used to perform Markov chain Monte Carlo sampling on the posterior distribution, with a sampling chain length of no less than 50,000 steps. The first 20% of the samples are discarded as pre-burning segments, and the contribution rate of each candidate pollution source is output. mean Standard deviation and 95% confidence interval .

[0031] Furthermore, the specific method by which the sediment BCR morphological constraint equation and the stable isotope mass balance equation are embedded as additional constraint terms into the Bayesian framework is as follows: The content constraint of the movable BCR state in the sediment is used as an additional likelihood term to reflect the cumulative contribution of candidate pollution sources to the movable heavy metals in the sediment: ;in, For the first The first candidate pollution source end-member spectrum The total proportion of acid-soluble and reducible forms of various heavy metals in sediments; These are the observed values ​​of the sediment columnar samples corresponding to the sampling points; This represents the corresponding measurement uncertainty; the weight of this constraint term is equal to the solid-liquid Spearman rank correlation coefficient. .

[0032] Furthermore, the stable isotope mass balance equation is used as an additional likelihood term: ;in, For the first The isotope ratios or isotope values ​​of the corresponding heavy metals in the end-member spectra of each candidate pollution source; These are the observed values ​​at the sampling points; This represents the uncertainty in isotope measurements.

[0033] The coupled posterior axiom is obtained by incorporating the BCR morphological constraint equation of the sediment and the stable isotopic mass balance equation into the log-likelihood function: ;in, This is a multi-media integrated observation dataset; and These are the sediment morphology constraint weights and isotope constraint weights, respectively, with values ​​ranging from [value range missing]. By introducing the BCR morphological constraints and isotopic mass balance constraints of the sediments, the liquid phase tracing results are simultaneously anchored in terms of solid phase morphology and isotopic characteristics, significantly reducing the posterior distribution width of the contribution rate.

[0034] Furthermore, the method also includes Monte Carlo uncertainty quantification and spatial distribution output of pollution source contribution rates: For each sampling point, a random perturbation following a normal distribution is applied to the observed concentration of heavy metals in groundwater, the concentration of candidate pollution source endmember spectra, the proportion of BCR speciation in sediments, and the isotope ratio. The standard deviation of the random perturbation is equal to the relative uncertainty of the corresponding measurement multiplied by the observed value.

[0035] Perform no fewer than 1000 Monte Carlo simulations, with each simulation running the Bayesian analysis process independently, and calculate the sample mean and 95% confidence interval of the contribution rate of each candidate pollution source: ;in, Total number of Monte Carlo simulations; For the first The simulation yielded the... The average contribution rate of each candidate pollution source.

[0036] Optimal contribution rate for all sampling points The spatial distribution field of the entire region is obtained by ordinary kriging interpolation, and the form of ordinary kriging interpolation is as follows: ;in, Points to be estimated The first Estimated contribution rates of each candidate pollution source; For the first The Kriging weights for each sampling point are determined by the variation function model and unbiased constraints. Solving the system of equations simultaneously yields the solution.

[0037] Will The connected components are identified as those with the first... The candidate pollution sources are identified as high-risk areas under primary control. The high-risk areas and the width of the 95% confidence interval are output together as the basis for the prevention and control of heavy metal pollution in groundwater in the target area and for tracing responsibility.

[0038] Compared with the prior art, the beneficial effects of this invention are: The triple characteristic ion screening mechanism proposed in this invention effectively filters out weak indicator ions, making the end-member matching of candidate pollution sources more accurate. The multi-dimensional characteristic fingerprint matrix integrates three types of evidence: water chemistry, solid phase morphology, and stable isotopes, enabling effective differentiation between acidic mine wastewater sources and tailings leaching sources with highly similar chemical characteristics within the same mining area. The number of identifiable pollution sources has been expanded from three types in the traditional PMF method to four types. After introducing solid phase BCR morphology constraints and isotope mass balance constraints, the average width of the 95% confidence interval of the contribution rate in the multi-media coupled Bayesian framework has been significantly compressed from 0.32–0.41 to 0.11, reducing uncertainty by approximately 65%–73%. The root mean square reconstruction error of the contribution rate is 0.21 mg / L, which is about 73% lower than the PMF method and about 86% lower than the single isotope mixture model. The consistency between pollution source types and field investigations reaches 92%, and the agreement rate between high-risk area identification and historically polluted sites reaches 88%. The complete analysis of 24 sampling points in the entire area takes about 10 minutes, meeting the requirements for engineering practicality. Attached Figure Description

[0039] Figure 1 This is a flowchart of the groundwater heavy metal pollution source analysis method based on characteristic ion screening according to the present invention. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention are described clearly and completely below. Obviously, the described embodiments are only a part of the embodiments of this invention, not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0041] This embodiment provides a method for analyzing heavy metal pollution sources in groundwater based on characteristic ion screening, using the analysis of heavy metal pollution sources in shallow groundwater within a 5-kilometer downstream area of ​​a large lead-zinc mine in South China as a specific scenario.

[0042] The target area is characterized by a lead-zinc mine and its tailings pond (north) that has been in operation for over 40 years upstream, a historical dust deposition area from a smelter (northwest) in the middle, and an agricultural planting area (southeast) that has long received substandard irrigation water downstream. The region contains three types of potential pollution sources: mining, smelting, and wastewater irrigation, making it a typical scenario for testing the multi-source apportionment capability of this method. In this embodiment, 24 shallow monitoring wells were deployed in the target area, arranged in a 500m × 500m grid according to the hydrogeological trend (northwest-southeast). Groundwater samples, aquifer sediment columnar samples, and surface soil samples were simultaneously collected from the corresponding 24 sampling points. The monitoring period was 12 months, covering three hydrological periods: high water season (April-June), normal water season (October-December), and low water season (January-March). The target heavy metals identified included arsenic, lead, cadmium, and zinc.

[0043] like Figure 1 The diagram shown is a flowchart of the groundwater heavy metal pollution source apportionment method based on characteristic ion screening in this embodiment. The method includes the following steps S1 to S4.

[0044] Step S1: Multi-media collaborative sampling and monitoring.

[0045] Groundwater samples, sediment column samples from aquifers corresponding to the well locations, and surface soil samples were collected from the monitoring wells in the target area. The concentrations of heavy metals, major anions and cations, pH, redox potential, and dissolved oxygen in the groundwater samples were measured. The total amount of heavy metals and the content of various forms of BCR in the sediment column samples were measured. The stable isotope ratios of lead, zinc, and copper in the surface soil and sediment samples were measured.

[0046] Specifically, monitoring wells are deployed according to the grid method or the direction of the pollution halo, and groundwater samples are collected from each monitoring well no less than three times to cover the wet season, normal season and dry season.

[0047] Columnar sediment samples were collected simultaneously at each monitoring well location or within a 20-meter radius of the well. The sediment samples could be collected in layers. The groundwater samples were tested for total concentrations of arsenic, lead, cadmium, chromium, copper, and zinc; cation concentrations of sodium, potassium, calcium, magnesium, and ammonium; anion concentrations of chloride, sulfate, nitrate, and bicarbonate; as well as pH, redox potential, dissolved oxygen, conductivity, and water temperature.

[0048] Isotope determination was performed using a multi-receiver inductively coupled plasma mass spectrometer. Three sets of lead isotope ratios, and zinc isotope values Copper isotope values .

[0049] In this embodiment, groundwater samples were collected using low-flow-rate Beller tubes in the middle section of the aquifer of each monitoring well, with 2 liters collected each time. Samples were collected once each during the wet season, normal season, and dry season, for a total of 24 wells × 3 times = 72 groundwater samples throughout the year.

[0050] Total heavy metal content was determined using inductively coupled plasma mass spectrometry (ICP-MS, detection limits: arsenic 0.1 μg / L, lead 0.05 μg / L, cadmium 0.02 μg / L, zinc 0.5 μg / L). Major anions and cations were determined using ion chromatography. pH, redox potential, dissolved oxygen, conductivity, and water temperature were measured on-site using a multi-parameter water quality analyzer.

[0051] The specific ranges of heavy metal concentrations in groundwater at the 24 sampling points in this embodiment are as follows: arsenic concentration... (median ), lead concentration (median ), cadmium concentration (median ), zinc concentration (median Among them, the arsenic concentration in well MW-03, located 300 meters downstream of the mine, was as high as […]. Lead concentration as high as The arsenic content far exceeds the Class III water limit specified in the "Groundwater Quality Standard" (GB / T14848—2017). ,lead (This is a heavily polluted sampling point.)

[0052] Sediment core samples were collected using single-stroke drilling, with one core sample taken from each monitoring well. The core samples were then cut into four layers: 0-50 cm, 50-100 cm, 100-200 cm, and below 200 cm (a total of four layers), resulting in 24 × 4 = 96 sediment layer samples. Topsoil samples were collected from within a 10 m × 10 m radius around each monitoring well using a five-point sampling method, resulting in 24 soil samples.

[0053] Sediment and soil samples were freeze-dried, passed through a 100-mesh sieve, and sent to the laboratory for analysis. The total amount of heavy metals was determined by ICP-MS after digestion with aqua regia-perchloric acid-hydrofluoric acid triacid. BCR continuous extraction is described in step S3.

[0054] For isotope determination, lead was chemically purified using a Sr-Spec resin column, and zinc and copper were purified using an AGMP-1 anion exchange resin column. The purified sample solutions were then analyzed by a multi-collector inductively coupled plasma mass spectrometer (MC-ICP-MS). , and The internal precision of lead isotope ratios is better than 0.005% (2σ), and the internal precision of zinc and copper isotope values ​​is better than 0.05‰ (2σ).

[0055] In this embodiment, a total of 72 groundwater samples × 16 indicators + 96 sediment stratification samples × 9 indicators + 24 soil samples × 7 indicators = approximately 1980 original test data points were obtained from the multi-media indicators corresponding to all 24 sampling points, constituting a complete dataset of multi-media collaborative sampling.

[0056] Step S2: Triple screening of characteristic ions in groundwater.

[0057] The ions in the groundwater sample were evaluated sequentially based on enrichment factor screening, ion ratio stability screening, and source region correlation screening to obtain characteristic ion combinations that have an indicative role for the heavy metal pollution source.

[0058] The specific method for screening enrichment factors is as follows: using aluminum or lithium as a reference element, the enrichment factors of each heavy metal in the groundwater sample, namely arsenic, lead, cadmium, chromium, copper, and zinc, are calculated respectively. ;in, For the first Enrichment factors for heavy metals; The first in the groundwater sample Concentration of certain heavy metals; This refers to the concentration of the reference element in the same water sample. and These are the ratios of heavy metal concentrations to reference element concentrations in the water samples from the sampling points and the background values, respectively; the background value is the arithmetic median of the concentrations from at least ten unpolluted sampling points in the upstream unpolluted area of ​​the target region.

[0059] Pollution source classification for each heavy metal is based on the following criteria: When It was determined to be of natural origin at that time; when It was determined to be a mild anthropogenic enrichment at that time; It was determined to be of significant anthropogenic origin and the heavy metal was included in the candidate set of characteristic ions.

[0060] In this embodiment, lithium (Li) was chosen as the reference element because its concentration is stable in natural groundwater systems and it is not easily affected by human pollution. Background values ​​were taken from the arithmetic median concentrations of 12 uncontaminated control sampling points (located in clean aquifers in mountainous areas more than 5 kilometers upstream of the mine) upstream of the northwest side of the target area, unaffected by lead-zinc mining. These values ​​were: arsenic... ,lead ,cadmium Zinc ,lithium Taking a water sample from the downstream MW-03 well during the high-water season as an example, the water measured from this sample... , , , , .

[0061] The enrichment factor formula is calculated item by item as follows: Arsenic enrichment factor Lead enrichment factor ; cadmium enrichment factor Zinc enrichment factor .

[0062] according to Based on the established criteria, all four heavy metals—arsenic, lead, cadmium, and zinc—at the MW-03 sampling point were determined to have significant anthropogenic origin and were included in the characteristic ion candidate set. After calculating the results for all 72 water samples from the 24 sampling points using the same method, 19 sampling points met the criteria for at least one heavy metal. This confirms that the target area as a whole is in a state of significant human-caused pollution.

[0063] The specific methods for the ion ratio stability screening and source region correlation screening are as follows: For the characteristic ion candidate set, at least four sets of indicator ion ratios are constructed, including arsenic to fluorine ratio, iron to manganese ratio, sulfate to chloride ratio, and nitrate to chloride ratio, and the coefficient of variation of each ratio at all sampling points is evaluated using the following formula: ;in, For the first The coefficient of variation of the group ion ratio; and The first The sample standard deviation and arithmetic mean of the group ion ratios at all sampling points; retain The ion ratio is used as a stability characteristic index; Pearson correlation analysis is performed on the concentration of each heavy metal in the candidate set of characteristic ions and the stability characteristic index, and only those that meet the criteria are retained. and The ion pairs are used as the final characteristic ion combination.

[0064] The characteristic ion combination is used to construct the groundwater chemical fingerprint vector: ;in, For the first The characteristic ion and the first Pearson correlation coefficients of characteristic ions in groundwater samples; This represents the total number of groundwater samples. , The first At the sampling point The, the The concentration or ratio of a characteristic ion; , This is the corresponding sample mean; This represents the significance probability of the corresponding hypothesis test.

[0065] In this embodiment, arsenic to fluoride ratios were constructed for all 72 water samples. Iron to manganese ratio sulfate to chloride ratio nitrate to chloride ratio There are a total of 4 groups of indicator ion ratios, which are calculated group by group according to the coefficient of variation formula: The sample mean of the ratio across all 72 water samples Sample standard deviation Therefore This is determined to be a stable characteristic indicator; ratio , , It is also determined to be a stable characteristic indicator; ratio , , This is determined to be a stable characteristic indicator; ratio , , Discard. Ultimately retain. , and Three sets of stable characteristic indicators.

[0066] Subsequently, the concentrations of four heavy metals were analyzed. The correlation coefficients were calculated by combining the above three sets of stable characteristic indicators with each of the four heavy metals in pairs using the Pearson correlation formula. and correlation coefficient , ,and correlation coefficient , This reflects the causes of acidic wastewater from mines; and correlation coefficient , This reflects the contribution of lead-zinc symbiotic minerals to dissolution; and correlation coefficient , This reflects the oxidation and release of zinc sphalerite; and Correlation coefficient , Because of dissatisfaction and And discard it.

[0067] The final combination of characteristic ions obtained through screening is as follows: There are a total of 7 characteristic indicators, and the combination of these characteristic ions is used to construct the chemical fingerprint vector of groundwater.

[0068] Step S3: Construction of multidimensional feature fingerprint matrix.

[0069] Based on the characteristic ion combination obtained in step S2, a groundwater chemical fingerprint vector is constructed; based on the BCR speciation content obtained in step S1, a sediment morphological fingerprint vector is constructed; and based on the stable isotope ratio obtained in step S1, an isotope fingerprint vector is constructed. The groundwater chemical fingerprint vector, the sediment morphological fingerprint vector, and the isotope fingerprint vector are Z-score normalized and then concatenated column by column to obtain a normalized multidimensional feature fingerprint matrix. The normalized multidimensional feature fingerprint matrix is ​​then matched with the candidate pollution source endmember spectrum using cosine similarity to select the set of candidate pollution sources to be analyzed.

[0070] The five-step continuous extraction of sediment BCR and the construction of sediment morphological fingerprint vectors are as follows: For each heavy metal in the sediment column sample—arsenic, lead, cadmium, chromium, copper, and zinc—a four-step sequential extraction process was performed, followed by a final step of complete digestion for analysis. The four-step sequential extraction process included: The first step involves extracting acid-soluble or exchangeable substances using acetic acid at a concentration of 0.11 mol / L. ; The second step involves extracting the reducible form of hydroxylamine hydrochloride at a concentration of 0.5 mol / L and a pH of 2. ; The third step involves using hydrogen peroxide at a concentration of 8.8 mol / L and ammonium acetate at a concentration of 1 mol / L to extract the oxidizable state. ; The fourth step involves microwave digestion of the residue with aqua regia to obtain the residue state. Finally, the total amount of heavy metals was determined by complete digestion with three acids. Definition of the first Mobility index of heavy metals: ;in, , , , The first The content of each form of heavy metal in the sediment column sample; when At that time, the first These heavy metals were identified as highly mobile and prioritized for inclusion in the pollution source apportionment system; for each sediment column corresponding to each sampling point, the composition of each heavy metal was determined. The morphological proportions of the sediments were assembled in order of heavy metal species to obtain the morphological fingerprint vector of the sediments. .

[0071] In this embodiment, using sediment samples from the 0-50cm layer of well MW-03 as an example, the contents of various arsenic forms were measured according to the four-step sequential extraction method of BCR. , , , .

[0072] The migration index is calculated using the following formula: .because Arsenic is classified as a highly mobile heavy metal and is therefore prioritized for inclusion in the pollution source apportionment system; similarly, it can be calculated that... (High transferability, included in the analysis) (High transferability, included in the analysis) (High mobility, included in the analysis).

[0073] The results indicate that upstream mining sources mainly release oxidized sulfide minerals through dissolution, while heavy metals in sediments are mainly in acid-soluble and reducible forms, exhibiting strong mobility and a tendency to be released secondary into groundwater.

[0074] For four heavy metals in sediment samples from the 0-50cm layer of well MW-03, sedimentary morphological fingerprint vectors were constructed according to the morphological splicing method: arsenic, lead, cadmium, and zinc were each... .

[0075] The four-dimensional morphological proportion vectors are concatenated in the order of arsenic, lead, cadmium, and zinc to obtain a 16-dimensional sediment morphological fingerprint vector: The sediment morphological fingerprint vector is the core component of the "solid phase dimension" in the subsequent multidimensional feature fingerprint matrix.

[0076] The construction of the isotopic fingerprint vector and its coupling with the sediment morphological fingerprint are as follows: Constructing the isotopic fingerprint vector of the sampling point: The lead isotope ratios are determined by a multi-receiver inductively coupled plasma mass spectrometer using the acid digestion solution of the groundwater sample and the corresponding sediment column sample.

[0077] Zinc isotope value Using IRMM-3702 as a reference standard, copper isotope values Using NIST-976 as a reference standard: ;in, For elements The ratio of the abundance of heavy isotopes to that of light isotopes; subscript and These represent the sample and the corresponding international reference standard, respectively; the morphological fingerprint vector of the sediment With the isotopic fingerprint vector The solid-liquid Spearman rank correlation coefficient is calculated as a coupling weight in the following manner, and the coupling weight is used as the weight coefficient of the solid-phase constraint equation in the Bayesian mixture analytical model: ;in, For the first Spearman rank correlation coefficients between groundwater concentrations of certain heavy metals and the content of BCR mobile states in corresponding sediments; For the first The rank difference between the two variables at each sampling point; only when And when the significance probability of the corresponding hypothesis test is less than 0.05, the first... The solid-state morphology constraint equations for heavy metals are incorporated into the Bayesian framework of step S4.

[0078] In this embodiment, the MW-03 well sample is still used as the example, and the measured lead isotope ratio is... , , Zinc isotope values Calculated using IRMM-3702 as the reference standard, the actual measurement in this embodiment... , Substitute into the definition formula for δ: .

[0079] Similarly, the actual measurement in this embodiment , Substituting into the calculation, we get The isotopic fingerprint vectors of the 0-50cm layers in well MW-03 were obtained by synthesis. This isotopic fingerprint is similar to the known lead isotopic characteristics (typical values) of lead-zinc ore. , The results are highly consistent, and the isotopic analysis initially points to a mine source.

[0080] For all 24 sampling points, the concentrations of heavy metals in groundwater and the corresponding BCR migration states in sediments were calculated using the Spearman rank correlation formula. The Spearman rank correlation coefficient between the contents was obtained. ( ), ( ), ( ), ( All values ​​were greater than 0.6 and significant. Therefore, the solid phase morphology constraint equations for the four heavy metals were all incorporated into the Bayesian framework, with corresponding coupling weights of 0.81, 0.74, 0.69, and 0.78 (weighting coefficients used for the sediment BCR morphology constraint equations).

[0081] The construction of the normalized multidimensional feature fingerprint matrix and its cosine similarity matching with the candidate pollution source end-member spectrum are specifically as follows: the groundwater chemical fingerprint vector Sediment morphological fingerprint vector With isotopic fingerprint vector After Z-score normalization, the vectors are concatenated column-wise to obtain the normalized multidimensional feature fingerprint vector: ;in, , , These are the results of Z-score normalization of the corresponding fingerprint vectors; all sampling points The normalized multidimensional feature fingerprint matrix is ​​obtained by stacking the rows.

[0082] For the aforementioned mining areas, electroplating and metal processing industrial parks, or wastewater irrigation agricultural areas, a candidate pollution source end-member spectrum library is pre-established. This library includes at least three categories: tailings leaching sources, smelting dust deposition sources, electroplating wastewater discharge sources, acidic mining wastewater sources, agricultural wastewater irrigation sources, and natural background sources. The cosine similarity between the normalized multidimensional feature fingerprint vector of each sampling point and the candidate pollution source end-member spectrum is calculated using the following formula: ;in, For the first One candidate pollution source end-member spectral vector; for Norm; reservation The pollution sources are used as a set of candidate pollution sources to be analyzed.

[0083] In this embodiment, the 0-50cm layer of well MW-03 is used as an example to calculate the groundwater chemical fingerprint vector. The measured concentration or ratio of the characteristic ion combination obtained from step S2, i.e. (The concentrations of arsenic, lead, cadmium, and zinc (in μg / L) and the ratios of As / F and Fe / Mn, respectively) (ratio), 7 dimensions in total; sediment morphological fingerprint vector See the aforementioned 16-dimensional vector; isotope fingerprint vector See the aforementioned 5-dimensional vector.

[0084] The three types of fingerprint vectors were each normalized using Z-score (each dimension was subtracted from the mean of all 72 water samples in that dimension and then divided by the sample standard deviation), and then concatenated column-wise to obtain the normalized multidimensional feature fingerprint vector. Dimensions: 7 + 16 + 5 = 28.

[0085] The candidate pollution source end-member spectrum library was established in advance based on the hydrogeological background and potential pollution source investigation results of the target area, including the following 5 categories of candidate pollution sources: ① acidic mine wastewater source (Typical lead-zinc mine oxidation leachate), its isotopic characteristics , ; ② Tailings leaching source (Tailings dam seepage water), isotopic characteristics , ③ Smelting dust settling sources (Atmospheric deposition and leaching from lead smelters), isotopic characteristics , ④ Agricultural wastewater irrigation sources (Fertilizer leaching + wastewater irrigation), isotopic characteristics , ⑤ Natural background source (Regional bedrock natural weathering), isotopic characteristics , .

[0086] Each endmember spectrum Also a 28-dimensional vector, it is labeled by historical survey data and typical values ​​from literature.

[0087] MW-03 well was calculated one by one using the cosine similarity formula. The cosine similarity with the five endmember spectra is obtained. , , , , .according to Based on the criteria, the set of candidate pollution sources to be analyzed in well MW-03 is as follows: (Acidic mine wastewater sources, tailings leaching sources, and smelting dust deposition sources) Agricultural wastewater irrigation sources and natural background sources were excluded due to low similarity. Candidate pollution source sets were screened for all 24 sampling points using the same method. For the target area as a whole, it was determined that four of the five candidate pollution sources—acidic mine wastewater sources, tailings leaching sources, smelting dust deposition sources, and agricultural wastewater irrigation sources—need to be included in the analysis (covering at least one of all 24 sampling points). Natural background source similarity is generally below 0.5 and is not included in the main analysis.

[0088] Step S4, multi-media coupled Bayesian hybrid analysis.

[0089] Using the mass conservation equation as the principal term of the likelihood function, and the sediment BCR morphological constraint equation and the stable isotope mass balance equation as additional constraint terms, a Bayesian framework was embedded. The Dirichlet distribution was used as the prior distribution of the contribution rate of each candidate pollution source. The Markov chain Monte Carlo method was used to sample the posterior distribution of the contribution rate of each candidate pollution source. The mean, standard deviation and 95% confidence interval of the contribution rate of each heavy metal candidate pollution source were output. The spatial distribution map of the contribution rate of each pollution source was obtained by ordinary Kriging interpolation.

[0090] The specific method of the multi-media coupled Bayesian hybrid analysis is as follows: for each sampling point, using the mass conservation equation as the main term of the likelihood function, a hybrid relationship between the observed concentration and the contribution of each candidate pollution source is constructed for each heavy metal: ;in, For the sampling point The observed concentrations of several heavy metals; For the first The contribution rate of each candidate pollution source; For the first The first candidate pollution source end-member spectrum The concentration of the end-members of the heavy metals; To conform to a mean of 0 and a standard deviation of The observation error follows a normal distribution; the contribution rates of candidate pollution sources satisfy the non-negativity and normalization constraints. .

[0091] Using the Dirichlet distribution as the joint prior for the contribution rates of each candidate pollution source: ;in, The prior concentration parameter has a range of values. Construct the log-likelihood function: ;in, The total number of heavy metals involved in the analysis is given; the Metropolis-Hastings algorithm is used to perform Markov chain Monte Carlo sampling on the posterior distribution, with a sampling chain length of no less than 50,000 steps. The first 20% of the samples are discarded as pre-burning segments, and the contribution rate of each candidate pollution source is output. mean Standard deviation and 95% confidence interval .

[0092] In this embodiment, a water sample from well MW-03 during its high-water season is used as an example to calculate the candidate pollution source set. (Acidic mine wastewater source) Tailings leaching source Smelting dust settling sources Agricultural wastewater irrigation sources ), heavy metals involved in the analysis (Arsenic, lead, cadmium, zinc). The end-member concentrations of each heavy metal in each end-member spectrum were calibrated according to historical survey data of the target area, as shown in Table 1.

[0093] Table 1 shows the end-member concentrations of each heavy metal in the end-member spectra, calibrated based on historical survey data of the target area. The standard deviation of the observation error is estimated at 5% of the measured value, i.e. , , , Dirichlet prior concentration parameter is taken (No prior information).

[0094] According to the log-likelihood formula, the observed concentration of well MW-03 Modeling hybrid relationships with endmember spectra. For example, for a candidate contribution rate combination. Arsenic concentration can be predicted by calculating using the law of conservation of mass. The residual from the observed value of 186 μg / L was 19.5 μg / L; the predicted lead concentration The residual from the observed value of 62 μg / L was 43.15 μg / L; the predicted cadmium concentration The residual from the observed value of 8.3 μg / L was 3.76 μg / L; the predicted zinc concentration The residual from the observed value of 1820 μg / L is 636.25 μg / L. Substituting this into the log-likelihood calculation yields... (The probability density with only this candidate vector is far from its maximum), indicating that this combination of contribution rates is not optimal. The Metropolis-Hastings sampler iteratively corrects itself accordingly, gradually approaching the high-probability region of the posterior distribution.

[0095] Markov chain Monte Carlo sampling was performed using the Metropolis-Hastings algorithm, with a chain length of 80,000 steps and a 20% pre-burn-in (i.e., discarding samples in the first 16,000 steps), resulting in 64,000 valid posterior samples. In this embodiment, after 80,000 steps of MCMC sampling, the posterior mean and 95% confidence interval of the contribution rate of each candidate pollution source in well MW-03 are as follows: Acidic mine wastewater source (95% CI: [0.52, 0.71]), tailings leaching source (95% CI: [0.13, 0.29]), Smelting dust settling source (95% CI: [0.05, 0.17]), agricultural wastewater irrigation sources (95% CI: [0.02, 0.10]), total 1.00, results indicate mineral source ( The total of 83% was the main source of heavy metal pollution in the MW-03 well, which is consistent with the actual situation of the well being located 300 meters downstream of the mine.

[0096] The specific way in which the sediment BCR morphological constraint equation and the stable isotope mass balance equation are embedded as additional constraint terms into the Bayesian framework is as follows: The content constraint of the movable BCR state in the sediment is used as an additional likelihood term to reflect the cumulative contribution of candidate pollution sources to the movable heavy metals in the sediment: ;in, For the first The first candidate pollution source end-member spectrum The total proportion of acid-soluble and reducible forms of various heavy metals in sediments; These are the observed values ​​of the sediment columnar samples corresponding to the sampling points; This represents the corresponding measurement uncertainty.

[0097] The weight of this constraint term is equal to the solid-liquid Spearman rank correlation coefficient. The stable isotope mass balance equation is used as an additional likelihood term: ;in, For the first The isotope ratios or isotope values ​​of the corresponding heavy metals in the end-member spectra of each candidate pollution source; These are the observed values ​​at the sampling points; This represents the uncertainty in isotope measurements.

[0098] The coupled posterior a posteriori (BPCR) was obtained by incorporating the BCR morphological constraint equation and the stable isotopic mass balance equation into the log-likelihood function. ;in, This is a multi-media integrated observation dataset; and These are the sediment morphology constraint weights and isotope constraint weights, respectively, with values ​​ranging from [value range missing]. By introducing the BCR morphological constraints and isotopic mass balance constraints of the sediments, the liquid phase tracing results are simultaneously anchored in terms of solid phase morphology and isotopic characteristics, significantly reducing the posterior distribution width of the contribution rate.

[0099] In this embodiment, sedimentary BCR morphological constraints and isotopic mass balance constraints are introduced into the MW-03 well sample. The sedimentary BCR is a mobile state. The observed values ​​(in the order of arsenic, lead, cadmium, and zinc) are: , , , .

[0100] Typical proportions of the corresponding transferable states in the four types of candidate source endmember spectra are: (Acidic mine wastewater source) ; (Tailings leaching source) ; (Source of smelting dust settling) ; (Agricultural wastewater irrigation source) BCR measurement uncertainty .

[0101] Isotope mass balance constraints and Two categories are representative, and four categories of candidate source end-member spectra Typical value Observed values Measurement uncertainty The sediment morphology constraint weights are taken as follows: Isotope constraint weights (Given that isotopic signals have stronger fingerprint characteristics in distinguishing source spectra).

[0102] After introducing dual solid-phase constraints and re-performing Metropolis-Hastings sampling, with the chain length remaining at 80,000 steps and a pre-burning rate of 20%, the posterior distribution of the contribution rate of well MW-03 in this embodiment is significantly tightened: acidic mine wastewater source (95% CI: [0.59, 0.71], confidence interval width narrowed from 0.19 for liquid-phase only constraints to 0.12, a reduction of approximately 37%), tailings leaching source (95% CI: [0.15, 0.25], confidence interval width narrowed from 0.16 to 0.10, a reduction of approximately 38%), smelting dust settling source (95% CI: [0.06, 0.14]), agricultural wastewater irrigation sources (95% CI: [0.02, 0.08]). The overall narrowing of the posterior distribution width confirms that multi-media coupling constraints have a significant uncertainty reduction effect compared to pure liquid-phase sourcing.

[0103] The method further includes Monte Carlo uncertainty quantification and spatial distribution output steps for pollution source contribution rates: For each sampling point, random perturbations following a normal distribution are applied to the observed concentrations of heavy metals in groundwater, the concentrations of candidate pollution source endmember spectra, the proportion of BCR speciation in sediments, and the isotope ratios. The standard deviation of the random perturbation is equal to the relative uncertainty of the corresponding measurement multiplied by the observed value. At least 1000 Monte Carlo simulations are performed, with each simulation running an independent Bayesian analytical process. The sample mean and 95% confidence interval of the contribution rate of each candidate pollution source are calculated. ;in, Total number of Monte Carlo simulations; For the first The simulation yielded the... The average contribution rate of each candidate pollution source; the optimal contribution rate for all sampling points. The spatial distribution field of the entire region is obtained by ordinary kriging interpolation, and the form of ordinary kriging interpolation is as follows: ;in, Points to be estimated The first Estimated contribution rates of each candidate pollution source; For the first The Kriging weights for each sampling point are determined by the variation function model and unbiased constraints. Solving the system simultaneously yields the result; The connected components are identified as those with the first... The candidate pollution sources are identified as high-risk areas under primary control. The high-risk areas and the width of the 95% confidence interval are output together as the basis for the prevention and control of heavy metal pollution in groundwater in the target area and for tracing responsibility.

[0104] In this embodiment, 1500 Monte Carlo simulations were performed on all 24 sampling points. A normal perturbation with a relative uncertainty of 5% was applied to the groundwater heavy metal concentration, a normal perturbation with a relative uncertainty of 10% was applied to the endmember spectrum concentration, a normal perturbation with an absolute uncertainty of 5% was applied to the proportion of BCR morphology in sediments, and a normal perturbation with a relative uncertainty of 0.005% was applied to the lead isotope ratio. Each simulation was run independently using Bayesian analysis (each MCMC chain length was 50,000 steps), resulting in a total of 24 × 1500 = 36,000 posterior distribution samples of contribution rates.

[0105] The final contribution rate of well MW-03, obtained from 1500 Monte Carlo simulations, stabilizes within the 95% confidence interval as follows: (95% CI: [0.58, 0.71]) (95% CI: [0.16, 0.26]) (95% CI: [0.06, 0.14]) (95% CI: [0.02, 0.09]). The mean contribution rates of the remaining 23 sampling points are distributed as follows: For the 8 sampling points MW-01 to MW-08 near the mine, the mean contribution rate of acidic mine wastewater source is... Between 0.45 0.68，矿山源贡献率 合计达0.65 0.85; the average contribution rate of smelting dust settling sources at seven sampling points (MW-09 to MW-15) located downwind of the smelter was 0.85. The value ranged from 0.30 to 0.45, indicating it was the primary controlling source in the area. Nine sampling points (MW-16 to MW-24) located in agricultural planting areas showed an average contribution rate from agricultural wastewater irrigation. Between 0.30 and 0.55.

[0106] The optimal contribution rate of 24 sampling points Spatial interpolation was performed using the ordinary Kriging interpolation formula. The variogram model employed an isotropic spherical model (range 800 meters, sill value 0.04, nugget value 0.005). The variogram model was then combined with unbiased constraints. Simultaneous solution of Kriging weights This yielded a continuous contribution rate distribution field for a 500m × 500m grid across the entire region. The connected area (the main control area of ​​acidic mine wastewater source) is approximately 2.1 square kilometers in the northern part of the target area, covering the area surrounding MW-01 to MW-08 downstream of the mine; The connected area (the main control area of ​​the smelting dust deposition source) is about 1.4 square kilometers in the northwest of the target area, which coincides perfectly with the downwind location of the smelter. The target area is a 1.8-square-kilometer southeastern region (the main control area of ​​agricultural wastewater irrigation sources), covering agricultural planting areas.

[0107] The high-risk area and the width of the 95% confidence interval are output together as a scientific basis for the prevention and control of heavy metal pollution in groundwater in the target area and for tracing responsibility.

[0108] To verify the effectiveness of the method in this embodiment, the inventors conducted comparative experiments using the following three schemes on a multi-media dataset of the same target area: Scheme A was the traditional positive definite matrix factorization (PMF) method, which performed source apportionment based solely on groundwater chemical data; Scheme B was the single isotope mixture model (IsoSource), which performed binary / ternary mixture analysis based solely on lead isotope ratios; and Scheme C was the multi-media coupled Bayesian mixture analysis method based on feature ion screening described in this embodiment. The three schemes were run independently on the same dataset of 24 sampling points, and the comparison metrics are shown in Table 2.

[0109] Table 2 Comparison of Evaluation Indicators for the Three Schemes As shown in Table 2: (1) In terms of pollution source identification capability, Scheme C can simultaneously identify 4 types of candidate pollution sources (acidic mine wastewater source, tailings leaching source, smelting dust deposition source, and agricultural sewage irrigation source), while Scheme A can only identify 3 types due to the similarity of water chemical fingerprints, and Scheme B can only identify 2-3 types due to the number of isotope endmembers. Scheme C's multi-source analysis capability is significantly better than the comparison schemes; (2) In terms of analysis accuracy, the root mean square reconstruction error of the contribution rate of Scheme C is 0.21 mg / L, which is about 73% lower than that of Scheme A and about 86% lower than that of Scheme B. The multi-media coupling constraint significantly improves the accuracy of analysis; (3) In terms of uncertainty quantification, the average width of the 95% confidence interval of the contribution rate of Scheme C is only 0.11, compared to 0.32 of Scheme A. The 0.41 of Scheme C was significantly reduced by about 65%~73%, thanks to the dual anchoring effect of solid phase morphology constraints and isotope mass balance constraints; (4) In terms of on-site consistency, the pollution source types identified by Scheme C had a consistency of 92% with the actual pollution source types known from the investigation of the target area, and the high-risk area identification had a consistency rate of 88% with historically polluted plots, which was much higher than Scheme A and Scheme B; (5) In terms of analysis time, Scheme C introduced 1500 Monte Carlo simulations and multi-media constraint solutions, and the analysis time for a single sampling point was about 25.6 seconds (under Intel Xeon Gold 6248R CPU, single thread), which was slightly higher than Scheme A and B, but the absolute time was still within the acceptable range for engineering (analysis of the entire dataset of 24 sampling points took about 10 minutes).

[0110] In summary, the method described in this embodiment has significant advantages over traditional PMF and single isotope mixture models in key indicators such as multi-source identification capability, analytical accuracy, uncertainty quantification, and field fit. It can meet the application needs of refined source apportionment and responsibility tracing in downstream multi-source superimposed heavy metal pollution scenarios in lead-zinc mining areas.

[0111] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for source apportionment of heavy metal pollution in groundwater based on characteristic ion screening, wherein the heavy metals include at least three of arsenic, lead, cadmium, chromium, copper, and zinc, characterized in that, The method includes: Groundwater samples, sediment columnar samples from aquifers corresponding to the location of the monitoring wells, and surface soil samples were collected from the target area. The total amount of heavy metals and the content of BCR speciation in the sediment columnar samples were determined, and the stable isotope ratios of lead, zinc, and copper in the surface soil and sediment samples were determined. Monitoring wells are deployed using a grid method, and groundwater samples are collected from each monitoring well no less than three times to cover the wet season, normal season and dry season, while sediment column samples from the aquifer are collected simultaneously. The groundwater sample testing items include the total concentration of arsenic, lead, cadmium, chromium, copper and zinc; the cation concentration of sodium, potassium, calcium, magnesium and ammonium; the anion concentration of chloride, sulfate, nitrate and bicarbonate; as well as pH value, redox potential, dissolved oxygen, conductivity and water temperature. Isotope determination includes Three sets of lead isotope ratios, and zinc isotope values Copper isotope values ; The ions in the groundwater sample were evaluated sequentially based on enrichment factor screening, ion ratio stability screening and source region correlation screening to obtain characteristic ion combinations that have an indicative role for the heavy metal pollution source. The specific method for enrichment factor screening is as follows: using lithium as a reference element, the enrichment factors of each heavy metal in the groundwater sample—arsenic, lead, cadmium, chromium, copper, and zinc—are calculated separately. ;in, For the first Enrichment factors for heavy metals; The first in the groundwater sample The concentration of certain heavy metals; This refers to the concentration of the reference element in the same water sample; and These are the ratios of heavy metal concentration to reference element concentration in the water samples from the sampling point and the background value, respectively. Pollution source indicative classification for each heavy metal: when It was determined to be of natural origin at that time; when It was determined to be a mild anthropogenic enrichment at that time; It was determined to be of significant anthropogenic origin, and the heavy metal was included in the candidate set of characteristic ions; The specific methods for screening ion ratio stability and source region correlation are as follows: For the aforementioned candidate set of characteristic ions, at least four sets of indicator ion ratios are constructed, including arsenic to fluorine ratio, iron to manganese ratio, sulfate to chloride ratio, and nitrate to chloride ratio, and the coefficient of variation of each set of ratios at all sampling points is evaluated. ;in, For the first The coefficient of variation of the group ion ratio; and The first The sample standard deviation and arithmetic mean of the group ion ratios at all sampling points; retain The ion ratio is used as a stability characteristic indicator; Pearson correlation analysis was performed on the concentrations of each heavy metal in the candidate set of characteristic ions and the stability characteristic indices, and only those meeting the criteria were retained. and The ion pairs are used as the final characteristic ion combinations to construct the groundwater chemical fingerprint vector: ;in, For the first The characteristic ion and the first Pearson correlation coefficients of characteristic ions in groundwater samples; This represents the total number of groundwater samples. , The first At the sampling point The, the The concentration or ratio of a characteristic ion; , This is the corresponding sample mean; This represents the significance probability of the corresponding hypothesis test. A groundwater chemical fingerprint vector is constructed based on characteristic ion combinations, a sediment morphological fingerprint vector is constructed based on BCR speciation content, and an isotopic fingerprint vector is constructed based on stable isotope ratios. After normalization, the vectors are concatenated column by column to obtain a normalized multidimensional feature fingerprint matrix. The normalized multidimensional feature fingerprint matrix is ​​matched with the candidate pollution source endmember spectrum by cosine similarity to screen out the set of candidate pollution sources to be analyzed. Using the mass conservation equation as the principal term of the likelihood function, and the sediment BCR morphological constraint equation and the stable isotope mass balance equation as additional constraint terms, a Bayesian framework is embedded. The Dirichlet distribution is used as the prior distribution of the contribution rate of each candidate pollution source. The Markov chain Monte Carlo method is used to sample the posterior distribution of the contribution rate of each candidate pollution source, and the mean, standard deviation and 95% confidence interval of the contribution rate of each candidate pollution source for each heavy metal are output.

2. The method according to claim 1, characterized in that, The method for constructing sediment morphological fingerprint vectors based on BCR speciation content is as follows: For each heavy metal (arsenic, lead, cadmium, chromium, copper, and zinc) in the sediment column sample, four sequential extraction steps are performed, followed by a final step of complete digestion. The four sequential extraction steps include: The first step involves extracting acid-soluble or exchangeable substances using acetic acid at a concentration of 0.11 mol / L. ; The second step involves extracting the reducible form of hydroxylamine hydrochloride at a concentration of 0.5 mol / L and a pH of 2. ; The third step involves using hydrogen peroxide at a concentration of 8.8 mol / L and ammonium acetate at a concentration of 1 mol / L in combination to extract the oxidizable form. ; The fourth step involves using aqua regia to microwave digest the residue to obtain the residue state. Finally, the total amount of heavy metals was determined by complete digestion with three acids. ; Definition of the first Mobility index of heavy metals: ;in, , , , The first The content of each form of heavy metal in the sediment column sample; when At that time, the first These heavy metals were identified as highly mobile heavy metals and included in the pollution source analysis system; For each sediment column corresponding to a sampling point, the heavy metals were... The morphological proportions of the sediments were assembled in order of heavy metal species to obtain the morphological fingerprint vector of the sediments. .

3. The method according to claim 2, characterized in that, Construct the isotopic fingerprint vector of the sampling point: The lead isotope ratios were determined by multi-receiver inductively coupled plasma mass spectrometry using the acid digestion solution of the groundwater sample and the corresponding sediment column sample. For the sediment morphological fingerprint vector With the isotopic fingerprint vector The solid-liquid Spearman rank correlation coefficient is calculated as a coupling weight, which is used as the weighting coefficient of the solid-phase constraint equation in the Bayesian mixture analytical model. ;in, For the first Spearman rank correlation coefficients between groundwater concentrations of certain heavy metals and the content of BCR mobile states in corresponding sediments; For the first At the sampling point, the first The rank difference between the groundwater concentration and the content of the corresponding BCR mobile state in sediments for a certain heavy metal; Only when And when the significance probability of the corresponding hypothesis test is less than 0.05, the first... The solid-state constraint equations for heavy metals are incorporated into a Bayesian framework.

4. The method according to claim 3, characterized in that, The groundwater chemical fingerprint vector Sediment morphological fingerprint vector With isotopic fingerprint vector After Z-score normalization, the vectors are concatenated column-wise to obtain the normalized multidimensional feature fingerprint vector: ;in, , , These are the results of Z-score normalization of the corresponding fingerprint vectors; all sampling points The normalized multidimensional feature fingerprint matrix is ​​obtained by stacking the rows. Calculate the cosine similarity between the normalized multidimensional feature fingerprint vector of each sampling point and the candidate pollution source endmember spectrum: ;in, For the first One candidate pollution source end-member spectral vector; retain The pollution sources are used as a set of candidate pollution sources to be analyzed.

5. The method according to claim 4, characterized in that, For each sampling point, using the mass conservation equation as the main term of the likelihood function, a mixed relationship between the observed concentration and the contribution of each candidate pollution source is constructed for each heavy metal: ;in, For the sampling point The observed concentrations of several heavy metals; For the first The contribution rate of each candidate pollution source; For the first The first candidate pollution source end-member spectrum The concentration of the end-members of the heavy metals; To conform to a mean of 0 and a standard deviation of The observation error of the normal distribution; The contribution rates of candidate pollution sources satisfy the non-negativity and normalization constraints: Using the Dirichlet distribution as the joint prior for the contribution rates of each candidate pollution source: ;in, The prior concentration parameter has a range of values. Construct the log-likelihood function: ;in, The total number of heavy metals involved in the analysis is given; the Metropolis-Hastings algorithm is used to perform Markov chain Monte Carlo sampling on the posterior distribution, with a sampling chain length of no less than 50,000 steps. The first 20% of the samples are discarded as pre-burning segments, and the contribution rate of each candidate pollution source is output. mean Standard deviation and 95% confidence interval .

6. The method according to claim 5, characterized in that, The BCR mobile state content constraint in the sediments is used as an additional likelihood term: ;in, For the first The first candidate pollution source end-member spectrum The total proportion of acid-soluble and reducible forms of various heavy metals in sediments; These are the observed values ​​of the sediment columnar samples corresponding to the sampling points; This represents the corresponding measurement uncertainty; the weight of this constraint term is equal to the solid-liquid Spearman rank correlation coefficient. ; The stable isotope mass balance equation is used as an additional likelihood term: ;in, For the first The isotope ratios or isotope values ​​of the corresponding heavy metals in the end-member spectra of each candidate pollution source; These are the observed values ​​at the sampling points; The uncertainty of isotope measurement; The coupled posterior axiom is obtained by incorporating the BCR morphological constraint equation of the sediment and the stable isotopic mass balance equation into the log-likelihood function: ;in, This is a multi-media integrated observation dataset; and These are the sediment morphology constraint weights and isotope constraint weights, respectively, with values ​​ranging from [value range missing]. .

Images