Atmospheric nitrogen sedimentation source analysis method and system for coupling stable isotopic tracing and Bayesian mixture model

Abstract

The invention discloses an atmospheric nitrogen deposition source analysis method and system for coupling stable isotopic tracing and a Bayesian mixture model, and belongs to the technical field of environmental monitoring and pollution traceability, and the method comprises the following steps: collecting atmospheric dry and wet deposition samples, and recording sampling metadata; carrying out automatic separation and purification to obtain to-be-detected nitrogen solutions in various forms, and measuring stable isotope values; integrating multi-source data to construct a dynamic source end member isotope feature database, inputting an improved Bayesian mixture model, and outputting a nitrogen source contribution proportion and an uncertainty interval through parameter optimization and polymorphic nitrogen differentiation analysis; and an LSTM prediction model is coupled, short-term and medium-and-long-term prediction is realized in combination with an emission scene, a result is output through spatio-temporal dynamic visualization, and an emission reduction decision suggestion is generated. According to the method, through automatic process, model optimization and multi-technology coupling, the analysis precision and efficiency are improved, and technical support is provided for atmospheric nitrogen pollution precise treatment, cross-regional collaborative management and control and ecological risk prevention and control.

Description

Technical Field

[0001] This invention relates to the field of environmental monitoring and pollution source tracing technology, and more specifically to a method and system for analyzing atmospheric nitrogen deposition sources by coupling stable isotope tracing with a Bayesian mixture model. Background Technology

[0002] Atmospheric nitrogen deposition is the process by which nitrogen-containing compounds in the atmosphere enter the Earth's surface through both dry and wet deposition. Excessive nitrogen deposition leads to ecological problems such as soil acidification, water eutrophication, and biodiversity loss. Accurate source tracing is crucial for developing regional air pollution prevention and control plans and implementing cross-regional collaborative management. Current technologies primarily rely on stable isotope tracing (EA-IRMS measurement of delta-carbon dioxide). 15 N, δ 18 The combined application of isotope-Bayesian (Iotope-Bayesian) coupling schemes with Bayesian mixture models (SIAR) offers theoretical advantages over traditional models, as the latter simulates the uncertainty contributed by quantified sources through MCMC. However, existing Iotope-Bayesian coupling schemes still face significant bottlenecks in practical engineering applications. Stable isotope tracing (δ) 15 N, δ 18 Although O determination is used for nitrogen source identification, there is overlap in isotopic characteristics between different nitrogen sources (e.g., NO from transportation and industry). x δ 15 N may be in the range of 5‰ to 12‰, and the organic nitrogen DON isotope database is missing, with a total nitrogen analysis deviation of 25% to 35%. Secondly, sampling, pretreatment and determination are completed manually, with a single sample cycle of 3 to 5 days, which cannot be adapted to high-density sampling over large areas and makes it difficult to capture spatiotemporal dynamics.

[0003] Furthermore, traditional mathematical models have limited analytical capabilities. Traditional methods such as chemical mass balance models and positive definite matrix factorization models heavily rely on complete and accurate emission source inventory data, resulting in significant errors in regions with incomplete inventories. They cannot distinguish between different forms of nitrogen from the same source and can only perform static analysis based on historical data, lacking the ability to dynamically predict future emission scenarios and thus failing to support forward-looking management decisions.

[0004] Therefore, how to provide a method and system for analyzing atmospheric nitrogen deposition sources that can deeply integrate stable isotope tracing and Bayesian hybrid models is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a method and system for atmospheric nitrogen deposition source apportionment that couples stable isotope tracing with a Bayesian mixture model. By deeply integrating stable isotope tracing with a Bayesian mixture model, it provides a method and system for atmospheric nitrogen deposition source apportionment that integrates high precision, high efficiency, dynamic prediction and decision support, providing strong technical support for the precise control of atmospheric nitrogen pollution and ecological environmental protection.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: On the one hand, this invention provides a method for analyzing atmospheric nitrogen deposition sources by coupling stable isotope tracing with a Bayesian mixture model, comprising: Collect dry and wet atmospheric deposition samples and record sampling metadata simultaneously; The sample was automatically separated and purified to obtain the test solution of nitrogen in various forms; The stable isotope values ​​of various forms of nitrogen were determined using the test solution to obtain the measurement data; Obtain emission source inventories, meteorological data, and land use data for the measurement area, and integrate the measurement data, emission source inventories, meteorological and land use data to construct a dynamically updated source-end-member isotope characteristic database; The integrated data is input into the improved Bayesian mixture model for probabilistic source apportionment, calculating the contribution ratio and uncertainty range of each preset nitrogen source, and optimizing the parameters of the measured data and the Bayesian mixture model based on the output and input data of the Bayesian mixture model. Based on the analysis results and combined with future emission scenarios, a coupled LSTM prediction model is used to make short-term and medium-to-long-term predictions of the source contribution ratio, and the prediction results are obtained. The analysis results and prediction results are dynamically visualized in time and space, and corresponding pollution reduction decision-making suggestions are automatically generated.

[0007] Preferably, dry and wet atmospheric deposition samples are collected, and sampling metadata is recorded simultaneously, including: A three-level sampling strategy was adopted, with several sampling points set up in the measurement area; Automatic samplers for precipitation and dustfall with automatic opening and closing functions were used for sampling. The synchronously recorded sampling metadata includes: latitude and longitude of the sampling point, altitude, sample type, sampling start and end time, precipitation, precipitation pH value, conductivity, temperature, humidity, wind speed, and wind direction.

[0008] Preferably, the automated separation and purification of the sample includes: NH4 in the sample is automatically separated using a fully automated ion analyzer via an ion exchange resin column. + NO3 - ; The concentration of DON was calculated using the difference method. The separated solution was automatically concentrated and purified using a rotary evaporator; Set up blank controls and standard samples for quality control; non-conforming samples will be automatically reprocessed.

[0009] Preferably, the optimization process of the improved Bayesian mixture model includes: The prior contribution probability of each nitrogen source is determined based on the emission source inventory; Map the accuracy level of the measured data to the model input weights; Dynamically adjust the fractionation coefficient parameters using sampled metadata; The Markov chain Monte Carlo simulation algorithm was used to obtain the posterior probability distribution of the contribution ratio of each nitrogen source through iterative calculation. When the width of the probability interval of the posterior distribution exceeds the preset value, the measurement data diagnosis is triggered, and the model parameters are corrected according to the diagnosis results. The diagnosis content includes the proportion of measurement data with data accuracy lower than the preset accuracy and the overlap of source end-member features.

[0010] Preferably, the probabilistic source apportionment includes differential apportionment of multiple nitrogen forms: For NH4 + -N, based on δ 15 N value distinguishes between agricultural fertilization, livestock and poultry farming and industrial ammonia escape sources; For NO3 - -N, combined with δ 15 N and δ 18 The dual indicators distinguish between atmospheric deposition, industrial emissions, traffic exhaust, and biomass combustion sources. For DON, based on the pre-defined organic nitrogen source endmember δ 15 Source contribution inversion was performed using N range and total nitrogen percentage constraints.

[0011] Preferably, the prediction via a coupled LSTM prediction model includes: The historical Bayesian posterior distribution data was used as the training set. Input parameters for future emission scenarios, including baseline scenario, strong emission reduction scenario, and moderate emission reduction scenario; Based on meteorological trend data, output the predicted source contribution values ​​with confidence intervals greater than the preset interval; A rolling training strategy is adopted, in which the model is retrained based on a preset period.

[0012] On the other hand, the present invention provides an atmospheric nitrogen deposition source apportionment system that couples stable isotope tracing with a Bayesian mixture model, comprising: An automated sampling module is used to collect dry and wet atmospheric deposition samples and simultaneously record sampling metadata. The pretreatment module is used to automatically separate and purify the sample to obtain the test solution of nitrogen in various forms; The isotope determination module is used to determine the stable isotope values ​​of various forms of nitrogen using the test solution, and obtain the measurement data. The data integration module is used to acquire emission source inventories, meteorological data, and land use data for the measurement area, and to integrate the measurement data, emission source inventories, meteorological and land use data to construct a dynamically updated source end-member isotope characteristic database. The analysis module is used to input the integrated data into the improved Bayesian mixture model, perform probabilistic source analysis, calculate the contribution ratio and uncertainty range of each preset nitrogen source, and optimize the parameters of the measured data and the Bayesian mixture model based on the output and input data of the Bayesian mixture model. The prediction module is used to make short-term and medium- to long-term predictions of the source contribution ratio based on the analysis results and future emission scenarios, through a coupled LSTM prediction model, and obtain the prediction results.

[0013] The visualization module is used to dynamically visualize the analysis results and prediction results in time and space, and automatically generate corresponding pollution reduction decision recommendations.

[0014] As can be seen from the above technical solutions, compared with the prior art, this invention discloses a method and system for atmospheric nitrogen deposition source apportionment that couples stable isotope tracing with a Bayesian mixture model. Through differential analysis of multiple forms of nitrogen and uncertainty quantification, the calculation error of the source contribution of ammonium nitrogen, nitrate nitrogen, and organic nitrogen is reduced to less than 10%, accurately distinguishing pollution sources. Furthermore, this invention supports updating the prior source metadata of the SIAR model in conjunction with emission scenarios, outputting short-term and medium-to-long-term source contribution probability distributions with a prediction error ≤15%, providing forward-looking support for policy formulation. Moreover, the Bayesian model coupled with an LSTM network realizes dynamic prediction capabilities from short-term to medium-to-long-term, enabling environmental management to shift from passive response to proactive planning. Attached Figure Description

[0015] 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 described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0016] Figure 1 This is a flowchart provided for the present invention; Figure 2 This is a schematic diagram of the structure provided by the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] This invention discloses a method for analyzing atmospheric nitrogen deposition sources by coupling stable isotope tracing with a Bayesian mixture model, such as... Figure 1 As shown, it includes: Collect dry and wet atmospheric deposition samples and record sampling metadata simultaneously; The sample was automatically separated and purified to obtain the test solution of nitrogen in various forms; The stable isotope values ​​of various forms of nitrogen were determined using the test solution to obtain measurement data, specifically including: The pretreated NH4 was automatically drawn in using an elemental analyzer-isotope mass spectrometer (EA-IRMS, such as Thermo Delta V). + NO3 - Solution and DON sample (DON is converted to NO3 by oxidation) - (Post-determination); Set measurement accuracy (δ) 15 N error ≤ 0.3‰, δ 18 (Error ≤ 0.3‰), single sample measurement time ≤ 30 minutes, which is 4-6 times more efficient than traditional manual operation; Based on measurement data and literature data, an isotopic characteristic database for six types of nitrogen sources was established (e.g., agricultural fertilization: δ¹⁸O⁻). 15 N = -10‰ ~ 6‰, δ 18 O = -5‰ ~ 25‰; Industrial emissions: δ 15 N = -5‰ ~ 15‰, δ 18 The isotope database has an isotope value range of 10‰ to 35‰ and supports dynamic updates based on regional characteristics (e.g., supplementing isotope characteristic values ​​after a new emission source is added to an industrial zone). An example of the established isotope database is shown in Table 1. Table 1. Examples of Isotope Databases

[0019] Obtain emission source inventories, meteorological data, and land use data for the measurement area, and integrate the measurement data, emission source inventories, meteorological data, and land use data to construct a dynamically updated source-end-member isotope characteristic database, specifically including: Using the Python Pandas library, data in different formats (such as txt files output by mass spectrometers and Excel files of emission inventories) are converted into a unified format (JSON), and outlier removal (such as δ) is performed. 15 N exceeding -20‰ or 40‰ is considered abnormal; missing values ​​are filled (using the mean of adjacent sampling points); By connecting to environmental monitoring platforms (such as the National Ambient Air Quality Monitoring Network) and meteorological platforms (such as the China Meteorological Administration) through API interfaces, the automatic updating of emission inventories and meteorological data can be achieved, ensuring data timeliness.

[0020] The integrated data is input into the improved Bayesian mixture model for probabilistic source apportionment, calculating the contribution ratio and uncertainty range of each preset nitrogen source, and optimizing the parameters of the measured data and the Bayesian mixture model based on the output and input data of the Bayesian mixture model. Based on the analysis results and combined with future emission scenarios, a coupled LSTM prediction model is used to make short-term and medium-to-long-term predictions of the source contribution ratio, and the prediction results are obtained. The system dynamically visualizes the analytical and predictive results in a spatiotemporal manner and automatically generates corresponding pollution reduction decision-making recommendations. The visualization function is developed based on WebGIS (OpenLayers) and ECharts, and its core functions include: Spatiotemporal distribution display: Heat map of nitrogen source contribution in different seasons in a certain county (with 95% confidence interval marked), such as "Agricultural source contribution in agricultural area in June 2024: 40% ± 4%"; Uncertainty display: probability density curves of nitrogen source contributions (e.g., industrial source contributions follow a normal distribution with a peak at 25%). Scenario Comparison Chart: Comparison of source contribution trends under different emission reduction scenarios (e.g., NO3 in the strong emission reduction vs. baseline scenario) - -N Industrial Source Contribution Curve); Intelligent decision-making suggestions: Based on SIAR analysis and prediction results, the system automatically generates: Key control areas: Areas where the source contributes more than 50% (e.g., "NH4 from agricultural sources in XX township"). + -N contributes 55%±3%, and should be prioritized for control); Emission reduction recommendations: Based on scenario forecasts, output the "optimal input-output ratio" solution, such as "installing SCR denitrification devices in industrial settings can reduce NO..." x The source contribution decreased by 15%, and the cost-benefit ratio was 1:8. Results Export: Supports exporting SIAR model results in R language format (for secondary analysis), source contribution data in Excel format (including confidence intervals), and analytical reports in PDF format (including charts and decision recommendations) to meet the archiving and reporting needs of environmental management departments.

[0021] Specifically, dry and wet atmospheric deposition samples are collected, and sampling metadata is recorded simultaneously, including: A three-tiered sampling strategy was adopted, setting up several sampling points in the measurement area. Specifically, a functional zone-gradient-density three-tiered sampling strategy was used to set up several sampling points in the target area, covering urban areas, industrial areas, agricultural areas, background areas, and boundary transition areas. The sampling point layout was based on the regional nitrogen emission characteristics, using a grid method to deploy the sampling points, and supporting GPS positioning and remote control.

[0022] Automatic samplers for precipitation and dustfall with automatic opening and closing functions were used for sampling. The synchronously recorded sampling metadata includes: latitude and longitude of the sampling point, altitude, sample type, sampling start and end time, precipitation, precipitation pH value, conductivity, and descriptions of atmospheric environmental conditions during the sampling period, such as temperature, humidity, wind speed, and wind direction.

[0023] Furthermore, automated separation and purification of samples includes: NH4 in samples is automatically separated using a fully automated ion analyzer (such as AA3) via an ion exchange resin column. + NO3 - ; The concentration of DON and total nitrogen TN-NH4 were calculated using the difference method. + -N-NO3 - -N); The separated solution was automatically concentrated and purified using a rotary evaporator; specifically, the separated NH4 was automatically concentrated using a rotary evaporator. + NO3 - The solution (concentrated to 10 mL) is used to remove impurities (such as heavy metal ions) to ensure the accuracy of isotope determination.

[0024] Set up a blank control (ultrapure water) and a standard sample (with known δ). 15 Quality control is performed using NH4Cl standard with N value, and non-conforming samples are automatically reprocessed.

[0025] Furthermore, the optimization process of the improved Bayesian mixture model includes: The sample isotope values, accuracy level, and corrected fractionation coefficient output from the data integration layer are used as inputs to the SIAR model. The prior contribution probability of each nitrogen source is set based on the emission source inventory (e.g., the prior probability of agricultural sources in agricultural areas = 50% ± 10%). The accuracy level of the measured data is mapped to the model input weights (weight of A-level data = 1.0, weight of B-level data = 0.8). The fractionation coefficient parameters are dynamically adjusted using sampled metadata (e.g., when the wind speed is >5m / s, the standard deviation of the fractionation coefficient is increased by 20%). The Markov chain Monte Carlo simulation algorithm is used to obtain the posterior probability distribution of the contribution ratio of each nitrogen source through iterative calculation. For samples with high isotope value stability (coefficient of variation CV < 10%), the initial iteration step size is set to 1000 times to quickly approach the convergence interval. For samples with low stability (CV > 15%), the initial step size is set to 500 times to ensure analytical accuracy and reduce the number of iterations by an average of 60%. When the standard deviation of the posterior distribution of 500 consecutive MCMC iterations is < 5% (which can be adjusted according to the analytical accuracy requirements), the sample is determined to have converged, and the iteration is automatically terminated.

[0026] When the width of the probability interval of the posterior distribution exceeds the preset value, the measurement data diagnosis is triggered, and the model parameters are corrected according to the diagnosis results. The diagnosis content includes the proportion of measurement data with data accuracy lower than the preset accuracy and the overlap of source end-member features.

[0027] Furthermore, probabilistic source apportionment includes differential apportionment of multiple forms of nitrogen: For NH4 + -N, based on δ 15 The N value distinguishes between agricultural fertilization, livestock and poultry farming, and industrial ammonia escape sources; specifically, it focuses on three types of sources: agricultural fertilization (NH3 volatilization), livestock and poultry farming (NH3 emissions), and industrial ammonia escape, using δ... 15 N is the core indicator (agricultural NH3: -10‰~0‰, industrial NH3: -5‰~5‰), and the contribution probability of each source is output through the SIAR model by combining the differences in fractionation coefficients. For NO3 - -N, combined with δ 15 N and δ 18 The dual indicators differentiate between atmospheric deposition, industrial emissions, traffic exhaust, and biomass combustion sources. Specifically, they integrate δ 15 N and δ 18 The O dual indicator distinguishes between atmospheric deposition, industrial emissions, traffic exhaust, and biomass combustion, for example, through δ... 18 O > 60‰ prioritizes atmospheric deposition, and then its contribution ratio is quantified by SIAR. For DON, based on the pre-defined organic nitrogen source endmember δ 15 Source contribution inversion was performed based on constraints of N range and total nitrogen percentage. Specifically, the source endmember δ of DON was set based on literature data. 15 The N range (atmospheric organic nitrogen: -15‰~10‰, biomass combustion organic nitrogen: -5‰~5‰), combined with the total nitrogen ratio constraint, uses the SIAR model to invert the source contribution of DON, filling the gap in existing technology.

[0028] In another embodiment, prediction via a coupled LSTM prediction model includes: During the model training phase, Bayesian analytical feature vectors from the past three years were used as input, along with actual pollution source emission data from the same period (such as industrial NOx). x Using emissions and agricultural fertilizer application as labels, an LSTM model is trained to learn the mapping relationship between "changes in pollution source emissions and contribution response". During the scenario prediction phase, users input future control scenario parameters (such as "Industrial NO"). x (Emissions reduced by 30% and agricultural fertilizer application reduced by 20%), the system automatically updates the input variables of the LSTM model and outputs the predicted mean of the contribution of pollution sources; In the Bayesian calibration phase, the mean of the LSTM predictions is used as the prior distribution of the Bayesian model, combined with the updated source-endmember parameters (such as industrial source δ). 15 (Adjust N range to 8‰~12‰), rerun the Bayesian model, and output the final prediction result with confidence interval (e.g., "Industrial source contribution will decrease to 20%±3%)", with a prediction error ≤10%; A rolling training strategy is adopted, in which the model is retrained based on a preset period.

[0029] On the other hand, this invention provides an atmospheric nitrogen deposition source apportionment system that couples stable isotope tracing with a Bayesian mixture model, such as... Figure 2 As shown, it includes: An automated sampling module is used to collect dry and wet atmospheric deposition samples and simultaneously record sampling metadata. The pretreatment module is used to automatically separate and purify the sample to obtain the test solution of nitrogen in various forms; The isotope determination module is used to determine the stable isotope values ​​of various forms of nitrogen using the test solution, and obtain the measurement data. The data integration module is used to acquire emission source inventories, meteorological data, and land use data for the measurement area, and to integrate the measurement data, emission source inventories, meteorological and land use data to construct a dynamically updated source end-member isotope characteristic database. The analysis module is used to input the integrated data into the improved Bayesian mixture model, perform probabilistic source analysis, calculate the contribution ratio and uncertainty range of each preset nitrogen source, and optimize the parameters of the measured data and the Bayesian mixture model based on the output and input data of the Bayesian mixture model. The prediction module is used to make short-term and medium- to long-term predictions of the source contribution ratio based on the analysis results and future emission scenarios, through a coupled LSTM prediction model, and obtain the prediction results.

[0030] The visualization module is used to dynamically visualize the analysis results and prediction results in time and space, and automatically generate corresponding pollution reduction decision recommendations.

[0031] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0032] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for source apportionment of atmospheric nitrogen deposition coupled with stable isotope tracing and a Bayesian mixture model, characterized in that, include: Collect dry and wet atmospheric deposition samples and record sampling metadata simultaneously; The sample was automatically separated and purified to obtain the test solution of nitrogen in various forms; The stable isotope values ​​of various forms of nitrogen were determined using the test solution to obtain the measurement data; Obtain emission source inventories, meteorological data, and land use data for the measurement area, and integrate the measurement data, emission source inventories, meteorological and land use data to construct a dynamically updated source-end-member isotope characteristic database; The integrated data is input into the improved Bayesian mixture model for probabilistic source apportionment, calculating the contribution ratio and uncertainty range of each preset nitrogen source, and optimizing the parameters of the measured data and the Bayesian mixture model based on the output and input data of the Bayesian mixture model. Based on the analysis results and combined with future emission scenarios, a coupled LSTM prediction model is used to make short-term and medium-to-long-term predictions of the source contribution ratio, and the prediction results are obtained. The analysis results and prediction results are dynamically visualized in time and space, and corresponding pollution reduction decision-making suggestions are automatically generated.

2. The method for analyzing atmospheric nitrogen deposition sources by coupling stable isotope tracing with a Bayesian mixture model as described in claim 1, characterized in that, Collect dry and wet atmospheric deposition samples and simultaneously record sampling metadata, including: A three-level sampling strategy was adopted, with several sampling points set up in the measurement area; Automatic samplers for precipitation and dustfall with automatic opening and closing functions were used for sampling. The synchronously recorded sampling metadata includes: latitude and longitude of the sampling point, altitude, sample type, sampling start and end time, precipitation, precipitation pH value, conductivity, temperature, humidity, wind speed, and wind direction.

3. The method for analyzing atmospheric nitrogen deposition sources by coupling stable isotope tracing with a Bayesian mixture model as described in claim 1, characterized in that, The automated separation and purification of samples includes: NH4 in the sample is automatically separated using a fully automated ion analyzer via an ion exchange resin column. + NO3 - ; The concentration of DON was calculated using the difference method. The separated solution was automatically concentrated and purified using a rotary evaporator; Set up blank controls and standard samples for quality control; non-conforming samples will be automatically reprocessed.

4. The method for analyzing atmospheric nitrogen deposition sources by coupling stable isotope tracing with a Bayesian mixture model according to claim 1, characterized in that, The optimization process of the improved Bayesian mixture model includes: The prior contribution probability of each nitrogen source is determined based on the emission source inventory; Map the accuracy level of the measured data to the model input weights; Dynamically adjust the fractionation coefficient parameters using sampled metadata; The Markov chain Monte Carlo simulation algorithm was used to obtain the posterior probability distribution of the contribution ratio of each nitrogen source through iterative calculation. When the width of the probability interval of the posterior distribution exceeds the preset value, the measurement data diagnosis is triggered, and the model parameters are corrected according to the diagnosis results. The diagnosis content includes the proportion of measurement data with data accuracy lower than the preset accuracy and the overlap of source end-member features.

5. The method for analyzing atmospheric nitrogen deposition sources by coupling stable isotope tracing with a Bayesian mixture model according to claim 1, characterized in that, The probabilistic source apportionment includes differential apportionment of multiple forms of nitrogen: For NH4 + -N, based on δ 15 N value distinguishes between agricultural fertilization, livestock and poultry farming and industrial ammonia escape sources; For NO3 - -N, combined with δ 15 N and δ 18 The dual indicators distinguish between atmospheric deposition, industrial emissions, traffic exhaust, and biomass combustion sources. For DON, based on the pre-defined organic nitrogen source endmember δ 15 Source contribution inversion was performed using N range and total nitrogen percentage constraints.

6. The method for analyzing atmospheric nitrogen deposition sources by coupling stable isotope tracing with a Bayesian mixture model according to claim 1, characterized in that, The prediction via a coupled LSTM prediction model includes: The historical Bayesian posterior distribution data was used as the training set. Input parameters for future emission scenarios, including baseline scenario, strong emission reduction scenario, and moderate emission reduction scenario; Based on meteorological trend data, output the predicted source contribution values ​​with confidence intervals greater than the preset interval; A rolling training strategy is adopted, in which the model is retrained based on a preset period.

7. A system for apportioning atmospheric nitrogen deposition sources coupled with stable isotope tracing and a Bayesian mixture model, characterized in that, include: An automated sampling module is used to collect dry and wet atmospheric deposition samples and simultaneously record sampling metadata. The pretreatment module is used to automatically separate and purify the sample to obtain the test solution of nitrogen in various forms; The isotope determination module is used to determine the stable isotope values ​​of various forms of nitrogen using the test solution, and obtain the measurement data. The data integration module is used to acquire emission source inventories, meteorological data, and land use data for the measurement area, and to integrate the measurement data, emission source inventories, meteorological and land use data to construct a dynamically updated source end-member isotope characteristic database. The analysis module is used to input the integrated data into the improved Bayesian mixture model, perform probabilistic source analysis, calculate the contribution ratio and uncertainty range of each preset nitrogen source, and optimize the parameters of the measured data and the Bayesian mixture model based on the output and input data of the Bayesian mixture model. The prediction module is used to make short-term and medium- to long-term predictions of the source contribution ratio based on the analysis results and future emission scenarios, through a coupled LSTM prediction model, and obtain the prediction results. The visualization module is used to dynamically visualize the analysis results and prediction results in time and space, and automatically generate corresponding pollution reduction decision recommendations.

Images