A method for assessing and analyzing the impact of key source emissions on population health.

Abstract

This invention discloses a method for assessing and analyzing the impact of key source emissions on population health, including the following steps: (1) establishing a key source emission characteristic database and clarifying the list of key sources; (2) establishing an exposed population characteristic database and generating a spatial distribution map of exposure scenarios; (3) a health risk assessment method to quantify the health risks caused by pollutants to different population groups in different exposure scenarios; and (4) assessment, analysis, prediction, simulation, and decision-making to identify key prevention and control areas of "high pollution emissions + highly sensitive populations," conduct differentiated scenario prediction and simulation, and propose targeted action recommendations. This invention establishes a full-chain correlation analysis method of "key source emissions - environmental concentration - exposed populations - health effects," identifies key prevention and control areas of "high pollution emissions + highly sensitive populations," and realizes quantitative assessment, prediction, and precise prevention and control of health risks from key source emissions, providing a scientific basis for "targeted management" and "protection of sensitive populations" decisions.

Description

Technical Field

[0001] This invention relates to the field of atmospheric environmental protection technology, specifically to a method for assessing and analyzing the impact of key source emissions on human health. Background Technology

[0002] With the acceleration of urbanization and industrialization, air pollution has spread from traditional industrial areas to urban centers, suburbs, and surrounding rural areas, exposing people of all ages and exhibiting characteristics of "high exposure and long-term duration." The pollutants emitted not only cause environmental problems such as reduced visibility and frequent acid rain, but also enter the human body through inhalation and skin contact, causing respiratory diseases, cardiovascular diseases, and even increasing the risk of chronic diseases such as cancer. Air pollution has become one of the major environmental problems threatening the ecological environment and human health.

[0003] Existing research largely focuses on the spatiotemporal distribution of common pollutant concentrations, source apportionment, and Air Quality Index (AQI) assessment. While these studies can reflect ambient air quality, they remain at the level of "pollution degree description" and do not delve into the relationship between characteristic pollutant emissions and human health effects. This often leads to the identification of uninhabited areas with high pollutant concentrations as high-risk areas, easily overlooking the pollution exposure risks in areas with concentrated sensitive populations, and neglecting the impact of pollution source emissions, especially industrial sources, on the environmental health risks of exposed populations.

[0004] Therefore, it is necessary to establish a full-chain correlation analysis method of "key source emissions - environmental concentration - exposed population - health effects" to identify the risk superposition area of ​​"high pollution emissions + highly sensitive population" and then formulate targeted prevention and control strategies. Summary of the Invention

[0005] The purpose of this application is to provide a method for assessing and analyzing the impact of key source emissions that incorporates population health effects, and to establish a full-chain analysis framework of "key source emissions - environmental concentration - exposed population - health effects" to achieve quantitative assessment, prediction and precise prevention and control of health risks from key source emissions.

[0006] This invention breaks down the data barriers between pollution sources and exposed populations by establishing a database of key source emission characteristics and a database of exposed population characteristics. It calculates the actual exposure dose to the population by combining a pollutant diffusion model, and then quantifies and assesses the health risks of pollution through a population health effect model, and predicts risk changes under different scenarios, providing a scientific basis for decision-making on "targeted management" and "protection of sensitive populations".

[0007] Furthermore, embodiments of the present invention provide a method for assessing and analyzing the impact of key source emissions incorporating population health effects, specifically including the following steps:

[0008] Step 1: Establishment of the Key Source Emission Characteristic Database: Through on-site investigation, document review, questionnaires, and data collection, combined with regional pollution characteristics and industry attributes, pollution sources with a pollution contribution rate higher than 10% and pollution sources from high-emission key industries in the "List of Key Controlled Industries" are included in the candidate list. Through cross-comparison, pollution sources with a pollution contribution rate that meets the standards but do not belong to key industries are eliminated, and the list of key sources is clearly defined. The entire chain of information on key sources, from "basic attributes" to "emission details", is obtained to construct the emission profile of key sources.

[0009] Furthermore, the core characteristic indicators mentioned in step 1 mainly cover basic attribute information of key sources, pollutant emission information, source emission mode and parameter information, and dynamic change information.

[0010] Step 2, Establishment of the Exposed Population Feature Database: Acquire remote sensing imagery data, identify building types and land use types based on deep learning models, and delineate potential exposure scenarios and regional boundaries. Combining data sources such as on-site surveys, questionnaires, and document review, obtain core characteristic information of the exposed population within the region, and generate a spatial distribution map of exposure scenarios using GIS maps.

[0011] Furthermore, the building types mentioned in step 2 should cover industrial buildings, commercial buildings, residences, schools, hospitals, nursing homes, public facilities, and other facilities, and the land use types should cover industrial land, agricultural land, forest land, construction land, water areas, and other land.

[0012] Furthermore, the deep learning model described in step 2 adopts a dual-branch U-Net model, which achieves a balance between global semantics and local details through an encoder-decoder architecture and skip connections, thereby enabling the delineation of potential exposure scenarios and region boundaries.

[0013] Furthermore, the core characteristic information of the exposed population mentioned in step 2 mainly includes the classification of the exposed population, its quantity and structural characteristics, and its spatial behavior characteristics.

[0014] Step 3, Health Risk Assessment Method: Using standardized source parameters extracted from the key source emission feature database as input, high-precision meteorological field data are integrated, and the spatial gridded concentration distribution of pollutants is simulated through pollutant diffusion models; the spatial distribution map of exposure scenarios is integrated, the exposure parameters of the corresponding populations in each scenario are identified, and the parameters are substituted into the health effect model to calculate the carcinogenic and non-carcinogenic health risk values ​​of pollutants caused by different exposure pathways to each exposure scenario and each population group.

[0015] Furthermore, the pollutant diffusion model simulation described in step 3 specifically includes the following steps:

[0016] The S1 Weather Research and Forecasting (WRF) model preprocessing delineates the simulation range centered on key source concentration areas. A three-layer nested grid design is used for grid configuration, with spatial resolutions of 10km×10km, 5km×5km, and 1km×1km. External data such as topographic elevation data, land use type data, and reanalysis meteorological data are input. Interpolation is used to interpolate the external data to each level of the grid. Meteorological field data is extracted and horizontally interpolated to each nested grid within the simulation range. The WRF model preprocessing system converts the data into initial field and boundary condition files in a specified format. Finally, the main program of the model is run to output the spatiotemporal meteorological field data required for subsequent pollutant diffusion simulation.

[0017] The S2 pollutant diffusion model simulation includes AERMOD, CALPUFF, and Flexpart models. The AERMOD model is used for short-distance near-surface diffusion, the CALPUFF model is used for long-distance cross-regional diffusion, and the Flexpart model is used for sudden emergency emission diffusion. The standardized source parameters extracted from the key source emission feature database are used as input, combined with high-precision spatiotemporal meteorological field data output by WRF. Through pollutant diffusion model simulation, the final output is spatially gridded pollutant concentration data.

[0018] Furthermore, the health effect model described in step 3 quantitatively assesses health risks, specifically including the following steps:

[0019] The S1 health effect assessment method uses the human health risk model recommended by the U.S. Environmental Protection Agency (EPA) to calculate the carcinogenic and non-carcinogenic health risks of key source emissions pollutants to different population groups through inhalation and skin contact exposure.

[0020] The daily average exposure via inhalation and skin contact is calculated as follows:

[0021]

[0022]

[0023] Among them, ADD inh This represents the average daily exposure to pollutant i via inhalation; ADD dermal C represents the average daily exposure to contaminant i via skin contact; i 1 represents the environmental concentration of pollutant i; IR represents the respiration rate; ET represents the exposure duration; EF represents the exposure frequency; ED represents the exposure years; BW represents body weight; AT represents the average exposure time; SA represents the skin contact area; Kp represents the skin permeability coefficient; and i represents the pollutant.

[0024] The non-carcinogenic risk calculation method for inhalation and skin contact is as follows:

[0025]

[0026]

[0027] Among them, HQ inh HQ dermal These are the non-carcinogenic risks of pollutant i via inhalation and skin contact; ADD inh This represents the average daily exposure to pollutant i via inhalation; ADD dermal RfC represents the average daily exposure to contaminant i via skin contact. i Reference concentration for inhalation of pollutant i; RfD i , where i is the skin reference dose of contaminant i; i is the contaminant.

[0028] The method for calculating the carcinogenic risk through inhalation and skin contact is as follows:

[0029]

[0030]

[0031] Among them, ELCR inh ELCR dermal The carcinogenic risks of pollutant i via inhalation and skin contact are respectively; ADD inh This represents the average daily exposure to pollutant i via inhalation; ADD dermal IUR represents the average daily exposure to contaminant i via skin contact. i Risk of inhalation of pollutant i per unit; SF i Let be the risk slope factor for pollutant i; where i is the pollutant.

[0032] S2 exposure parameters are determined by identifying population characteristics, potential pollutants, exposure source intensity, exposure pathways, and exposure cycles for each group, based on the established key source emission characteristic database and exposed population characteristic database.

[0033] The S3 toxicity parameters were determined based on the Integrated Risk Information System database of the U.S. Environmental Protection Agency (EPA), the World Health Organization (WHO) toxicology database, and the toxicity parameter database of the Chinese Academy of Environmental Sciences, obtaining the corresponding parameter data of IUR (inhalation unit risk), SF (slope factor), RfC (inhalation reference concentration), and RfD (dermal reference dose).

[0034] S4 health risk quantification: The non-carcinogenic risk (HQ) of a single pathway reflects the degree of harm of that pathway. If HQ < 1, the risk is acceptable; if HQ ≥ 1, a non-carcinogenic health hazard may exist. The total hazard from combined exposure via multiple pathways / contaminants is the sum of the individual HQs, i.e., HI = ∑HQi. If HI ≥ 1, the cumulative effect of combined exposure needs to be considered. The total carcinogenic risk (ELCR) is calculated as ∑ELCR. i The maximum acceptable level of 1×10⁻⁶ recommended by the Dutch Ministry of the Environment and the Swedish Environmental Protection Agency was adopted. -6 As a basis for health risk assessment.

[0035] S5 uncertainty analysis employs a combination of sensitivity analysis and Monte Carlo simulation to assess the quality of the results.

[0036] Step 4, Assessment, Analysis, Prediction, and Simulation Decision Making: Based on the established health risk assessment method, key prevention and control areas with "high pollution emissions + highly sensitive populations" are identified through current status assessment; differentiated scenarios are formulated for key prevention and control areas, and the health effects of each scenario are compared through prediction and simulation; finally, the optimal solution is selected from the simulation results of each scenario, and it is transformed into specific control policies, and the implementation effect is verified through subsequent core indicator assessment and analysis.

[0037] Furthermore, the implementation effect verification process described in step 4 firstly clarifies the core verification indicators, using the compliance rate of key source emission concentrations in key prevention and control areas, the reduction rate of carcinogenic / non-carcinogenic risks, and the reduction rate of the proportion of high-risk populations as core indicators. Secondly, it quantifies the judgment criteria, referring to the US EPA health risk standards, the "Technical Guidelines for Environmental Health Risk Assessment," and the pre-set targets of my country's policies, and sets corresponding thresholds for each indicator. Finally, it conducts indicator verification and effect judgment. If all core indicators meet the standards, the implementation effect is judged to be good, and the entire control plan is archived as experience for promotion. If any indicator fails to meet the standards, the implementation effect is judged to be poor, the data sources of each input module are checked, the scenario plan measures are adjusted accordingly, and a second verification is conducted until the control objectives are met.

[0038] In summary, the present invention has the following effects:

[0039] This invention provides a method for assessing and analyzing the impact of key source emissions on population health. By establishing a key source emission characteristic database and an exposed population characteristic database, the data barriers between pollution sources and exposed populations are broken down. By effectively combining key source pollution emission diffusion simulation with exposed population health effect models, a full-chain correlation analysis method of "key source emissions - environmental concentration - exposed population - health effect" is built. This method identifies key prevention and control areas of "high pollution emissions + highly sensitive populations", enabling quantitative assessment, prediction, and precise prevention and control of health risks from key source emissions, and providing a scientific basis for decision-making such as "targeted management" and "protection of sensitive populations". Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of this application 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 some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 The flowchart below shows the overall process of a method for assessing and analyzing the impact of key source emissions on population health, as provided in an embodiment of the present invention.

[0042] Figure 2 A flowchart illustrating the steps for establishing a key source emission characteristic library according to an embodiment of the present invention.

[0043] Figure 3 A flowchart illustrating the steps for establishing the exposed population feature database according to an embodiment of the present invention.

[0044] Figure 4 This is a flowchart of the health risk assessment method according to an embodiment of the present invention. Detailed Implementation

[0045] 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.

[0046] The specific implementation steps of the present invention will be described in detail below with reference to specific embodiments:

[0047] like Figures 1-4 The diagram illustrates the overall process of a key source emission impact assessment and analysis method incorporating population health effects, provided by an embodiment of the present invention. The process mainly includes the following steps:

[0048] Step 1: Establishment of the Key Source Emission Characteristic Database: Through on-site investigation, document review, questionnaires, and data collection, combined with regional pollution characteristics and industry attributes, pollution sources with a pollution contribution rate higher than 10% and pollution sources in high-emission key industries listed in the "Key Controlled Industry Directory" are included in the candidate list. Through cross-comparison, pollution sources with a pollution contribution rate that meets the standards but do not belong to key industries are eliminated, and the list of key sources is clarified. The entire chain of information on key sources, from "basic attributes" to "emission details", is obtained to construct the emission profile of key sources. The core characteristic indicators mainly cover the basic attribute information of key sources, pollutant emission information, source emission mode and parameter information, and dynamic change information.

[0049] Furthermore, the basic attribute information of key sources in step 1 should include the name of the key source, unique identifier, geographical location (latitude and longitude), industry, administrative region, detailed address, and discharge permit, serving as the identity identifier for the emission profile, used to locate key sources, and associate spatial and management attributes.

[0050] Furthermore, the pollutant emission information in step 1 should include the production and discharge stages, types of pollutants emitted, emission concentrations, emission volumes, pollutant forms, and toxicity parameters, serving as core input data for subsequent health risk assessments. Specifically, the types of emitted pollutants should clearly distinguish between conventional pollutants and characteristic pollutants. Characteristic pollutants generally exhibit strong health-damaging effects, severe environmental pollution, and a wide range of impacts on the population, making the excess health risks they pose to exposed individuals particularly noteworthy.

[0051] Furthermore, the source emission methods and parameter information in step 1 should include emission form, emission period, and emission pattern to support the simulation of key source pollutant emission diffusion and improve the accuracy of exposure concentration calculation. Specifically, emission form should include both organized and unorganized emissions. Organized emissions are considered point sources, requiring the collection of parameters such as outlet height, outlet inner diameter, flue gas temperature, and flue gas velocity. Unorganized emissions are considered area sources, requiring the collection of parameters such as release height, length, width, and the angle to due north.

[0052] Furthermore, the dynamic change information in step 1 should include changes in production capacity, process adjustments, emission reduction measures, and corresponding changes in emissions, in order to support scenario prediction and simulation decision-making.

[0053] Step 2, Establishment of the Exposed Population Feature Database: Acquire remote sensing imagery of the target area, automatically identify building types and land use types based on a deep learning model, and initially delineate potential exposure scenarios and corresponding regional boundaries. Combining data sources such as on-site surveys, questionnaires, and literature reviews, obtain core characteristic information of the exposed population within the region, mainly including the classification, quantity and structural characteristics, and spatial behavioral characteristics of the exposed population. Combine this with Geographic Information System (GIS) map overlay to generate a spatial distribution map of the exposure scenarios.

[0054] Furthermore, in step 2, the building types should cover industrial buildings, commercial buildings, residences, schools, hospitals, nursing homes, public facilities, and other facilities, and the land use types should cover industrial land, agricultural land, forest land, construction land, water areas, and other land.

[0055] Furthermore, the delineation of potential exposure scenarios and region boundaries in step 2 is mainly based on the U-Net model. A balance between global semantic information and local detail information is achieved through an encoder-decoder architecture and skip connections. The encoder is responsible for downsampling to extract multi-scale features, the decoder is responsible for upsampling to restore image resolution, and skip connections pass shallow features from the encoder to the decoder. Specifically, this includes the following steps:

[0056] Step S1 involves constructing the training set. The training set is built using remote sensing images of the target area as input, and preprocessed through radiometric normalization, band fusion, and data augmentation. Radiometric normalization employs dark target subtraction to eliminate atmospheric scattering interference, making the texture features of buildings and land boundaries clearer. Band fusion uses hue-saturation-brightness transformation to fuse multispectral and panchromatic bands, enriching the image's spectral information and helping to distinguish similar-looking but different land features. Data augmentation expands the training sample size through random rotation, horizontal flipping, and brightness perturbation, avoiding model overfitting and ensuring stable, high-precision recognition results.

[0057] Step S2 involves a dual-branch U-Net architecture design, employing an improved U-Net model architecture of "shared encoder + dual-task decoder." First, the pre-processed remote sensing image is input. The shared encoder uses a residual network structure, extracting multi-scale features through multi-scale convolution and downsampling. Initial feature mapping is performed using 3×3 convolution kernels, and feature extraction is achieved through four sets of residual blocks. After each residual block, 2×2 max pooling is used to halve the feature map size. Simultaneously, the multi-scale features output from each level (shallow features store edges / texture, deep features store semantics) are synchronously transmitted to the decoder via skip connections, enabling the simultaneous extraction of common features of buildings and land use types. First, to avoid redundant feature extraction; second, the dual decoders are optimized for different task characteristics. The building decoder uses "3×3 device convolution upsampling" and the land use decoder uses "bilinear interpolation and 1×1 convolution to adjust channel upsampling" to restore the feature map size. After each upsampling stage, feature alignment, channel compression, and element-wise addition are used to fuse the features of the i-th layer of the encoder and the i-th layer of the decoder to complete the skip connection. Then, an attention mechanism is embedded to output the two classification masks respectively, solving the problem of loss of detail information in the traditional downsampling process. Finally, the error of the two classification tasks is optimized simultaneously by combining the multi-task loss function to avoid misjudgment of the range due to blurred boundaries.

[0058] Furthermore, the attention mechanism in step S2 is tailored to the differences in task characteristics by embedding channel attention and spatial attention mechanisms respectively. A channel attention module is embedded in the building decoder, performing global average pooling on the feature map of a certain level of the decoder, compressing the spatial information of each channel into a channel descriptor, learning channel weights to generate attention coefficients, and multiplying the attention coefficients element-wise with the channels of the original feature map to strengthen the channel response of the building's iconic features. A spatial attention module is embedded in the land use decoder, performing max pooling and average pooling on the channel dimension of the feature map of a certain level of the decoder to generate a spatial weight map, which is then multiplied pixel-wise with the original feature map to strengthen the consistency features within land use types. Through dynamic weight allocation, it adaptively focuses on key feature regions, eliminating the need to retrain the model for different regions, while suppressing irrelevant noise and solving the problem of pixel misjudgment at region edges.

[0059] Furthermore, the multi-task loss function in step S2 adopts a "weighted fusion" architecture for task-differentiated loss design and dynamic weight allocation, and the formula is defined as follows:

[0060] Loss = α×Loss1 + (1-α)×Loss2

[0061] Where Loss is the total task loss, Loss1 is the building type classification loss, Loss2 is the land use type classification loss, and α∈[0.5,0.7] is the weight coefficient used to balance the priorities of the two types of tasks.

[0062] The building type classification loss, Loss1, uses weighted cross-entropy loss, defined by the formula:

[0063]

[0064] Where N is the total number of pixels i in the remote sensing image, M is the total number of building category numbers c, and W c Y represents the weight of building category. i,c P is the true label for pixel i belonging to category c. i,c To predict the probability that pixel i belongs to category c, where c is the building category number and i is a pixel in the remote sensing image.

[0065] Land use type classification loss Loss2 adopts Dice loss, which is defined by the formula:

[0066]

[0067] Where N is the total number of pixels i in the remote sensing image, M is the total number of land use type numbers k, and Y i,k For pixel i to be the true label of type k, P i,kTo predict the probability that pixel i belongs to type k, ε is a smoothing term used to avoid cases where the denominator is 0, k is the land use type number, and i is a pixel in the remote sensing image.

[0068] Step S3, post-processing of the inference stage, introduces a Conditional Random Field (CRF) to refine the boundaries and uses the correlation between pixels to correct the boundary ambiguity problem of the initial classification mask; then, it integrates point of interest (POI) data and land use planning vector map to correct misclassification of the classification results after CRF processing, and finally achieves high-precision classification of buildings and land use types.

[0069] Step S3, which introduces a Conditional Random Field (CRF) to refine the boundaries, is a post-processing module in the U-Net model's inference stage. It compensates for the model's shortcomings in local detail recognition and enhances the spatial correlation between pixels to improve the reasonableness of the classification results. The initial classification results are used as the base label input, while remote sensing imagery data is introduced as variables. Label dependencies are constructed by defining an energy function, which mainly includes unary and binary potential terms. The unary potential term retains the high-confidence classification results from the initial U-Net predictions, ensuring the accuracy of core region classification. The binary potential term establishes class relationships between pixels, constraining the class consistency of adjacent pixels. By minimizing the energy function to solve for the globally optimal label configuration, a classification mask with clear boundaries and continuous classes can be output.

[0070] The Points of Interest (POI) data in step S3 is crucial spatial data for correcting classification errors. It primarily consists of: spatial location information, including latitude and longitude coordinates and the administrative region it belongs to, ensuring coordinate matching with remote sensing imagery and GIS maps; attribute information, including the identifier, name, classification label, and detailed address of the geographic entity, used to clarify the function of the geographic entity; and extended attribute information, including entity size parameters and feature markers, which can further improve the accuracy of the correction. The main correction process is as follows: first, coordinate transformation, duplicate entity deduplication, and classification label standardization are performed on the multi-source POI data to ensure data consistency; then, the spatial distance between each POI and the pixel is calculated, a distance threshold is set, and pixels within the distance range are determined as "related pixels." If the classification result of a related pixel is inconsistent with the POI classification label, and the confidence score of the POI is higher than 0.8, correction is triggered, and the classification results of the related pixel and its surrounding pixels are corrected to the type corresponding to the POI label, thereby correcting the classification misclassification problem that still exists after CRF processing.

[0071] Furthermore, in step 2, the exposed population is divided into groups based on the average values ​​in the "Handbook of Exposure Parameters for Chinese Population" and the National Physical Fitness Monitoring Bulletin published by the General Administration of Sport of China. The exposed population in the region is divided into children (<6 years old), youth (6 to ≤17 years old), adults (18 to <60 years old), young elderly (60 to <80 years old), and very old elderly (≥80 years old). Each group is further divided into males and females.

[0072] Furthermore, the number and structural characteristics of the exposed population in step 2 are determined by data sources such as on-site surveys and population censuses, to statistically analyze the total number of exposed people in the region, the number of people in each exposed group, and their proportion.

[0073] Furthermore, in step 2, the spatial behavioral characteristics are determined based on the defined potential exposure scenarios and regional boundaries. Data sources such as on-site surveys, questionnaires, and literature reviews are used to identify the core distribution areas of the exposed population and to statistically analyze the index parameters of the exposed population (including height, weight, age, and gender), as well as the frequency and timing of outdoor activities.

[0074] Furthermore, in step 2, the Geographic Information System (GIS) overlay technology first performs unified preprocessing on the input potential exposure scene boundary data, exposed population characteristic data, and basic geographic data. Spatial coordinates are unified using the internationally recognized WGS84 coordinate system. Scene boundary data is converted from raster classification masks to vector polygon data. Population characteristic data is stored as attribute tables with key fields and associated with the vector polygon data. The resolution is adapted to the highest precision. Then, the overlay is performed in two layers. The first layer uses the scene boundary vector polygon as the base layer and overlays the basic geographic layer, preserving the complete geometric shape of the scene and associating it with the basic geographic data attributes. The second layer uses the scene boundary vector polygon as the target layer and uses the population characteristic data as the connecting layer, binding the population characteristic data to the corresponding scene boundary. Finally, a spatial distribution map of the exposure scene is generated.

[0075] Step 3, Health Risk Assessment Method: Using standardized source parameters extracted from the key source emission feature database as input, high-precision meteorological field data are integrated, and the spatial gridded concentration distribution of pollutants is simulated through pollutant diffusion models; the spatial distribution map of exposure scenarios is integrated, the exposure parameters of the corresponding populations in each scenario are identified, and the parameters are substituted into the health effect model to calculate the carcinogenic and non-carcinogenic health risk values ​​of pollutants caused by different exposure pathways to each exposure scenario and each population group.

[0076] Furthermore, the pollutant diffusion simulation in step 3 mainly involves the following steps:

[0077] Step S1, Weather Research and Forecasting (WRF) model preprocessing, is a crucial preliminary step providing high-precision spatiotemporal meteorological fields for subsequent pollutant diffusion simulation. The simulation range is delineated centered on key source concentration areas, and a three-layer nested grid design is used for grid configuration, with spatial resolutions of 10km×10km, 5km×5km, and 1km×1km. External data such as topographic elevation data, land use type data, and reanalysis meteorological data are input. The external data are interpolated to each level of the grid using interpolation methods. Meteorological field data is extracted and horizontally interpolated to each nested grid within the simulation range. The WRF model preprocessing system converts the data into initial field and boundary condition files in a specified format. Finally, the main program of the model is run to output the spatiotemporal meteorological field data required for subsequent pollutant diffusion simulation.

[0078] Step S2 involves simulating pollutant diffusion patterns. The optimal diffusion model is selected based on the simulation scale, key source type, and pollutant migration characteristics. The main diffusion models include the AERMOD model, CALPUFF model, or Flexpart model. For near-surface short-distance diffusion, the AERMOD model is used. This model is suitable for key sources that are elevated point sources or small-area sources, focusing on near-surface turbulent diffusion (without long-distance cross-regional transport). For long-distance cross-regional diffusion, the CALPUFF model is used. This model is suitable for key sources that are large elevated point sources or regional area sources, and can simulate long-distance pollutant transport and dry / wet deposition processes. For sudden emergency emission diffusion, the Flexpart model is used. This model is suitable for key sources that are sudden intermittent emissions, and can quickly simulate pollutant diffusion trajectories and short-term high-concentration areas. However, these models are not limited to the aforementioned diffusion models. Other models, such as ADMS, CMAQ, and AUSTAL2000, are extended based on scenario specificity. ADMS focuses on complex urban micro-scale terrain, adapting to pollutant diffusion in densely built-up areas. CMAQ targets regional scales, simulating pollutant chemical transformation and secondary pollution, suitable for cross-provincial and municipal complex pollution assessments. AUSTAL2000 is suitable for complex terrain and special sources, quantifying the impact of terrain shielding and temperature inversion effects on pollution diffusion. All models are based on fundamental atmospheric diffusion theories (turbulent diffusion theory, law of conservation of mass, and gradient transport theory), using standardized source parameters extracted from a key source emission characteristic database as input. Data format conversion is performed according to model requirements, and high-precision meteorological field data output from the WRF model is integrated. The selected diffusion model is run according to configured parameters, ultimately outputting spatially gridded pollutant concentration data, providing quantitative concentration input for subsequent health risk assessments.

[0079] Furthermore, the health effect model in step 3, which quantifies health risks, mainly involves the following steps:

[0080] Step S1, the health effect model, uses the human health risk model recommended by the US Environmental Protection Agency (US EPA) as its core framework. The main pathways for pollutants emitted from key sources to enter the human body include inhalation and skin contact. The carcinogenic and non-carcinogenic health risks of pollutants to each group of people through these pathways are calculated.

[0081] The daily average exposure via inhalation and skin contact is calculated as follows:

[0082]

[0083]

[0084] Among them, ADD inh This represents the average daily exposure to pollutant i via inhalation; ADD dermal C represents the average daily exposure to contaminant i via skin contact; i 1 represents the environmental concentration of pollutant i; IR represents the respiration rate; ET represents the exposure duration; EF represents the exposure frequency; ED represents the exposure years; BW represents body weight; AT represents the average exposure time; SA represents the skin contact area; Kp represents the skin permeability coefficient; and i represents the pollutant.

[0085] The non-carcinogenic risk calculation method for inhalation and skin contact is as follows:

[0086]

[0087]

[0088] Among them, HQ inh HQ dermal These are the non-carcinogenic risks of pollutant i via inhalation and skin contact; ADD inh This represents the average daily exposure to pollutant i via inhalation; ADD dermal RfC represents the average daily exposure to contaminant i via skin contact. i Reference concentration for inhalation of pollutant i; RfD i , where i is the skin reference dose of contaminant i; i is the contaminant.

[0089] The method for calculating the carcinogenic risk through inhalation and skin contact is as follows:

[0090]

[0091]

[0092] Among them, ELCR inh ELCR dermal The carcinogenic risks of pollutant i via inhalation and skin contact are respectively; ADDinh This represents the average daily exposure to pollutant i via inhalation; ADD dermal IUR represents the average daily exposure to contaminants through skin contact. i Risk of inhalation of pollutant i per unit; SF i Let be the risk slope factor for pollutant i; where i is the pollutant.

[0093] Step S2 involves determining exposure parameters. Based on the established key source emission characteristic database and exposed population characteristic database, the characteristic parameters of each group of people in the exposure scenario, the pollutants they may be exposed to, the intensity of the exposure source, the exposure route, and the exposure period are identified and extracted.

[0094] Step S3 determines the toxicity parameters. RfC (inhalation reference concentration) and RfD (dermal reference dose) reflect the safety threshold for no harmful effects after long-term exposure in the population. IUR (inhalation unit risk) and SF (risk slope factor) reflect the carcinogenicity of a unit dose. The corresponding pollutant parameter data can be obtained from the US Environmental Protection Agency (EPA) Integrated Risk Information System database, the World Health Organization (WHO) Toxicology Database, and the Chinese Academy of Environmental Sciences Toxicology Parameter Database.

[0095] Step S4: Health risk quantification. The non-carcinogenic risk (HQ) of a single pathway reflects the degree of harm of that pathway. If HQ < 1, the risk is acceptable; if HQ ≥ 1, a non-carcinogenic health hazard may exist. The total hazard from combined exposure to multiple pathways / contaminants is the sum of the individual HQs, i.e., HI = ∑HQ. i If HI ≥ 1, the cumulative effect of combined exposures needs to be considered; Total carcinogenic risk ELCR = ∑ELCR i The maximum acceptable level of 1×10⁻⁶ recommended by the Dutch Ministry of the Environment and the Swedish Environmental Protection Agency was adopted. -6 As a basis for health risk assessment.

[0096] Step S5 employs a combination of sensitivity analysis and Monte Carlo simulation to perform uncertainty analysis.

[0097] ① Based on the entire process of health risk quantification, the core parameter Xi to be analyzed is statistically analyzed, as follows:

[0098] Exposure source parameters: including potential pollutants and pollutant emission rates, derived from the key source emission characteristic database, with fluctuation ranges referenced to the "Environmental Monitoring Data Rounding Rules", and a disturbance ratio of δ=±10%;

[0099] Environmental parameters: These include pollutant concentrations, derived from grid averages in diffusion model simulations, with fluctuation ranges referenced to the "Environmental Monitoring Data Rounding Rules," and a perturbation ratio of δ = ±10%.

[0100] Exposure parameters: Population characteristic parameters for each group, including weight, respiratory rate, and skin contact area, are derived from the sampling mean of the exposed population characteristic database, and the perturbation ratio is taken as δ=±5% based on the statistical coefficient of variation of population physiological parameters;

[0101] Toxicity parameters: Carcinogenic parameters include inhalation unit risk (IUR) and risk slope factor (SF), while non-carcinogenic parameters include inhalation reference concentration (RfC) and dermal reference dose (RfD). The perturbation ratio is taken as δ = ±15%, referring to the US EPA's labeling range for uncertainty in toxicity data.

[0102] ② Based on the parameter list, sensitivity analysis is used to systematically perturb a single input parameter while fixing other parameters. This allows for the selection of "critical parameters" with high sensitivity coefficients and the elimination of "non-critical parameters" with extremely low sensitivity coefficients, reducing the dimensionality of subsequent Monte Carlo simulations. The sensitivity coefficient (SSi) is calculated using the following formula:

[0103]

[0104] Where Y0=f(X) 1,0 ,...,X n,0 ): The output result of health risk quantification when all parameters are taken as baseline values;

[0105] X i,0 Parameter X i The baseline value;

[0106] X i =X i,0 ×(1±δ): The disturbance value of parameter Xi, where δ is the disturbance ratio;

[0107] Y i =f(X 1,0 ,...,Xi,...,X n,0 ): Only X i Quantitative output of health risk during disturbances;

[0108] i: Parameter number.

[0109] The larger the absolute value of the sensitivity coefficient (SSi), the more significant the impact of the parameter on the quantification of health risk. We set |SSi| ≥ 0.2 as the key parameter and excluded |SSi| < 0.2 as the non-key parameter.

[0110] ③ Based on the selected k key parameters (denoted as X1,...,X...) k(k≤n) Using Monte Carlo simulation to focus on the synergistic effect of these parameters, the probability distribution type and parameters (mean μ, standard deviation σ) of each key parameter are set according to their actual fluctuation characteristics. The simulation is set to 1000 iterations. In each iteration, one value is randomly selected from the probability distribution of each key parameter and combined into a set of input parameters. This set of parameters is substituted into the health risk model to calculate and output a health risk quantification result. The sampling is repeated 1000 times to obtain 1000 health risk output results (i.e., HQ1~HQ). 1000 ELCR1~ELCR 1000 ), which generates the probability distribution (mean, standard deviation, 95% confidence interval) of health risk outcomes through statistical analysis.

[0111] ④ Finally, the linear correlation strength between key parameters and risk outcomes is quantified using the Pearson correlation coefficient, and the contribution ratio of each parameter to the total uncertainty is clarified using the normalized contribution rate. The specific calculation process is as follows:

[0112] Measuring key parameter X i The degree of linear correlation with the risk outcome Y:

[0113]

[0114] Where, r i : Pearson correlation coefficient of the i-th key parameter;

[0115] X i,m : The m-th sampled value of the i-th key parameter;

[0116] : The mean of 1000 samples of the i-th key parameter;

[0117] Y m : The output result of health risk quantification corresponding to the m-th sampling;

[0118] The mean of the quantitative output results of 1000 health risk samplings;

[0119] m: Sampling number;

[0120] i: Key parameter number;

[0121] N: Total number of Monte Carlo simulations, which is 1000.

[0122] The closer |ri| is to 1, the stronger the linear correlation between parameter Xi and risk outcome Y.

[0123] The square of the correlation coefficients of each parameter is equal to r. i 2 Then divide by all parameters r i2 The summation, i.e., the normalization process, yields the contribution rate of each parameter to the total uncertainty:

[0124]

[0125] Where Ci: the contribution rate of the i-th key parameter to the total uncertainty;

[0126] r i 2 The square of the correlation coefficient of the i-th key parameter;

[0127] r j 2 The square of the correlation coefficient of the j-th key parameter;

[0128] k: Number of key parameters;

[0129] i: Key parameter number.

[0130] The top three parameters in terms of contribution rate are included in the "key monitoring list". Their uncertainty should be reduced through online monitoring, model calibration and toxicity data updates. If the contribution rate of a parameter is greater than 40%, it is recommended that control measures be prioritized for that parameter.

[0131] Step 4, Assessment, Analysis, Prediction, and Simulation Decision Making: Based on the established health risk assessment method, key prevention and control areas with "high pollution emissions + highly sensitive populations" are identified through current status assessment; differentiated scenarios are formulated for key prevention and control areas, and the health effects of each scenario are compared through prediction and simulation; finally, the optimal solution is selected from the simulation results of each scenario, and it is transformed into specific control policies, and the implementation effect is verified through subsequent core indicator assessment and analysis.

[0132] Furthermore, the current status assessment process in step 4, based on the established health risk assessment methodology, uses standardized source parameters extracted from the key source emission characteristic database as input, integrates high-precision meteorological field data, simulates the spatial gridded concentration distribution of pollutants through pollutant diffusion models, generates a dynamic diffusion trend map of key source pollution, and quantitatively characterizes the regional pollution exceedance multiple and the contribution ratio of each key source; integrates the spatial distribution map of exposure scenarios, clarifies the exposure parameters of the corresponding populations for each scenario, substitutes them into the health effect model, calculates the carcinogenic and non-carcinogenic risk values ​​of pollutants through different exposure pathways to each exposure scenario and each population group, statistically analyzes the differences in exposure levels of each scenario and the proportion of high-risk populations, analyzes the risk characteristics of sensitive populations, calculates the risk contribution ratio of each exposure pathway, and outputs a list of high-risk areas, key pollution sources, and sensitive populations.

[0133] Furthermore, in step 4, the simulation and prediction process uses the current status assessment results as the baseline scenario and formulates differentiated scenarios for key prevention and control areas, covering control scenarios (including technological emission reduction, structural adjustment, and emergency control) and extreme scenarios (including extreme weather and sudden pollution events). Based on the parameters of each scenario, the key source emission characteristic database is updated, and pollutant concentration changes are predicted using pollutant diffusion models. Combined with the exposed population characteristic database, the risk changes and health effects of each scenario are quantified using health risk assessment methods. Finally, a comparison table of the contribution ratio of key sources, carcinogenic / non-carcinogenic risk values, and the proportion of high-risk populations under each scenario is output, along with early warnings of key risk thresholds and targeted action recommendations.

[0134] Furthermore, the optimal solution described in step 4, which is a combination of control strategies determined through multi-dimensional screening of the simulation results of various scenarios, is based on the principle of ensuring the health benefits of the population while balancing control feasibility and risk adaptability. Firstly, regarding the health effect dimension, it requires that the reduction rate of carcinogenic or non-carcinogenic risks in key prevention and control areas be ≥ a preset threshold (e.g., ≥30%), and the reduction rate of the proportion of high-risk populations be ≥ the target value (e.g., ≥25%). Secondly, regarding control feasibility, it requires that the input-output ratio of technological emission reduction measures be reasonable, the socio-economic impact of structural adjustments be controllable, and the efficiency of emergency control implementation meet standards (e.g., response time ≤2 hours). Thirdly, regarding risk adaptability, it prioritizes measures for the top 3 or >40% key parameters contributing to the uncertainty analysis mentioned above. Typical solutions include a combination of multiple measures such as technological emission reduction, population protection, and emergency control.

[0135] Furthermore, in step 4, the implementation effect verification process first clarifies the core verification indicators, using the compliance rate of key source emission concentrations in key prevention and control areas, the reduction rate of carcinogenic / non-carcinogenic risks, and the reduction rate of the proportion of high-risk populations as core indicators. Second, it quantifies the judgment criteria, referring to the US EPA health risk standards, the "Technical Guidelines for Environmental Health Risk Assessment," and the pre-set targets of my country's policies, and sets corresponding thresholds for each indicator. Finally, it conducts indicator verification and effect judgment. If all core indicators meet the standards, the implementation effect is judged to be good, and the entire control plan is archived as experience for promotion. If any indicator fails to meet the standards, the effect is judged to be poor, the data sources of each input module are checked, the scenario plan measures are adjusted accordingly, and a second verification is conducted until the control objectives are met.

[0136] If any indicator fails to meet the standard, the spatial coordinates of multiple key pollution emission sources are determined based on the established key source emission profile. Combined with the simulation results of their pollution emission types and pollutant diffusion patterns, high-risk population gathering areas in the surrounding area are selected as information collection locations. Information collection-protection deployment structures are deployed at each information collection location. The structures have external information collectors and internal multiple types of pollution protection devices. The core characteristic information of the exposed population in the area is entered into the cloud system through the information collectors. After the system calls the established exposed population characteristic database to complete the information verification, it matches the protection needs based on the entered information and deploys appropriate pollution protection devices to the exposed population.

[0137] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for assessing and analyzing the impact of key source emissions incorporating population health effects, characterized in that, Specifically, the following steps are included: Step 1: Establishment of a key source emission characteristic database; combining regional pollution characteristics and industry attributes, screen and clarify the list of key sources and construct key source emission profiles to obtain standardized source parameters; Step 2: Establishment of the exposed population feature database; delineate potential exposure scenarios and area boundaries based on building and land use types, obtain core feature information of exposed populations within the area, and generate a spatial distribution map of exposure scenarios; Step 3: Health risk assessment; Using standardized source parameters extracted from the key source emission feature database as input, the spatial gridded concentration distribution of pollutants is simulated; the spatial distribution map of exposure scenarios is integrated, the exposure parameters of the corresponding populations in each scenario are identified, and the parameters are substituted into the health effect model to calculate the health risk values ​​caused by pollutants through different exposure pathways to each exposure scenario and each population group. Step 4: Evaluation, analysis, prediction, and simulation decision-making; Based on the established health risk assessment methodology, key prevention and control areas with high pollution emissions and highly sensitive populations are identified through current status assessment. Differentiated scenarios are developed for key prevention and control areas, and the health effects of each scenario are compared through prediction simulation. Finally, the optimal solution is selected from the simulation results of each scenario and transformed into specific control policies. The implementation effect is verified through subsequent evaluation and analysis of core indicators.

2. The method for assessing and analyzing the impact of key source emissions incorporating population health effects as described in claim 1, characterized in that, The establishment of the key source emission characteristic database in step 1 specifically includes the following steps: S1: Through on-site inspections, document review, questionnaires, and data collection, combined with regional pollution characteristics and industry attributes, pollution sources with a pollution contribution rate of more than 10% are included in the candidate list. Through cross-comparison, pollution sources with a pollution contribution rate that meets the standard but do not belong to key industries are eliminated, and the list of key sources is clarified. S2: Obtain full-chain information on the list of key sources. The core feature indicators cover the basic attribute information of key sources, pollutant emission information, source emission mode and parameter information, and dynamic change information to construct a profile of key source emissions.

3. The method for assessing and analyzing the impact of key source emissions incorporating population health effects as described in claim 1, characterized in that, Step 2, the establishment of the exposed population feature database, specifically includes the following steps: S1: First, acquire remote sensing image data, and then use a dual-branch U-Net deep learning model to automatically identify building types and land use types, and delineate potential exposure scenarios and area boundaries. S2: Then, obtain the core characteristic information of the exposed population in the region, mainly including the classification of the exposed population, the quantity and structural characteristics, and the spatial behavior characteristics, and generate a spatial distribution map of the exposure scene by combining it with GIS map overlay.

4. The method for assessing and analyzing the impact of key source emissions incorporating population health effects as described in claim 1, characterized in that, The potential exposure scenarios and region boundary delineation in step 2 are mainly based on the U-Net model. The balance between global semantic information and local detail information is achieved through the encoder-decoder architecture and skip connections. The encoder is responsible for downsampling to extract multi-scale features, the decoder is responsible for upsampling to restore image resolution, and the skip connections pass the shallow features of the encoder to the decoder. Specifically, the following steps are included: Step S1, training set construction: The training set is constructed based on remote sensing images of the target area as input, and after preprocessing operations such as radiometric normalization, band fusion and data augmentation. Step S2: Dual-branch U-Net architecture design. An improved U-Net model architecture with a shared encoder and dual-task decoder is adopted. First, the pre-processed remote sensing image is input. The shared encoder adopts a residual network structure, extracting multi-scale features through multi-scale convolution and downsampling. Initial feature mapping is performed through 3×3 convolution kernels, and feature extraction is achieved through 4 sets of residual blocks. After each residual block, 2×2 max pooling is used to halve the feature map size. At the same time, the multi-scale features output from each level are synchronously transmitted to the decoder through skip connections. Second, the dual decoders are optimized for different task characteristics. The building decoder uses 3×3 device convolution upsampling, and the land use decoder uses "bilinear interpolation and 1×1 convolution to adjust channel upsampling" to restore the feature map size. After each upsampling, feature alignment, channel compression, and element-wise addition are performed to fuse the features of the i-th layer of the encoder and the i-th layer of the decoder to complete the skip connection. Then, an attention mechanism is embedded to output the two classification masks respectively. Finally, the error of the two classification tasks is synchronously optimized by combining a multi-task loss function. Step S3, post-processing of the inference stage, introduces a conditional random field to refine the boundary, and uses the correlation between pixels to correct the boundary ambiguity of the initial classification mask; then, it integrates the point of interest data and the land use planning vector map to correct misclassification of the classification results after processing by the conditional random field, and finally achieves high-precision classification of buildings and land use types.

5. The method for assessing and analyzing the impact of key source emissions incorporating population health effects as described in claim 4, characterized in that, The attention mechanism in step S2 embeds a channel attention module into the building decoder, performs global average pooling on the feature map of a certain level of the decoder, compresses the spatial information of each channel into a channel descriptor, learns the channel weights to generate attention coefficients, and multiplies the attention coefficients element-wise with the channels of the original feature map to enhance the channel response of the building's iconic features. A spatial attention module is embedded in the land use decoder. Max pooling and average pooling are performed on the feature map of a certain level of the decoder to generate a spatial weight map. The spatial weight map is then multiplied pixel by pixel with the original feature map to enhance the consistency features within the land use type and suppress misjudged pixels at the edge of the region.

6. The method for assessing and analyzing the impact of key source emissions incorporating population health effects as described in claim 4, characterized in that, The multi-task loss function in step S2 employs a weighted fusion architecture for task-differentiated loss design and dynamic weight allocation, and is defined as follows: Loss = α×Loss1 + (1-α)×Loss2; Where Loss is the total task loss, Loss1 is the building type classification loss, Loss2 is the land use type classification loss, and α∈[0.5,0.7] is the weight coefficient used to balance the priorities of the two types of tasks; The building type classification loss, Loss1, uses weighted cross-entropy loss, defined by the formula: ； Where N is the total number of pixels i in the remote sensing image, M is the total number of building category numbers c, and W c Y represents the weight of building category. i,c P is the true label for pixel i belonging to category c. i,c To predict the probability that pixel i belongs to category c; c is the building category number, and i is a pixel in the remote sensing image; Land use type classification loss Loss2 adopts Dice loss, which is defined by the formula: ； Where N is the total number of pixels i in the remote sensing image, M is the total number of land use type numbers k, and Y i,k For pixel i to be the true label of type k, P i,k To predict the probability that pixel i belongs to type k, ε is a smoothing term used to avoid cases where the denominator is 0, k is the land use type number, and i is a pixel in the remote sensing image.

7. The method for assessing and analyzing the impact of key source emissions incorporating population health effects as described in claim 4, characterized in that, The introduction of a conditional random field to refine the boundary in step S3 is a post-processing module in the inference stage of the U-Net model. It takes the initial classification result as the base label input and introduces remote sensing image observation data as variables. The label dependency relationship is constructed by defining an energy function, which mainly includes a univariate potential term and a binary potential term. The univariate potential term retains the high-confidence classification results in the initial prediction of U-Net, while the binary potential term establishes the class association between pixels and constrains the class consistency of adjacent pixels. The globally optimal label configuration is solved by minimizing the energy function, and a classification mask with clear boundaries and continuous classes is output.

8. The method for assessing and analyzing the impact of key source emissions incorporating population health effects as described in claim 1, characterized in that, The health risk assessment method in step 3 specifically includes the following steps: S1: Pollutant diffusion simulation, using standardized source parameters extracted from the key source emission feature database as input, combined with pollutant diffusion models and high-precision meteorological field data, to simulate the spatial concentration distribution of pollutants; S2: Exposure information is clear, and a spatial distribution map of exposure scenarios is overlaid to identify the characteristic parameters of exposed populations, possible pollutants, exposure source intensity, exposure pathways, and exposure periods in each scenario; S3: Health risk quantification, based on the established health effect assessment method, combined with exposure information and toxicity parameters, to quantify the health risks caused by pollutants through different exposure routes to different exposure scenarios and population groups.

9. The method for assessing and analyzing the impact of key source emissions incorporating population health effects as described in claim 8, characterized in that, The pollutant diffusion simulation in step S1 specifically includes the following steps: S1: WRF model preprocessing, the simulation range is defined with the key source concentration area as the center, and a three-layer nested grid design is used for grid configuration, with spatial resolutions of 10km×10km, 5km×5km, and 1km×1km respectively. External source data is input, and the external source data is interpolated to each level of grid using the interpolation method. Meteorological field data is extracted and horizontally interpolated to each nested grid of the simulation range. The WRF model preprocessing system is used to convert it into initial field and boundary condition files in a specified format. Finally, the main program of the model is run to output the spatiotemporal meteorological field data required for subsequent pollutant diffusion simulation. S2: Pollutant diffusion model simulation. It takes standardized source parameters extracted from the key source emission feature library as input, combines high-precision spatiotemporal meteorological field data output by WRF model, and finally outputs spatially gridded pollutant concentration data through pollutant diffusion model simulation.

10. The method for assessing and analyzing the impact of key source emissions incorporating population health effects as described in claim 1, characterized in that, In step 4, if any indicator fails to meet the standard, the spatial coordinates of multiple key pollution emission sources are determined based on the established key source emission profile. Combined with the simulation results of their pollution emission types and pollutant diffusion patterns, high-risk population gathering areas in the surrounding area are selected as information collection locations. Information collection and protection deployment structures are set up at each information collection location. The structures have external information collectors and internal multiple types of pollution protection devices. After the core characteristic information of the exposed population in the area is entered into the cloud system through the information collectors, the system calls the established exposed population characteristic database to complete the information verification. Based on the entered information, the protection needs are matched, and suitable pollution protection devices are deployed to the exposed population.

Images