A method and system for evaluating the diffusion of heavy metal pollution in a plot

Abstract

The application discloses a kind of land heavy metal pollution diffusion evaluation method and system, it is related to heavy metal pollution evaluation technical field.A kind of land heavy metal pollution diffusion evaluation system, including have: data acquisition module, quality calibration module, model construction module, pollution source inversion module, parallel simulation module, credibility field construction module, risk assessment module, dynamic updating module, visual output module and database module.The present application is through data quality calibration, pollution source probability inversion, multiple diffusion path generation, credibility field construction, credibility field and concentration field coupling evaluation and dynamic updating mechanism, constructs from data acquisition, quality evaluation, uncertainty modeling, credibility quantification to risk classification evaluation, dynamic updating complete technical chain, realizes to consider data quality fully, complete characterization diffusion process uncertainty, quantification evaluation result credibility and support dynamic updating.

Description

Technical Field

[0001] This invention relates to the field of heavy metal pollution assessment technology, and in particular to a method and system for assessing the spread of heavy metal pollution in a land plot. Background Technology

[0002] Heavy metal pollution is a prominent problem facing soil and groundwater environments, posing a potential threat to the ecological environment and human health. Conducting heavy metal pollution diffusion assessments of sites is of great significance for pollution risk management and remediation plan design.

[0003] Traditional methods for assessing the spread of heavy metal pollution mainly rely on field monitoring and numerical simulation. Field monitoring methods are limited by sampling costs and conditions, resulting in limited spatial data coverage and discontinuous time series. While numerical simulation methods can predict pollution spread trends, model parameters are often difficult to obtain accurately, leading to significant uncertainties. In recent years, methods combining monitoring data with numerical simulation have emerged, such as optimizing model parameters through data assimilation techniques or using Bayesian inversion to perform probabilistic estimation of pollution sources.

[0004] However, existing technologies still have the following shortcomings: First, the differences in the quality of monitoring data are not fully considered, and the reliability of data from different sources varies, affecting the objectivity of the assessment results; second, the uncertainty of the pollution diffusion process is not fully characterized, making it difficult to generate multiple candidate paths that reflect the uncertainty of diffusion; third, the risk assessment results lack credible quantitative indicators, and risk classification based solely on concentration may lead to decision-making bias; and fourth, there is a lack of an effective dynamic update mechanism, making it difficult to integrate new data into existing assessment results.

[0005] To address the aforementioned technical issues, there is an urgent need in this field for a method and system for assessing the diffusion of heavy metal pollution in land parcels that can fully consider data quality, comprehensively characterize the uncertainty of the diffusion process, quantify the reliability of assessment results, and support dynamic updates. Summary of the Invention

[0006] The core technical problem solved by this invention is that existing methods for assessing the diffusion of heavy metal pollution in land parcels lack a quantitative characterization of the reliability of the assessment results, leading to incomplete decision-making basis. To achieve the above objective, the technical solution adopted by this invention is as follows:

[0007] A method for assessing the diffusion of heavy metal pollution in a land plot includes the following steps:

[0008] Step S1: For the target site, collect multi-source heavy metal monitoring data covering different spatial locations, depth levels, and time scales. The multi-source heavy metal monitoring data includes historical survey data, field sampling data, online monitoring data, and remote sensing interpretation data. Perform quality calibration on the multi-source heavy metal monitoring data. Calculate the data quality weight corresponding to each monitoring data based on one or more quality indicators among the monitoring data's collection time, collection method, sample type, storage conditions, detection method, detection accuracy, or data integrity.

[0009] Step S2: Based on the quality-calibrated multi-source heavy metal monitoring data, construct a plot heavy metal migration and transformation model coupled with soil physicochemical properties; use Bayesian inversion method or ensemble Kalman filter method, combined with the data quality weight, to dynamically invert one or more parameters of the pollution source location, intensity, release time or release mode, construct a probability distribution model of pollution source intensity, and obtain the posterior probability distribution of pollution source parameters.

[0010] Step S3: Under the premise of considering the physicochemical constraints of soil texture, groundwater level, hydraulic gradient, soil porosity, permeability coefficient, adsorption coefficient and degradation coefficient, random sampling is performed from the probability distribution model of the pollution source intensity to generate multiple sets of pollution source parameter samples; each set of pollution source parameter samples is used as input to drive the plot heavy metal migration and transformation model to perform forward simulation, generating multiple candidate paths for heavy metal pollution diffusion corresponding to different pollution source parameter samples, each candidate path including the spatiotemporal concentration distribution of heavy metals in soil and groundwater;

[0011] Step S4: Combining the data quality weights, the path consistency among multiple diffusion candidate paths, and the path stability of each diffusion candidate path in iterative calculation, a diffusion credibility field is constructed to quantify the reliability of pollution diffusion results. The credibility field is a three-dimensional spatial distribution field corresponding to the spatial location of the target plot, used to characterize the reliability of pollution diffusion simulation results at different spatial locations.

[0012] Step S5: Spatially superimpose and couple the diffusion confidence field output in step S4 with the pollution concentration field output in step S3 to calculate the risk level of each spatial location; the risk level considers both the pollution concentration level and the confidence level of the simulation results, and divides the plot into at least two risk types, such as high-risk-high confidence area, high-risk-low confidence area, low-risk-high confidence area, or low-risk-low confidence area, to achieve a graded assessment of the diffusion risk of heavy metal pollution in the plot;

[0013] Step S6: After acquiring new monitoring data, the new monitoring data is integrated with historical monitoring data. By repeating steps S1 to S5, the pollution diffusion assessment results are dynamically updated and their accuracy is gradually improved.

[0014] As a preferred embodiment of the present invention, the quality calibration in step S1 specifically includes:

[0015] Sub-step S11: Construct a multi-dimensional quality indicator system, which includes time-dimensional indicators, spatial-dimensional indicators, methodological-dimensional indicators, and analytical-dimensional indicators.

[0016] Sub-step S12: For each monitoring data point, calculate its quality score in each dimension.

[0017] Sub-step S13 uses a weighted summation method to merge the quality scores of each dimension into a comprehensive quality score;

[0018] Sub-step S14: Based on the comprehensive quality score, normalization or hierarchical assignment is used to determine the data quality weight of each monitoring data. The data quality weight ranges from 0 to 1, and the higher the data quality, the closer the weight is to 1.

[0019] As a preferred embodiment of the present invention, the step S2 of constructing the probability distribution model of pollution source intensity specifically includes:

[0020] Sub-step S21: Construct a priori distribution of pollution source parameters, the priori distribution being determined based on the historical land use, production process, and pollution discharge records;

[0021] Sub-step S22: Establish the likelihood function between the observation data and the pollution source parameters. The data quality weight is introduced into the likelihood function as an adjustment factor for the observation error. The higher the data quality weight, the greater the contribution of the corresponding observation data to the likelihood function.

[0022] Sub-step S23: Using the Markov chain Monte Carlo method or the sequential data assimilation method, combined with the prior distribution and the likelihood function, solve for the posterior probability distribution of the pollution source parameters.

[0023] Sub-step S24 involves performing statistical feature analysis on the posterior probability distribution to obtain the mean, variance, and quantiles of the pollution source parameters, as well as the correlation information between the parameters.

[0024] As a preferred embodiment of the present invention, step S3, which involves generating multiple candidate diffusion paths for heavy metal pollution corresponding to different pollution source parameter samples, specifically includes:

[0025] Sub-step S31: Randomly sample from the posterior probability distribution of pollution source parameters obtained in step S2 to generate N sets of pollution source parameter samples, where N is the preset number of samples;

[0026] Sub-step S32: Construct a numerical model for the migration and transformation of heavy metals in a plot of land that considers heterogeneous soil media. The numerical model includes convection-dispersion equations, adsorption-desorption equations, and chemical reaction kinetic equations.

[0027] Sub-step S33: Input each set of pollution source parameter samples as boundary conditions into the numerical model, run N-fold numerical simulation in parallel, and obtain N sets of spatiotemporal distribution results of pollution concentration.

[0028] Sub-step S34 involves statistically analyzing the spatiotemporal distribution results of the N groups of pollution concentrations, calculating the mean, median, quantile, and variance of the concentration at each spatial location and time point, thus forming the pollution concentration field and its uncertainty distribution.

[0029] As a preferred embodiment of the present invention, the construction of a diffusion confidence field for quantifying the reliability of pollution diffusion results in step S4 specifically includes:

[0030] Sub-step S41: Based on the data quality weights, a spatial interpolation method or geostatistical method is used to generate a constraint strength field of the monitoring data on the diffusion path. The constraint strength field is used to characterize the ability of the monitoring data to constrain the simulation results of the surrounding area.

[0031] Sub-step S42: Calculate the spatial distribution similarity among the multiple candidate paths generated in step S3 to obtain a diffusion path consistency field; the spatial distribution similarity is quantified by one of the correlation coefficient, mutual information, or Euclidean distance of the concentration values ​​of each candidate path at the same spatial location;

[0032] Sub-step S43 evaluates the convergence of the multiple candidate paths generated in step S3 in the iterative calculation to obtain the path stability field; the convergence is quantified by comparing the differences in simulation results under different sampling numbers or analyzing the trend of the variance of the simulation results with the number of iterations.

[0033] Sub-step S44 involves using weighted fusion, fuzzy comprehensive evaluation, or machine learning methods to fuse the constraint strength field, diffusion path consistency field, and path stability field through multi-source information fusion to construct the diffusion credibility field.

[0034] Sub-step S45 involves post-processing the diffusion confidence field, including spatial smoothing, threshold segmentation, or confidence interval calibration, to obtain the final confidence spatial distribution.

[0035] As a preferred embodiment of the present invention, the risk classification assessment in step S5 specifically includes:

[0036] Sub-step S51: Spatially align the pollution concentration field and the diffusion confidence field to obtain the concentration value C(x,y,z) and confidence value R(x,y,z) at each geographical location;

[0037] Sub-step S52: Set the concentration threshold C_th and the confidence threshold R_th, define the position where the concentration value is higher than C_th as the high concentration region, and define the position where the confidence value is higher than R_th as the high confidence region;

[0038] Sub-step S53: Based on the concentration threshold and confidence threshold, the land parcel is divided into four risk types: high-risk-high confidence area, high-risk-low confidence area, low-risk-high confidence area and low-risk-low confidence area;

[0039] Sub-step S54: For high-risk-low confidence areas, output supplementary monitoring recommendations; for high-risk-high confidence areas, output priority control recommendations; for low-risk-low confidence areas, output continuous monitoring recommendations; for low-risk-high confidence areas, output recommendations to reduce monitoring frequency.

[0040] Sub-step S55 generates a risk distribution map of heavy metal pollution diffusion in the site. The risk distribution map uses different colors or patterns to mark areas with different risk types, and includes a description of the risk level and control recommendations for each area.

[0041] As a preferred embodiment of the present invention, the dynamic update in step S6 specifically includes:

[0042] Sub-step S61: Obtain newly added monitoring data and perform quality calibration on the newly added monitoring data to obtain the data quality weight of the newly added data;

[0043] Sub-step S62: Merge the newly added monitoring data with the historical monitoring data to form an updated monitoring dataset;

[0044] Sub-step S63: Based on the updated monitoring dataset, the posterior probability distribution of pollution source parameters obtained in the previous round of assessment is used as the initial prior distribution for this round of inversion. Steps S2 to S5 are executed again to obtain the updated pollution source probability distribution, updated diffusion candidate paths, updated confidence field, and updated risk classification assessment results.

[0045] Sub-step S64: Evaluate the magnitude of change and convergence trend of the evaluation results before and after the update. When the magnitude of change of the evaluation results in multiple consecutive updates is less than a preset threshold, it is determined that the evaluation results have reached a stable state.

[0046] Sub-step S65 records and outputs the evolution trajectory of the assessment results during the dynamic update process, which is used to analyze the spatiotemporal evolution of heavy metal pollution in the plot.

[0047] A land plot heavy metal pollution diffusion assessment system, comprising:

[0048] The data acquisition module is used to collect multi-source heavy metal monitoring data of the target site, including historical survey data interface, field sampling data interface, online monitoring data interface and remote sensing interpretation data interface;

[0049] The quality calibration module, connected to the data acquisition module, is used to calibrate the quality of the acquired multi-source heavy metal monitoring data and calculate the data quality weight corresponding to each monitoring data based on a multi-dimensional quality index system.

[0050] The model building module is used to construct a plot heavy metal migration and transformation model that couples soil physicochemical properties. The model includes a convection-diffusion sub-model, an adsorption-desorption sub-model, and a chemical reaction kinetics sub-model.

[0051] The pollution source inversion module, connected to the quality calibration module and the model building module, is used to construct a probability distribution model of pollution source intensity by using the Bayesian inversion method or the ensemble Kalman filter method, combined with the data quality weights.

[0052] The parallel simulation module, connected to the pollution source inversion module and the model building module, is used to sample and generate multiple sets of pollution source parameter samples from the pollution source probability distribution model, and drive the heavy metal migration and transformation model of the plot in parallel to perform forward simulation, generating multiple candidate paths for heavy metal pollution diffusion.

[0053] The credibility field construction module, connected to the quality calibration module and the parallel simulation module, is used to construct a credibility field for heavy metal pollution diffusion by integrating the data quality weights, the path consistency among multiple diffusion candidate paths, and the path stability of each diffusion candidate path.

[0054] The risk assessment module, connected to the parallel simulation module and the confidence field construction module, is used to spatially couple the diffusion confidence field with the pollution concentration field, calculate the risk level of each spatial location, and generate a distribution map of heavy metal pollution diffusion risk of the plot.

[0055] The dynamic update module, connected to the data acquisition module, is used to trigger each module to repeatedly execute its function after new monitoring data is acquired, so as to realize the dynamic update of the pollution diffusion assessment results.

[0056] The visualization output module, connected to the risk assessment module and the dynamic update module, is used to output the pollution concentration field, confidence field, risk classification results, and dynamic update trajectory in a graphical manner.

[0057] The database module is used to store historical monitoring data, model parameters, intermediate calculation results, and final evaluation results.

[0058] As a preferred embodiment of the present invention, it further includes:

[0059] The uncertainty analysis module, connected to the parallel simulation module, is used to perform statistical analysis on the multiple diffusion candidate paths and calculate the probability distribution, confidence interval, and probability of exceeding the standard of pollution concentration at each spatial location.

[0060] The sensitivity analysis module, connected to the model building module, is used to analyze the sensitivity of model parameters to simulation results and identify key parameters affecting pollution diffusion.

[0061] The scheme optimization module, connected to the risk assessment module, is used to generate optimized monitoring site layout schemes and pollution control schemes based on the risk classification assessment results.

[0062] The report generation module is used to automatically generate a heavy metal pollution diffusion assessment report for a site, including an explanation of the assessment method, a description of the data source, model parameter settings, assessment result charts, and control recommendations.

[0063] The present invention has the following advantages:

[0064] This invention calibrates the quality of multi-source heavy metal monitoring data and calculates the data quality weight corresponding to each monitoring data based on multi-dimensional quality indicators such as acquisition time, acquisition method, and detection accuracy. This quantitatively introduces the reliability of the data into the evaluation process. This invention can objectively reflect the quality differences of data from different sources, avoid the interference of low-quality data on the evaluation results, provide a more reliable input basis for subsequent inversion and simulation, and significantly improve the objectivity of the evaluation results.

[0065] This invention constructs a probability distribution model of pollution source intensity by employing Bayesian inversion or ensemble Kalman filtering methods, combined with data quality weights, to obtain the posterior probability distribution of pollution source parameters. This invention can fully characterize the uncertainties of parameters such as the location, intensity, and release time of pollution sources, providing a scientific basis for the probability prediction of diffusion processes, and enabling the assessment results to reflect the inherent uncertainties in the pollution source inversion process.

[0066] This invention generates multiple sets of pollution source parameter samples by randomly sampling from a pollution source probability distribution model, drives a migration and transformation model to perform forward simulation, and obtains multiple candidate paths for the diffusion of heavy metal pollution. This invention intuitively reflects the uncertainty of the diffusion process through multiple candidate paths, and constructs a pollution concentration field and its uncertainty distribution containing information such as concentration mean, quantiles, and variance through statistical analysis of multiple sets of simulation results.

[0067] This invention constructs a diffusion credibility field for quantifying the reliability of pollution diffusion results by comprehensively considering data quality weights, path consistency among multiple diffusion candidate paths, and path stability of each diffusion candidate path. This credibility field is a three-dimensional spatial distribution field, which can intuitively characterize the reliability of pollution diffusion simulation results at different spatial locations, thus solving the long-standing problem of difficulty in quantifying the "credibility" of evaluation results.

[0068] This invention spatially superimposes and couples the diffusion confidence field with the pollution concentration field, while considering both the level of pollution concentration and the confidence level of the simulation results, dividing the land parcel into various risk types such as high-risk-high confidence area, high-risk-low confidence area, low-risk-high confidence area, and low-risk-low confidence area. This invention outputs differentiated monitoring or control recommendations for different risk types, providing a more scientific and accurate basis for risk control decisions.

[0069] This invention establishes a dynamic update mechanism. After acquiring new monitoring data, the new data is integrated with historical data, and the assessment process is re-executed based on the updated dataset. The posterior probability distribution of pollution source parameters obtained in the previous round of assessment is used as the initial prior distribution for this round of inversion, thus achieving effective knowledge transfer. Through convergence trend analysis of multiple consecutive rounds of updates, it can be determined whether the assessment results have reached a stable state. By recording the evolution trajectory of the assessment results, data support is provided for analyzing the spatiotemporal evolution of heavy metal pollution in land parcels, enabling the assessment system to have the ability to continuously learn and improve accuracy. Attached Figure Description

[0070] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are only schematic diagrams of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0071] Figure 1 This is a schematic diagram of a land plot heavy metal pollution diffusion assessment system used in an embodiment of the present invention. Detailed Implementation

[0072] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this invention, and not all embodiments. 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.

[0073] Example 1: A method for assessing the diffusion of heavy metal pollution in a land plot, comprising the following steps:

[0074] Step S1: For the target site, collect multi-source heavy metal monitoring data covering different spatial locations, depth levels, and time scales. The multi-source heavy metal monitoring data includes historical survey data, field sampling data, online monitoring data, and remote sensing interpretation data. Perform quality calibration on the multi-source heavy metal monitoring data. Calculate the data quality weight corresponding to each monitoring data based on one or more quality indicators among the monitoring data's collection time, collection method, sample type, storage conditions, detection method, detection accuracy, or data integrity.

[0075] Multi-source heavy metal monitoring data are used to characterize the distribution of heavy metal content in different historical periods and spatial locations of land parcels. They serve as the basic data set for constructing migration and transformation models and retrieving pollution sources.

[0076] In a complete testing case, the target site was the former site of an abandoned chemical plant that had previously engaged in electroplating production, posing a risk of contamination with heavy metals such as lead and cadmium. To comprehensively understand the site's contamination status, multi-source data were collected using the following methods:

[0077] The historical survey data was obtained by collecting soil environmental survey data from the past three surveys of the site, including the survey reports from 2005, 2010, and 2015. Heavy metal concentration data from 120 historical sampling points were extracted, along with corresponding borehole columnar sections and hydrogeological parameter test records. Each historical data record includes information such as sampling point coordinates, sampling depth, sampling time, analysis method, and test results.

[0078] The on-site sampling data was obtained as follows: supplementary sampling was organized at the current time, with 50 sampling points set up at a grid density of 40m × 40m using a grid method. At each point, soil samples were collected from the surface (0-0.5m), middle (0.5-2.0m), and deep (2.0-4.0m) layers. Simultaneously, 10 monitoring wells were set up to collect groundwater samples. A total of 160 soil samples and 10 water samples were obtained. Lead and cadmium content were analyzed according to the national standard method (GB / T17141-1997), and information such as coordinates, depth, sampling time, preservation conditions, detection method, detection limit, and precision for each sample was recorded.

[0079] The online monitoring data was acquired by deploying four sets of soil heavy metal online monitoring probes at the downstream boundary of the site and near suspected pollution sources. These probes employed ion-selective electrode technology to monitor the concentrations of lead and cadmium in the soil solution in real time, with a sampling frequency of once per hour. The data was transmitted to the data center in real time via a wireless communication module. Each online monitoring data point included a timestamp, concentration value, and temperature compensation value.

[0080] The remote sensing interpretation data was acquired by obtaining high-resolution satellite imagery (WorldView-3, 0.3-meter resolution) of the site over the past 5 years. Potential distribution areas of heavy metal pollution in the soil were then retrieved using vegetation stress indices (such as the Normalized Difference Vegetation Index, NDVI), generating a vector layer of the suspected pollution area. Each pixel corresponds to a retrieved concentration value and a retrieved error.

[0081] After data collection is completed, the quality of the collected multi-source heavy metal monitoring data is calibrated. Based on one or more quality indicators of the monitoring data, such as collection time, collection method, sample type, storage conditions, detection method, detection accuracy or data integrity, the data quality weight corresponding to each monitoring data is calculated.

[0082] Quality calibration specifically includes the following sub-steps:

[0083] Sub-step S11: Construct a multi-dimensional quality indicator system. This system includes time-dimensional indicators, spatial-dimensional indicators, methodological-dimensional indicators, and analytical-dimensional indicators. Time-dimensional indicators include the time elapsed since sampling and the timeliness level; spatial-dimensional indicators include positioning accuracy (e.g., whether differential GPS was used) and the completeness of the sampling depth record; methodological-dimensional indicators include the standardization of the sampling method (whether it complies with national standards) and sample preservation conditions (whether it is stored at low temperatures and analyzed within a specified time); analytical-dimensional indicators include the standardization of the detection method (whether it is a national standard method), the accuracy of the detection instrument, the detection limit, and the spiked recovery rate.

[0084] Sub-step S12: For each monitoring data point, calculate its quality score in each dimension. For example, if the sampling date is 18 years ago, the time score is set to 0.3; if handheld GPS positioning is used with a coordinate accuracy of about 10 meters, the spatial score is set to 0.8; if the sampling method was a commonly used method at the time but no longer meets current standards, the method score is set to 0.6; if the detection method is a national standard method but has average precision, the analysis score is set to 0.7.

[0085] Sub-step S13 uses a weighted summation method to merge the quality scores of each dimension into a comprehensive quality score.

[0086] The preset weights for each dimension are: time dimension 0.3, space dimension 0.2, method dimension 0.2, and analysis dimension 0.3. Therefore, the overall quality score for this historical data is: 0.3×0.3+0.2×0.8+0.2×0.6+0.3×0.7=0.09+0.16+0.12+0.21=0.58.

[0087] Sub-step S14 involves determining the data quality weight of each monitoring data point based on the comprehensive quality score using a normalization process. Since the comprehensive quality score is already in the range of 0 to 1, it can be directly used as the data quality weight; that is, the data quality weight of this historical data is 0.58. Other data are calculated using the same method. For example, field sampling data, due to its recent sampling time and standardized sampling method, can achieve a comprehensive quality score of 0.95, and its data quality weight is 0.95. The data quality weights of all monitoring data are stored in the database for later use.

[0088] Step S2: Based on the quality-calibrated multi-source heavy metal monitoring data, construct a plot heavy metal migration and transformation model coupled with soil physicochemical properties; using Bayesian inversion method or ensemble Kalman filter method, combined with the data quality weights, dynamically invert one or more parameters of pollution source location, intensity, release time or release mode, construct a probability distribution model of pollution source intensity, and obtain the posterior probability distribution of pollution source parameters.

[0089] After data quality calibration, a heavy metal migration and transformation model for the site was constructed based on the calibrated data. This model is a numerical model coupled with soil physicochemical properties. A two-dimensional profile model was established using HYDRUS-2D software, considering water movement and solute transport in the unsaturated zone, and coupling the adsorption-desorption processes of lead and cadmium (using Freundlich isotherms). Model parameters include soil texture, porosity, permeability coefficient, dispersion, and partition coefficient, which were initially set based on field geotechnical tests and empirical values ​​from literature.

[0090] A Bayesian inversion method, combined with data quality weights, was used to dynamically invert pollution source parameters. These parameters included the pollution source location coordinates (x, z), release intensity (mg / L), release onset time, and release duration.

[0091] Constructing a probability distribution model of pollution source intensity specifically includes the following sub-steps:

[0092] Sub-step S21: Construct the prior distribution of pollution source parameters. This prior distribution is determined based on the site's historical land use, production processes, and discharge records. According to historical data, the site was formerly an electroplating workshop, and the pollution source is likely located within the workshop area. Therefore, the prior distribution of the source location coordinates (x, z) is set to uniform, with the x-direction ranging from 50 to 150 meters east-west along the workshop area, and the z-direction ranging from the surface to 2 meters underground. The prior distribution of release intensity is set to a log-normal distribution with a mean of 100 mg / L and a standard deviation of 50 mg / L. The prior distribution of release onset time is set to uniform, ranging from 1980 to 2000. The prior distribution of release duration is set to uniform, ranging from 1 to 20 years.

[0093] Sub-step S22 establishes the likelihood function between the observed data and the pollution source parameters. The observed data consists of heavy metal concentrations from all historical and new sampling points, totaling M data points. The data quality weight is introduced into the likelihood function as an adjustment factor for the observation error; the higher the data quality weight, the greater the contribution of the corresponding observed data to the likelihood function.

[0094] Specifically, let the model prediction value f(θ) be the simulated concentration at each monitoring point given the pollution source parameter θ, and the observation vector be... The observation error ε=yf(θ) follows a multivariate normal distribution, and the covariance matrix Σ is a diagonal matrix. diagonal elements ,in Let the basic error variance be (10 mg / kg)^2. Let Variance be the observation error variance corresponding to the i-th monitoring data. Let be the data quality weight for the i-th monitoring data. The likelihood function is: .

[0095] Sub-step S23 employs the Markov chain Monte Carlo method, combining the prior distribution and likelihood function to solve for the posterior probability distribution of the pollution source parameters. Specifically, the Metropolis-Hastings algorithm is used, running four Markov chains, each iterating 20,000 times, with the first 10,000 iterations discarded as a pre-burning period. In each iteration, candidate parameters are generated from the proposal distribution, the acceptance probability is calculated, and a decision is made on whether to accept them. The final posterior sample set contains approximately 40,000 valid samples.

[0096] Sub-step S24 involves performing statistical feature analysis on the posterior probability distribution to obtain the mean, variance, and quantiles of the pollution source parameters, as well as the correlation information between the parameters.

[0097] The calculated posterior mean of the x-coordinate of the source location is 98 meters, and the 95% confidence interval is [85, 112] meters.

[0098] The posterior mean of the z-coordinate is 1.2 meters, and the 95% confidence interval is [0.8, 1.8] meters;

[0099] The mean release strength was 85 mg / L, and the standard deviation was 12 mg / L.

[0100] The posterior mean of the release start time is 1995, with a 95% confidence interval of [1988, 2000].

[0101] The posterior mean of the release duration is 8 years, with a 95% confidence interval of [5, 12] years.

[0102] Parameter correlation analysis showed that release intensity was negatively correlated with duration.

[0103] Step S3: Under the premise of considering the physicochemical constraints of soil texture, groundwater level, hydraulic gradient, soil porosity, permeability coefficient, adsorption coefficient and degradation coefficient, random sampling is performed from the probability distribution model of the pollution source intensity to generate multiple sets of pollution source parameter samples; each set of pollution source parameter samples is used as input to drive the heavy metal migration and transformation model of the plot to perform forward simulation, generating multiple candidate paths for heavy metal pollution diffusion corresponding to different pollution source parameter samples, each candidate path including the spatiotemporal concentration distribution of heavy metals in soil and groundwater;

[0104] After solving the posterior distribution of pollution source parameters, multiple sets of parameter samples are randomly sampled from the distribution to drive the model to perform forward simulation and generate multiple candidate pollution diffusion paths.

[0105] Generating multiple candidate diffusion pathways for heavy metal pollution corresponding to different pollution source parameter samples specifically includes the following sub-steps:

[0106] Sub-step S31 involves randomly sampling from the posterior probability distribution of the pollution source parameters obtained in step S2 to generate N sets of pollution source parameter samples. N is preset to 500 sets. 500 parameter vectors are randomly selected (with replacement) from the posterior sample set. Each set includes the source location, release intensity, release start time, and release duration.

[0107] Sub-step S32 involves constructing a numerical model for the migration and transformation of heavy metals in a plot of land, considering heterogeneous soil media. The numerical model includes convection-dispersion equations and adsorption-desorption equations.

[0108] The model was created in HYDRUS-2D, with the plot measuring 200 meters horizontally and 10 meters vertically, and divided into 100×50 grids.

[0109] The soil is divided into three layers: the top layer (0-1 meters) is fill soil with a permeability coefficient of 1.0 m / day; the middle layer (1-5 meters) is silty clay with a permeability coefficient of 0.1 m / day; and the deep layer (5-10 meters) is sandy soil with a permeability coefficient of 5.0 m / day.

[0110] The adsorption parameters were determined using the Freundlich equation, with the partition coefficient Kd for lead set at 0.5 L / kg and Kd for cadmium set at 0.2 L / kg.

[0111] In sub-step S33, each group of pollution source parameter samples is input into the numerical model as boundary conditions, and N numerical simulations are run in parallel to obtain N groups of spatiotemporal distribution results of pollution concentration.

[0112] 500 simulation tasks are run in parallel on the computing cluster using MPI. Each task simulates the period from 1940 (assuming the start of industrial activity) to 2023, outputting the concentration distribution at the end of each year. Each task outputs the concentration values ​​on a three-dimensional grid (x,z,t).

[0113] Sub-step S34 involves statistically analyzing the N sets of spatiotemporal distribution results of pollution concentrations, calculating the mean, median, quantiles, and variance of concentration at each spatial location and time point, thus constructing the pollution concentration field and its uncertainty distribution. For each spatial grid point, the mean, median, 5th percentile, 95th percentile, and variance of 500 simulation results are calculated.

[0114] For example, at a depth of 1.5 meters and 20 meters below the source region, the mean lead concentration reaches 120 mg / kg, with a 95% confidence interval of [80, 180] mg / kg. The mean values ​​of all grid points are combined to form a pollution concentration field, and the quantile range is used as a representation of uncertainty.

[0115] Step S4: Combining the data quality weights, path consistency among multiple diffusion candidate paths, and path stability of each diffusion candidate path in iterative calculations, a diffusion credibility field is constructed to quantify the reliability of pollution diffusion results. The credibility field is a three-dimensional spatial distribution field corresponding to the spatial location of the target site, used to characterize the reliability of pollution diffusion simulation results at different spatial locations.

[0116] After obtaining multiple candidate paths, a diffusion credibility field is constructed by combining data quality weights, path consistency, and path stability.

[0117] Constructing a diffusion confidence field for quantifying the reliability of pollution diffusion outcomes includes the following sub-steps:

[0118] Sub-step S41: Based on the data quality weights, a spatial interpolation method is used to generate a constraint strength field of the monitoring data on the diffusion path.

[0119] Ordinary Kriging interpolation is employed, using the data quality weights of each monitoring point as the main variables for three-dimensional spatial interpolation. First, the empirical variogram is calculated, and a theoretical variogram model is fitted. Then, Kriging estimation is performed on each unmonitored grid point to obtain the constraint strength value. Monitoring points with higher weights have higher constraint strength in their surrounding areas.

[0120] Sub-step S42 calculates the spatial distribution similarity among the multiple candidate paths generated in step S3 to obtain a diffusion path consistency field. For each spatial grid point, the coefficient of variation (CV) = standard deviation / mean of the concentration values ​​of 500 candidate paths is calculated. The smaller the coefficient of variation, the more consistent the simulation results of each candidate path at that point. A path consistency index is defined. , The path consistency index represents the spatial location (x, y, z) over time t, and its value range is mapped to 0~1. Let be the standard deviation of the concentration values ​​of N candidate paths at this location; ε is the average concentration value of N candidate paths at this location; ε is a very small positive number used to avoid division by zero error caused by a mean of zero; C_consistency = 1 when CV is 0, and the larger the CV, the smaller the C_consistency. The diffusion path consistency field is calculated for all grid points.

[0121] Sub-step S43: Evaluate the convergence of the multiple candidate paths generated in step S3 in the iterative calculation to obtain the path stability field.

[0122] The 500 candidate paths were randomly divided into two groups of 250 each, and the mean concentrations μ1 and μ2 of the two groups at each spatial point were calculated.

[0123] Define relative differences ε is set to 0.01 mg / kg to prevent division by zero.

[0124] Define stability index The smaller d is, the closer S_stability is to 1. The path stability field is calculated for all grid points.

[0125] Sub-step S44: Using a weighted fusion method, the constraint strength field, diffusion path consistency field, and path stability field are fused from multiple sources to construct the diffusion credibility field.

[0126] The weights are set as γ1=0.3, γ2=0.4, and γ3=0.3, and the confidence value of each grid point is calculated, with the value ranging from 0 to 1.

[0127] Sub-step S45 involves post-processing the diffusion confidence field, including spatial smoothing and threshold segmentation. A Gaussian filter is used for spatial smoothing to eliminate local outliers. Then, threshold segmentation is performed, dividing the confidence values ​​into three levels: high confidence region R ≥ 0.7, medium confidence region 0.4 ≤ R < 0.7, and low confidence region R < 0.4. For example, in areas with dense monitoring points and high data quality, the confidence level can reach 0.9; in areas with no monitoring points and discrete simulation results, the confidence level is only 0.3.

[0128] Step S5: Spatially superimpose and couple the diffusion confidence field output in Step S4 with the pollution concentration field output in Step S3 to calculate the risk level of each spatial location; the risk level considers both the pollution concentration level and the confidence level of the simulation results, and divides the plot into at least two risk types, such as high-risk-high confidence area, high-risk-low confidence area, low-risk-high confidence area, or low-risk-low confidence area, to achieve a graded assessment of the diffusion risk of heavy metal pollution in the plot;

[0129] By spatially superimposing the diffusion confidence field and the pollution concentration field, risk classification assessment can be achieved.

[0130] The risk grading assessment specifically includes the following sub-steps:

[0131] Sub-step S51 involves spatially aligning the pollution concentration field and the diffusion confidence field to obtain the concentration and confidence values ​​at each geographical location. The 95th percentile of the 2023 pollution concentration field is taken as a conservative estimate of the concentration value, and it is mapped to the confidence field on the same spatial grid.

[0132] Sub-step S52 sets the concentration threshold C_th and the confidence threshold R_th. According to the "Soil Environmental Quality Standard for Construction Land Soil Pollution Risk Control" GB36600-2018, the screening value for lead in Class I land use is 400 mg / kg, so C_th is set to 400 mg / kg. The confidence threshold is set to R_th = 0.6 based on engineering experience. Locations with concentration values ​​higher than 400 mg / kg are defined as high-concentration areas, and locations with confidence values ​​higher than 0.6 are defined as high-confidence areas.

[0133] Sub-step S53: Based on the concentration threshold and confidence threshold, the land parcel is divided into four risk types: high-risk-high confidence area (C≥400 and R≥0.6), high-risk-low confidence area (C≥400 and R<0.6), low-risk-high confidence area (C<400 and R≥0.6), and low-risk-low confidence area (C<400 and R<0.6).

[0134] In sub-step S54, for high-risk-low confidence areas, supplementary monitoring recommendations are output, suggesting that sampling points be deployed more densely in the area to reduce uncertainty; for high-risk-high confidence areas, priority control recommendations are output, suggesting measures such as isolation, solidification and stabilization, or excavation and repair are recommended; for low-risk-low confidence areas, continuous monitoring recommendations are output, suggesting that the regular monitoring frequency be maintained to accumulate data; for low-risk-high confidence areas, recommendations to reduce the monitoring frequency are output, suggesting that the monitoring cycle be appropriately extended.

[0135] Sub-step S55 generates a risk distribution map of heavy metal pollution diffusion from the site. A two-dimensional risk distribution map is generated using GIS tools, with red representing high-risk to high-confidence areas, orange representing high-risk to low-confidence areas, yellow representing low-risk to low-confidence areas, and green representing low-risk to high-confidence areas. The map includes a legend, area statistics for each area, risk level descriptions, and corresponding control recommendations. For example, the high-risk to high-confidence area is mainly located near the former electroplating workshop and within 20 meters downstream, covering an area of ​​approximately 2000 square meters; the high-risk to low-confidence area is distributed in the lateral diffusion zone of the source area, covering an area of ​​approximately 800 square meters.

[0136] Step S6: After acquiring new monitoring data, merge the new monitoring data with historical monitoring data. By repeating steps S1 to S5, the pollution diffusion assessment results are dynamically updated and the accuracy is gradually improved.

[0137] After acquiring new data in subsequent monitoring, the evaluation results are continuously optimized through a dynamic update process.

[0138] Dynamic updates specifically include the following sub-steps:

[0139] Sub-step S61: Acquire newly added monitoring data and perform quality calibration on the newly added monitoring data to obtain the data quality weight of the newly added data. Three months later, acquire newly added online monitoring data (hourly data for 90 consecutive days) and one supplementary sampling data (20 new sampling points). Perform quality calibration on the newly added data according to the method in step S1 and calculate the data quality weight of each newly added data point.

[0140] Sub-step S62 merges the newly added monitoring data with the historical monitoring data to form an updated monitoring dataset. The newly added online monitoring data is averaged daily as new data points and added to the original monitoring dataset along with the supplementary sampling data to form an updated dataset containing more data points.

[0141] Sub-step S63, based on the updated monitoring dataset, uses the posterior probability distribution of pollution source parameters obtained in the previous round of assessment as the initial prior distribution for this round of inversion, and re-executes steps S2 to S5 to obtain the updated pollution source probability distribution, updated diffusion candidate paths, updated confidence field, and updated risk classification assessment results. When rerunning the MCMC inversion, using the posterior mean from the previous round as the initial value can accelerate convergence. The updated pollution source parameters have reduced uncertainty, and the standard deviation of release intensity has decreased to 8 mg / L.

[0142] Sub-step S64 assesses the magnitude and convergence trend of the changes in the assessment results before and after the update. The root mean square error (RMSE) of the pollution concentration field before and after the update is calculated; the RMSE for this update is 3.2%, which is less than the preset threshold of 5%. The area change of the high-risk, low-confidence region is calculated, shrinking from 800 square meters to 300 square meters. When the magnitude of change in the assessment results over multiple consecutive updates is less than the preset threshold, the assessment results are considered to have reached a stable state.

[0143] Sub-step S65 records and outputs the evolution trajectory of the assessment results during the dynamic update process. The average pollution source parameters, high-risk area area, and confidence field distribution of each update are recorded chronologically and plotted as change curves. For example, the area of ​​the high-risk area gradually converges over time, reflecting a gradual improvement in assessment accuracy. These evolution trajectories can be used to analyze the spatiotemporal evolution of heavy metal pollution in land parcels, providing a basis for adjusting long-term monitoring and control strategies.

[0144] Example 2, a land plot heavy metal pollution diffusion assessment system, see [link to example]. Figure 1 As shown, it includes the following modules:

[0145] Data Acquisition Module: Used to collect multi-source heavy metal monitoring data for the target site, including interfaces for historical survey data, field sampling data, online monitoring data, and remote sensing interpretation data. The historical survey data interface supports importing historical databases in formats such as Excel and Access; the field sampling data interface provides a data entry interface, supporting manual entry or batch import of sampling data; the online monitoring data interface receives monitoring probe data in real time via the MQTT protocol; the remote sensing interpretation data interface supports importing remote sensing images and interpretation results in formats such as GeoTIFF and Shapefile. Each interface automatically extracts metadata from the data, including sampling time, coordinates, depth, and analysis method.

[0146] Quality Calibration Module: Connected to the data acquisition module, this module calibrates the quality of acquired multi-source heavy metal monitoring data, calculating the data quality weight for each monitoring data point based on a multi-dimensional quality indicator system. This module includes algorithms for sub-steps S11 to S14, encompassing the definition of the multi-dimensional quality indicator system, a scoring rule library for each dimension, a weighted summation formula, and a weight normalization method. Users can customize the weights and scoring criteria for each dimension through the configuration interface. The module outputs the data quality weight for each monitoring data point and stores it in the database.

[0147] Model Building Module: Used to construct a plot heavy metal migration and transformation model that couples soil physicochemical properties. This module provides a graphical modeling interface, supporting the setting of soil stratification, hydrogeological parameters (porosity, permeability, dispersion, etc.), chemical parameters (adsorption coefficient, degradation coefficient, etc.), as well as boundary conditions and initial conditions. The module has a built-in database of adsorption parameters for common heavy metals, which users can call or customize. The module can generate input files for numerical simulation software such as HYDRUS and FEFLOW for subsequent simulation use.

[0148] The pollution source inversion module, connected to the quality calibration and model building modules, is used to construct a probability distribution model of pollution source intensity using the Bayesian inversion method and the data quality weights. This module implements sub-steps S21 to S24, including a prior distribution setting interface, likelihood function construction (automatically adjusting observation errors using data quality weights), the MCMC sampling algorithm (with built-in Metropolis-Hastings and adaptive Metropolis algorithms), and posterior statistical analysis functions. The module outputs a posterior sample set of pollution source parameters and statistical feature charts.

[0149] The parallel simulation module, connected to the pollution source inversion module and the model building module, is used to sample and generate multiple sets of pollution source parameter samples from the pollution source probability distribution model, and to drive the heavy metal migration and transformation model of the land parcel in parallel to perform forward simulation, generating multiple candidate paths for heavy metal pollution diffusion. This module supports parallel computing on computing clusters or cloud platforms, automatically distributing tasks to multiple computing nodes to implement sub-steps S31 to S34. The module outputs N sets of concentration field data and their statistical results, including the concentration mean field, quantile field, and variance field.

[0150] The credibility field construction module, connected to the quality calibration module and the parallel simulation module, is used to construct a credibility field for heavy metal pollution diffusion by integrating the data quality weights, path consistency among multiple diffusion candidate paths, and path stability of each diffusion candidate path. This module implements sub-steps S41 to S45, including a spatial interpolation algorithm library (Kriging, inverse distance weighting, etc.), coefficient of variation calculation, group comparison algorithm, weighted fusion function, and post-processing functions (smoothing filtering, threshold segmentation). The module outputs three-dimensional credibility field data, supporting visual preview.

[0151] Risk Assessment Module: Connected to the parallel simulation module and the confidence field construction module, this module spatially couples the diffusion confidence field with the pollution concentration field to calculate the risk level at each spatial location and generate a distribution map of heavy metal pollution diffusion risk for the site. This module implements sub-steps S51 to S55, including a threshold setting interface, a regional division algorithm, a rule base for generating control recommendations, and GIS mapping functions. The module outputs a risk distribution map (supporting multiple formats such as Shapefile, GeoTIFF, and PNG) and statistical reports.

[0152] Dynamic Update Module: Connected to the data acquisition module, this module triggers each module to repeatedly execute its function after new monitoring data is acquired, thereby dynamically updating the pollution diffusion assessment results. This module implements sub-steps S61 to S65, including triggering new data quality calibration, dataset merging, automatically starting a new round of assessment, convergence judgment (calculating the magnitude of change and comparing it with a preset threshold), and evolution trajectory recording (storing key indicators from each update into a time-series database).

[0153] Visualization Output Module: Connected to the risk assessment and dynamic update modules, this module graphically outputs pollution concentration fields, confidence fields, risk classification results, and dynamic update trajectories. It supports various formats, including 2D contour maps, 3D stereoscopic renderings, time-series animations, and interactive maps. Users can browse and interact with these formats through a web interface.

[0154] Database module: Used to store historical monitoring data, model parameters, intermediate calculation results (such as posterior samples and candidate paths), and final evaluation results (concentration field, confidence field, risk map). It uses a PostgreSQL database with PostGIS spatial extensions, supporting spatial data querying and version management. All data is stored by timestamp and version number, supporting historical traceability.

[0155] Uncertainty Analysis Module: Connected to the parallel simulation module, this module performs statistical analysis on the multiple candidate diffusion paths, calculating the probability distribution, confidence interval, and probability of exceeding limits for pollution concentrations at each spatial location. It generates a probability distribution map of exceeding limits, displaying the probability that the concentration at each location exceeds the risk screening value.

[0156] Sensitivity Analysis Module: Connected to the model building module, this module analyzes the sensitivity of model parameters to simulation results and identifies key parameters affecting pollution diffusion. It employs the Sobol global sensitivity analysis method to rank parameters such as permeability coefficient and adsorption coefficient by sensitivity, outputting a sensitivity index.

[0157] The scheme optimization module, connected to the risk assessment module, generates optimized monitoring site placement and pollution control schemes based on the risk grading assessment results. Monitoring site placement optimization employs a spatial simulated annealing algorithm to minimize placement costs while maximizing information availability. Control scheme optimization combines multiple objectives such as remediation costs and risk reduction effectiveness, using the NSGA-II algorithm to solve for the Pareto front and output recommended schemes.

[0158] Report Generation Module: This module automatically generates a heavy metal pollution diffusion assessment report for a given site. The report includes an explanation of the assessment methodology, a list of data sources and their weighting statistics, model parameter settings, pollution source inversion results (posterior distribution charts), pollution concentration field and confidence field charts, risk classification maps and regional area statistics, control recommendations, and dynamic update records. The report can be exported in PDF, Word, or HTML format.

[0159] 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 assessing the diffusion of heavy metal pollution in a land plot, characterized in that, Includes the following steps: Step S1: For the target plot, collect multi-source heavy metal monitoring data covering different spatial locations, different depth levels, and different time scales. The multi-source heavy metal monitoring data includes historical survey data, field sampling data, online monitoring data, and remote sensing interpretation data. The quality of the multi-source heavy metal monitoring data is calibrated, and the data quality weight corresponding to each monitoring data is calculated based on one or more quality indicators among the monitoring data acquisition time, acquisition method, sample type, storage conditions, detection method, detection accuracy or data integrity. Step S2: Based on the quality-calibrated multi-source heavy metal monitoring data, construct a plot heavy metal migration and transformation model coupled with soil physicochemical properties; use Bayesian inversion method or ensemble Kalman filter method, combined with the data quality weight, to dynamically invert one or more parameters of the pollution source location, intensity, release time or release mode, construct a probability distribution model of pollution source intensity, and obtain the posterior probability distribution of pollution source parameters. Step S3: Under the premise of considering the physicochemical constraints of soil texture, groundwater level, hydraulic gradient, soil porosity, permeability coefficient, adsorption coefficient and degradation coefficient, random sampling is performed from the probability distribution model of the pollution source intensity to generate multiple sets of pollution source parameter samples; each set of pollution source parameter samples is used as input to drive the plot heavy metal migration and transformation model to perform forward simulation, generating multiple candidate paths for heavy metal pollution diffusion corresponding to different pollution source parameter samples, each candidate path including the spatiotemporal concentration distribution of heavy metals in soil and groundwater; Step S4: Combining the data quality weights, the path consistency among multiple diffusion candidate paths, and the path stability of each diffusion candidate path in iterative calculation, a diffusion credibility field is constructed to quantify the reliability of pollution diffusion results. The credibility field is a three-dimensional spatial distribution field corresponding to the spatial location of the target plot, used to characterize the reliability of pollution diffusion simulation results at different spatial locations. Step S5: Spatially superimpose and couple the diffusion confidence field output in step S4 with the pollution concentration field output in step S3 to calculate the risk level at each spatial location. The risk level considers both the level of pollution concentration and the level of confidence of the simulation results, and divides the plot into at least two risk types, such as high-risk-high confidence area, high-risk-low confidence area, low-risk-high confidence area or low-risk-low confidence area, to achieve a graded assessment of the risk of heavy metal pollution diffusion of the plot. Step S6: After acquiring new monitoring data, the new monitoring data is integrated with historical monitoring data. By repeating steps S1 to S5, the pollution diffusion assessment results are dynamically updated and their accuracy is gradually improved.

2. The method for assessing the diffusion of heavy metal pollution in a land parcel according to claim 1, characterized in that, The quality calibration mentioned in step S1 specifically includes: Sub-step S11: Construct a multi-dimensional quality indicator system, which includes time-dimensional indicators, spatial-dimensional indicators, methodological-dimensional indicators, and analytical-dimensional indicators. Sub-step S12: For each monitoring data point, calculate its quality score in each dimension. Sub-step S13 uses a weighted summation method to merge the quality scores of each dimension into a comprehensive quality score; Sub-step S14: Based on the comprehensive quality score, normalization or hierarchical assignment is used to determine the data quality weight of each monitoring data. The data quality weight ranges from 0 to 1, and the higher the data quality, the closer the weight is to 1.

3. The method for assessing the diffusion of heavy metal pollution in a land parcel according to claim 1, characterized in that, Step S2, which involves constructing a probability distribution model for pollution source intensity, specifically includes: Sub-step S21: Construct a priori distribution of pollution source parameters, the priori distribution being determined based on the historical land use, production process, and pollution discharge records; Sub-step S22: Establish the likelihood function between the observation data and the pollution source parameters. The data quality weight is introduced into the likelihood function as an adjustment factor for the observation error. The higher the data quality weight, the greater the contribution of the corresponding observation data to the likelihood function. Sub-step S23: Using the Markov chain Monte Carlo method or the sequential data assimilation method, combined with the prior distribution and the likelihood function, solve for the posterior probability distribution of the pollution source parameters. Sub-step S24 involves performing statistical feature analysis on the posterior probability distribution to obtain the mean, variance, and quantiles of the pollution source parameters, as well as the correlation information between the parameters.

4. The method for assessing the diffusion of heavy metal pollution in a land parcel according to claim 1, characterized in that, Step S3, which involves generating multiple candidate diffusion paths for heavy metal pollution corresponding to different pollution source parameter samples, specifically includes: Sub-step S31: Randomly sample from the posterior probability distribution of pollution source parameters obtained in step S2 to generate N sets of pollution source parameter samples, where N is the preset number of samples; Sub-step S32: Construct a numerical model for the migration and transformation of heavy metals in a plot of land that considers heterogeneous soil media. The numerical model includes convection-dispersion equations, adsorption-desorption equations, and chemical reaction kinetic equations. Sub-step S33: Input each set of pollution source parameter samples as boundary conditions into the numerical model, run N-fold numerical simulation in parallel, and obtain N sets of spatiotemporal distribution results of pollution concentration. Sub-step S34 involves statistically analyzing the spatiotemporal distribution results of the N groups of pollution concentrations, calculating the mean, median, quantile, and variance of the concentration at each spatial location and time point, thus forming the pollution concentration field and its uncertainty distribution.

5. The method for assessing the diffusion of heavy metal pollution in a land parcel according to claim 1, characterized in that, Step S4, which involves constructing a diffusion confidence field to quantify the reliability of pollution diffusion results, specifically includes: Sub-step S41: Based on the data quality weights, a spatial interpolation method or geostatistical method is used to generate a constraint strength field of the monitoring data on the diffusion path. The constraint strength field is used to characterize the ability of the monitoring data to constrain the simulation results of the surrounding area. Sub-step S42: Calculate the spatial distribution similarity among the multiple candidate paths generated in step S3 to obtain a diffusion path consistency field; the spatial distribution similarity is quantified by one of the correlation coefficient, mutual information, or Euclidean distance of the concentration values ​​of each candidate path at the same spatial location; Sub-step S43 evaluates the convergence of the multiple candidate paths generated in step S3 in the iterative calculation to obtain the path stability field; the convergence is quantified by comparing the differences in simulation results under different sampling numbers or analyzing the trend of the variance of the simulation results with the number of iterations. Sub-step S44 involves using weighted fusion, fuzzy comprehensive evaluation, or machine learning methods to fuse the constraint strength field, diffusion path consistency field, and path stability field through multi-source information fusion to construct the diffusion credibility field. Sub-step S45 involves post-processing the diffusion confidence field, including spatial smoothing, threshold segmentation, or confidence interval calibration, to obtain the final confidence spatial distribution.

6. The method for assessing the diffusion of heavy metal pollution in a land parcel according to claim 1, characterized in that, The risk grading assessment described in step S5 specifically includes: Sub-step S51: Spatially align the pollution concentration field and the diffusion confidence field to obtain the concentration value C(x,y,z) and confidence value R(x,y,z) at each geographical location; Sub-step S52: Set the concentration threshold C_th and the confidence threshold R_th, define the position where the concentration value is higher than C_th as the high concentration region, and define the position where the confidence value is higher than R_th as the high confidence region; Sub-step S53: Based on the concentration threshold and confidence threshold, the land parcel is divided into four risk types: high-risk-high confidence area, high-risk-low confidence area, low-risk-high confidence area and low-risk-low confidence area; Sub-step S54: For high-risk-low confidence areas, output supplementary monitoring recommendations; for high-risk-high confidence areas, output priority control recommendations; for low-risk-low confidence areas, output continuous monitoring recommendations; for low-risk-high confidence areas, output recommendations to reduce monitoring frequency. Sub-step S55 generates a risk distribution map of heavy metal pollution diffusion in the site. The risk distribution map uses different colors or patterns to mark areas with different risk types, and includes a description of the risk level and control recommendations for each area.

7. The method for assessing the diffusion of heavy metal pollution in a land parcel according to claim 1, characterized in that, The dynamic update mentioned in step S6 specifically includes: Sub-step S61: Obtain newly added monitoring data and perform quality calibration on the newly added monitoring data to obtain the data quality weight of the newly added data; Sub-step S62: Merge the newly added monitoring data with the historical monitoring data to form an updated monitoring dataset; Sub-step S63: Based on the updated monitoring dataset, the posterior probability distribution of pollution source parameters obtained in the previous round of assessment is used as the initial prior distribution for this round of inversion. Steps S2 to S5 are executed again to obtain the updated pollution source probability distribution, updated diffusion candidate paths, updated confidence field, and updated risk classification assessment results. Sub-step S64: Evaluate the magnitude of change and convergence trend of the evaluation results before and after the update. When the magnitude of change of the evaluation results in multiple consecutive updates is less than a preset threshold, it is determined that the evaluation results have reached a stable state. Sub-step S65 records and outputs the evolution trajectory of the assessment results during the dynamic update process, which is used to analyze the spatiotemporal evolution of heavy metal pollution in the plot.

8. A land plot heavy metal pollution diffusion assessment system, characterized in that, The system applies a method for assessing the diffusion of heavy metal pollution in a land parcel as described in any one of claims 1 to 7, including: The data acquisition module is used to collect multi-source heavy metal monitoring data of the target site, including historical survey data interface, field sampling data interface, online monitoring data interface and remote sensing interpretation data interface; The quality calibration module, connected to the data acquisition module, is used to calibrate the quality of the acquired multi-source heavy metal monitoring data and calculate the data quality weight corresponding to each monitoring data based on a multi-dimensional quality index system. The model building module is used to construct a plot heavy metal migration and transformation model that couples soil physicochemical properties. The model includes a convection-diffusion sub-model, an adsorption-desorption sub-model, and a chemical reaction kinetics sub-model. The pollution source inversion module, connected to the quality calibration module and the model building module, is used to construct a probability distribution model of pollution source intensity by using the Bayesian inversion method or the ensemble Kalman filter method, combined with the data quality weights. The parallel simulation module, connected to the pollution source inversion module and the model building module, is used to sample and generate multiple sets of pollution source parameter samples from the pollution source probability distribution model, and drive the heavy metal migration and transformation model of the plot in parallel to perform forward simulation, generating multiple candidate paths for heavy metal pollution diffusion. The credibility field construction module, connected to the quality calibration module and the parallel simulation module, is used to construct a credibility field for heavy metal pollution diffusion by integrating the data quality weights, the path consistency among multiple diffusion candidate paths, and the path stability of each diffusion candidate path. The risk assessment module, connected to the parallel simulation module and the confidence field construction module, is used to spatially couple the diffusion confidence field with the pollution concentration field, calculate the risk level of each spatial location, and generate a distribution map of heavy metal pollution diffusion risk of the plot. The dynamic update module, connected to the data acquisition module, is used to trigger each module to repeatedly execute its function after new monitoring data is acquired, so as to realize the dynamic update of the pollution diffusion assessment results. The visualization output module, connected to the risk assessment module and the dynamic update module, is used to output the pollution concentration field, confidence field, risk classification results, and dynamic update trajectory in a graphical manner. The database module is used to store historical monitoring data, model parameters, intermediate calculation results, and final evaluation results.

9. The land plot heavy metal pollution diffusion assessment system according to claim 8, characterized in that, Also includes: The uncertainty analysis module, connected to the parallel simulation module, is used to perform statistical analysis on the multiple diffusion candidate paths and calculate the probability distribution, confidence interval, and probability of exceeding the standard of pollution concentration at each spatial location. The sensitivity analysis module, connected to the model building module, is used to analyze the sensitivity of model parameters to simulation results and identify key parameters affecting pollution diffusion. The scheme optimization module, connected to the risk assessment module, is used to generate optimized monitoring site layout schemes and pollution control schemes based on the risk classification assessment results. The report generation module is used to automatically generate a heavy metal pollution diffusion assessment report for a site, including an explanation of the assessment method, a description of the data source, model parameter settings, assessment result charts, and control recommendations.

Images