Multi-source satellite remote sensing cooperative atmospheric pollution monitoring and dynamic tracing method and system

Abstract

The invention belongs to the technical field of environmental science, and relates to a multi-source satellite remote sensing cooperative atmospheric pollution monitoring and dynamic tracing method and system. The method comprises the following steps: collecting, processing and screening multi-source data; constructing a pollutant diffusion model, and generating pollutant concentration distribution data; dynamically adjusting the diffusion coefficient and attenuation rate of the pollutant diffusion model, and generating a pollutant concentration distribution thermodynamic diagram; performing space matching, and iteratively updating the diffusion coefficient and the attenuation rate of the pollutant diffusion model until matching parameters meet set conditions; and verifying pollutant concentration distribution data in the pollutant concentration distribution thermodynamic diagram. According to the invention, parameters of the pollutant diffusion model can be dynamically adjusted and optimized based on real-time meteorological data, the accuracy of pollutant concentration distribution data generated by the pollutant diffusion model is ensured, and real-time tracking and accurate positioning of a pollution source are realized; the monitoring precision is obviously improved; the cooperative monitoring efficiency is improved, and comprehensive monitoring of atmospheric pollutants is realized.

Description

Technical Field

[0001] The present invention belongs to the field of environmental science and technology, and specifically relates to a method and system for collaborative air pollution monitoring and dynamic source tracing using multi-source satellite remote sensing. Background Art

[0002] Air pollution monitoring, as an important part of environmental protection, faces many challenges. Although traditional ground monitoring stations can provide high-precision data, their coverage is limited and the construction cost is high, making it impossible to comprehensively reflect the overall situation of air pollution. Although single satellite remote sensing technology has the ability to observe large areas, it is limited by problems such as resolution, incomplete coverage of pollutant types, and meteorological condition interference, resulting in deviations in monitoring results and difficulty in meeting the monitoring requirements of high precision and real-time.

[0003] In the prior art, the industrial point source monitoring method based on statistical Z-test and atmospheric diffusion model estimates the emission source rate by fitting the plume model. Although it improves the monitoring accuracy to a certain extent, it fails to solve the problem of multi-pollutant collaborative inversion and has poor adaptability to complex terrain and meteorological conditions. Existing multi-source satellite remote sensing integrated monitoring systems can invert multiple pollutants, but they are insufficient in dynamic source tracing and fail to integrate real-time tracking and precise positioning of pollution sources. In addition, existing air pollution monitoring methods rely on fixed historical data and cannot adjust the wind direction and diffusion conditions in real time, resulting in large errors in pollution source positioning and limited dynamic adjustment ability, making it difficult to adapt to complex and changeable air pollution situations. Summary of the Invention

[0004] To solve the above technical problems, the present invention provides a method and system for collaborative air pollution monitoring and dynamic source tracing using multi-source satellite remote sensing.

[0005] In a first aspect, the present invention provides a method for collaborative air pollution monitoring and dynamic source tracing using multi-source satellite remote sensing, including:

[0006] Multi-source data collection, processing, and screening; the multi-source data includes pollution source location data, pollution emission intensity data, and meteorological data, as well as pollutant concentration observation data in satellite remote sensing data;

[0007] Construct a pollutant diffusion model, input the pollution source location data, pollution emission intensity data, and real-time meteorological data into the pollutant diffusion model to generate pollutant concentration distribution data;

[0008] Use a numerical simulation method to input multi-source data, dynamically adjust the diffusion coefficient and attenuation rate of the pollutant diffusion model, and generate a pollutant concentration distribution heat map;

[0009] Spatially match the heat map of pollutant concentration distribution with the observed pollutant concentration data, and iteratively update the diffusion coefficient and attenuation rate of the pollutant diffusion model until the matching parameters meet the set conditions;

[0010] Construct a deep learning model to fuse multi-source data to obtain a source probability distribution model. Use an unmanned terminal to scan the target area corresponding to the heat map of pollutant concentration distribution, calculate the source probability distribution data using the source probability distribution model, and verify the pollutant concentration distribution data in the heat map of pollutant concentration distribution.

[0011] In a second aspect, the present invention provides a multi-source satellite remote sensing collaborative air pollution monitoring and dynamic source tracing system, including a data acquisition and processing unit, a model construction and data generation unit, an adjustment and heat map generation unit, a matching and iteration unit, and a fusion and verification unit;

[0012] The data acquisition and processing unit is used for multi-source data acquisition, processing, and screening; the multi-source data includes source location data, pollution emission intensity data, and meteorological data, as well as observed pollutant concentration data in satellite remote sensing data;

[0013] The model construction and data generation unit is used for constructing a pollutant diffusion model, inputting the source location data, pollution emission intensity data, and real-time meteorological data into the pollutant diffusion model, and generating pollutant concentration distribution data;

[0014] The adjustment and heat map generation unit is used for inputting multi-source data by means of numerical simulation, dynamically adjusting the diffusion coefficient and attenuation rate of the pollutant diffusion model, and generating a heat map of pollutant concentration distribution;

[0015] The matching and iteration unit is used for spatially matching the heat map of pollutant concentration distribution with the observed pollutant concentration data, and iteratively updating the diffusion coefficient and attenuation rate of the pollutant diffusion model until the matching parameters meet the set conditions;

[0016] The fusion and verification unit is used for constructing a deep learning model to fuse multi-source data to obtain a source probability distribution model, using an unmanned terminal to scan the target area corresponding to the heat map of pollutant concentration distribution, calculating the source probability distribution data using the source probability distribution model, and verifying the pollutant concentration distribution data in the heat map of pollutant concentration distribution.

[0017] Based on the above technical solutions, the present invention can also be improved as follows.

[0018] Further, collect multi-source data through hyperspectral satellites, geostationary meteorological satellites, and ground lidar; the real-time meteorological data includes real-time wind speed data, real-time humidity data, and real-time temperature data.

[0019] Furthermore, the multi-source data is processed, including: performing spatial smoothing on the pollutant concentration observation data and the multi-source data, calculating the signal-to-noise ratio, setting a signal-to-noise ratio threshold, and filtering the pollutant concentration observation data and the multi-source data according to the magnitude relationship between the signal-to-noise ratio and the signal-to-noise ratio threshold.

[0020] Furthermore, the multi-source data is screened, including: using the Z-test method based on satellite remote sensing data to determine the area where the pollutant emission index is greater than the set threshold as the target area.

[0021] Furthermore, the numerical simulation method is to perform simulation using the Gaussian diffusion model or the Lagrangian particle model.

[0022] Furthermore, the matching parameter is to minimize the root mean square error or the spatial similarity coefficient; performing spatial matching between the pollutant concentration distribution heat map and the pollutant concentration observation data, including: aligning the grids of the pollutant concentration distribution heat map with the grids of the pollutant concentration observation data, and calculating the root mean square error or the spatial similarity coefficient of the values at the same geographical location; setting the condition as the minimum root mean square error or the spatial similarity coefficient being greater than the set threshold.

[0023] Furthermore, the spatial similarity coefficient is the Pearson correlation coefficient; let X i be the concentration value of the heat map, be the average value of the concentration values of the heat map, Y i be the satellite observation value, be the average value of the satellite observation values, and the Pearson correlation coefficient be R, then:

[0024]

[0025] Furthermore, a deep learning model is constructed to fuse the multi-source data to obtain a pollution source probability distribution model. An unmanned terminal is used to scan the target area corresponding to the pollutant concentration distribution heat map, and the pollution source probability distribution data is calculated using the pollution source probability distribution model to verify the pollutant concentration distribution data in the pollutant concentration distribution heat map, including:

[0026] Constructing a sample data set containing multi-source data;

[0027] Unifying the multi-source data to the same time resolution and spatial grid and performing feature extraction. After feature splicing, spatio-temporal feature parameters are obtained, and a spatio-temporal feature parameter data set is constructed. The spatio-temporal feature parameter data set is divided into a training set and a validation set;

[0028] Presetting the parameters of the deep learning model, inputting the spatio-temporal feature parameters in the training set into the deep learning model in chronological order, and using the deep learning model to calculate the probability of the occurrence of a pollution source at the next time node;

[0029] Update the parameters of the deep learning model by optimizing the loss function through gradient descent;

[0030] Calculate the performance metrics of the deep learning model using the validation set of the spatio-temporal feature parameter dataset, so that the performance metrics of the deep learning model meet the set values, and obtain the pollution source probability distribution model;

[0031] Use an unmanned terminal to scan the target area corresponding to the pollutant concentration distribution heat map to obtain multi-source data and perform feature extraction, and input the pollution source probability distribution model to obtain the pollution source probability distribution data to verify the pollutant concentration distribution data in the pollutant concentration distribution heat map.

[0032] Further, the pollutant diffusion model is a Gaussian plume model or a Lagrangian particle model.

[0033] The beneficial effects of the present invention are: The present invention can dynamically adjust and optimize the parameters of the pollutant diffusion model based on real-time meteorological data, ensure the accuracy of the pollutant concentration distribution data generated by the pollutant diffusion model, can reflect the diffusion situation of atmospheric pollutants in real time, and realize the real-time tracking and precise positioning of pollution sources; The present invention adopts a dynamic pollutant diffusion model to reduce the pollution source positioning error from the kilometer level to the hundred-meter level, significantly improving the monitoring accuracy; The multi-source data collaboration shortens the response time from 24 hours to 4 hours, can monitor multiple pollutant parameters at the same time, including volatile organic compounds that are difficult to detect by traditional methods, overcomes the problem that single satellite data cannot cover multiple pollutant types, meets the requirements of real-time monitoring, improves the collaborative monitoring efficiency, and realizes the comprehensive monitoring of atmospheric pollutants. Description of the Drawings

[0034] Figure 1 It is the schematic diagram of the multi-source satellite remote sensing collaborative air pollution monitoring and dynamic tracing method provided by Embodiment 1 of the present invention;

[0035] Figure 2 It is the principle flow chart for constructing the pollution source probability distribution model;

[0036] Figure 3 It is the system block diagram of the multi-source satellite remote sensing collaborative air pollution monitoring and dynamic tracing system provided by Embodiment 2 of the present invention. Detailed Embodiments

[0037] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Usually, the components of the embodiments of the present invention described and illustrated herein can be arranged and designed in various different configurations.

[0038] Embodiment 1

[0039] As an embodiment, as shown in the appendix Figure 1 To solve the above technical problems, this embodiment provides a multi-source satellite remote sensing collaborative air pollution monitoring and dynamic source tracing method, including:

[0040] Multi-source data collection, processing and screening; the multi-source data includes pollution source location data, pollution emission intensity data and meteorological data, as well as pollutant concentration observation data in satellite remote sensing data;

[0041] Construct a pollutant diffusion model, input the pollution source location data, pollution emission intensity data and real-time meteorological data into the pollutant diffusion model to generate pollutant concentration distribution data;

[0042] Adopt a numerical simulation method to input multi-source data, dynamically adjust the diffusion coefficient and attenuation rate of the pollutant diffusion model to generate a pollutant concentration distribution heat map;

[0043] Perform spatial matching on the pollutant concentration distribution heat map and the pollutant concentration observation data, and iteratively update the diffusion coefficient and attenuation rate of the pollutant diffusion model until the matching parameters meet the set conditions;

[0044] Construct a deep learning model to fuse multi-source data to obtain a pollution source probability distribution model, use an unmanned terminal to scan the target area corresponding to the pollutant concentration distribution heat map, calculate the pollution source probability distribution data using the pollution source probability distribution model, and verify the pollutant concentration distribution data in the pollutant concentration distribution heat map.

[0045] Integrate hyperspectral satellites (such as Sentinel-5P), geostationary meteorological satellites (such as FY-4A) and ground lidar data to achieve synchronous monitoring of various pollutants such as PM2.5, NO2 and VOCs, and improve the accuracy and timeliness of data.

[0046] Optionally, collect multi-source data through hyperspectral satellites, geostationary meteorological satellites and ground lidar; the real-time meteorological data includes real-time wind speed data, real-time humidity data and real-time temperature data.

[0047] Optionally, process the multi-source data, including: perform spatial smoothing processing on the pollutant concentration observation data and the multi-source data, calculate the signal-to-noise ratio, set the signal-to-noise ratio threshold, and perform filtering processing on the pollutant concentration observation data and the multi-source data according to the size of the signal-to-noise ratio and the signal-to-noise ratio threshold.

[0048] Performing spatial smoothing processing on satellite data and calculating the signal-to-noise ratio can filter out noises such as cloud interference and improve data quality.

[0049] Optionally, filter the multi-source data, including: using the Z-test method based on satellite remote sensing data to determine the area where the pollutant emission intensity is greater than the set threshold as the target area.

[0050] Through the statistical method of Z-test and satellite remote sensing data (such as AOD and NO2 column concentration thresholds), the emission areas with pollutant emission intensity higher than the set threshold can be quickly locked, providing reliable multi-source data of the target area for the pollutant diffusion model. At the same time, it is beneficial to reduce the calculation amount, avoid high-precision modeling of the entire area, improve the efficiency and model accuracy. The screened pollutant source location data can be used as the input parameters of the pollutant diffusion model, directly affecting the simulation of the pollutant diffusion path. Taking the screening results (pollutant source location data, pollution emission intensity data and real-time meteorological data) as the input parameters of the pollutant diffusion model, the pollutant diffusion model is driven by combining meteorological data to generate pollutant concentration distribution data, realizing the dynamic simulation from the pollutant source to the diffusion path.

[0051] The real-time wind direction determines the diffusion direction, and the temperature affects the pollutant decay rate. Adjust the diffusion rate according to the pollutant type (such as PM2.5 and SO2, etc.).

[0052] Optionally, the numerical simulation method is to use the Gaussian diffusion model or the Lagrangian particle model for simulation.

[0053] Optionally, the matching parameter is to minimize the root mean square error or the spatial similarity coefficient; perform spatial matching between the pollutant concentration distribution heat map and the pollutant concentration observation data, including: aligning the grids of the pollutant concentration distribution heat map with the grids of the pollutant concentration observation data, and calculating the root mean square error or the spatial similarity coefficient of the values at the same geographical location; the set condition is that the root mean square error is the smallest or the spatial similarity coefficient is greater than the set threshold.

[0054] Optionally, the spatial similarity coefficient is the Pearson correlation coefficient; let X i be the concentration value of the heat map, be the average value of the concentration values of the heat map, Y i be the satellite observation value, be the average value of the satellite observation values, and the Pearson correlation coefficient is R, then:

[0055]

[0056] The spatial similarity coefficient is used to quantify the matching degree between the pollutant concentration distribution heat map and the pollutant concentration observation data. Generally, the spatial similarity coefficient is greater than or equal to 0.85.

[0057] Optionally, as shown in the appendix Figure 2As shown in the figure, a deep learning model is constructed to fuse multi-source data to obtain a pollution source probability distribution model. An unmanned terminal is used to scan the target area corresponding to the pollutant concentration distribution heat map, and the pollution source probability distribution data is calculated using the pollution source probability distribution model to verify the pollutant concentration distribution data in the pollutant concentration distribution heat map, including:

[0058] Construct a sample data set containing multi-source data;

[0059] Unify the multi-source data to the same time resolution and spatial grid and perform feature extraction. After feature splicing, spatio-temporal feature parameters are obtained, a spatio-temporal feature parameter data set is constructed, and the spatio-temporal feature parameter data set is divided into a training set and a validation set;

[0060] Preset the parameters of the deep learning model, input the spatio-temporal feature parameters in the training set into the deep learning model in chronological order, and use the deep learning model to calculate the probability of the appearance of a pollution source at the next time node;

[0061] Update the parameters of the deep learning model by optimizing the loss function through gradient descent;

[0062] Use the validation set of the spatio-temporal feature parameter data set to calculate the performance metrics of the deep learning model, so that the performance metrics of the deep learning model meet the set values, and obtain the pollution source probability distribution model;

[0063] An unmanned terminal is used to scan the target area corresponding to the pollutant concentration distribution heat map to obtain multi-source data and perform feature extraction, and the pollution source probability distribution data is input into the pollution source probability distribution model to verify the pollutant concentration distribution data in the pollutant concentration distribution heat map.

[0064] Use the deep learning model to fuse multi-source data to generate a pollution source probability distribution map. Combining the unmanned terminal inspection can achieve meter-level resolution verification of high-probability areas and realize the precise positioning of pollution sources. Deep learning models include recurrent convolutional neural network models and random forest models, etc.

[0065] Optionally, the pollutant diffusion model is a Gaussian plume model or a Lagrangian particle model.

[0066] The pollutant diffusion model can reflect the diffusion of atmospheric pollutants in real time and provide a reliable basis for pollution source positioning.

[0067] The present invention can dynamically adjust and optimize the parameters of the pollutant diffusion model based on real-time meteorological data, ensure the accuracy of the pollutant concentration distribution data generated by the pollutant diffusion model, can reflect the diffusion of atmospheric pollutants in real time, and realize the real-time tracking and precise positioning of pollution sources.

[0068] The present invention adopts a dynamic pollutant diffusion model to reduce the source location error of pollutants from the kilometer level to the hundred-meter level, significantly improving the monitoring accuracy; the multi-source data collaboration shortens the response time from 24 hours to 4 hours, can simultaneously monitor multiple pollutant parameters, including volatile organic compounds that are difficult to detect by traditional methods, overcomes the problem that single satellite data cannot cover multiple pollutant types, meets the requirements of real-time monitoring, improves the collaborative monitoring efficiency, and realizes the comprehensive monitoring of atmospheric pollutants.

[0069] Embodiment 2

[0070] Based on the same principle as the method shown in Embodiment 1 of the present invention, as shown in the appendix Figure 3 In the embodiments of the present invention, a multi-source satellite remote sensing collaborative air pollution monitoring and dynamic tracing system, a data acquisition and processing unit, a model construction and data generation unit, an adjustment and heat map generation unit, a matching and iteration unit, and a fusion and verification unit are also provided;

[0071] The data acquisition and processing unit is used for multi-source data acquisition, processing and screening; the multi-source data includes source location data of pollution sources, pollution emission intensity data and meteorological data, as well as pollutant concentration observation data in satellite remote sensing data;

[0072] The model construction and data generation unit is used for constructing a pollutant diffusion model, inputting the source location data of pollution sources, pollution emission intensity data and real-time meteorological data into the pollutant diffusion model, and generating pollutant concentration distribution data;

[0073] The adjustment and heat map generation unit is used for inputting multi-source data by means of numerical simulation, dynamically adjusting the diffusion coefficient and attenuation rate of the pollutant diffusion model, and generating a pollutant concentration distribution heat map;

[0074] The matching and iteration unit is used for spatially matching the pollutant concentration distribution heat map with the pollutant concentration observation data, and iteratively updating the diffusion coefficient and attenuation rate of the pollutant diffusion model until the matching parameters meet the set conditions;

[0075] The fusion and verification unit is used for constructing a deep learning model to fuse multi-source data to obtain a source probability distribution model, using an unmanned terminal to scan the target area corresponding to the pollutant concentration distribution heat map, calculating the source probability distribution data by using the source probability distribution model, and verifying the pollutant concentration distribution data in the pollutant concentration distribution heat map.

[0076] Optionally, multi-source data is collected by a hyperspectral satellite, a geostationary meteorological satellite and a ground lidar; the real-time meteorological data includes real-time wind speed data, real-time humidity data and real-time temperature data.

[0077] Optionally, process multi-source data, including: performing spatial smoothing on the pollutant concentration observation data and the multi-source data, calculating the signal-to-noise ratio, setting a signal-to-noise ratio threshold, and filtering the pollutant concentration observation data and the multi-source data according to the magnitude relationship between the signal-to-noise ratio and the signal-to-noise ratio threshold.

[0078] Optionally, screen the multi-source data, including: using the Z-test method based on satellite remote sensing data to determine the area where the pollutant emission index is greater than the set threshold as the target area.

[0079] Optionally, the numerical simulation method is to use the Gaussian diffusion model or the Lagrangian particle model for simulation.

[0080] Optionally, the matching parameter is the root mean square error or the spatial similarity coefficient minimized; perform spatial matching between the pollutant concentration distribution heat map and the pollutant concentration observation data, including: aligning the grids of the pollutant concentration distribution heat map with the grids of the pollutant concentration observation data, and calculating the root mean square error or the spatial similarity coefficient of the values at the same geographical location; the set condition is that the root mean square error is the smallest or the spatial similarity coefficient is greater than the set threshold.

[0081] Optionally, the spatial similarity coefficient is the Pearson correlation coefficient; let X i be the concentration value of the heat map, be the average value of the concentration values of the heat map, Y i be the satellite observation value, be the average value of the satellite observation values, and the Pearson correlation coefficient be R, then:

[0082]

[0083] Optionally, construct a deep learning model to fuse multi-source data to obtain a pollution source probability distribution model, use an unmanned terminal to scan the target area corresponding to the pollutant concentration distribution heat map, calculate the pollution source probability distribution data using the pollution source probability distribution model, and verify the pollutant concentration distribution data in the pollutant concentration distribution heat map, including:

[0084] Construct a sample data set containing multi-source data;

[0085] Unify the multi-source data to the same time resolution and spatial grid and perform feature extraction. After feature splicing, obtain spatio-temporal feature parameters, construct a spatio-temporal feature parameter data set, and divide the spatio-temporal feature parameter data set into a training set and a validation set;

[0086] Preset the parameters of the deep learning model, input the spatio-temporal feature parameters in the training set into the deep learning model in chronological order, and use the deep learning model to calculate the probability of the occurrence of a pollution source at the next time node;

[0087] Updating the parameters of the deep learning model by optimizing the loss function through gradient descent;

[0088] Calculating the performance metrics of the deep learning model using the validation set of the spatio-temporal feature parameter dataset, so that the performance metrics of the deep learning model meet the set values, and obtaining the pollution source probability distribution model;

[0089] Using an unmanned terminal to scan the target area corresponding to the pollutant concentration distribution heat map to obtain multi-source data and perform feature extraction, and inputting the pollution source probability distribution model to obtain the pollution source probability distribution data to verify the pollutant concentration distribution data in the pollutant concentration distribution heat map.

[0090] Optionally, the pollutant diffusion model is a Gaussian plume model or a Lagrangian particle model.

[0091] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A multi-source satellite remote sensing coordinated atmospheric pollution monitoring and dynamic source tracing method, characterized in that: include: Multi-source data collection, processing and screening; Multi-source data include pollution source location data, pollution emission intensity data and meteorological data, as well as pollutant concentration observation data from satellite remote sensing data; Construct a pollutant diffusion model, input the pollution source location data, pollution emission intensity data and real-time meteorological data into the pollutant diffusion model to generate pollutant concentration distribution data; The numerical simulation method is used to input multi-source data, dynamically adjust the diffusion coefficient and attenuation rate of the pollutant diffusion model, and generate a thermal map of pollutant concentration distribution; Spatially match the pollutant concentration distribution thermodynamic map with the pollutant concentration observation data, and iteratively update the diffusion coefficient and attenuation rate of the pollutant diffusion model until the matching parameters meet the set conditions; A deep learning model is constructed to fuse multi-source data to obtain a pollution source probability distribution model. An unmanned terminal is used to scan the target area corresponding to the pollutant concentration distribution heat map. The pollution source probability distribution model is used to calculate the pollution source probability distribution data, and the pollutant concentration distribution data in the pollutant concentration distribution heat map is verified.

2. The multi-source satellite remote sensing coordinated atmospheric pollution monitoring and dynamic source tracing method according to claim 1 is characterized in that: Multi-source data is collected through hyperspectral satellites, geostationary meteorological satellites and ground-based lidar; real-time meteorological data includes real-time wind speed data, real-time humidity data and real-time temperature data.

3. The multi-source satellite remote sensing coordinated atmospheric pollution monitoring and dynamic source tracing method according to claim 1 is characterized in that: Processing multi-source data includes: spatially smoothing pollutant concentration observation data and multi-source data, calculating signal-to-noise ratio, setting signal-to-noise ratio threshold, and filtering pollutant concentration observation data and multi-source data according to the signal-to-noise ratio and the signal-to-noise ratio threshold.

4. The multi-source satellite remote sensing coordinated air pollution monitoring and dynamic source tracing method according to claim 1 is characterized in that: Screening of multi-source data includes: using the Z test method based on satellite remote sensing data to determine areas where pollutant emission indicators are greater than set thresholds as target areas.

5. The multi-source satellite remote sensing coordinated atmospheric pollution monitoring and dynamic source tracing method according to claim 1 is characterized in that: The numerical simulation method is to use Gaussian diffusion model or Lagrangian particle model for simulation.

6. The multi-source satellite remote sensing coordinated air pollution monitoring and dynamic source tracing method according to claim 1 is characterized in that: The matching parameter is to minimize the root mean square error or the spatial similarity coefficient; Spatial matching of the pollutant concentration distribution heat map with the pollutant concentration observation data includes: aligning the grid of the pollutant concentration distribution heat map with the grid of the pollutant concentration observation data, calculating the root mean square error or spatial similarity coefficient of the values ​​at the same geographical location; setting the condition that the root mean square error is minimum or the spatial similarity coefficient is greater than a set threshold.

7. The multi-source satellite remote sensing coordinated air pollution monitoring and dynamic source tracing method according to claim 6 is characterized in that: The spatial similarity coefficient is the Pearson correlation coefficient; let X i is the heat map concentration value, is the average value of the heat map concentration, Y i is the satellite observation value, is the average value of satellite observations, and the Pearson correlation coefficient is R, then:

8. The multi-source satellite remote sensing coordinated air pollution monitoring and dynamic source tracing method according to claim 6 is characterized in that: A deep learning model is constructed to fuse multi-source data to obtain a pollution source probability distribution model. An unmanned terminal is used to scan the target area corresponding to the pollutant concentration distribution heat map. The pollution source probability distribution model is used to calculate the pollution source probability distribution data. The pollutant concentration distribution data in the pollutant concentration distribution heat map is verified, including: Construct a sample dataset containing multi-source data; Unify multi-source data to the same temporal resolution and spatial grid and perform feature extraction. After feature splicing, obtain spatiotemporal feature parameters, construct a spatiotemporal feature parameter dataset, and divide the spatiotemporal feature parameter dataset into a training set and a validation set. Preset the parameters of the deep learning model, input the spatiotemporal feature parameters in the training set into the deep learning model in chronological order, and use the deep learning model to calculate the probability of the pollution source appearing at the next time node; Update the parameters of the deep learning model by optimizing the loss function through gradient descent; The performance index of the deep learning model is calculated using the validation set of the spatiotemporal characteristic parameter data set, so that the performance index of the deep learning model meets the set value and the pollution source probability distribution model is obtained; An unmanned terminal is used to scan the target area corresponding to the pollutant concentration distribution heat map to obtain multi-source data and perform feature extraction. The pollution source probability distribution model is input to obtain the pollution source probability distribution data to verify the pollutant concentration distribution data in the pollutant concentration distribution heat map.

9. The multi-source satellite remote sensing coordinated air pollution monitoring and dynamic source tracing method according to claim 1 is characterized in that: The pollutant diffusion model is a Gaussian plume model or a Lagrangian particle model.

10. Multi-source satellite remote sensing coordinated air pollution monitoring and dynamic tracing system, characterized by: It includes data acquisition and processing unit, model building and data generation unit, adjustment and heat map generation unit, matching and iteration unit, and fusion and verification unit; Data acquisition and processing unit, used for multi-source data acquisition, processing and screening; multi-source data includes pollution source location data, pollution emission intensity data and meteorological data, as well as pollutant concentration observation data from satellite remote sensing data; The model building and data generation unit is used to build a pollutant diffusion model, input the pollution source location data, pollution emission intensity data and real-time meteorological data into the pollutant diffusion model, and generate pollutant concentration distribution data; An adjustment and heat map generation unit, which is used to input multi-source data using a numerical simulation method, dynamically adjust the diffusion coefficient and decay rate of the pollutant diffusion model, and generate a pollutant concentration distribution heat map; The matching and iteration unit is used to spatially match the pollutant concentration distribution thermodynamic map with the pollutant concentration observation data, and iteratively update the diffusion coefficient and attenuation rate of the pollutant diffusion model until the matching parameters meet the set conditions; The fusion and verification unit is used to build a deep learning model to fuse multi-source data to obtain a pollution source probability distribution model, use an unmanned terminal to scan the target area corresponding to the pollutant concentration distribution heat map, use the pollution source probability distribution model to calculate the pollution source probability distribution data, and verify the pollutant concentration distribution data in the pollutant concentration distribution heat map.

Images