Regional water pollutant source analysis method and system

Abstract

The invention discloses a regional water body pollutant source analysis method and system, and the method comprises the steps: calculating a corresponding scene feature index based on the input of multi-dimensional data, and carrying out the scene type judgment. Then, constructing a steady-state fugacity model of a four-medium system comprising a water phase (W), sediments (S), land soil (L) and ambient air (A), and establishing a mass balance equation set of the steady-state fugacity model; and in combination with a scene classification result, solving the model equation by adopting a Bayesian inversion method, and simultaneously inverting the emission source intensity, the non-advection input flux and the advection input concentration of each medium on the basis of the migration matrix and the fugacity capacity of the model equation. The system comprises a data acquisition module, an index calculation module, a scene classification module, an equation construction module and an inversion module. By using the method, the flux of each medium discharge source and advection input is synchronously inverted, and pollution traceability is realized. The method can be widely applied to the field of pollution analysis.

Description

Technical Field

[0001] This invention relates to the field of pollution analysis, and in particular to a method and system for source apportionment of pollutants in regional water bodies. Background Technology

[0002] With economic development and urbanization, the types and sources of water pollutants are becoming increasingly diversified. Relying solely on emission inventories is no longer sufficient to reveal the true pollution load. Only by first clarifying the sources of pollutants can subsequent treatment measures (such as upgrading wastewater treatment, reducing agricultural waste, dredging sediment, and ecological restoration) be truly targeted. Therefore, regional water pollutant source apportionment is the fundamental technical support for achieving precise pollution control, scientific decision-making, and effective supervision, and it occupies an irreplaceable core position in the modern water environment governance system.

[0003] Most existing technologies are multivariate statistical models represented by positive definite matrix factorization (PMF). Although widely used, they are essentially data-driven "black box" models that lack descriptions of the physicochemical processes such as the migration and transformation of pollutants between different environmental media, resulting in unclear physical meaning of the analysis results. Summary of the Invention

[0004] In view of this, in order to address the technical problem that existing water pollutant source apportionment methods lack characterization of the interaction processes of pollutants between different environmental media, leading to inaccurate source tracing results, the present invention proposes a regional water pollutant source apportionment method, which includes the following steps: Based on multidimensional data input, corresponding scene feature indicators are calculated and used to determine scene categories. Subsequently, a steady-state fugacity model of a four-media system including water (W), sediment (S), terrestrial soil (L), and ambient air (A) is constructed, and its mass balance equations are established. Combining the scene classification results, a Bayesian inversion method is used to solve the model equations. Based on the migration matrix and fugacity capacity of the aforementioned model equations, the emission source intensity, non-advection input flux, and advection input concentration of each medium are simultaneously inverted.

[0005] Furthermore, the scenario classification results include three categories: high pollutant migration-persistence comprehensive index and low system heterogeneity index; high pollutant migration-persistence comprehensive index and high system heterogeneity index; and low pollutant migration-persistence comprehensive index.

[0006] Based on the above method, in a second aspect, the present invention also proposes a regional water pollutant source apportionment system, which includes a data acquisition module, an index calculation module, a scenario classification module, an equation construction module, and an inversion module.

[0007] Based on the above scheme, this invention provides a method and system for source apportionment of pollutants in regional water bodies. It integrates a three-level steady-state fugacity model with a clear physicochemical mechanism with powerful Bayesian statistical inference, thereby overcoming the fundamental defects of pure statistical models, such as "black box" operation and unclear physical meaning. At the same time, it solves the problem that traditional fugacity models cannot quantify the uncertainty of results in inversion applications. Secondly, by introducing a scenario index system (TPI, SHI) and an adaptive branching mechanism based on it, this invention can intelligently identify the characteristics of different pollutant-environment systems and dynamically adjust the model complexity and inversion strategy (such as fully coupled inversion and local source fine inversion). While ensuring the integrity of the physical mechanism, it improves the computational efficiency and the robustness of the model under different application scenarios. Attached Figure Description

[0008] Figure 1 This is a flowchart of the steps of a method for source apportionment of pollutants in regional water bodies according to the present invention; Figure 2 This is a schematic diagram comparing the measured and predicted concentrations of pollutants in different environmental media according to the present invention; where (A) is α-HCH and (B) is β-HCH. Figure 3 This is a schematic diagram illustrating the water sources of α-HCH and β-HCH in a specific embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the fugacity and fugacity capacity inversion of pollutants in α-HCH according to a specific embodiment of the present invention; Figure 5 This is a schematic diagram illustrating the fugacity and fugacity capacity inversion of pollutants in β-HCH according to a specific embodiment of the present invention; Figure 6 This is a schematic diagram illustrating the fugacity distribution of pollutants in different media according to a specific embodiment of the present invention; Figure 7 This is a schematic diagram illustrating the pollution flux (D value) entering the water body according to a specific embodiment of the present invention; wherein (A) α-HCH; (B) β-HCH; Figure 8 This is a schematic diagram of the migration flux (D value) of α-HCH in a specific embodiment of the present invention; Figure 9 This is a schematic diagram of the migration flux (D value) of β-HCH in a specific embodiment of the present invention; Figure 10 This is a schematic diagram illustrating the background source concentrations of α-HCH and β-HCH in a specific embodiment of the present invention; Figure 11 This is a specific embodiment of the present invention showing the background concentration inversion of α-HCH and β-HCH in sediments and soil; Figure 12This describes the relative error in predicting the concentrations of α-HCH and β-HCH in different media in specific embodiments of the present invention. Figure 13 This is a schematic diagram of the posterior distribution of emission sources in a specific embodiment of the present invention; wherein (A) α-HCH; (B) β-HCH; Figure 14 This is a schematic diagram of the posterior distribution of water input in a specific embodiment of the present invention; wherein (A) α-HCH; (B) β-HCH; Figure 15 This is a schematic diagram of the a posteriori distribution of water advection input concentration in a specific embodiment of the present invention; wherein (A) water body α-HCH; (B) water body β-HCH; Figure 16 This is a schematic diagram of the posterior distribution of air advection input concentration in a specific embodiment of the present invention; wherein (C) air α-HCH; (D) air β-HCH; Figure 17 This is a schematic diagram of the posterior distribution of sediment background concentration in a specific embodiment of the present invention; wherein (A) sediment α-HCH; (B) sediment β-HCH; Figure 18 This is a schematic diagram of the posterior distribution of soil background concentration in a specific embodiment of the present invention; wherein (C) soil α-HCH; (D) soil β-HCH. Detailed Implementation

[0009] In addition to the technical problems mentioned in the background section, existing water pollutant source apportionment methods also employ chemical mass balance and isotope tracing methods. However, the former relies on accurate source composition spectra, while the latter is limited by high costs and applicability to specific pollutants.

[0010] To address these shortcomings, the present invention aims to provide a multi-source apportionment method for regional water pollutants that couples a Level III steady-state fugacity model with Bayesian inversion. Its primary objective is to establish a source apportionment framework with a clear physical mechanism, accurately describing multi-media interaction processes through the fugacity model. The core objective is to introduce Bayesian statistical inference to achieve simultaneous flux inversion for emission sources and advection inputs of various media, and to quantitatively output the uncertainty of the results in the form of a posteriori distribution. This constructs a robust source tracing tool that reflects the physical essence, possesses anti-interference capabilities, and quantifies risks, providing a scientific basis for precise watershed management.

[0011] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0012] Reference Figure 1 This is a schematic flowchart of an optional example of the regional water pollutant source apportionment method proposed in this invention. The method can be applied to computer equipment, and the apportionment method proposed in this embodiment may include, but is not limited to, the following steps: Step S1: Obtain multidimensional data; Specifically, the geographical boundaries of the study area must first be defined. For the four media within the study area—water, sediment, soil, and atmosphere—basic environmental parameters (such as the area of ​​the study area and atmospheric volume) must be obtained. The chemical properties of pollutants must be obtained. The observed concentrations must be obtained through monitoring and analysis or by collecting historical monitoring data. The prior parameters of the model must be obtained. The data distribution type (such as normal distribution) and parameter values ​​of the above data must be obtained through statistical inference.

[0013] In the data preparation phase, a scenario index calculation module is added. To overcome the technical shortcomings of existing technologies that rely on subjective experience in model selection and cannot achieve an adaptive balance between computational efficiency and model accuracy, the following scenario indices (TPI, SHI) are proposed, which automatically calculate two types of scenario indices based on input environmental parameters and chemical properties: Pollutant Migration-Persistence Composite Index (TPI): This index is used to predict the overall behavior trend of pollutants in the environment. Its calculation method is shown in formula (1), and its classification is shown in Table 1.

[0014] (1) in, The octanol-water partition coefficient. It is the half-life.

[0015] Table 1. TPI Classification System heterogeneity index (SHI): It is the product of the coefficient of variation of the medium volume and the area ratio of the soil-air / water-air interface. It is used to quantify the spatial structural complexity of the environmental medium in the study area. Its calculation method is shown in formula (2), and its classification is shown in Table 2.

[0016] (2) Table 2 SHI Classification These two indicators will provide a basis for decision-making in subsequent model selection and parameterization. First, they are used to improve computational efficiency, that is, to avoid unnecessary full-model computation in complex scenarios (high SHI). Second, they are used to enhance model stability, that is, to provide early warning for highly heterogeneous systems and initiate stabilization measures such as regularization.

[0017] Step S2: Calculate scenario indicators based on multidimensional data; Step S3: Classify scenarios based on scenario metrics; Specifically, Scenario A (high TPI, low SHI): This is determined to be the migration of persistent pollutants in a homogeneous environment. In this case, the system preferentially employs a complete four-medium coupled inversion model (i.e., the aforementioned basic model) and enables Bayesian inversion in the full parameter space to accurately capture the details of cross-medium exchange. Scenario B (high TPI, high SHI): This is determined to be the migration of persistent pollutants in a heterogeneous and complex environment. To avoid over-parameterization and enhance stability, the system enables a model dimensionality reduction branch. This branch, based on sensitivity analysis results, will... For migration processes with a medium flux contribution rate less than a threshold (e.g., <1%), the D value is fixed as the prior expected value. Inversion is performed only for key sources and migration processes to reduce computational complexity. Case C (Low TPI): The pollutant is determined to be easily degradable or have weak mobility. The system determines that it is insensitive to contributions from distant or long-term sources, therefore a local source-dominated branch is activated. This branch focuses on direct discharge into water bodies (E... W ) and adjacent media (sediments E) S Terrestrial soil E L A fine inversion is performed on the source strength of (C), while the atmospheric advection input (C) is... in,A For distant sources, stronger prior constraints are applied (such as reducing the standard deviation of the prior distribution).

[0018] Step S4: Construct the mass balance equation for the four-medium fugacity model; Specifically, for the study area, the method for constructing the steady-state fugacity model of the four-media system—aquatic phase (W), sediment (S), terrestrial soil (L), and ambient air (A)—is as follows: Fundamental fugacity equation: The principle of fugacity and its basic equation are shown in formula (3): (3) in, as medium Fugacity (Pa); as medium The concentration (mol / m³); as medium Fugacity capacity (mol / m³ / Pa).

[0019] For fugacity capacity ( Its calculation model is as follows: Fugacity capacity of water body: Z W = 1 / H, where H is the Henry's constant (Pa·m³ / mol); Air fugacity capacity: Z A= 1 / (R·T), where R=8.314 J / (mol·K) and T=298 K.

[0020] Sediment fugacity capacity: Z S = K oc ·Z W ·foc S ·ρ S K oc The adsorption coefficient of pollutants on organic carbon in sediments (m) 3 / kg), foc S Organic carbon fraction (%), ρ S The medium density of the sediment (kg / m³) 3 ).

[0021] Soil fugacity capacity: Z L = K oc ·Z W ·foc L ·ρ L K oc The organic carbon adsorption coefficient of pollutants in soil (m 3 / kg), foc S Organic carbon fraction (%), ρ S The medium density of the soil (kg / m³) 3 ).

[0022] Construction of the mass balance equation for the four-medium fugacity model: For a four-medium system consisting of water (W), sediment (S), soil (L), and air (A), the mass balance equation for the four-medium steady-state fugacity model is expressed in matrix form, as shown in formula (4): (4) in, Here is the migration matrix. For emission source vectors, The fugacity vector. This is the external input vector. The mass balance equations for each medium are shown in Table 3: Table 3 Mass Balance Equations for Four Media as medium The advection output flux (mol / Pa / h).

[0023] Step S5: Based on the scene classification results, perform Bayesian inversion on the mass balance equation of the four-medium fugacity model to obtain analytical results.

[0024] Specifically, the Bayesian inversion algorithm is constructed as follows: Posterior distribution form: For the mass balance equation in Table 3, the basic formula for the posterior distribution of its Bayesian inversion is shown in formula (5): (5) in, The vector of parameters to be estimated; For observation datasets; It is a posterior distribution; Let θ be the likelihood function; P(θ) be the prior distribution. Evidence (normalization constant). E is the amount of exogenous pollutant input; Q in,W C represents the amount of external pollutants input into the water body. in,W The concentration of pollutants is introduced into the water body through advection; C in,A The concentration of pollutants is introduced by atmospheric advection; This represents the background concentration of sediments. This represents the background concentration in terrestrial soil.

[0025] Parameter space definition: The complete parameter vector is shown in formula (6): (6) in, (Emission source intensity); (Non-adaptive input); (Advection input concentration); (Background pollutant content). Its parameter constraints are: ; ; .

[0026] Prior distribution model: The hierarchical prior structure is shown in formula (7): (7) The prior distributions of the parameters in formula (7) are as follows: The prior distribution of emission sources is shown in formula (8): (8) in, The mean, The standard deviation is denoted as .

[0027] The probability density (PDF) function is shown in equation (9): (9) For standard normal PDFs, It is a standard normal CDF.

[0028] The prior distribution of the input parameters is shown in formula (10): (10) Prior hyperparameter setting: The prior hyperparameter setting is shown in formula (11): (11) Likelihood function model: The observation error model is shown in formula (12): (12) in, To predict concentration (mol / m 3 ).

[0029] The complete likelihood function is shown in equation (13): (13) in, To predict concentration (mol / m 3 ).

[0030] The concentration predicted by the model is calculated as shown in formula (14): (14) Where Z is the fugacity capacity (mol / Pa / m³) 3 ).

[0031] (15) in, ; .

[0032] The complete form of the posterior distribution: The non-normalized posterior distribution is shown in formula (16): (16) The log-posterior distribution is shown in formula (17): (17) Bayesian computational model: Markov chain-Monte Carlo (MCMC) sampling algorithm: MCMC uses the No-U-TurnSampler (NUTS) algorithm, as shown in formula (18): (18) in, Indicates the step size parameter; Represents the maximum tree depth; Acceptance probability: That is, using the NUTS algorithm from time 1000... status (Parameter vector), generation time status (A new state generated by the NUTS algorithm).

[0033] Transfer nucleus: the nuclear transfer parameters involved in formula (17) That is, describing from State transition The probability law of the state is shown in formula (19): (19) in, Conditional transition distribution for each dimension.

[0034] Generation calculation model: The posterior prediction distribution is shown in formula (20): (20) The posterior distribution of fugacity is shown in Equation (21): (twenty one) The posterior distribution of migration flux in each medium is shown in Equation (22): (twenty two) In some feasible embodiments, enhanced dynamic regularization processing is also included, specifically including: Based on the original dynamic regularization, a transition matrix-based approach is introduced. Adaptive regularization strategy for analysis: Perform eigenvalue decomposition, if its smallest eigenvalue is... ( To preset tolerance, such as or If the condition number is increased, then diagonal elements are added. This improvement can more effectively improve the matrix condition number and accelerate the convergence of linear equations in the MCMC sampling process.

[0035] In some feasible embodiments, parallelization and preheating strategies for the MCMC sampling process are also included, specifically including: To address the efficiency issue in high-dimensional parameter space sampling, the MCMC sampling process is improved. For cases A and B, a multi-chain parallel sampling technique is employed, utilizing modern multi-core processor architectures to run multiple MCMC chains simultaneously. In the initial sampling phase, a staged warm-up strategy is introduced: the first stage uses smaller `adapt_delta` (e.g., 0.8) and `max_treedepth` (e.g., 8) for rapid exploration to locate the approximate region of the posterior distribution; the second stage reverts to standard parameters (`adapt_delta=0.95`, `max_treedepth=12`) for fine-tuning. This strategy effectively reduces computation time while maintaining sampling quality.

[0036] Improvements to the computation process in this embodiment (such as enhanced dynamic regularization and parallel preheating sampling) further ensure the feasibility of the method in large-scale, high-dimensional problems.

[0037] In some feasible embodiments, model validation and result analysis are also included, specifically: Model Validation: The posterior predicted value (PPP-value) is used to check the model as shown in formula (23): (twenty three) in, For a given type and data ( Under the condition of ), the predictive statistic ( ) greater than the observed statistic ( The probability of ).

[0038] The Bayesian p-value is shown in formula (24): (twenty four) The information criteria adopted are the Deviation Information Criterion (DIC) and the Watanabe-Akaike Information Criterion (WAIC). The DIC is shown in formula (25): (25) in, .

[0039] WAIC is shown in formula (26): (26) Uncertainty Quantification Model: Posterior Interval Estimation: 95% Confidence Interval is ,in, For the posterior sample Quantiles.

[0040] The posterior correlation coefficient matrix is ​​shown in formula (27): (27) in, This means that for a given set of data ( In the case of ), two parameters ( and The strength of the linear correlation between variables j and k, i.e., the correlation coefficient between variables j and k; Indicates that in the given data ( Under the condition of ), parameters and covariance; and These respectively represent the given data ( Under the condition of ), parameters and The variance.

[0041] The posterior coefficient of variation is shown in formula (28): (28) in, The posterior coefficient of variation quantifies a specific parameter based on existing data. The ratio of the dispersion of the parameter estimate to its mean is used to evaluate the stability and reliability of the parameter estimate. For given data ( Under the condition of ), parameters The expected value.

[0042] Multi-pollutant Bayesian hierarchical model: Hierarchical prior structure: for pollutants Its hierarchical prior structure is shown in formula (29): (29) in, To be with pollutants The relevant parameters follow a multivariate normal distribution with a mean of 1. The covariance matrix is ; It follows a normal distribution with a mean of 0 and a variance of 10 squared (i.e., 100). It follows the inverse Wishart distribution, with parameters as follows: and .

[0043] Joint posterior distribution ( As shown in formula (30): (30) Convergence diagnostic statistic: Potential scaling factor ( As shown in formula (31): (31) in, For the intra-chain variance, Let V be the inter-chain variance.

[0044] Effective sample size ( The calculation method is shown in formula (32): (32) in, This is the autocorrelation function.

[0045] This invention not only characterizes the entire migration and transformation process of pollutants among multiple media such as water, sediment, soil, and air, achieving a physical mechanism-driven analysis of pollution sources, but also its Bayesian inversion framework can synchronously and robustly estimate various emission sources and input fluxes, and provide the confidence interval of each source contribution value in the form of a complete posterior probability distribution, realizing the leap from "point estimation" to "probability assessment". It can also efficiently achieve synchronous analysis of multiple pollutants, providing uncertainty quantification information for risk management and technical support for source tracing methods for regional water pollution prevention and control.

[0046] In some feasible embodiments, step S1 specifically includes: the system collecting and organizing various parameters and observation data required for modeling, specifically including key information such as environmental parameters, chemical properties, migration parameters, actual observed concentrations, and prior parameters. These data are shown in the table below. By integrating the multidimensional data covered in the above table, a comprehensive and consistent data foundation is provided for subsequent model construction and parameter estimation.

[0047] Table 4. Physicochemical properties of compounds Table 5 Environmental parameters Table 6 Mass transfer process parameters (transport_parameters) Table 7. Observed concentrations of compounds in environmental media. Table 8 Prior parameters In some feasible embodiments, step S2 specifically includes: The calculation results of the scenario indicators are shown in Tables 9 and 10.

[0048] Table 9. TPI Calculation Process and Results As shown in Table 9, the total pollutant concentrations (TPIs) of both HCH isomers are greater than the "high TPI" threshold (as shown in Table 1), thus classifying them as high TPI pollutants. This indicates that they possess high environmental persistence and significant cross-media migration capabilities, exhibiting complex fate behavior in the water-sediment-soil-air system. Among them, β-HCH is more persistent than α-HCH, remaining longer in soil and sediments, but its atmospheric migration is relatively weaker.

[0049] Table 10 SHI Calculation Results The calculation process for each parameter is as follows: The volume mean (μ) = (3.42E+08 + 8.55E+06 + 2.12E+09 + 2.47E+12) / 4 = 6.18E+11; The standard deviation of volume (σ) = 1.23E+12; the coefficient of variation of volume (σ / μ) = 1.23E+12 / 6.18E+11 = 1.99.

[0050] The results showed that the SHI value of 193.2 was significantly higher than the typical threshold (see Table 2), indicating that the environmental media in the study area had high heterogeneity and complexity in spatial distribution.

[0051] In some feasible embodiments, step S3 specifically includes: Based on the above indicators, the system automatically determines the situation: Judgment result: (High TPI) (High TPI), SHI = 193.2 (High SHI), meets scenario B: persistent pollutants migrate in heterogeneous and complex environments.

[0052] The corresponding branching strategy is "model dimensionality reduction and branch activation," meaning the system automatically performs sensitivity analysis to identify key migration processes. Key process identification (based on D-value calculation): Atmosphere-water exchange: Water-sediment diffusion: ; Water body advection output: ; Example of a non-critical process: Sediment-water resuspension: (<1% threshold); Wet deposition into soil: (<1% threshold); Dimensional reduction operation: , For migration processes with an equal flux contribution rate of <1%, the D value is fixed as the prior expected value, and inversion is performed only for key source terms and major migration processes.

[0053] In some feasible embodiments, step S4 specifically includes: By establishing a mass balance equation based on matrix expression, constructing a migration matrix, and using a dynamic regularization algorithm to solve the singular matrix problem, the fugacity capacity (Z) and transmission flux (D) of each medium are calculated.

[0054] The mass balance equation for the four-medium steady-state fugacity model is expressed in matrix form: ; migration matrix Construction: For a four-media system consisting of water (W), sediment (S), terrestrial soil (L), and ambient air (A), a 4×4 migration matrix is ​​first defined as follows: The specific calculations for each element are as follows: Table 11 Formulas for calculating the migration processes (D value) of α-HCH and β-HCH between different phases In some feasible embodiments, step S5 specifically includes: For the matrix equation of the migration matrix, MCMC sampling is performed using the Stan platform to obtain the posterior distribution: The parameter space is defined as follows: in, as medium The emission source intensity (mol / h); This indicates non-advection input of water (mol / h); This indicates the advection input concentration of water (mol / m³). Indicates the advection input concentration of air (mol / m³); Indicates the background concentration of sediments (mol / m³); This indicates the background concentration of terrestrial soil (mol / m³).

[0055] The Bayesian posterior distribution is shown in the following equation: Its likelihood function is shown below: Its prior distribution parameters are as follows: The model predicts the concentration as follows: Dynamic regularization algorithm: Before using the enhanced dynamic regularization algorithm, the transition matrix D matrix The condition number is calculated as: cond(D) matrix ) ≈ 1.2E+15 > Tolerance ε tol =1E-8.

[0056] An enhanced dynamic regularization algorithm is used to improve the matrix condition number by iteratively increasing diagonal elements, thus ensuring the numerical stability of the linear equation system. Regularization process: Initial regularization: diag(D matrix ) = diag(D matrix ) + 1E-3; Spectral analysis: Minimum eigenvalue .

[0057] MCMC sampling algorithm: Initialization strategy: Each chain independently samples initial values ​​from the prior distribution. The distribution of core parameters is as follows: A phased warm-up strategy is adopted: Phase 1: rapid exploration (first 500 iterations) to quickly locate the approximate area of ​​the posterior distribution and avoid lingering near local extrema in the early stage; Phase 2: fine sampling (last 2500 iterations) to perform fine sampling in the located area to ensure the quality of the posterior distribution.

[0058] Flux matrix calculation algorithm: Instantaneous flux calculation: For each posterior sample : For each source medium src and target medium tgt: Average flux matrix: Algorithm for water source contribution analysis: Source flux extraction: The following formula is a code example based on the R language: Background_inflow = Q_in_W, including background input + wastewater discharge; Advection_inflow = Q_advection_W, advection input; Soil_runoff = flux[:, L, W], soil runoff; Sediment_release = flux[:, S, W], sediment release; Atm_deposition = flux[:, A, W], atmospheric deposition; Contribution ratio calculation: The following formula is a code example based on the R language: Total input = Background_inflow + Advection_inflow + Soil_runoff + Sediment_release + Atm_deposition Source contribution ratio = (Total throughput from each source / Total input) × 100% Model validation algorithm: The prediction accuracy evaluation is shown below: Acceptable criteria: relative error ≤ 30%; uncertainty quantification: 95% confidence interval: [q2.5, q97.5]; coefficient of variation: CV = σ / μ × 100%; Further analysis of the inversion results: Using Bayesian inference methods, the Bayesian Markov Chain Monte Carlo (MCMC) method is employed for parameter inversion, simultaneously estimating α-HCH and β-HCH emission sources, non-advection inputs, and advection input concentrations, as well as background pollutant concentrations in soil and sediments; and achieving simultaneous source apportionment and uncertainty quantification for multiple pollutants.

[0059] The multi-pollutant simultaneous analysis algorithm adopts a parallel processing framework and performs the following steps for each pollutant: 1) prepare pollutant-specific parameters; 2) run Bayesian inversion; 3) calculate flux matrix and source contribution; 4) generate visualization results.

[0060] The measured and predicted concentrations of α-HCH and β-HCH in different environmental media are shown in Table 12. A comparison of the measured and predicted concentrations is shown below. Figure 2 As shown in Table 13, the inversion results of input fluxes from different sources in the regional water bodies are presented in Table 14. Figure 3 As shown in Table 14, the fugacity values ​​(f) and fugacity capacities (Z) of α-HCH and β-HCH in different phases (environmental media) are inverted. Figure 4 and Figure 5 As shown, its fugacity distribution in different media is as follows: Figure 6 As shown. The pollution flux entering the water body based on Bayesian inversion (D value, mol / Pa / h) is as follows. Figure 7 As shown; the migration flux (D value) of α-HCH and β-HCH between different phases is as follows. Figure 8 and Figure 9 As shown in Table 15, the exogenous input amounts of α-HCH and β-HCH in different media in the region and their background concentrations in different media are shown in Table 15. Figure 10 As shown, the background concentration inversion in sediments and soils is as follows: Figure 11 As shown.

[0061] Table 12 Measured and predicted concentrations in different environmental media Table 13 Bayesian inversion results: Input fluxes from different sources in the regional water body Table 14 Bayesian inversion results: Fugacity values ​​(f) and fugacity capacity (Z) between different phases Table 15 Bayesian Inversion Results: External Input Amount and Background Pollutant Concentration (mol / m³) 3 ) Model validation and posterior parameter distribution analysis: Phase error between predicted and measured concentrations of α-HCH and β-HCH in different media, as shown below. Figure 12 As shown in Table 12, the results indicate that the relative error of α-HCH in all media is less than 30%; except for air β-HCH, whose relative error is 44.1%, the relative error of β-HCH in other media is less than 30%. These results demonstrate that the method of this invention has reliable prediction results. The posterior distribution of emission sources for α-HCH and β-HCH is shown in Table 12. Figure 13 As shown; the posterior distribution of water input is as follows Figure 14 As shown; the posterior distribution of advection input concentrations in water and air is as follows. Figure 15 and Figure 16 As shown; the posterior distribution of sediment and soil background concentrations is as follows. Figure 17 and Figure 18 As shown.

[0062] A regional water pollutant source apportionment system, comprising: The data acquisition module is used to execute step S1; the index calculation module is used to execute step S2; the scene classification module is used to execute step S3; the equation construction module is used to execute step S4; and the inversion module is used to execute step S5.

[0063] The content of the above method embodiments is applicable to this system embodiment. The specific functions implemented in this system embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0064] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.

Claims

1. A method for source apportionment of regional water body pollutants, characterized in that, The method comprises the following steps: acquiring multi-dimensional data; calculating a scenario index according to the multi-dimensional data; making a situation judgment according to the scenario index to obtain a scenario classification result; constructing a mass balance equation of a four-medium advection-diffusion model; performing Bayesian inversion on the mass balance equation of the four-medium advection-diffusion model in combination with the scenario classification result to obtain an analysis result.

2. The method according to claim 1, wherein, The scenario index comprises: a comprehensive index of pollutant migration and persistence: wherein, represents the contaminant migration-persistence integrated index, represents the octanol-water partition coefficient, represents the half-life of the contaminant in the water medium, represents the half-life of the contaminant in the soil medium, represents the half-life of the contaminant in the air medium; a system heterogeneity index. wherein, represents a system heterogeneity index, represents a volume standard deviation, represents a volume of water medium, represents a volume of sediment medium, represents a volume of soil medium, represents a volume of air medium, represents an air-land-soil contact area represents a volume mean, represents an air-water contact area.

3. The method according to claim 2, wherein, The mass balance equation of the four-medium advection-diffusion model is expressed in a matrix form and is specifically as follows: wherein, is a migration matrix, denotes a diagonal matrix, E is a source vector of emissions, f is a fugacity vector, is an external input vector.

4. The method of claim 1, wherein, The step of performing Bayesian inversion on the mass balance equation of the four-medium advection-diffusion model in combination with the scenario classification result specifically comprises: when the comprehensive index of pollutant migration and persistence is greater than a first threshold value and the system heterogeneity index is less than a second threshold value, enabling Bayesian inversion in a full parameter space; when the comprehensive index of pollutant migration and persistence is greater than the first threshold value and the system heterogeneity index is greater than the second threshold value, fixing a migration flux of a migration process as a priori expectation value and performing inversion only on a key source and the migration process; when the comprehensive index of pollutant migration and persistence is less than a third threshold value, performing inversion on source strengths of water body direct discharge and adjacent media by using a first grade of priori constraint and performing inversion on a remote source by using a second grade of priori constraint.

5. The method of claim 1, wherein, The step of performing Bayesian inversion on the mass balance equation of the four-medium advection-diffusion model specifically comprises: defining a migration matrix in the mass balance equation of the four-medium advection-diffusion model; defining a parameter space; taking source strengths, input fluxes and concentrations of each medium in the mass balance equation as unknown parameters and assigning a priori distribution to the unknown parameters; establishing a likelihood function of observation concentration by taking the mass balance equation as a constraint to obtain a parameter posterior distribution; obtaining posterior samples by random sampling and solving the mass balance equation by using each sample to obtain posterior statistical results of advection and flux; analyzing water pollution sources according to the posterior statistical results of the flux.

6. The method according to claim 5, wherein, The method further comprises: verifying and quantifying uncertainty of the analysis result.

7. The method according to claim 5, wherein, An expression of the parameter space is as follows: wherein, represents the emission source strength for a medium being water phase; represents the emission source strength for a medium being sediment; represents the emission source strength for a medium being terrestrial soil; represents the emission source strength for a medium being ambient air; represents the water body non-advective input; represents the water body advective input concentration; represents the air advective input concentration; represents the sediment background concentration; represents the terrestrial soil background concentration.

8. The method of claim 5, wherein, An expression of flux calculation on each posterior sample is as follows: For each posterior sample i: wherein, represents the system matrix corresponding to the posterior sample, which is composed of a fixed transfer matrix and a diagonal matrix ; represents the emission source strength vector in the posterior sample; represents the external input vector corresponding to the posterior sample; represents the non-advection input flux into the water body in the posterior sample; represents the advection input flux into the water body through advection in the posterior sample; represents the advection input flux into the air through advection in the posterior sample; represents the divergence vector obtained by solving the linear equation set in the posterior sample; represents a numerical operation for solving the linear equation set; for each source medium src and target medium tgt: in, Indicates the first In each posterior sample, from the source medium to target medium The magnitude of the net migration flux of pollutants; Represents the migration matrix The middle is located in the first line, number Column elements; Indicates the first In each posterior sample, the source medium Fugacity value; Indicates the first In the posterior sample, the target medium The fugacity value.

9. The method according to claim 8, wherein, The method further comprises: introducing a dynamic regularization algorithm for regularization processing before solving a linear equation.

10. A regional water body contaminant source resolution system, comprising: The method comprises: a data acquisition module configured to acquire multi-dimensional data; an index calculation module configured to calculate a scenario index according to the multi-dimensional data; a scenario classification module configured to make a situation judgment according to the scenario index to obtain a scenario classification result; an equation construction module configured to construct a mass balance equation of a four-medium advection-diffusion model; an inversion module configured to perform Bayesian inversion on the mass balance equation of the four-medium advection-diffusion model in combination with the scenario classification result to obtain an analysis result.

Images