Urban composite environment health risk assessment and intervention method and system

Abstract

The invention relates to the technical field of urban intelligent management and control, and provides an urban composite environment health risk assessment and intervention method and system, and the method comprises the steps: obtaining the geographic space data and ground observation data of an urban region; constructing a collaborative environmental stress index, and calculating an SESI predicted value; performing grid construction on the urban area to construct a graph structure, performing feature learning by using a GNN algorithm, predicting a node SESI value, and generating a collaborative stress map; constructing a comprehensive vulnerability evaluation model, calculating a vulnerability index, and identifying a high-vulnerability region; fuzzy comprehensive evaluation and TOPSIS methods are fused, intervention measures of the high-fragility area are sorted, and an intervention scheme is generated. According to the method, the nonlinear interaction modeling and the high-resolution graph neural network are fused to realize accurate recognition of the composite environment health risk, and the intervention strategy is optimized through the multi-criterion decision model, so that scientific regulation and dynamic optimization of the urban intervention scheme are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of urban intelligent management and control, in particular to a method and system for assessing and intervening in urban complex environmental health risks. BACKGROUND

[0002] The statements in this section merely provide background information related to the present application and do not necessarily constitute prior art.

[0003] With the intensification of global climate change and the acceleration of urbanization, the urban environment is facing unprecedented complex ecological pressure, among which the superimposed influence of high temperature heat wave and air pollution is the most significant. In recent years, the frequency and duration of high temperature weather have been rising, while ozone (O3), fine particulate matter (PM2.5) and other air pollutants are more likely to exceed the standard under high temperature conditions in summer, forming a typical "heat pollution" complex stress scenario. This complex stress not only brings serious interference to residents' daily life, but also poses a significant threat to human health, especially to sensitive groups such as the elderly, children and patients with chronic diseases, whose mortality and hospitalization rates increase significantly. Therefore, under the background of urban health governance and fine management, it has become an urgent need in the field of public health and urban planning to build a scientific and comprehensive complex environmental health risk assessment mechanism.

[0004] However, the existing complex environmental health risk assessment methods generally have limitations in methodology, which are difficult to truly reflect the complexity and population differences of complex risks. First, in terms of risk exposure modeling, most current studies only treat high temperature and air pollution as two independent variables for simple superposition, ignoring the nonlinear synergistic mechanism between the two in biological effects, i.e., the health hazard enhancement effect of "1+1>2", leading to distorted risk assessment results. Second, traditional assessment methods mostly rely on sparse fixed monitoring site data for spatial interpolation, which is difficult to accurately depict the environmental heterogeneity at the micro scale (such as street, community level) within the city, resulting in low assessment accuracy and inability to provide effective support for fine urban management and intervention measures. In terms of health risk bearing capacity assessment, existing methods usually stay at the physical level of environmental exposure, only considering the absolute value of temperature or pollutant concentration, ignoring population susceptibility and social adaptability, which leads to the fact that the assessment results cannot reflect the actual vulnerability of different regions and populations in the face of complex stress, thus reducing the practicality and scientificity of the assessment tool in the formulation of urban control measures, resource allocation and intervention priority determination. SUMMARY

[0005] The present application proposes a city composite environmental health risk assessment and intervention method and system to solve the above problems, which realizes accurate identification of composite environmental health risk by combining nonlinear interaction modeling and high-resolution graph neural network, and optimizes intervention strategies through a multi-criteria decision model, thereby realizing scientific regulation and dynamic optimization of city intervention schemes.

[0006] In order to achieve the above purpose, the present application adopts the following technical solutions: One or more embodiments provide a city composite environmental health risk assessment and intervention method, comprising the following steps: Obtain geographic spatial data and ground observation data of a target city area; Nonlinear interaction modeling is performed for the synergistic enhancement effect of high temperature and air pollution, a synergistic environmental stress index is constructed, and SESI prediction values are calculated according to the obtained ground observation data; The target city spatial area is divided into a grid structure; the node features and adjacency relationships in the graph structure are learned and aggregated by using a GNN algorithm, the SESI prediction values of each grid node are predicted, and a synergistic environmental stress map is obtained; Modeling is performed on environmental exposure, population sensitivity and social adaptability to construct a comprehensive vulnerability assessment model, based on the synergistic environmental stress map, the comprehensive vulnerability index of each node of the target city spatial area is calculated, a comprehensive vulnerability level map of the city space for the population is generated, and a high vulnerability area is identified; The high vulnerability area is taken as a decision unit, a multi-criteria decision method combining fuzzy comprehensive evaluation and TOPSIS is used to prioritize the intervention measures for the high vulnerability area, and an intervention action plan is generated.

[0007] One or more embodiments provide a city composite environmental health risk assessment and intervention system, comprising: A data management module configured to obtain geographic spatial data and ground observation data of a target city area; A synergistic effect calculation module configured to perform nonlinear interaction modeling for the synergistic enhancement effect of high temperature and air pollution, construct a synergistic environmental stress index, and calculate SESI prediction values according to the obtained ground observation data; A spatial prediction modeling module configured to divide the target city spatial area into a grid structure; the node features and adjacency relationships in the graph structure are learned and aggregated by using a GNN algorithm, the SESI prediction values of each grid node are predicted, and a synergistic environmental stress map is obtained; The vulnerability assessment module is configured to model environmental exposure, population sensitivity, and social adaptability in a coupled manner, construct a comprehensive vulnerability assessment model, calculate the comprehensive vulnerability index of each node in the target urban spatial area based on a collaborative environmental stress map, generate a comprehensive vulnerability level map of urban space oriented towards the population, and identify high-vulnerability areas. The decision support module is configured to use highly vulnerable areas as decision-making units, integrate fuzzy comprehensive evaluation and TOPSIS multi-criteria decision-making methods, prioritize intervention measures for highly vulnerable areas, and generate intervention action plans.

[0008] An electronic device includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, perform the steps in the above-described method for assessing and intervening in health risks in a complex urban environment.

[0009] A computer-readable storage medium for storing computer instructions, which, when executed by a processor, complete the steps in the above-described method for assessing and intervening in health risks in a complex urban environment. Compared with the prior art, the beneficial effects of the present invention are as follows: Compared to the linear superposition assumption of high temperature and pollution risks in existing technologies, the method of this invention accurately reflects the nonlinear synergistic mechanism between the two in terms of health risks by introducing a nonlinear interaction model. This effectively overcomes the problem of ignoring the "1+1>2" enhancement effect and improves the realism of risk assessment. Simultaneously, the spatial modeling framework based on hybrid graph neural networks achieves high-resolution risk prediction at the microscale (e.g., street and community levels), significantly outperforming traditional interpolation methods that rely on sparse monitoring station data, thus improving the ability to identify spatial heterogeneity. In terms of health risk tolerance assessment, this method integrates population heterogeneity and regional socioeconomic differences (e.g., medical resources, response capabilities), shifting from a single environmental exposure model to a complex vulnerability analysis, making the results closer to actual stress levels and enhancing the application value of the assessment model in urban governance. By integrating fuzzy comprehensive evaluation and the TOPSIS decision-making model, it also achieves the scientific ranking of intervention programs, enhancing the quantitative basis for the formulation of control measures.

[0010] The advantages of the present invention, as well as its additional advantages, will be described in detail in the following specific embodiments. Attached Figure Description

[0011] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute a limitation thereof.

[0012] Figure 1This is a flowchart of a method for assessing and intervening in urban complex environmental health risks according to Embodiment 1 of the present invention; Figure 2 This is a schematic diagram illustrating the construction process of the Cooperative Environmental Stress Index (SESI) in Embodiment 1 of the present invention; Figure 3 This is a schematic diagram of the Geo-GNN model architecture in Embodiment 1 of the present invention; Figure 4 This is a schematic diagram of the Integrated Vulnerability Assessment (IVAM) process in Embodiment 1 of the present invention; Figure 5 This is a flowchart illustrating the intervention priority ranking decision-making process of Embodiment 1 of the present invention; Figure 6 This is a system structure block diagram of Embodiment 2 of the present invention; Figure 7 This is a sample map of a visual intervention action in an example urban area according to Embodiment 1 of the present invention. Detailed Implementation

[0013] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0014] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0015] It should be noted that the terminology used herein is for describing particular embodiments only and is not intended to limit the exemplary embodiments of the present invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof. It should be noted that, without conflict, the various embodiments and features within those embodiments can be combined with each other. The embodiments will now be described in detail with reference to the accompanying drawings.

[0016] Example 1 In one or more of the technical solutions disclosed in the embodiments, such as Figures 1 to 7 As shown, a method for assessing and intervening in urban complex environmental health risks includes the following steps: Step 1: Acquisition and Spatiotemporal Alignment of Multi-Source Heterogeneous Data: Acquire geospatial data and ground observation data of the target urban area and perform spatiotemporal alignment processing; among which, ground observation data includes thermal comfort observation data, air pollutant concentration observation data, and regional meteorological data; Step 2: Quantitative modeling of the synergistic effect of combined environmental risks: Nonlinear interaction modeling is performed for the synergistic enhancement effect of high temperature and air pollution, and the Synergistic Environmental Stress Index (SESI) is constructed. The value of the Synergistic Environmental Stress Index is calculated based on the obtained ground observation data. Specifically, the synergistic environmental stress index value is calculated based on thermal comfort observation data and air pollutant concentration observation data from ground observation data; Step 3: Generation of a seamless global risk map based on geographic artificial intelligence: The target city spatial area is divided into grids to construct a graph structure; the GNN algorithm is used to learn and aggregate spatial features of the node features and adjacency relationships in the graph structure, and the SESI prediction value of each grid node is predicted to obtain a collaborative environmental stress map. Step 4, Spatial Vulnerability Assessment: Modeling is performed by coupling environmental exposure, population sensitivity and social adaptability to construct an Integrated Vulnerability Assessment Model (IVAM). Based on the collaborative environmental stress map, the integrated vulnerability index of each node in the target urban spatial area is calculated to generate a population-oriented integrated vulnerability level map of urban space and identify high-vulnerability areas. Step 5: Prioritizing Intervention Measures Based on Multi-Criterion Decision Analysis: Using the high-vulnerability areas identified in Step 4 as decision-making units, the multi-criteria decision-making methods of fuzzy comprehensive evaluation and TOPSIS are integrated to prioritize intervention measures in high-vulnerability areas, generate intervention action plans, and provide quantitative decision support for urban management. In this embodiment, firstly, multi-source ground observation data, including thermal comfort, air pollution concentration, and meteorological factors, are collected and combined with geospatial information to construct a multi-dimensional data input set. Then, a nonlinear interaction function `fint` is introduced to model the synergistic enhancement effect between high temperature and air pollution. By quantifying the interaction between the two in terms of physiological and environmental stress, a Synergistic Environmental Stress Index (SESI) is constructed. To achieve spatial risk assessment at the urban scale, the target urban area is gridded at a fixed resolution to construct a graph structure. A Hybrid Geo-GNN combined with an attention mechanism is applied to jointly model the spatial adjacency relationships and feature information of nodes in the graph, outputting the predicted SESI values ​​for each grid to form a high-resolution synergistic stress map. Subsequently, indicators such as exposure level, population sensitivity (e.g., age, basic health status), and social adaptability (e.g., medical resources, infrastructure, response capacity) are introduced to construct a Comprehensive Vulnerability Assessment Model (IVAM) to assess the vulnerability level of the population in the face of combined environmental stresses, forming a vulnerability level map. Finally, a combination of fuzzy comprehensive evaluation and TOPSIS method was applied to identify highly vulnerable areas to prioritize multiple criteria, formulate a priority sequence of intervention actions, and achieve dual coupling regulation of spatial accuracy and population differences.

[0017] Compared to the linear superposition assumption of high temperature and pollution risks in existing technologies, this method introduces a nonlinear interaction model to accurately reflect the nonlinear synergistic mechanism between the two in terms of health risks. This effectively overcomes the problem of ignoring the "1+1>2" enhancement effect and improves the realism of risk assessment. Simultaneously, a spatial modeling framework based on hybrid graph neural networks enables high-resolution risk prediction at the microscale (e.g., street and community levels), significantly outperforming traditional interpolation methods that rely on sparse monitoring station data, thus improving the ability to identify spatial heterogeneity. In terms of health risk tolerance assessment, this method integrates population heterogeneity and regional socioeconomic differences (e.g., medical resources, response capabilities), shifting from a single environmental exposure model to a complex vulnerability analysis. This makes the results closer to actual stress levels and enhances the application value of the assessment model in urban governance. By integrating fuzzy comprehensive evaluation and the TOPSIS decision-making model, it also enables the scientific ranking of intervention programs, enhancing the quantitative basis for the formulation of control measures.

[0018] In step 1, thermal comfort observation data, air pollutant concentration observation data, and regional meteorological data are ground-based observation data, which can be collected by setting up relevant measuring instruments; Thermal comfort observation data includes air temperature, relative humidity, wind speed, and black sphere temperature in the target urban area. The measured meteorological parameters are input into professional thermal environment assessment software to calculate the Universal Thermal Climate Index (UTCI) for each sampling point, which serves as a unified indicator for measuring thermal comfort.

[0019] Air pollutant concentration monitoring data can include PM2.5 concentration, PM10 concentration, etc. Regional meteorological data can be obtained from observational data released by official meteorological observatories and meteorological data released by open meteorological data platforms; In step 1, geospatial data includes urban morphology data, land cover data, topographic and location data, and socioeconomic data. The specific method for collecting geospatial data is as follows: Urban morphology data can be obtained by extracting building heights using building vector data and aerial imagery, and calculating urban morphology parameters for each spatial grid based on a spatial analysis model. Specifically, building heights can be extracted using building vector data and high-resolution aerial imagery. The target urban area under study can then be gridded, and spatial analysis models (such as the SEBE model plugin) can be used to calculate urban morphology parameters such as building density, floor area ratio, and sky visibility factor (SVF) for each grid in batches. Land cover data can be classified and indexed based on satellite imagery to obtain the land cover ratio. Specifically, by processing satellite imagery (such as Sentinel-2), supervised classification and index calculations (such as NDVI, NDWI, NDBI) are used to extract the vegetation, water body, and impervious surface cover ratios for each grid.

[0020] Topographic and location data are extracted using digital elevation models (such as ASTER GDEM) to obtain the average elevation and slope for each grid. Socioeconomic data are obtained based on street-level population structure data and points of interest information from statistical and API data; population structure data includes the proportion of various population groups, especially the proportion of the elderly and children; points of interest information may include public areas such as schools, hospitals, and parks. All numerical parameters used for modeling are normalized to eliminate the influence of dimensions and provide a standardized data foundation for subsequent model training. The standardized data are then registered according to latitude and longitude coordinates and timestamps to obtain multi-source heterogeneous data. Each multi-source heterogeneous data record includes: geographic coordinates of the sampling point, observation timestamp, UTCI value (thermal comfort index), PM2.5 concentration value, and regional meteorological data. In some embodiments, in step 2, the Synergistic Environmental Stress Index (SESI) includes a nonlinear interaction function term for simulating the synergistic enhancement effect of high temperature and air pollution, and a weighted superposition term of a single environmental indicator; the formula for calculating the Synergistic Environmental Stress Index (SESI) is as follows:

[0021] in, is the index value at position i; Tnorm,i and Pnorm,i are the normalized thermal comfort index and air pollutant concentration index at position i, respectively; α and β are objective weights; γ is a synergistic effect adjustment coefficient. It is a non-linear interactive function.

[0022] Furthermore, the nonlinear interaction function fint can be any of the following functions: 1) Product form: fint = Tnorm,i × Pnorm,i; 2) Modify the Arrhenius form: ; 3) Gaussian kernel function form: ; Where Ea is the apparent activation energy, R is the gas constant, Tabs,i is the absolute temperature, ct and cp are the central parameters, and σ is the bandwidth coefficient. The above functional form can be selected and optimized according to the actual stress response characteristics to more accurately characterize the synergistic mechanism of high temperature and pollution.

[0023] Entropy weighting is an objective weighting method that determines the weight of indicators based on the degree of variation in the data of each indicator. The smaller the information entropy of an indicator, the greater the degree of variation in its value, the more information it provides, and the greater its weight should be in the comprehensive evaluation. Furthermore, the weight coefficients α and β in the collaborative environmental stress index can be calculated based on the entropy weighting method (EWM), and the specific calculation process is as follows: Step 11: Construct the original data matrix: Assuming there are m observation samples and n evaluation indicators, construct the original data matrix X; In this embodiment, there are 300 sampling points, n=2, representing thermal comfort index and air pollutant concentration index, respectively. The observed value of the j-th index of the i-th sample is denoted as x. ij .

[0024] Step 12: Data standardization processing: To eliminate the influence of different dimensions and orders of magnitude of the indicators, the original data is standardized. Since all indicators are positive (the larger the value, the more severe the stress), the max-min normalization method can be used: ; in, These are the standardized values; max(xj) and min(xj) are the maximum and minimum values ​​of the j-th indicator in all samples, respectively. The standardized data is the T mentioned earlier. norm,i and P norm,i .

[0025] Step 13: Calculate the sample weight: Calculate the proportion p of the i-th sample under the j-th indicator. ij This step calculates the contribution of each sample to a specific indicator, using the following formula: ; Step 14: Calculate information entropy: Calculate the information entropy e of the j-th indicator. j Information entropy e j This reflects the dispersion of the j-th indicator, e j The smaller the value, the greater the variability of the indicator and the more information it contains. The calculation formula is: ; Where, constant , where m is the number of samples. To ensure ln(p ij It is meaningful when p ij When = 0, define p ij ln(p ij = 0.

[0026] Step 15: Calculate the information entropy difference coefficient. :d j With information entropy e j Inversely proportional, d j The larger the value, the greater the importance of the indicator for evaluation. The calculation formula is: ; Step 16: Based on the obtained information entropy difference coefficient Calculate the indicator weights: By normalizing the difference coefficients of each indicator, the final weight coefficients are obtained: The above process is performed on the data of the two indicators respectively to obtain the corresponding weight coefficients α and β; In step 3, the target urban spatial area is divided into grids to construct a graph structure. The constructed graph structure includes grid nodes and edges. Geospatial data is assigned to each grid node, and the adjacency relationship between nodes is defined as an edge. In the above implementation, when addressing the problem of organizing spatial data into a graph by dividing the target urban spatial area into a grid, it is first clarified that the grid nodes in the graph structure correspond to specific locations within the urban spatial area. These locations are endowed with rich geospatial data, such as topography, land use type, and population density. The adjacency relationships between nodes, as edges, reflect the spatial correlation between different locations. This correlation can be based on proximity or complex connections based on factors such as function and transportation. In this way, the spatial data is effectively organized into a graph structure, providing a solid foundation for the subsequent learning tasks performed by the graph neural network.

[0027] Based on the graph structure described above, a machine learning model framework, namely a geographic artificial intelligence model, is constructed. In this embodiment, a hybrid geographic neural network model (Hybrid Geo-GNN) is used to construct the geographic artificial intelligence model, and the SESI value of each grid node is predicted based on this hybrid geographic neural network model. In some embodiments, the Hybrid Geo-GNN model includes: The backbone network of a graph neural network (GNN) is used to learn and aggregate the spatial structure features of each node and its neighborhood, extract features, and obtain a preliminary spatial representation of each node's information in order to capture local spatial autocorrelation. The attention mechanism module is used to dynamically assign weights to the input features of the neighboring nodes of different nodes based on the features output by the graph neural network (GNN) backbone. Residual connections and an output layer are used to map the aggregated features to the final SESI prediction.

[0028] A method for constructing a hybrid geographic neural network model, using the GNN algorithm to learn and aggregate spatial features of nodes and adjacency relationships in a graph structure, and predicting the SESI value of each grid node to obtain a collaborative environmental stress map includes the following steps: Step 31: Obtain geospatial data and ground observation data of some sampling points in the target area, and calculate the SESI value based on the calculation formula of the Cooperative Environmental Stress Index (SESI); Step 32: Use geospatial data as input and the calculated SESI value as a label to train the hybrid geographic neural network model, and obtain the trained hybrid geographic neural network model. Step 33: Obtain geospatial data of the entire target city and input it into the trained hybrid geographic neural network model to obtain the SESI values ​​of each location point in the entire target city. Step 34: Using a GIS system, add the SESI values ​​to the map of the target city to obtain a collaborative environmental stress map; GIS is short for Geographic Information System, a comprehensive information system used for collecting, storing, managing, analyzing, displaying, and outputting geospatial information.

[0029] Furthermore, in the hybrid geographic neural network model, predicting the SESI value based on the input geospatial data includes the following steps: Step 331: Through the graph neural network (GNN) backbone network, aggregate and extract features from the information of each node and its neighboring nodes to obtain the preliminary spatial representation of each node's information, i.e., the extracted spatial features; Specifically, in the backbone of a graph neural network (GNN), the following GNN operations are performed: Neighbor information aggregation: Each node receives and aggregates the feature information of its neighboring nodes; Spatial structure feature extraction: Spatial features are extracted from the aggregated information of each node through multi-layer graph convolution (GATConv) to capture local spatial autocorrelation; After multiple layers of stacked GNN operations, a preliminary spatial representation of the information of each node is obtained; Step 332: The attention mechanism module performs feature weighting. Through the attention mechanism module, different weights are assigned to neighboring nodes, and the initial spatial representations of the nodes are weighted and fused to obtain node feature representations that enhance the ability to express spatial heterogeneity, including: Step 3321: Calculate attention weights: Perform attention operations on the nodes and assign different attention weights to each neighbor node to reflect its influence on the target node; Step 3322: Based on the attention weights, perform weighted fusion of the spatial features of different nodes, output the weighted feature vector, and enhance the model's ability to express local differences.

[0030] In this step, the specific calculation formula for node feature update can be expressed as:

[0031] Where N(i) is the set of neighbors of node i. It is the feature of neighbor j at the k level. It is a learnable weight matrix, and σ is a non-linear activation function (such as ReLU). The key is the attention coefficient. It represents the importance of neighbor j's information to node i, and is calculated by a learnable attention network based on the features of nodes i and j.

[0032] Step 333: Input the weighted fused node features into the residual connection structure and perform weighted fusion with the preliminary spatial representation of each node information in the original input; Step 334: The fused features of the residual connection output are mapped to the corresponding SESI prediction values ​​through the fully connected output layer.

[0033] In the above implementation, a synergistic environmental stress index (SESI) incorporating nonlinear interaction terms is first constructed to quantify the synergistic aggravation effect of high temperature and air pollution. Then, a geographic artificial intelligence model is used to integrate multi-source geospatial data to generate a high-resolution synergistic environmental stress map covering the entire region. This enables an explicit spatial characterization of multi-dimensional environmental pressures in the complex urban geographical environment. By introducing graph neural networks and attention mechanisms, the nonlinear interaction patterns and spatial heterogeneity response characteristics of high temperature and pollutants in different regions are effectively captured. The embedding of residual structures further improves the model's convergence and prediction stability, making the SESI values ​​exhibit significant gradient differences and clustering patterns at the urban block scale. The final generated synergistic environmental stress map can accurately identify high-stress hotspots, providing a scientific basis for urban public health risk early warning and climate adaptation planning.

[0034] Step 4 involves modeling environmental exposure, population sensitivity, and social adaptability to construct a comprehensive vulnerability assessment model (IVAM). Based on a collaborative environmental stress map, the process of calculating the comprehensive vulnerability index for each node in the target urban spatial area includes the following steps: Step 41: The density of sensitive POIs (such as schools and hospitals) calculated through kernel density analysis is weighted and superimposed with the gridded population age structure ratio data to obtain the population sensitivity value of the grid nodes in the target city. The specific process is as follows: Step 411: Obtain POI data, including data on activity spaces representing sensitive populations, such as medical facilities, schools, and elderly care institutions; Step 412: Perform kernel density analysis to calculate the distribution intensity of sensitivity POIs in each grid region; Step 413: Obtain grid population age structure data, such as the proportion of elderly people and children; Step 414: Weighted overlay and fusion of the distribution intensity of sensitive POIs with population structure data, each grid node has a population sensitivity value Sj, which is used to construct a sensitivity layer S; Step 42: By comprehensively evaluating indicators such as green space coverage, accessibility of public service facilities, and regional economic level through the indicator system method, the social adaptability index value ACj of the target city grid node is obtained; Social adaptability modeling, calculating the adaptability value ACj of grid nodes, includes the following steps: Step 421: Collect adaptation index data, such as: Vegetation coverage; Accessibility of public service facilities (such as large hospitals and subway stations); Regional socioeconomic level (such as average house price); Step 422: Construct a multi-indicator comprehensive evaluation model and normalize different types of adaptive indicators; Step 423: Use the AHP method to determine the weights of each fitness index, and generate the social fitness value ACj for each grid node by weighted summation, finally obtaining the fitness layer AC; Step 43: Couple environmental exposure, population sensitivity and social adaptability to form a model and construct an integrated vulnerability assessment model (IVAM). Calculate the integrated vulnerability index based on the obtained social adaptability index value, population sensitivity index value and SESI predicted value. The expression for the constructed Integrated Vulnerability Assessment Model (IVAM) is as follows:

[0035] in, The comprehensive vulnerability index of grid node j in the target city; The environmental exposure of the target city grid node j is the SESI value obtained in step 3. ACj represents the population sensitivity index value of grid node j in the target city; ACj represents the social adaptability index value of grid node j in the target city. The superscripts we, ws, wa of each indicator value represent the weight of each dimension, which can be determined by expert scoring or analytic hierarchy process (AHP). In step 5, the identified high-vulnerability areas are used as decision-making units. The method integrates fuzzy comprehensive evaluation and TOPSIS multi-criteria decision-making methods to prioritize intervention measures for these high-vulnerability areas and generate intervention action plans. This process includes the following steps: Step 51: For each decision-making unit, set fuzzy criteria to conduct fuzzy evaluation, obtain the evaluation values ​​of each area of ​​the target city under different evaluation criteria, and obtain the fuzzy evaluation matrix. Specifically, the criteria for setting fuzzy evaluation include: the urgency of intervention, the effectiveness of intervention, and the feasibility of implementation; For the decision-making area, fuzzy scoring of each criterion is carried out by organizing experts or using model assignment; using the fuzzy comprehensive evaluation method, fuzzy linguistic information (such as "higher" and "medium") is converted into quantitative scores to obtain the evaluation values ​​of each area of ​​the target city under different criteria, and a fuzzy evaluation matrix is ​​constructed as the fuzzy evaluation result; Step 52: Use the fuzzy evaluation matrix as input to the TOPSIS model, calculate the distance between each highly vulnerable area and the optimal and worst intervention options, and obtain the priority ranking of each intervention option.

[0036] Among them, the optimal intervention plan and the worst intervention plan are the positive ideal solution and the negative ideal solution in the TOPSIS method; The fuzzy comprehensive evaluation results obtained in step 51 are used as input to the TOPSIS model to rank the intervention priorities of each highly vulnerable area. The core of this process is to determine the ranking by calculating the relative proximity of each area to a theoretical "optimal reference point" (positive ideal solution) and a theoretical "worst reference point" (negative ideal solution). The specific process is as follows: Step 521: Construct positive and negative ideal solutions; the positive ideal solution (A+) consists of the optimal value under each fuzzy evaluation criterion in the decision matrix; the negative ideal solution (A-) consists of the worst value under each fuzzy evaluation criterion. Step 522: For each highly vulnerable region, calculate the Euclidean distance between its evaluation value vector and the positive ideal solution (A+) and the negative ideal solution (A-), denoted as […]. and : ; ; in, This represents the evaluation value of the i-th highly vulnerable region under the j-th evaluation index; This represents the positive ideal value (the best indicator value in all regions) for this indicator; it also represents the negative ideal value (the worst indicator value in all regions) for this indicator.

[0037] Step 523: Calculate and sort the relative proximity scores; relative proximity scores The closer the value is to 1, the closer the region is to the ideal state, and the higher its intervention priority. Finally, the intervention priorities are determined by sorting the values ​​in descending order of Ci.

[0038] To make the technical solution of this embodiment clearer, a specific embodiment will be described below.

[0039] This embodiment takes a city as an example to assess the synergistic risks of its "urban air quality" and "heat island effect" and to determine the priority areas for intervention.

[0040] First, following steps 1 to 4, relevant geographical and meteorological data were collected and processed to construct and calculate the Air Quality Index (AQI) and the Urban Heat Island Effect Index (HIEI). Their weights were determined using the entropy weight method (e.g., α=0.6, β=0.4), and then the Cooperative Environmental Stress Index (SESI) for the entire city was calculated.

[0041] Then, proceed to step 5, identifying the top three regions (regions A, B, and C) with the highest SESI values ​​as high-vulnerability regions. Next: In step 51, through expert fuzzy evaluation, a decision matrix is ​​obtained for the three regions under the three criteria of "intervention urgency", "intervention effectiveness" and "implementation feasibility", as shown in Table 1; Table 1 Decision Matrix;

[0042] In step 52, using the above matrix as input, perform TOPSIS sorting: 1) Determine the positive ideal solution A+ = {0.9, 0.8, 0.8} and the negative ideal solution A- = {0.6, 0.6, 0.4}.

[0043] 2) Calculate the Euclidean distance from each region to A+ and A-. and .

[0044] 3) Calculate the relative closeness C i Value. Calculated result: C for region A. i The value is 0.51, and the C value of region B is... i The value is 0.60, and the C value of region C is... i It is 0.30.

[0045] 4) According to C i The values ​​are sorted to obtain the final intervention priority as: Region B > Region A > Region C.

[0046] This embodiment comprehensively considers the synergistic effects of environmental stress, the heterogeneity of urban space, and the vulnerability of the population. It can scientifically and accurately assess the complex environmental risks in cities, and while improving the accuracy of the assessment, it provides a scientific priority ranking for intervention decisions.

[0047] To illustrate the implementation process and effects of the above-mentioned urban intervention and control methods, a simulation experiment was conducted to verify the results. The composite environmental risk assessment and intervention were carried out using a northern city as an example. The specific details are as follows: In this implementation example, multi-source heterogeneous data covering the city center were obtained through on-site measurements, remote sensing image interpretation, and acquisition from open data platforms, providing a data foundation for subsequent modeling.

[0048] Optionally, the ground observation data can be acquired as follows: On a typical sunny, hot, and calm summer day in July 2023, a measurement team will use a multi-functional measuring instrument (such as the Testo 480) and an aerosol monitor (such as the TSI DustTrak II 8530) to conduct simultaneous measurements at 10:00 AM, 2:00 PM, and 6:00 PM at 100 sampling points covering different local climate zones (LCZs). The data will include air temperature, humidity, wind speed, black sphere temperature, and real-time PM2.5 concentration. The measured meteorological parameters will be input into software to calculate the Universal Thermal Climate Index (UTCI). This will ultimately yield a ground-value dataset of 300 sets containing precise geographic coordinates, timestamps, UTCI values, and PM2.5 concentrations.

[0049] In this example, the geospatial data is obtained as follows: 1) Urban morphology data: Building heights are extracted using building vector data and high-resolution aerial images. The study area is then gridded (e.g., 100mm×100mm). Spatial analysis models (e.g., the SEBE model plugin) are used to batch calculate a total of 20 urban morphology parameters for each grid, including building density, floor area ratio, and sky visibility factor (SVF).

[0050] 2) Land cover data: By processing satellite imagery (such as Sentinel-2), supervised classification and index calculation (such as NDVI, NDWI, NDBI) are used to extract the vegetation, water body and impervious surface cover ratios for each grid.

[0051] 3) Topographic and location data: Using digital elevation models (such as ASTER GDEM), extract the average elevation and slope of each grid.

[0052] 4) Socioeconomic data: Obtain street-level population structure data (especially the elderly and children) and point of interest (POI) data (such as schools, hospitals, parks, etc.) through statistical yearbooks and map APIs.

[0053] All the data acquired were standardized to the same coordinate system (e.g., WGS_1984_UTM_Zone_50N) and resampled to a uniform spatial resolution (e.g., 100m). All numerical parameters used for modeling were subjected to max-min normalization to eliminate the influence of dimensions.

[0054] The next step is to construct a Synergistic Environmental Stress Index (SESI) that comprehensively considers the combined effects of thermal environment and air pollution, quantifies the combined impact of thermal comfort and air pollution, and especially their synergistic enhancement effect.

[0055] First, thermal comfort indices (such as UTCI) and air pollution indices (such as PM2.5 concentration) are selected. Then, the entropy weight method (EWM) is applied to objectively calculate the weights of the two indices. In this embodiment, 300 sets of measured data are used to construct a 300*2 matrix, and the weights of UTCI (α=0.58) and PM2.5 (β=0.42) are calculated.

[0056] To reflect the synergistic effect, a nonlinear interaction term is introduced into the calculation of the synergistic environmental stress index based on linear superposition. In this embodiment, a product form is selected as the interaction function, and the synergistic effect adjustment coefficient γ = 0.3 is calibrated according to relevant epidemiological study results. Finally, the formula for calculating the Cooperative Environmental Stress Index (SESI) is as follows:

[0057] Where SESIi is the exponent value at position i; and Here, represents the normalized UTCI value and PM2.5 concentration at location i, respectively. Substituting the measured data from 100 observation points into this formula, the true SESI value is calculated and used as the label for the next step of training the hybrid geographic neural network model.

[0058] Reference Figure 3 To address the issue of sparse observation points, this step utilizes a hybrid geographic neural network model to construct a geographic artificial intelligence model, integrating geospatial features to predict the SESI distribution across the entire urban area.

[0059] First, the study area is divided into a 100m × 100m grid (approximately 1,000,000 grids), with each grid considered a graph node. Spatial adjacency relationships between nodes (such as Queen adjacency) are defined as graph edges. The feature vector of each node consists of 25 geospatial parameters (urban morphology, land cover, topography) obtained in step 1. In other implementations, graph edges can be defined not only by spatial adjacency relationships but also by functional associations (such as two business district nodes connected) or transportation connectivity (such as nodes connected along major roads) to construct more complex spatial relationship graphs.

[0060] Optionally, the Geo-GNN model uses graph attention network convolutional (GATConv) layers to learn and aggregate the spatial features of each node and its neighborhood to capture the spatial autocorrelation of geographical phenomena. The core advantage of GAT lies in assigning different weights to different neighboring nodes through a self-attention mechanism, thereby capturing spatial heterogeneity more precisely.

[0061] The calculated SESI ground truth values ​​of 100 observation points were used as training labels, and the training and validation sets were divided in an 8:2 ratio. Through model training, the coefficient of determination (R²) between the predicted and true values ​​on the validation set reached 0.91, indicating that the model has high prediction accuracy.

[0062] Optionally, the trained model can be applied to all 1,000,000 grid nodes to predict the SESI value for each grid and visualize it in GIS software to generate a high-resolution map of the cooperative environmental stress index distribution.

[0063] Reference Figure 4 The purpose of assessing spatial vulnerability is to identify areas and populations that are not only severely stressed by the environment but also have weak resilience.

[0064] The Integrated Vulnerability Assessment Model (IVAM) couples three dimensions: Exposure (E), Sensitivity (S), and Adaptability (AC). Its calculation formula is as follows: ; Where Vj is the comprehensive vulnerability index of spatial unit j; we, ws, and wa are the weights of each dimension, which can be determined by expert scoring or the analytic hierarchy process (AHP). In this embodiment, we=0.5, ws=0.3, and wa=0.2 are determined by expert scoring.

[0065] Optionally, the methods for generating layers in each dimension are as follows: 1) Exposure (E) layer: Use the SESI distribution map generated in step S300 directly.

[0066] 2) Sensitivity (S) layer: Kernel density analysis is performed on sensitive POIs such as schools and hospitals, and the data is weighted and overlaid with gridded data on the proportion of elderly and young populations.

[0067] 3) Adaptability (AC) layer: This layer comprehensively assesses multiple indicators such as vegetation coverage, distance to large hospitals, and regional socioeconomic level (e.g., average house price as a proxy indicator), and obtains the weighted sum using the AHP method.

[0068] Then, the IVAM formula is applied in the GIS raster calculator to calculate the comprehensive vulnerability index of each grid, and it is divided into 5 levels according to the natural breakpoint method to generate a comprehensive vulnerability level map.

[0069] Reference Figure 5 Areas with a vulnerability level of "very high" (such as 10 communities) were selected as decision-making units. An evaluation system was constructed that includes three primary criteria ("urgency", "effectiveness" and "feasibility") and multiple secondary criteria.

[0070] Optionally, the decision-making model employs a combined approach of AHP, fuzzy comprehensive evaluation, and TOPSIS. First, the weights of each criterion are determined using AHP. Then, experts are invited to conduct fuzzy evaluations of the performance of different alternative interventions (such as increasing parks, street spraying, and building greening) for each community under each secondary criterion. Finally, the fuzzy evaluation results are clarified and used as input to the TOPSIS model to calculate the relative proximity Ci for each alternative; a higher value indicates a higher priority.

[0071] Optional, refer to Figure 7 The ranking results are visualized on a map to generate an urban environmental intervention action map. The highest priority areas are highlighted, and interactive information boxes display their specific problems, vulnerability rankings, recommended interventions, and their priority scores.

[0072] In summary, this embodiment first constructs the SESI index, which includes nonlinear interaction terms, to scientifically quantify the synergistic effect of heat and pollution. Secondly, it utilizes a Geo-GNN model to integrate multi-source geographic data, achieving high-precision global risk mapping. Then, it constructs an IVAM model to accurately identify hotspot areas with high overall vulnerability. Finally, through a multi-criteria decision-making model, it provides a quantitative and scientific priority ranking of interventions for urban governance. This method comprehensively improves the refinement of urban complex environmental risk assessment and the scientific rigor of decision support.

[0073] Example 2 Based on Example 1, this example provides a system for assessing and intervening in urban complex environmental health risks, including: The data management module is configured to acquire geospatial data and ground observation data for the target urban area; The synergistic effect calculation module is configured to perform nonlinear interactive modeling of the synergistic enhancement effect of high temperature and air pollution, construct a synergistic environmental stress index, and calculate the SESI prediction value based on the acquired ground observation data. The spatial prediction modeling module is configured to divide the target city spatial area into a grid and construct a graph structure; use the GNN algorithm to learn and aggregate spatial features of the node features and adjacency relationships in the graph structure, predict the SESI prediction value of each grid node, and obtain a collaborative environmental stress map. The vulnerability assessment module is configured to model environmental exposure, population sensitivity, and social adaptability in a coupled manner, construct a comprehensive vulnerability assessment model, calculate the comprehensive vulnerability index of each node in the target urban spatial area based on a collaborative environmental stress map, generate a comprehensive vulnerability level map of urban space oriented towards the population, and identify high-vulnerability areas. The decision support module is configured to use highly vulnerable areas as decision-making units, integrate fuzzy comprehensive evaluation and TOPSIS multi-criteria decision-making methods, prioritize intervention measures for highly vulnerable areas, and generate intervention action plans.

[0074] The visualization and interaction module is configured to display the output results of each module in the form of maps, charts, and reports, and to provide users with interactive query and scenario simulation functions.

[0075] It should be noted that each module in this embodiment corresponds one-to-one with each step in embodiment 1, and their specific implementation process is the same, so it will not be repeated here.

[0076] Example 3 Based on Embodiment 1, this embodiment provides an electronic device, including a memory and a processor, as well as computer instructions stored in the memory and running on the processor. When the computer instructions are executed by the processor, they complete the steps in the method for assessing and intervening in the health risks of a complex urban environment as described in Embodiment 1.

[0077] Example 4 Based on Embodiment 1, this embodiment provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, they complete the steps in the method for assessing and intervening in urban complex environmental health risks described in Embodiment 1.

[0078] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0079] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for assessing and intervening in urban complex environmental health risks, characterized in that, Includes the following steps: Acquire geospatial data and ground observation data for the target urban area; Nonlinear interaction modeling was used to investigate the synergistic enhancement effect of high temperature and air pollution, and a synergistic environmental stress index (SESI) was constructed. The predicted value of SESI was calculated based on the obtained ground observation data. The target city spatial area is divided into grids to construct a graph structure; the GNN algorithm is used to learn and aggregate spatial features of the nodes and adjacency relationships in the graph structure, and the SESI prediction value of each grid node is predicted to obtain a collaborative environmental stress map. By modeling environmental exposure, population sensitivity, and social adaptability, a comprehensive vulnerability assessment model is constructed. Based on a collaborative environmental stress map, the comprehensive vulnerability index of each node in the target urban spatial area is calculated, generating a population-oriented comprehensive vulnerability level map of urban space and identifying high-vulnerability areas. By using highly vulnerable areas as decision-making units, and integrating fuzzy comprehensive evaluation and TOPSIS multi-criteria decision-making methods, intervention measures for highly vulnerable areas are prioritized and intervention action plans are generated.

2. The method for assessing and intervening in urban complex environmental health risks as described in claim 1, characterized in that: The synergistic environmental stress index includes a nonlinear interaction function term used to simulate the synergistic enhancement effect of high temperature and air pollution, and a weighted superposition term of individual environmental indicators.

3. The method for assessing and intervening in urban complex environmental health risks as described in claim 2, characterized in that: The formula for calculating the Cooperative Environmental Stress Index (SESI) is as follows: ； in, is the index value at location i; Tnorm,i and Pnorm,i are the normalized thermal comfort index and air pollutant concentration index at location i, respectively; α and β are weights; γ is a synergistic effect adjustment coefficient. It is a non-linear interactive function.

4. The method for assessing and intervening in urban complex environmental health risks as described in claim 1, characterized in that: The target urban spatial area is divided into grids and constructed into a graph structure. The constructed graph structure includes grid nodes and edges. Geospatial data is assigned to each grid node, and the adjacency relationship between nodes is defined as an edge.

5. The method for assessing and intervening in urban complex environmental health risks as described in claim 1, characterized in that: A method for constructing a hybrid geographic neural network model, using the GNN algorithm to learn and aggregate spatial features of nodes and adjacency relationships in a graph structure, and predicting the SESI prediction value of each grid node to obtain a collaborative environmental stress map includes the following steps: Geospatial data and ground observation data of some sampling points in the target area are obtained, and the SESI value is calculated based on the formula of the Cooperative Environmental Stress Index (SESI). Geospatial data is used as input, and the calculated SESI value is used as a label. The input is fed into the hybrid geospatial neural network model for training, resulting in the trained hybrid geospatial neural network model. Obtain geospatial data of the entire target city and input it into the trained hybrid geographic neural network model to obtain the SESI values ​​of each location point in the entire target city. By using a GIS system, the SESI values ​​are added to the map of the target city to obtain a collaborative environmental stress map.

6. The method for assessing and intervening in urban complex environmental health risks as described in claim 1, characterized in that: In a hybrid geographic neural network model, predicting SESI values ​​based on input geospatial data involves the following steps: By using the graph neural network backbone, the information of each node and its neighboring nodes is aggregated and features are extracted to obtain a preliminary spatial representation of the information of each node, namely the extracted spatial features. The attention mechanism module assigns different weights to neighboring nodes, and the initial spatial representations of the nodes are weighted and fused to obtain node feature representations that enhance the ability to express spatial heterogeneity. The fused features are output through residual connections, and the node features are mapped to the corresponding SESI prediction values ​​based on the fully connected output layer.

7. The method for assessing and intervening in urban complex environmental health risks as described in claim 1, characterized in that: The process of modeling environmental exposure, population sensitivity, and social adaptability to construct an integrated vulnerability assessment model (IVAM), and calculating the integrated vulnerability index of each node in the target urban spatial area based on a collaborative environmental stress map, includes the following steps: The population sensitivity values ​​of grid nodes in the target city can be obtained by coupling the sensitivity POI density calculated through kernel density analysis with gridded population age structure data. ; By comprehensively evaluating indicators such as green space coverage, accessibility of public service facilities, and regional economic level using the indicator system method, the social adaptability index value ACj of the target city grid node is obtained. By modeling environmental exposure, population sensitivity, and social adaptability, a comprehensive vulnerability assessment model (IVAM) is constructed, expressed as: ； in, The comprehensive vulnerability index of grid node j in the target city; The environmental exposure of grid node j in the target city is represented by its SESI value. The population sensitivity index value for the target city grid node; ACj represents the social adaptability index value of grid node j in the target city. The comprehensive vulnerability index is calculated based on the obtained social adaptability index value, population sensitivity index value, and SESI predicted value.

8. A system for assessing and intervening in urban complex environmental health risks, characterized in that, include: The data management module is configured to acquire geospatial data and ground observation data for the target urban area; The synergistic effect calculation module is configured to perform nonlinear interactive modeling of the synergistic enhancement effect of high temperature and air pollution, construct a synergistic environmental stress index, and calculate the SESI prediction value based on the acquired ground observation data. The spatial prediction modeling module is configured to divide the target city spatial area into a grid and construct a graph structure; use the GNN algorithm to learn and aggregate spatial features of the node features and adjacency relationships in the graph structure, predict the SESI prediction value of each grid node, and obtain a collaborative environmental stress map. The vulnerability assessment module is configured to model environmental exposure, population sensitivity, and social adaptability in a coupled manner, construct a comprehensive vulnerability assessment model, calculate the comprehensive vulnerability index of each node in the target urban spatial area based on a collaborative environmental stress map, generate a comprehensive vulnerability level map of urban space oriented towards the population, and identify high-vulnerability areas. The decision support module is configured to use highly vulnerable areas as decision-making units, integrate fuzzy comprehensive evaluation and TOPSIS multi-criteria decision-making methods, prioritize intervention measures for highly vulnerable areas, and generate intervention action plans.

9. An electronic device, characterized in that, It includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, which, when executed by the processor, perform the steps in the method for assessing and intervening in the health risks of a complex urban environment as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, Used to store computer instructions, which, when executed by a processor, complete the steps in the method for assessing and intervening in the complex urban environment health risks as described in any one of claims 1-7.

Images