Nitrogen pollutant tracing method and system based on PMF and targeted detection and medium

Abstract

The invention discloses a nitrogen pollutant traceability method and system based on PMF and targeted detection and a medium. Relates to the technical field of pollution traceability. The method is improved on the basis of traditional traceability, pollution source quantitative source analysis is carried out on the basis of PMF software and main ion and nitrogen pollutant concentration samples, the preliminary situation of pollutants is obtained, then Bayesian mixture model analysis is carried out in combination with nitrogen and oxygen double-isotope samples of monitoring point pollution sources and nitrogen and oxygen double-isotope samples of related pollution sources, and the pollutant quantitative source analysis result is obtained. An accurate nitrogen pollutant traceability result is obtained, and the accuracy of the traceability result is remarkably improved; meanwhile, the traceability workload of the Bayesian mixture model is reduced through PMF software analysis, and the manpower and material resource cost required for obtaining the pollution source double-isotope data is greatly reduced; through organic combination of the two models, a complete traceability process from preliminary identification to accurate quantification is realized, and a more reliable scientific basis is provided for drainage basin nitrogen pollution treatment.

Description

Technical Field

[0001] The present invention relates to the technical field of pollution source tracing, and in particular to a nitrogen pollutant source tracing method, system and medium based on PMF and targeted detection. Background Art

[0002] Nitrogen pollutants (such as nitrates and ammonia nitrogen) enter surface waters through agricultural runoff, domestic sewage, and industrial wastewater, causing eutrophication, algal blooms, and ecosystem degradation. Furthermore, nitrogen pollution exhibits significant spatiotemporal heterogeneity. For example, nitrate concentrations in agricultural areas increase significantly during fertilization periods, while urban areas are affected by seasonal sewage discharges. However, traditional monitoring methods struggle to dynamically capture spatiotemporal variations in pollution, and existing control measures often rely on end-of-pipe treatment, lacking the ability to accurately identify pollution sources and conduct dynamic control, resulting in limited control effectiveness.

[0003] Current nitrogen pollutant source tracing technologies mainly include receptor models (such as positive definite matrix factor analysis PMF), isotope mixing models (such as SIAR) and numerical simulations (such as diffusion models). Positive definite matrix factor analysis PMF identifies the main controlling factors by decomposing the pollutant concentration matrix, but is sensitive to environmental interference (such as land use type and hydrological conditions) and needs to be combined with stable isotopes (δ 15 N, δ 18 O) Improve accuracy. Traditional numerical models rely on parameters such as longitudinal and transverse diffusion coefficients, but these parameters are difficult to obtain in practice and vary greatly across regions, resulting in insufficient universality. In addition, existing methods are mostly based on static data and lack the dynamic fusion of multi-source data, making it difficult to quantify the spatiotemporal uncertainty of pollution source contributions.

[0004] Therefore, it is urgent to develop a new traceability method to achieve high-precision analysis and dynamic management of nitrogen pollutants in surface water. Summary of the Invention

[0005] The technical problem to be solved by the present invention is that the traditional nitrogen pollutant source tracing method is sensitive to environmental interference, and is mostly based on static data, lacks dynamic fusion of multi-source data, and is difficult to quantify the spatiotemporal uncertainty of the contribution of pollution sources; the purpose of the present invention is to provide a nitrogen pollutant source tracing method, system and medium based on PMF and targeted detection, and to improve the method on the basis of traditional source tracing. First, quantitative source analysis of pollution sources is performed based on PMF software and main ion and nitrogen pollutant concentration samples to obtain preliminary information on pollutants, and then Bayesian mixed model analysis is performed on the nitrogen and oxygen dual isotope samples of the monitoring point pollution source and the nitrogen and oxygen dual isotope samples of related pollution sources to obtain accurate nitrogen pollutant source tracing results and improve the accuracy of the source tracing results; at the same time, the PMF analysis model is used to reduce the tracing workload of the Bayesian mixed model, greatly reducing the workload and manpower and material costs of the Bayesian mixed model pollution source dual isotope data collection process; the organic combination of the PMF analysis model and the Bayesian mixed model realizes a complete and accurate tracing process from preliminary identification to precise quantification, providing a more reliable scientific basis for basin nitrogen pollution control.

[0006] The present invention is achieved through the following technical solutions: This solution provides a nitrogen pollutant source tracing method based on PMF and targeted detection, including: Collect concentration samples of major ions and nitrogen pollutants in the target watershed and nitrogen and oxygen dual isotope samples from pollution sources at monitoring points; Quantitative source apportionment of pollution sources based on PMF software and samples of major ion and nitrogen pollutant concentrations; Based on the results of quantitative source analysis of pollution sources, relevant pollution sources are determined, and nitrogen and oxygen dual isotope samples of relevant pollution sources are collected; Bayesian mixed model analysis is performed on the nitrogen and oxygen dual isotope samples of pollution sources at monitoring points and nitrogen and oxygen dual isotope samples of relevant pollution sources to obtain the nitrogen pollutant tracing results.

[0007] A further optimization scheme is to perform quantitative source apportionment of pollution sources based on the PMF software and samples of major ion and nitrogen pollutant concentrations; including methods: Construct concentration data matrix and uncertainty data matrix based on the concentration samples of major ions and nitrogen pollutants; Import the concentration data matrix and uncertainty data matrix into the PMF analysis model respectively. After configuring the signal-to-noise ratio level and the number of pollution sources, run the PMF analysis model to obtain the initial pollution source contribution rate. On the premise that the PMF analysis model's fitting degree, residual value and objective function Q all meet the standards, the number of pollution sources and error coefficient are adjusted to optimize the PMF analysis model and obtain the optimal pollution source contribution rate.

[0008] A further optimization scheme is to construct a concentration data matrix and an uncertainty data matrix based on the main ion and nitrogen pollutant concentration samples; including the following methods: Obtaining the composition and concentration of the main ion and nitrogen pollutant samples, and constructing a concentration data matrix with the composition of the main ion and nitrogen pollutant samples as columns; If the concentration of major ions and nitrogen pollutants in sample i is c≤MDL, then according to formula u ij =5MDL / 6 to calculate the uncertainty u of component j in sample i ij Otherwise, according to the formula Calculate the uncertainty u of component j in sample i ij ; Wherein, MDL represents the concentration of sample i measured by the element measuring instrument; EF represents the error coefficient; An uncertainty data matrix is constructed based on all uncertainty data.

[0009] A further optimized solution is that the method for obtaining the initial pollution source contribution rate includes: Determine whether there are missing values in the concentration data matrix. If so, fill the corresponding missing values with H = 0.5MDL; Configure the signal-to-noise ratio level and the number of pollution sources; the number of pollution sources must be ≥ 2; Randomly select the starting point and perform iterative calculation according to the following formula: ; Where, X ij G represents the concentration data matrix of component j in sample i; ik represents the factor contribution matrix of factor k in sample i; F kj represents the factor spectrum matrix of factor k in sample i; E ij represents the residual of component j in sample i; p represents the total number of factors resolved; The iterative calculation includes the following steps: ij and the non-negative matrix G ik and F kj As a constraint, is the objective function, and the objective function is minimized; where m represents the total number of samples; n represents the total number of components in the sample.

[0010] A further optimization scheme is that the PMF analysis model fitting degree is calculated according to the following formula: ; ; Where, represents the predicted value of monitoring point i; represents the measurement value of monitoring point i; represents the average measurement value of monitoring point i; n represents the number of monitoring points, R 2 Indicates the PMF analysis model fit.

[0011] A further optimization scheme is to perform Bayesian mixed model analysis on the nitrogen and oxygen dual isotope samples of the pollution source at the monitoring point and the nitrogen and oxygen dual isotope samples of related pollution sources to obtain the nitrogen pollutant source tracing results; including the following methods: Preprocessing the nitrogen and oxygen dual isotope samples of the monitoring point pollution source and the nitrogen and oxygen dual isotope samples of related pollution sources; the preprocessing includes the following steps: calculating the average value and standard deviation of the nitrogen and oxygen dual isotope data of the same pollution source, and merging multiple pollution sources into typical pollution sources based on the similarity of isotope characteristics; A Bayesian mixture model was constructed, and the pre-processed nitrogen and oxygen dual isotope samples of the monitoring point pollution source and the nitrogen and oxygen dual isotope samples of related pollution sources were input into the Bayesian mixture model to estimate the contribution rate of the pollution source to the nitrogen and oxygen isotopes of the monitoring point.

[0012] A further optimization scheme is that the method for constructing the Bayesian mixture model includes: Construct a hierarchical Bayesian network including the prior distribution of contribution and the isotope mixing process; The initial assumption of the contribution ratio of pollution sources in the prior distribution of contribution P ( f ), using Dirichlet distribution as the contribution ratio of pollution sources f i Prior distribution of : ; α represents the parameters of the Dirichlet distribution, α =[1, 1, …,1], indicating uniform prior; The probability density function of the contribution prior distribution is: Where: f = [ f 1, f 2,…, f k ] represents the contribution ratio of pollution sources, satisfying ; α = [ α 1, α 2, …, α k ] is the parameter of Dirichlet distribution, k represents the total number of parameters, when α i =1, the Dirichlet distribution degenerates into a uniform distribution, indicating that there is no prior preference for the contribution ratio of pollution sources; when α i >1, the Dirichlet distribution tends to make the contribution ratio more uniform; when α i When <1, the Dirichlet distribution tends to make some contribution ratios close to 0 or 1; The isotope mixing process is represented by a likelihood function, which is used to predict the contribution ratio of a given pollution source. f i The probability of the observed data occurring under the condition of The posterior distribution of contribution is inferred based on the hierarchical Bayesian network, and samples are drawn from the posterior distribution of contribution based on the MCMC sampling method to estimate the contribution rate distribution of pollution sources.

[0013] A further optimization scheme is that the likelihood function includes: ; ; ; in, Indicates surface water point j of δ 15 N Predicted value, Indicates surface water monitoring points j of δ 15 N Predicted value; and Respectively indicate the observed 15 N and 18 O isotope ratio; represents the proportion of the i-th source in the mixture; and Represents the i-th source 15 N and 18 The true isotope ratio of O; represents the variance of the observed data; represents the likelihood function; and Represents the jth monitoring point 15 N and 18 The predicted standard deviation of O; k represents the total number of pollution sources; and n represents the number of samples.

[0014] This solution also provides a nitrogen pollutant source tracing system based on PMF and targeted detection, which is used to implement the above-mentioned nitrogen pollutant source tracing method based on PMF and targeted detection. The system includes: The acquisition module is used to collect concentration samples of major ions and nitrogen pollutants in the target watershed and nitrogen and oxygen dual isotope samples from pollution sources at monitoring points; The first analysis module is used to perform quantitative source analysis of pollution sources based on PMF software and samples of major ion and nitrogen pollutant concentrations; The second analysis module is used to determine the relevant pollution sources based on the quantitative source analysis results of the pollution sources, and collect nitrogen and oxygen dual isotope samples from the relevant pollution sources; Bayesian mixed model analysis is performed on the nitrogen and oxygen dual isotope samples of the monitoring point pollution sources and the nitrogen and oxygen dual isotope samples of the relevant pollution sources to obtain the nitrogen pollutant tracing results.

[0015] This solution provides a computer-readable medium on which a computer program is stored. The computer program is executed by a processor to implement the nitrogen pollutant tracing method based on PMF and targeted detection as described above.

[0016] Compared with the prior art, the present invention has the following advantages and beneficial effects: The present invention provides a nitrogen pollutant source tracing method, system and medium based on PMF and targeted detection; based on the traditional source tracing, the method is improved, firstly, based on the PMF software and the main ion and nitrogen pollutant concentration samples, the pollution source quantitative source analysis is carried out to obtain the preliminary situation of the pollutants, and then the nitrogen and oxygen dual isotope samples of the monitoring point pollution source and the nitrogen and oxygen dual isotope samples of the related pollution sources are combined to perform Bayesian mixed model analysis to obtain accurate nitrogen pollutant source tracing results, which significantly improves the accuracy of the source tracing results; at the same time, the source tracing workload of the Bayesian mixed model is reduced through PMF software analysis, and the manpower and material costs required for obtaining dual isotope data of the pollution source are greatly reduced; the organic combination of the two models realizes the complete source tracing process from preliminary identification to precise quantification, and provides a more reliable scientific basis for the nitrogen pollution control in the watershed. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the following briefly introduces the drawings required for use in the examples. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be considered as limiting the scope. A person of ordinary skill in the art can also derive other relevant drawings based on these drawings without inventive effort. In the drawings: Figure 1 This is a flow chart of the nitrogen pollutant source tracing method based on PMF and targeted detection; Figure 2 Schematic diagram of the nitrogen pollutant source tracing principle based on PMF and targeted detection; Figure 3 This is a schematic diagram of the sampling point distribution in River a; Figure 4 Schematic diagram of the percentage contribution of PMF factors in periods 1 to 4 in river basin a; Figure 5 This is a schematic diagram of the contribution ratio of each factor in period 1 to 4 of river basin a; Figure 6This is a schematic diagram of the contribution rate of each pollution source to the monitoring points in River B from November 2023 to May 2024; Figure 7 This is a schematic diagram of the contribution rate of each pollution source to the monitoring points in the b river basin from July 2024 to August 2024; Figure 8 This is a schematic diagram of the change in contribution rate of each pollution source from upstream to downstream in River B from November 2023 to May 2024; Figure 9 This is a schematic diagram of the change in contribution rate of each pollution source from upstream to downstream in River B from July 2024 to August 2024; Figure 10 Schematic diagram of the contribution rate changes of monitoring points 18, 01, 02 and 03 in river basin b; Figure 11 This is a schematic diagram of the contribution rate changes of monitoring points 20, 05, 06 and 19 in the b river basin; Figure 12 This is a schematic diagram of the contribution rate changes of monitoring points 07, 08, 09 and 10 in the b river basin; Figure 13 This is a schematic diagram of the contribution rate changes of monitoring points 11, 12, 13 and 14 in River Basin b; Figure 14 Schematic diagram of the contribution rate changes of monitoring points 15, 16 and 17 in River Basin B. DETAILED DESCRIPTION

[0018] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with examples and drawings. The exemplary embodiments of the present invention and their descriptions are only used to explain the present invention and are not intended to limit the present invention.

[0019] Traditional nitrogen pollutant source tracing methods are sensitive to environmental interference and are mostly based on static data. They lack the dynamic fusion of multi-source data and make it difficult to quantify the spatiotemporal uncertainty of pollution source contributions. In view of this, this solution provides the following embodiments to solve the above technical problems.

[0020] Example 1: This example provides a nitrogen pollutant source tracing method based on PMF and targeted detection, such as Figure 1 and Figure 2 As shown, including: Step 1: Collect concentration samples of major ions and nitrogen pollutants in the target watershed and nitrogen and oxygen dual isotope samples from pollution sources at monitoring points; Step 2: Perform quantitative source apportionment of pollution sources based on PMF software (EPA PMF 5.0 software was used in this example) and samples of major ion and nitrogen pollutant concentrations. This step specifically includes the following methods: S21, constructing a concentration data matrix and an uncertainty data matrix based on the concentration samples of major ions and nitrogen pollutants; this step specifically includes the following methods: S211, obtaining the composition and concentration of the main ion and nitrogen pollutant samples, and constructing a concentration data matrix with the composition of the main ion and nitrogen pollutant samples as columns; S212, if the concentration c of the main ion and nitrogen pollutant sample i is less than or equal to MDL, then according to formula u ij =5MDL / 6 to calculate the uncertainty u of component j in sample i ij Otherwise, according to the formula Calculate the uncertainty u of component j in sample i ij ; Wherein, MDL represents the concentration of sample i measured by the element measuring instrument; EF represents the error coefficient; S213, constructing an uncertainty data matrix based on all uncertainty data.

[0021] S22, import the concentration data matrix and uncertainty data matrix into the PMF analysis model respectively, configure the signal-to-noise ratio level and the number of pollution sources, and run the PMF analysis model to obtain the initial pollution source contribution rate; The method for obtaining the initial pollution source contribution rate includes: Determine whether there are missing values in the concentration data matrix. If so, fill the corresponding missing values with H = 0.5MDL; Configure the signal-to-noise ratio level and the number of pollution sources; the number of pollution sources must be ≥ 2. When configuring the signal-to-noise ratio level, set the category of each component based on the signal-to-noise ratio (S / N) strength. The setting rule is: when the signal-to-noise ratio S / N ≤ 0.5, set it to the Bad level; when 0.5 < S / N ≤ 1.0, set it to the Weak level; when S / N > 1.0, set it to the Strong level.

[0022] Randomly select the starting point and perform iterative calculation according to the following formula: ; Where, X ij G represents the concentration data matrix of component j in sample i; ik represents the factor contribution matrix of factor k in sample i; F kj represents the factor spectrum matrix of factor k in sample i; E ij represents the residual of component j in sample i; p represents the total number of factors resolved; The iterative calculation includes the following steps: ij and the non-negative matrix G ik and F kj As a constraint, is the objective function, and the objective function is minimized; where m represents the total number of samples; n represents the total number of components in the sample.

[0023] S23, under the premise that the PMF analysis model fit, residual value and objective function Q all meet the standards, adjust the number of pollution sources and error coefficient to optimize the PMF analysis model to obtain the optimal pollution source contribution rate. Specifically, in this step, the PMF analysis model fit should be as close to 1 as possible, the residual value should be kept between -3 and 3, and Q / Q exp Less than 2, Q exp That is the expected value of Q.

[0024] The PMF analysis model fit is calculated according to the following formula: ; ; Where, represents the predicted value of monitoring point i; represents the measurement value of monitoring point i; represents the average measurement value of monitoring point i; n represents the number of monitoring points, R 2 Indicates the PMF analysis model fit.

[0025] Step 3: Determine the relevant pollution sources based on the results of the quantitative source analysis of the pollution sources, and collect nitrogen and oxygen dual isotope samples from the relevant pollution sources; combine the nitrogen and oxygen dual isotope samples of the monitoring point pollution sources with the nitrogen and oxygen dual isotope samples of the relevant pollution sources to perform Bayesian mixed model analysis to obtain the nitrogen pollutant tracing results. In this step, the nitrogen and oxygen dual isotope samples of the monitoring point pollution sources and the nitrogen and oxygen dual isotope samples of the relevant pollution sources are combined to perform Bayesian mixed model analysis to obtain the nitrogen pollutant tracing results; including the following methods: S31, preprocessing the nitrogen and oxygen dual isotope samples of the monitoring point pollution source and the nitrogen and oxygen dual isotope samples of related pollution sources; the preprocessing includes the following steps: calculating the average value and standard deviation of the nitrogen and oxygen dual isotope data of the same pollution source, and merging multiple pollution sources into typical pollution sources based on the similarity of isotope characteristics; S32, constructing a Bayesian mixture model, inputting the pre-processed nitrogen and oxygen dual isotope samples of the monitoring point pollution source and the nitrogen and oxygen dual isotope samples of related pollution sources into the Bayesian mixture model to estimate the contribution rate of the pollution source to the nitrogen and oxygen isotopes of the monitoring point.

[0026] In this step, the Bayesian mixture model construction method includes: S321, constructing a hierarchical Bayesian network including the prior distribution of contribution and the isotope mixing process; The initial assumption of the contribution ratio of pollution sources in the prior distribution of contribution P ( f ), using Dirichlet distribution as the contribution ratio of pollution sources f iPrior distribution of : ; α represents the parameters of the Dirichlet distribution, α =[1, 1, …,1], indicating uniform prior; The probability density function of the contribution prior distribution is: Where: f = [ f 1, f 2,…, f k ] represents the contribution ratio of pollution sources, satisfying ; α = [ α 1, α 2, …, α k ] is the parameter of Dirichlet distribution, k represents the total number of parameters, when α i =1, the Dirichlet distribution degenerates into a uniform distribution, indicating that there is no prior preference for the contribution ratio of pollution sources; when α i >1, the Dirichlet distribution tends to make the contribution ratio more uniform; when α i When <1, the Dirichlet distribution tends to make some contribution ratios close to 0 or 1; The isotope mixing process is represented by a likelihood function, which is used to predict the contribution ratio of a given pollution source. f i The probability of the observed data occurring under the condition of In this step, the likelihood function includes: ; ; ; in, Indicates surface water point j of δ 15 N Predicted value, Indicates surface water monitoring points j of δ 15 N Predicted value; and Respectively indicate the observed 15 N and 18 O isotope ratio; represents the proportion of the i-th source in the mixture; and Represents the i-th source 15 N and 18 The true isotope ratio of O; represents the variance of the observed data; represents the likelihood function; and Represents the jth monitoring point 15 N and 18 The predicted standard deviation of O; k represents the total number of pollution sources; and n represents the number of samples.

[0027] The role of the likelihood function in the Bayesian mixture model is to quantify the degree of match between the observed data (isotope values of surface water points) and the predicted data (weighted average based on the contribution ratio of pollution sources).

[0028] S322, infer the posterior distribution of contribution based on the hierarchical Bayesian network, extract samples from the posterior distribution of contribution based on the MCMC sampling method, and estimate the distribution of pollution source contribution rates.

[0029] Finally, the distribution of pollution source contribution rates is visualized.

[0030] Example 2 This embodiment provides a nitrogen pollutant source tracing system based on PMF and targeted detection, which is used to implement the nitrogen pollutant source tracing method based on PMF and targeted detection described in Example 1. The system includes: The acquisition module is used to collect concentration samples of major ions and nitrogen pollutants in the target watershed and nitrogen and oxygen dual isotope samples from pollution sources at monitoring points; The first analysis module is used to perform quantitative source analysis of pollution sources based on PMF software and samples of major ion and nitrogen pollutant concentrations; The second analysis module is used to determine the relevant pollution sources based on the quantitative source analysis results of the pollution sources, and collect nitrogen and oxygen dual isotope samples from the relevant pollution sources; Bayesian mixed model analysis is performed on the nitrogen and oxygen dual isotope samples of the monitoring point pollution sources and the nitrogen and oxygen dual isotope samples of the relevant pollution sources to obtain the nitrogen pollutant tracing results.

[0031] Example 3 This embodiment provides a computer-readable medium having a computer program stored thereon. The computer program is executed by a processor to implement the nitrogen pollutant source tracing method based on PMF and targeted detection as described in Example 1. Specifically, the following steps are performed: Step 1: Collect concentration samples of major ions and nitrogen pollutants in the target watershed and nitrogen and oxygen dual isotope samples from pollution sources at monitoring points; Step 2: Quantitative source apportionment of pollution sources based on PMF software and samples of major ion and nitrogen pollutant concentrations; Step three: Determine the relevant pollution sources based on the results of quantitative source analysis of pollution sources, and collect nitrogen and oxygen dual isotope samples from the relevant pollution sources; combine the nitrogen and oxygen dual isotope samples of the monitoring point pollution sources and the nitrogen and oxygen dual isotope samples of the relevant pollution sources to perform Bayesian mixed model analysis to obtain the nitrogen pollutant tracing results.

[0032] Example 4 This embodiment specifically describes the method of the present invention in detail using A plain and a river basin: Conduct field surveys in the A Plain a River Basin and collect samples of major ion and nitrogen pollutant concentrations. Samples from the a River Basin are available at Figure 3 , a total of 4 periods of data were collected in the a river basin, of which the first period was the dry season, the second period was the normal water period after the dry season, the third period was the wet season, and the fourth period was the normal water period after the wet season; a total of 29 sampling points were arranged in the a river basin, which are numbered 1-29 in the figure.

[0033] The indicators tested and experimented on the samples of major ion and nitrogen pollutant concentrations are: K + 、Na 2+ , Ca 2+ Mg 2+ 、HCO3 2– 、SO4 2– 、Cl – , COD, total nitrogen, ammonia nitrogen, nitrate nitrogen, nitrite nitrogen, other forms of nitrogen, total nitrogen, dissolved total phosphorus, dissolved phosphate, dissolved organic phosphorus and particulate phosphorus.

[0034] A concentration data matrix was constructed based on the concentration samples of major ions and nitrogen pollutants; the uncertainty was calculated based on the relationship between the concentration data of each component and the detection limit (MDL), and an uncertainty data matrix was constructed. Finally, the concentration data matrix and the uncertainty data matrix were imported into the EPA PMF 5.0 software respectively, the "missing value indicator" was set to -999, the signal-to-noise ratio (S / N) level was set, and the optimal number of pollution sources for the PMF analysis model of the four water periods was determined to be 4, 4, 4, and 3 respectively. After running the model, after obtaining the initial pollution source contribution rate results, different numbers of pollution sources and error coefficients were tried to optimize the model results: ensure that the model fit of each component is as close to 1 as possible, ensure that the residual value is between -3 and 3, and Q / Q exp If the value is less than 2, the quantitative source analysis results of the pollution sources are obtained, and a stacked bar chart of the contribution rate of the pollution sources to each chemical component and the total contribution rate of each pollution source to the chemical components of surface water are drawn. Figure 4 and Figure 5As shown, in the major waterways of River A, the main sources of various indicators of concern are agricultural fertilizer, sewage, and dissolved minerals. Agricultural fertilizer contributes the most. The composition of different pollution sources varies significantly, with agricultural fertilizer contributing between 16% and 75%. This indicates that, for most periods, non-point source pollution from agricultural fertilizer inflow remains the primary source of pollution in this important waterway of River A. In addition to agricultural sources, domestic sewage and manure also contribute to a certain extent over time, with a relatively stable contribution of approximately 13% to 35%.

[0035] The PMF analysis model identifies pollution sources primarily as agricultural fertilizers, sewage, and mineral dissolution, and the results are highly reasonable. However, the sources of nitrogen pollution in surface water are diverse, and the cross-mixing involved is extremely complex. Different anthropogenic sources may have similar contribution characteristics, and some typical pollution sources cannot be accurately identified and analyzed, such as aquaculture, regular discharge from sewage treatment plants, and industry, which are common in the basin area. The Bayesian mixture model is a necessary optimization method for PMF analysis results to address errors in PMF caused by data noise or source overlap. Based on the pollution sources analyzed by PMF, we obtain dual-isotope data of the relevant pollution sources and use the Bayesian mixture model to further optimize the pollution source tracing results.

[0036] During the early field survey and sampling process of the B River Basin in Plain A, samples from the B River Basin were also obtained. The nitrogen and oxygen double isotope data of the monitoring points and pollution sources showed that the land use type in the B River Basin is mainly farmland, followed by construction land, which is an area with intensive human activities.

[0037] Nitrogen and oxygen dual isotope samples from monitoring point pollution sources and related pollution sources were divided into nine categories based on similar characteristics, including aquaculture, litter, industry, sewage discharge (i.e., qualified sewage discharge after treatment), domestic sewage, crop farming, shrimp ponds, rainwater, and fish ponds. Their means and standard deviations were calculated simultaneously. The Bayesian mixture model is used to further analyze the nitrogen and oxygen isotope source: the Bayesian mixture model is implemented by Python programming packages such as pandas, numpy, pymc, arviz, and seaborn. Through isotope mass conservation and Bayesian statistical inference, it can more reasonably determine the pollution source. The calculated pollution source contribution rate for each period is as follows: Figure 6 and Figure 7 As shown, Figure 6 This is a schematic diagram of the contribution rate of each pollution source to the monitoring points in River B from November 2023 to May 2024; Figure 7This is a schematic diagram of the change in the contribution rate of pollution sources from upstream to downstream in the b river basin from July 2024 to August 2024; the monitoring points set up in this area mainly include 01-22 as shown in the figure; generally speaking, the water quality of this basin is mainly affected by the dual influence of sewage treatment plant discharge and domestic sewage, and shrimp pond farming activities also show a significant impact in local areas. The sections affected by sewage discharge are mainly reflected in monitoring points 01, 16, and 17; while domestic sewage mainly affects monitoring point 12 and monitoring point 03. The schematic diagram of the change in the contribution rate of each pollution source from upstream to downstream in the b river basin from November 2023 to August 2024 is shown in the figure. Figure 8 and Figure 9 As shown in Figure 2, water pollution in River B has significant spatial heterogeneity and temporal dynamics. Sewage discharge, as the primary source of pollution, has a contribution rate that is "high in the middle reaches and relatively low at both ends"; the impact of domestic sewage shows a linear trend of increasing along the course. The impact of shrimp pond farming has obvious seasonal characteristics, mainly concentrated in the active period of aquaculture (November). From November 2023 to late August 2024, as shown in Figure 2, the contribution rate of sewage discharge is "high in the middle reaches and relatively low at both ends"; the impact of domestic sewage shows a linear trend of increasing along the course. The impact of shrimp pond farming has obvious seasonal characteristics, mainly concentrated in the active period of aquaculture (November). From November 2023 to late August 2024, as shown in Figure 2, the contribution rate of domestic sewage discharge is "high in the middle reaches and relatively low at both ends"; ... Figure 10 and 14 As shown in Figure 2, water quality near each monitoring point in River Basin B is consistently controlled by domestic sewage and wastewater discharge. The contribution of sewage discharge has increased significantly at most sites since spring, indicating that changes in rainfall have a significant impact on the contribution of pollution sources.

[0038] This approach used the PMF analysis model to preliminarily identify agricultural fertilizers as the primary source of nitrogen pollution in the basin (contributing 16% to 75%). Based on this, dual-isotope data from relevant pollution sources was obtained and a Bayesian mixture model was used to refine the source characteristics. This optimization process further confirmed that sewage discharge and domestic wastewater were the key sources of pollution affecting the basin's water quality. This optimization process effectively addressed the PMF analysis model's analytical bias when isotope signatures of pollution sources overlapped, significantly improving the accuracy of the source tracing results and significantly reducing the labor and material costs required to obtain dual-isotope data from pollution sources. The organic combination of the two models enabled a complete source tracing process from preliminary identification to precise quantification, providing a more reliable scientific basis for nitrogen pollution control in the basin.

[0039] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method 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 in the scope of protection of the present invention.

Claims

1. A nitrogen pollutant source tracing method based on PMF and targeted detection, characterized in that: include: Collect concentration samples of major ions and nitrogen pollutants in the target watershed and nitrogen and oxygen dual isotope samples from pollution sources at monitoring points; Quantitative source apportionment of pollution sources based on PMF software and samples of major ion and nitrogen pollutant concentrations; Based on the results of quantitative source analysis of pollution sources, relevant pollution sources are determined, and nitrogen and oxygen dual isotope samples of relevant pollution sources are collected; Bayesian mixed model analysis is performed on the nitrogen and oxygen dual isotope samples of pollution sources at monitoring points and nitrogen and oxygen dual isotope samples of relevant pollution sources to obtain the nitrogen pollutant tracing results.

2. The nitrogen pollutant source tracing method based on PMF and targeted detection according to claim 1 is characterized in that: The method of performing quantitative source apportionment of pollution sources based on PMF software and samples of major ion and nitrogen pollutant concentrations includes: Construct concentration data matrix and uncertainty data matrix based on the concentration samples of major ions and nitrogen pollutants; Import the concentration data matrix and uncertainty data matrix into the PMF analysis model respectively. After configuring the signal-to-noise ratio level and the number of pollution sources, run the PMF analysis model to obtain the initial pollution source contribution rate. On the premise that the PMF analysis model's fitting degree, residual value and objective function Q all meet the standards, the number of pollution sources and error coefficient are adjusted to optimize the PMF analysis model and obtain the optimal pollution source contribution rate.

3. The nitrogen pollutant source tracing method based on PMF and targeted detection according to claim 2 is characterized in that: The method of constructing a concentration data matrix and an uncertainty data matrix based on the main ion and nitrogen pollutant concentration samples includes: Obtaining the composition and concentration of the main ion and nitrogen pollutant samples, and constructing a concentration data matrix with the composition of the main ion and nitrogen pollutant samples as columns; If the concentration of major ions and nitrogen pollutants in sample i is c≤MDL, then according to formula u ij =5MDL / 6 to calculate the uncertainty u of component j in sample i ij Otherwise, according to the formula Calculate the uncertainty u of component j in sample i ij ; Wherein, MDL represents the concentration of sample i measured by the element measuring instrument; EF represents the error coefficient; An uncertainty data matrix is constructed based on all uncertainty data.

4. The nitrogen pollutant source tracing method based on PMF and targeted detection according to claim 2, characterized in that: The method for obtaining the initial pollution source contribution rate includes: Determine whether there are missing values in the concentration data matrix. If so, fill the corresponding missing values with H = 0.5MDL; MDL represents the concentration of sample i measured by the element measurement instrument; H represents the missing value; Configure the signal-to-noise ratio level and the number of pollution sources; the number of pollution sources must be ≥ 2; Randomly select the starting point and perform iterative calculation according to the following formula: ； Where, X ij G represents the concentration data matrix of component j in sample i; ik represents the factor contribution matrix of factor k in sample i; F kj represents the factor spectrum matrix of factor k in sample i; E ij represents the residual of component j in sample i; p represents the total number of factors resolved; The iterative calculation includes the following steps: ij and the non-negative matrix G ik and F kj As a constraint, is the objective function, minimizing the objective function Q; where m represents the total number of samples; and n represents the total number of components in the sample.

5. The nitrogen pollutant source tracing method based on PMF and targeted detection according to claim 2, characterized in that: The PMF analysis model fit is calculated according to the following formula: ； ； Where, represents the predicted value of the monitoring point; represents the measurement value of monitoring point i; represents the average measurement value of monitoring point i; n represents the number of monitoring points, R 2 Indicates the PMF analysis model fit; G ik represents the factor contribution matrix of factor k in sample i; F kj Represents the factor spectrum matrix of factor k in sample i.

6. The nitrogen pollutant source tracing method based on PMF and targeted detection according to claim 1 is characterized in that: The Bayesian mixed model analysis is performed on the nitrogen and oxygen dual isotope samples of the pollution source at the monitoring point and the nitrogen and oxygen dual isotope samples of the related pollution sources to obtain the nitrogen pollutant tracing results; Included methods: Preprocessing the nitrogen and oxygen dual isotope samples of the monitoring point pollution source and the nitrogen and oxygen dual isotope samples of related pollution sources; the preprocessing includes the following steps: calculating the average value and standard deviation of the nitrogen and oxygen dual isotope data of the same pollution source, and merging multiple pollution sources into typical pollution sources based on the similarity of isotope characteristics; A Bayesian mixture model was constructed, and the pre-processed nitrogen and oxygen dual isotope samples of the monitoring point pollution source and the nitrogen and oxygen dual isotope samples of related pollution sources were input into the Bayesian mixture model to estimate the contribution rate of the pollution source to the nitrogen and oxygen isotopes of the monitoring point.

7. The nitrogen pollutant source tracing method based on PMF and targeted detection according to claim 6, characterized in that: The Bayesian mixture model construction method includes: Construct a hierarchical Bayesian network including the prior distribution of contribution and the isotope mixing process; The initial assumption of the contribution ratio of pollution sources in the prior distribution of contribution P ( f ), using Dirichlet distribution as the contribution ratio of pollution sources f i Prior distribution of : ; α represents the parameters of the Dirichlet distribution, α =[1, 1, …,1], indicating uniform prior; The probability density function of the contribution prior distribution is: Where: f = [ f 1, f 2,…, f k ] represents the contribution ratio of pollution sources, satisfying ; α = [ α 1, α 2, …, α k ] is the parameter of Dirichlet distribution, k represents the total number of parameters, when α i =1, the Dirichlet distribution degenerates into a uniform distribution, indicating that there is no prior preference for the contribution ratio of pollution sources; when α i >1, the Dirichlet distribution tends to make the contribution ratio more uniform; when α i When < 1, the Dirichlet distribution tends to make some contribution ratios close to 0 or 1; The isotope mixing process is represented by a likelihood function, which is used to predict the contribution ratio of a given pollution source. f i The probability of the observed data occurring under the condition of The posterior distribution of contribution is inferred based on the hierarchical Bayesian network, and samples are drawn from the posterior distribution of contribution based on the MCMC sampling method to estimate the contribution rate distribution of pollution sources.

8. The nitrogen pollutant source tracing method based on PMF and targeted detection according to claim 7 is characterized in that: The likelihood function includes: ； ； ； in, Indicates surface water point j of δ 15 N Predicted value; and Respectively indicate the observed 15 N and 18 O isotope ratio; represents the proportion of the i-th source in the mixture; and Represents the i-th source 15 N and 18 The true isotope ratio of O; represents the variance of the observed data; represents the likelihood function; and Represents the jth monitoring point 15 N and 18 The predicted standard deviation of O; k represents the total number of pollution sources; n represents the number of samples; N() represents the probability density function.

9. The nitrogen pollutant source tracing system based on PMF and targeted detection is characterized by: For implementing the nitrogen pollutant source tracing method based on PMF and targeted detection according to any one of claims 1 to 8, the system comprises: The acquisition module is used to collect concentration samples of major ions and nitrogen pollutants in the target watershed and nitrogen and oxygen dual isotope samples from pollution sources at monitoring points; The first analysis module is used to perform quantitative source analysis of pollution sources based on PMF software and samples of major ion and nitrogen pollutant concentrations; The second analysis module is used to determine the relevant pollution sources based on the quantitative source analysis results of the pollution sources, and collect nitrogen and oxygen dual isotope samples from the relevant pollution sources; Bayesian mixed model analysis is performed on the nitrogen and oxygen dual isotope samples of the monitoring point pollution sources and the nitrogen and oxygen dual isotope samples of the relevant pollution sources to obtain the nitrogen pollutant tracing results.

10. A computer-readable medium having a computer program stored thereon, characterized in that: The computer program is executed by a processor to implement the nitrogen pollutant source tracing method based on PMF and targeted detection as described in any one of claims 1 to 8.

Images