Urban heat risk network-based spatial form optimization method, device and equipment

Abstract

The present application relates to the field of urban planning and management, and particularly relates to a spatial form optimization method, device and equipment based on a city heat risk network, the method comprising: acquiring multi-source remote sensing data of a city, and constructing a spatial basic data set; identifying a risk surface, a corridor and an obstacle point based on the spatial basic data set, and constructing a high / low high heat risk network; extracting multi-dimensional landscape features of each network component, and training an optimal prediction model; extracting key form threshold values of each landscape feature in different network components; and generating spatial form optimization strategies corresponding to different network components in the high / low heat risk network. The present application accurately identifies the spatial structure of the city heat risk through a geographic information system and landscape ecology, and constructs independent high / low heat risk networks, while combining an interpretable machine learning to quantize form control threshold values, so as to convert an abstract heat environment improvement target into an operable planning and design index, and to realize accurate management of different components in the high / low heat risk network.

Description

Technical Field

[0001] This invention relates to the field of urban planning and management, and in particular to a method, apparatus and equipment for spatial morphology optimization based on urban thermal risk networks. Background Technology

[0002] Currently, research on mitigating urban heat islands largely focuses on analyzing the cooling effects of single landscape elements (such as water bodies and green spaces) or constructing cold island networks based on the "patch-corridor-matrix" theory. However, existing methods still have significant limitations: First, the construction of traditional cold island networks relies heavily on surface temperature as a physical indicator, lacking comprehensive consideration of the population dimension. This fails to organically combine thermal hazards with thermal exposure and vulnerability, potentially leading to "resource misallocation" in mitigation strategies—over-allocating cooling measures in low-population, high-exposure areas while neglecting high-risk areas with higher temperatures and population densities. Second, existing analyses of driving mechanisms often assume that urban thermal environment driving factors are spatially homogeneous and linear, ignoring the heterogeneous influence of different spatial functions. This fails to effectively distinguish the essential differences in dominant driving factors and thresholds among surfaces as thermal risks, corridors as heat transport channels, and barriers as thermal barriers. Therefore, there is an urgent need for a comprehensive evaluation framework that can integrate "thermal hazards, thermal exposure, and thermal vulnerability," and a technical method that can construct high / low thermal risk networks based on this framework, while accurately analyzing the nonlinear morphological thresholds under different network functions, in order to support the precision and effectiveness of urban thermal risk regulation. Summary of the Invention

[0003] This invention provides a spatial morphology optimization method, apparatus, and equipment based on urban thermal risk networks. It solves the technical problems in the existing technology, such as the lack of consideration of population dimension in the construction of heat island or cold island networks and the neglect of spatial heterogeneity of spatial functions on dominant factors and thresholds. By constructing high and low thermal risk networks and combining interpretable machine learning, it can achieve precise management of different components in high and low thermal risk networks, including surfaces, corridors, and obstacle points.

[0004] A first aspect of this invention provides a spatial morphology optimization method based on urban thermal risk networks, comprising the following steps:

[0005] Step 1: Acquire multi-source remote sensing data of the city, and preprocess and standardize the multi-source remote sensing data to construct a spatial basic dataset;

[0006] Step 2: Based on the aforementioned spatial basic dataset, perform morphological spatial pattern analysis to identify risk surfaces, calculate the comprehensive resistance surface, and identify corridors and obstacle points based on the minimum cumulative resistance model and circuit theory, and construct independent high thermal risk networks and low-high thermal risk networks.

[0007] Step 3: Extract multidimensional landscape features of each network component and train the optimal prediction model. The input of the optimal prediction model is multidimensional landscape features, and the output is surface temperature. The network components include the surface, corridors and obstacle points of the high / low heat risk network.

[0008] Step 4: Calculate the feature contribution of each landscape feature in the optimal prediction model using the SHAP model, and establish a SHAP local dependency graph to extract the key morphological thresholds of each landscape feature in different network components.

[0009] Step 5: Generate spatial morphology optimization strategies for different network components in high / low thermal risk networks based on the key morphology thresholds.

[0010] A second aspect of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the spatial morphology optimization method based on urban thermal risk networks described above.

[0011] A third aspect of the present invention provides a spatial morphology optimization device based on an urban thermal risk network, comprising a computer-readable storage medium and a processor, wherein the processor executes a computer program on the computer-readable storage medium to implement the steps of the spatial morphology optimization method based on an urban thermal risk network described above.

[0012] A fourth aspect of this invention provides a spatial morphology optimization device based on an urban thermal risk network, comprising a data acquisition module, a network construction module, a model optimization module, a parsing module, and a strategy generation module.

[0013] The data acquisition module is used to acquire multi-source remote sensing data of the city, and to preprocess and standardize the multi-source remote sensing data to construct a spatial basic dataset.

[0014] The network construction module is used to perform morphological spatial pattern analysis to identify risk surfaces based on the spatial basic dataset, calculate the comprehensive resistance surface, and identify corridors and obstacle points based on the minimum cumulative resistance model and circuit theory, and construct independent high thermal risk networks and low-high thermal risk networks.

[0015] The model optimization module is used to extract multidimensional landscape features of each network component and train the optimal prediction model. The input of the optimal prediction model is multidimensional landscape features, and the output is surface temperature. The network components include the surface, corridors, and obstacle points of the high / low thermal risk network.

[0016] The parsing module is used to calculate the feature contribution of each landscape feature in the optimal prediction model using the SHAP model, and to establish a SHAP local dependency graph to extract the key morphological thresholds of each landscape feature in different network components.

[0017] The strategy generation module is used to generate spatial morphology optimization strategies for different network components in high / low thermal risk networks based on the key morphology threshold.

[0018] The beneficial effects of this invention are as follows: This invention provides a spatial morphology optimization method, apparatus, and device based on urban thermal risk networks, which have the following beneficial effects compared with the prior art:

[0019] (1) It has realized the transformation from simply mitigating the physical heat island to managing heat risks from a human-centered perspective. By integrating population distribution and elderly population data to construct a "heat hazard-heat exposure-heat vulnerability" framework, the identified heat risk network is more in line with the actual health needs of the city, avoiding the resource misallocation problem of only focusing on physical cooling while ignoring the exposure of the population.

[0020] (2) An analytical framework of “thermal environment spatial function dependence” was proposed. High thermal risk network and low thermal risk network were constructed respectively, and the differential driving mechanism of morphological factors in different network components, namely surfaces, corridors and obstacle points, was revealed. This overcame the limitations of traditional research that assumed that the driving factors of urban thermal environment were homogeneous and linear.

[0021] (3) By adopting interpretable machine learning technology, not only can high-precision surface temperature prediction be achieved in different network components, but more importantly, specific morphological control thresholds are quantified, transforming abstract thermal environment improvement goals into operable planning and design indicators, such as specific building density and tree height, providing scientific and quantitative decision-making basis for urban renewal, climate adaptability planning and resilient city construction.

[0022] To make the above-mentioned objects, features and advantages of the invention more apparent and understandable, preferred embodiments of the invention are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0023] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 A schematic flowchart of the spatial morphology optimization method provided in an embodiment of the present invention;

[0025] Figure 2a This is the key core of the high-heat-risk zone in the embodiments of the present invention;

[0026] Figure 2bThis is the key core of the low-heat risk zone in the embodiments of the present invention;

[0027] Figure 2c This is the resistance surface of the high-heat-risk area in the embodiments of the present invention;

[0028] Figure 2d This is the resistance surface in the low-heat-risk zone of this invention embodiment;

[0029] Figure 3a This is a high-thermal-risk network according to an embodiment of the present invention;

[0030] Figure 3b This is a low-thermal-risk network according to an embodiment of the present invention;

[0031] Figure 4 This is a diagram illustrating the global importance and local interpretation of the surface, corridors, and obstacle points in the high-heat-risk area according to an embodiment of the present invention.

[0032] Figure 5 This is a diagram illustrating the global feature importance and local interpretation of surfaces, corridors, and obstacle points in the low-heat-risk area according to an embodiment of the present invention.

[0033] Figure 6 This is a local dependency map of the dominant landscape features of surfaces, corridors, and obstacle points in high / low heat risk zones in an embodiment of the present invention;

[0034] Figure 7 This is a schematic diagram of the spatial morphology optimization device provided in an embodiment of the present invention;

[0035] Figure 8 This is a schematic diagram of the spatial morphology optimization device provided in an embodiment of the present invention. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without inventive effort are within the scope of protection of this invention.

[0037] It should be noted that, unless otherwise specified, the various features in the embodiments of this invention can be combined with each other, all of which are within the protection scope of this invention. Furthermore, although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than the module division in the device or the order in the flowchart. Moreover, the terms "first," "second," and "third" used in this invention do not limit the data or execution order, but only distinguish identical or similar items with essentially the same function and effect.

[0038] Figure 1 This is a flowchart illustrating a spatial morphology optimization method based on an urban thermal risk network, as provided in Example 1. Figure 1 As shown, it includes the following steps:

[0039] Step 1: Acquire multi-source remote sensing data of the city, and preprocess and standardize the multi-source remote sensing data to construct a spatial basic dataset;

[0040] Step 2: Based on the aforementioned spatial basic dataset, perform morphological spatial pattern analysis to identify risk surfaces, calculate the comprehensive resistance surface, and identify corridors and obstacle points based on the minimum cumulative resistance model and circuit theory, and construct independent high thermal risk networks and low-high thermal risk networks.

[0041] Step 3: Extract multidimensional landscape features from each network component and train the optimal prediction model. The input of the optimal prediction model is the multidimensional landscape features, and the output is the surface temperature. The network components include the surface, corridors, and obstacle points of the high / low thermal risk network.

[0042] Step 4: Calculate the feature contribution of each landscape feature in the optimal prediction model using the SHAP model, and establish a SHAP local dependency graph to extract the key morphological thresholds of each landscape feature in different network components.

[0043] Step 5: Generate spatial morphology optimization strategies for different network components in high / low thermal risk networks based on the key morphology thresholds.

[0044] The above embodiments provide a spatial morphology optimization method based on urban thermal risk networks. By using geographic information systems and landscape ecology to accurately identify the spatial structure of urban thermal risks and construct independent high / low thermal risk networks, and by combining interpretable machine learning to quantify specific morphological control thresholds, the abstract thermal environment improvement goals are transformed into operable planning and design indicators, enabling precise management of different components such as surfaces, corridors and obstacle points in high / low thermal risk networks.

[0045] The following specific embodiments will be used to describe each step of the above method in detail.

[0046] In one specific embodiment, the multi-source remote sensing data includes land surface temperature data, socioeconomic data, and environmental characteristic data. A multi-source remote sensing database is constructed using these data, providing a solid data foundation for subsequent analysis. Specifically, step 1, constructing the spatial foundation dataset, includes the following steps:

[0047] Step 101: Select initial thermal infrared band images that meet preset conditions in Google Earth Engine, and fill in the missing areas in the thermal infrared band images using a preset interpolation strategy to obtain continuous surface temperature data.

[0048] For example, a preferred embodiment may use Landsat 8 OLI / TIRS remote sensing imagery as the primary data source. Considering that the urban heat island effect is most pronounced in summer, the specific operation involves selecting image sets from Google Earth Engine (GEE cloud platform) showing typical summer months (July-August) for a city with cloud cover below a preset value, such as below 20%. To ensure data continuity and integrity, for areas covered by clouds or with poor quality, a spatiotemporal interpolation strategy can be used first. This involves selecting image data from nearby years and times for initial filling, and then using co-kriging interpolation for areas with missing data. In co-kriging interpolation, environmental factors such as Normalized Difference Vegetation Index (NDVI), DEM, and slope are selected as covariates. Based on the strong correlation between surface temperature and these environmental factors, the interpolation accuracy is improved, ensuring the spatial continuity of thermal environment data.

[0049] Then, step 102 is executed to obtain socioeconomic data and environmental feature data, and spatial registration, resampling and regional statistical methods are used to integrate the surface temperature data, socioeconomic data and environmental feature data to the same spatial resolution, such as a 100m×100m fishing net grid, to construct a spatial basic dataset.

[0050] For example, the socioeconomic data includes road density data, thermal exposure data (such as population distribution data), and / or thermal vulnerability data (such as distribution data of the elderly population aged 65 and over); the environmental feature data includes at least one of topographic data, local climate zone maps, building morphology data, three-dimensional tree data, land cover data, and spectral indices.

[0051] In a preferred embodiment, road data can be obtained from OpenStreetMap to calculate road density and characterize anthropogenic heat emission intensity; a DEM can be obtained from the Space Shuttle Radar Topographic Mapping (SRTM) and slope can be calculated from the DEM as topographic data; population distribution data at 100m and distribution data of the elderly population aged 65 and above can be obtained from WorldPop to characterize heat exposure and thermal vulnerability; a Local Climate Zone (LCZ) map is used to systematically divide the urban surface into 10 built-up areas and 7 natural cover areas, aiming to provide high-resolution and physically meaningful basic data for urban climate, environment, and energy research; building morphology data comes from 3D-GloBFP, which has developed 31 ensemble models based on multi-source data and compared with heights in Google Earth Street, with an R² of 0.85; three-dimensional tree data, including tree height, comes from the canopy height map published by the Meta and World Resources Institute, and the dataset is created using machine learning algorithms based on Maxar satellite; albedo data is based on Landsat. The imagery was analyzed using a band conversion model. Land cover data was derived from a land cover type dataset, which has an overall classification accuracy of 84.35% ± 0.92%, meeting the required accuracy. Spectral indices were calculated from the bands of the preprocessed Landsat 8 OLI imagery in GEE. Standardizing and integrating these data yielded a complete spatial foundation dataset.

[0052] In a preferred embodiment, step 2 starts from the risk perspective of "thermal hazard-thermal exposure-thermal vulnerability". First, bivariate local spatial autocorrelation analysis is introduced to identify potential high-thermal-risk areas and potential low-thermal-risk areas. Morphological spatial pattern analysis, area screening and connectivity analysis methods are used to determine high-thermal-risk surfaces and low-thermal-risk surfaces respectively. Then, a comprehensive resistance surface containing three dimensions of thermal hazard, thermal exposure and thermal vulnerability is constructed. Based on the minimum cumulative resistance model and circuit theory, high-thermal-risk corridors connecting the high-thermal-risk surfaces and low-thermal-risk corridors connecting the low-thermal-risk sources are identified, and obstacle points in the network are identified, thereby constructing independent high-thermal-risk networks and low-thermal-risk networks.

[0053] For example, calculations can be performed in spatial statistics software such as GeoDa to identify surfaces with high / low thermal risk networks, including the following steps:

[0054] Step 201: Construct two sets of spatial correlations: thermal hazard-thermal exposure and thermal hazard-thermal vulnerability, and generate a local bivariate clustering map. Specifically, the degree of thermal hazard can be characterized by surface temperature, the degree of thermal exposure by population size, and the degree of thermal vulnerability by the number of elderly people, such as those over 65 years old. Therefore, when constructing the two sets of spatial correlations, one set is surface temperature-population size, and the other set is surface temperature-number of people over 65 years old.

[0055] Then, step 202 is executed, identifying potential high-heat-risk areas and potential low-heat-risk areas using the local bivariate clustering map. The potential high-heat-risk areas are the intersection of "high heat hazard - high heat exposure" and "high heat hazard - high heat vulnerability," while the low-heat-risk areas are the intersection of "low heat hazard - low heat exposure" and "low heat hazard - low heat vulnerability." For example, the intersection of "high surface temperature - high total population concentration area" and "high surface temperature - high elderly population concentration area" yields a "potential high-heat-risk area." These areas are both heat island centers and densely populated areas, exhibiting the highest heat risk. Simultaneously, the intersection of "low surface temperature - low total population concentration area" and "low surface temperature - low elderly population concentration area" yields a "potential low-heat-risk area." These areas are typically large bodies of water or forests; although sparsely populated, they serve as important cooling sources and have significant ecological value in mitigating urban thermal conditions.

[0056] In the specific calculation process, the bivariate local Moran's I index is used to calculate the spatial correlation characteristics between surface temperature and total population, and between surface temperature and elderly population, in order to identify potential high-heat-risk areas and potential low-heat-risk areas. The formula for calculating the bivariate local Moran's I index is as follows:

[0057]

[0058] Among them, I i Let x be the bivariate local Moran exponent for grid i. i Let y be the LST in grid i. i This represents the population size or the number of elderly people in grid i. and These are the mean of LST and the mean of the population size or the number of elderly people, respectively. ij Let be the spatial weight matrix for grids i and j.

[0059] Finally, step 203 is executed, in which the potential high-heat-risk area and the potential low-heat-risk area are converted into binary raster images respectively. After extracting the core patches through morphological spatial pattern analysis, the interfering patches are filtered out through the area screening mechanism, and the preferred core patches are extracted through the possible connectivity index and / or the overall connectivity index to generate high-heat-risk surfaces and low-heat-risk surfaces.

[0060] Specifically, in the binary raster image, the foreground represents potential high / low heat risk areas, while the background represents other elements. Morphological spatial pattern analysis is performed using Guidos Toolbox software. Based on mathematical morphology principles, the morphological spatial pattern analysis divides foreground pixels into seven landscape elements: core area, isolated areas, aquifer areas, edge areas, bridging areas, ring areas, and branch areas. This embodiment focuses on the core area, as it is the primary carrier of high or low heat risk. Simultaneously, to eliminate interference from fragmented patches and identify nodes with global influence, an area filtering mechanism is introduced to remove areas with small sizes, such as less than 0.1 km². 2 The remaining core patches were then analyzed using Conefor 2.6 software to calculate the Possible Connectivity Index (PC) and Overall Connectivity Index (IIC) of the remaining core patches. This evaluated the importance of each core patch in maintaining network connectivity, and selected the top 50 patches (those with higher overall importance) as high-heat-risk surfaces and low-heat-risk surfaces, respectively. Figures 2a-2b As shown.

[0061] The formulas for calculating the Possible Connectivity Index (PC) and the Overall Connectivity Index (IIC) are as follows:

[0062]

[0063]

[0064] Where n is the number of high or low heat risk surfaces, a i and a j It is the area of ​​surfaces i and j with high or low heat risk, nl ij A is the number of shortest paths between surfaces i and j, which are either high- or low-heat-risk surfaces. L It is the maximum area of ​​the surface with high or low heat risk. It represents the maximum product probability of all paths between surfaces i and j, which are either high or low thermal risk surfaces.

[0065] In an exemplary, preferred embodiment, identifying corridors and obstacle points in a high / low thermal risk network includes the following steps:

[0066] Step 204: Construct an evaluation framework based on "thermal hazard - thermal exposure - thermal vulnerability," and calculate the resistance values ​​of high / low thermal risk surfaces through weighted superposition to generate a comprehensive high / low thermal risk resistance surface, such as... Figures 2c-2dAs shown. The thermal hazard indicators include local climate zone type, road density, DEM (Digital Elevation Model), and slope; the thermal exposure indicator is population size; and the thermal vulnerability indicator is the number of people aged 65 and above. For high-heat-risk networks, the significance is to simulate the diffusion path of "heat risk" in the city. Therefore, factors promoting heat risk are assigned low resistance, while factors hindering heat risk are assigned high resistance. Furthermore, the resistance value for local climate zone type is set based on its surface heat island intensity.

[0067] For low-heat-risk networks, the significance is to identify transmission paths of low-heat-risk. Therefore, the resistance assignment logic is the opposite of that for high-heat-risk networks. That is, factors that promote the diffusion of the cold island effect are assigned low resistance values, while factors that hinder the flow of cold air are assigned high resistance values. Then, the resistance values ​​of high / low-heat-risk surfaces are calculated by a weighted superposition method to generate high-heat-risk comprehensive resistance surfaces and low-heat-risk comprehensive resistance surfaces that include three dimensions: thermal hazard, thermal exposure, and thermal vulnerability.

[0068] The specific resistance values ​​for high and low heat risk surfaces are shown in Table 1.

[0069] Table 1 Resistance values ​​of high / low thermal risk surfaces

[0070]

[0071]

[0072] Then, step 205 is executed, calculating the minimum cost path on the corresponding comprehensive resistance surface based on the minimum cumulative resistance model. Specifically, Linkage Mapper and Circuitscape software can be used for calculation to identify the corridors connecting high-heat-risk surfaces and the low-heat-risk corridors connecting the low-heat-risk surfaces. The calculation formula for the minimum cumulative resistance model is as follows:

[0073]

[0074] Where MCR is the minimum cumulative resistance model, D ij R is the distance between surfaces i and j, which are considered high or low thermal risk surfaces. i It is the connection resistance of surface i, which has a high or low thermal risk.

[0075] Step 206: Using the Barrier Mapper tool, barrier points on the high-thermal-risk surfaces and the low-thermal-risk surfaces are identified, thereby constructing a high / low dual thermal-risk network containing surfaces, corridors, and barrier points. This provides a spatial basis for subsequent zoned policy implementation. Figure 3a and 3b As shown.

[0076] In a preferred embodiment, in order to analyze the driving mechanisms behind different network components, step 3 extracts multi-dimensional landscape features within each component and constructs the optimal prediction model, including the following steps:

[0077] Step 301: Use a fishing net tool to extract multidimensional landscape features of surfaces, corridors and obstacle points in high / low heat risk networks. The multidimensional landscape features include two-dimensional landscape indicators and three-dimensional urban morphology indicators.

[0078] Specifically, the two-dimensional landscape indices include: landscape percentage (PLAND, representing the dominance of a certain type of landscape), average patch area (AREA_MN, representing fragmentation), edge density (ED, representing shape complexity and edge effect), similarity adjacency percentage (PLADJ, representing clustering), connectivity index (COHESION), normalized difference vegetation index (NDVI, representing vegetation cover), normalized difference building index (NDBI, representing impermeability intensity), and modified normalized difference water index (MNDWI, representing water information), etc.

[0079] The three-dimensional urban morphology indicators include: building density (BD, representing the degree of building clustering), average building height (MBH, representing the average building height), standard deviation of building height (BH_SD, representing the diversity of building height), floor area ratio (FAR, land use intensity), sky visibility (SVF, representing the openness of buildings), average tree height (MTH, representing the average tree height), standard deviation of tree height (MTH_SD, representing the diversity of tree height), albedo (reflectance of the land surface to solar radiation), etc. These indicators can more comprehensively depict the impact of the city's three-dimensional geometry on the thermal environment.

[0080] Step 302: Construct multiple machine learning prediction models based on the surface, corridors and obstacle points in the high / low thermal risk network, and perform hyperparameter tuning. Then, traverse the corresponding preset parameter space through grid search and cross-validation to select the optimal hyperparameter combination of each machine learning model on a specific dataset. Select the machine learning model with the highest accuracy as the optimal prediction model based on preset evaluation indicators. The input of the optimal prediction model is multidimensional landscape features and the output is surface temperature.

[0081] Specifically, Random Forest, Extreme Gradient Boosting Tree (XGBoost), and K-Nearest Neighbors (KNN) algorithms were selected to construct the machine learning prediction model. During training, to improve the prediction accuracy and generalization ability of the model, the preferred embodiment performed hyperparameter tuning on the three algorithms: Random Forest, K-Nearest Neighbors, and Extreme Gradient Boosting Tree. For the Random Forest algorithm, its parameter grid mainly covers the number of trees (n_estimators), the maximum depth of the trees (max_depth), and the minimum number of samples required for the leaf nodes (min_samples_leaf). The search range of n_estimators was set to 100 to 500 with a step size of 100, while the search ranges of max_depth and min_samples_leaf were both 2 to 10 with a step size of 2.

[0082] For the KNN algorithm, the parameter space includes the number of neighbors (n_neighbors), the distance metric, the weights, the search algorithm, and the leaf size (leaf_size). The value of n_neighbors ranges from 1 to 30, covering multiple key nodes; the metric parameter tested Euclidean distance, Manhattan distance, Minkowski distance, and Chebyshev distance; the weights parameter compared uniform weighting and distance-weighted weighting; the algorithm parameter considered four strategies: automatic selection, ball tree, KD tree, and brute-force search; and the leaf_size search range was set to 10 to 50, with a step size of 10.

[0083] For the XGBoost algorithm, parameter optimization focuses on the number of base learners (n_estimators), the maximum tree depth (max_depth), the learning rate (learning_rate), the subsample ratio (subsample), and the minimum loss reduction required for node splitting (gamma). The search range of n_estimators is set to 200 to 1000 with a step size of 200; the value of max_depth ranges from 2 to 10 with a step size of 2; the learning rate covers multiple gradients from 0.01 to 0.3; and the values ​​of subsample and gamma both range from 0.2 to 1 with a step size of 0.2.

[0084] Then, a grid search combined with cross-validation was used to traverse the parameter space to select the optimal hyperparameter combinations for each algorithm on the specific dataset. The coefficient of determination (R²) and root mean square error (RMSE) were used as model evaluation metrics. The hyperparameter results are shown in Table 2. XGBoost showed the highest accuracy in predicting the impact of landscape features on land surface temperature and was therefore used in subsequent analyses.

[0085] Table 2 Comparison of different machine learning prediction models

[0086]

[0087]

[0088] Traditional feature importance assessment can only provide a global overview of how important a feature is, without indicating whether the feature has a "positive" or "negative" impact, let alone revealing the specific nonlinear threshold of the feature. This invention introduces the Shapley Additive Explanations (SHAP) method to extract key morphological thresholds for various landscape features across different network components, specifically including the following steps:

[0089] Step 401: Using the SHAP model and based on the Explainer, calculate the SHAP value of each landscape feature in the optimal prediction model, i.e., the feature contribution. The SHAP value represents the nonlinear marginal contribution of the feature to the surface temperature: a positive value indicates that the feature has a positive contribution to the surface temperature (warming effect), and a negative value indicates that the feature has a negative contribution to the surface temperature (cooling effect).

[0090] Step 402: Summarize the average absolute SHAP values ​​of all samples to demonstrate the global importance ranking and local interpretation of each landscape feature across different network components, such as... Figure 4 and Figure 5 As shown;

[0091] Step 403: Establish a SHAP local dependency graph, and based on the SHAP local dependency graph, analyze the key morphological thresholds of each landscape feature in different network components. The key morphological threshold is the inflection point where the influence of a landscape feature on surface temperature changes from positive to negative or from negative to positive; that is, the critical threshold at which the feature produces a warming or cooling effect. In a specific embodiment, such as... Figure 6 As shown, the horizontal axis represents the actual numerical values ​​of landscape features, and the vertical axis represents the corresponding SHAP values. The system identifies the horizontal axis feature values ​​when the SHAP value changes from negative to positive (or from positive to negative). For example, when the building density exceeds 26%, the SHAP value changes from negative to positive, indicating that 26% is the critical point where building density begins to produce a significant heat accumulation effect. Through this method, the present invention can accurately analyze the specific control ranges of each dominant factor in different network components.

[0092] In a preferred embodiment, the spatial morphology optimization strategy generated in step 5 specifically involves: based on the key morphology threshold and combined with the role mechanism of different landscape features in physical thermodynamics and urban ecology, following the governance path of "spatial identification-factor diagnosis-threshold control", different spatial morphology optimization strategies are formulated for the surfaces, corridors and obstacle points of high-heat-risk networks and low-heat-risk networks.

[0093] Firstly, for high-heat-risk networks, the core of governance lies in suppressing heat generation and accumulation and blocking the spread of heat flow. For high-heat-risk surfaces, which are the core of continuous heat generation and accumulation, the control strategy needs to comprehensively cover building form, vegetation structure, and landscape configuration. Specifically, based on the nonlinear threshold identified by the SHAP dependency graph, building density exceeding 41.59% and normalized difference building index exceeding 0.07 are defined as critical points for accelerated warming. Therefore, development intensity must be strictly limited in such areas to prevent excessive expansion of impermeable surfaces. Regarding the negative correlation between sky openness and surface temperature within a specific range, a significant radiative cooling effect is observed when sky openness is below 0.45. Based on this, it is recommended to optimize building layout, such as reducing building height or increasing building spacing to improve mutual shading, thereby maintaining a lower sky openness to enhance long-wave radiative cooling.

[0094] In terms of urban greening, a two-way threshold effect of average tree height was identified. Low vegetation between 1.8 meters and 14.9 meters can have a warming effect due to insufficient shading and potential obstruction of ventilation. Only when the average tree height exceeds 15 meters does it show a strong cooling function. Based on this, it is proposed that tall trees should be planted instead of low shrubs on high heat risk surfaces, and the percentage of impermeable surface landscape should be controlled below 83% to alleviate heat accumulation.

[0095] For high-heat-risk corridors, the sensitivity to construction intensity is significantly higher than that of surface areas, and the temperature increase threshold of building density is reduced to 26.72%, indicating that stricter density control needs to be implemented in this area. At the same time, given the stable cooling effect exhibited when the modified normalized difference water index is greater than 0.10 and the water area ratio is greater than 0, the introduction of blue infrastructure such as wetlands or linear water bodies in the corridor is established as a priority strategy to cut off heat flow transmission through evaporative cooling effect.

[0096] For high-heat-risk areas, the focus of morphological management should be on optimizing the edge structure. Based on the finding that an impermeable surface edge density of less than 160 is more conducive to cooling, it is recommended to integrate fragmented construction land to form a simpler, more cohesive morphological structure, and combine this with humidity-enhancing measures to alleviate local hotspots.

[0097] Secondly, for low-heat-risk networks, the core of governance lies in protecting the stability of cold sources and maintaining unobstructed cold air circulation paths. For low-heat-risk surfaces, which are the core function of the cold island, the primary task is to maintain their integrity. Given that a normalized difference building index exceeding 0.74 will lead to a significant warming effect, this threshold is set as a red line for ecological protection, and high-intensity development and construction activities are strictly prohibited in such areas.

[0098] For low-heat-risk corridors connecting cold islands, the focus is on maintaining their natural attributes and landscape connectivity. Based on the cooling correlation when the normalized difference building index is less than -0.02 and the albedo is less than 0.09, the emphasis is on maintaining natural surface cover and the application of high-reflectivity materials, while controlling the density of impermeable surface edges to less than 120, so as to ensure appropriate edge complexity to maintain ecological functions.

[0099] For low-heat-risk areas, defensive control strategies must be implemented. Based on the threshold ranges of a normalized difference building index of less than -0.02, impermeable landscape percentage of less than 49%, edge density of less than 240, and sky openness of less than 0.26, the area is required to maintain spatial openness, restrict construction activities, and promote a centralized rather than decentralized development model. In particular, the threshold of a building density exceeding 27.13% that triggers warming has been established as an insurmountable rigid constraint for development in the area.

[0100] Ultimately, through the aforementioned threshold-based nonlinear precision intervention, the abstract thermal environment driving mechanism is transformed into specific spatial planning parameters. This involves implementing "density reduction, shape optimization, and tall tree planting" on high-heat-risk surfaces, "blue sky enhancement and heat insulation" in high-heat-risk corridors, "edge aggregation" at high-heat-risk obstacle points, and "strict red-line control and open maintenance" throughout the low-heat-risk region. This constructs a hierarchical, categorized, and scientifically quantified urban thermal environment resilience enhancement scheme.

[0101] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0102] This invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the spatial morphology optimization method based on urban thermal risk networks described above.

[0103] Figure 7 This is a schematic diagram of the spatial morphology optimization device based on urban thermal risk network provided in the embodiment, such as... Figure 7 As shown, it includes a data acquisition module 100, a network construction module 200, a model optimization module 300, a parsing module 400, and a strategy generation module 500.

[0104] The data acquisition module 100 is used to acquire multi-source remote sensing data of the city, and to preprocess and standardize the multi-source remote sensing data to construct a spatial basic dataset.

[0105] The network construction module 200 is used to perform morphological spatial pattern analysis based on the spatial basic dataset to identify risk surfaces, calculate the comprehensive resistance surface, and identify corridors and obstacle points based on the minimum cumulative resistance model and circuit theory, and construct independent high thermal risk networks and low-high thermal risk networks.

[0106] The model optimization module 300 is used to extract multidimensional landscape features of each network component and train the optimal prediction model. The input of the optimal prediction model is multidimensional landscape features, and the output is surface temperature. The network components include the surface, corridors, and obstacle points of the high / low thermal risk network.

[0107] The parsing module 400 is used to calculate the feature contribution of each landscape feature in the optimal prediction model using the SHAP model, and to establish a SHAP local dependency graph to extract the key morphological thresholds of each landscape feature in different network components.

[0108] The strategy generation module 500 is used to generate spatial morphology optimization strategies for different network components in high / low thermal risk networks based on the key morphology threshold.

[0109] The above embodiments provide a spatial morphology optimization device based on urban thermal risk networks. By using geographic information systems and landscape ecology, the spatial structure of urban thermal risks is accurately identified, and independent high / low thermal risk networks are constructed. At the same time, interpretable machine learning is combined to quantify specific morphological control thresholds, transforming abstract thermal environment improvement goals into operable planning and design indicators, and realizing precise management of different components such as surfaces, corridors, and obstacle points in high / low thermal risk networks.

[0110] In a preferred embodiment, the data acquisition module 100 specifically includes:

[0111] The first data acquisition unit is used to filter initial thermal infrared band images that meet preset conditions in Google Earth Engine, and fill in the missing areas in the thermal infrared band images through a preset interpolation strategy to obtain continuous surface temperature data.

[0112] The second data acquisition unit is used to acquire socio-economic data and environmental characteristic data, and to integrate the surface temperature data, socio-economic data and environmental characteristic data into a fishing net grid with the same spatial resolution using spatial registration, resampling and regional statistical methods to construct a spatial basic dataset.

[0113] In a preferred embodiment, the network construction module 200 includes a surface recognition unit, which specifically includes:

[0114] The building blocks are used to construct two sets of spatial relationships: thermal hazard-thermal exposure and thermal hazard-thermal vulnerability, and generate local bivariate clustering graphs.

[0115] The first identification unit is used to identify potential high-heat-risk areas and potential low-heat-risk areas through the local bivariate clustering map. The potential high-heat-risk areas are the intersection of "high-heat hazard - high-heat exposure" and "high-heat hazard - high-heat vulnerability", and the low-heat-risk areas are the intersection of "low-heat hazard - low-heat exposure" and "low-heat hazard - low-heat vulnerability".

[0116] The second identification unit is used to convert the potential high-heat-risk area and the potential low-heat-risk area into binary raster images respectively, extract core patches through morphological spatial pattern analysis, filter interfering patches through area screening mechanism, and extract preferred core patches through possible connectivity index and / or overall connectivity index to generate high-heat-risk surface and low-heat-risk surface.

[0117] In a preferred embodiment, the network construction module 200 includes a corridor identification unit, the corridor identification unit comprising:

[0118] The first calculation unit is used to construct an evaluation framework based on "thermal hazard-thermal exposure-thermal vulnerability" and calculate the resistance values ​​of high / low thermal risk surfaces by weighted superposition to generate a comprehensive resistance surface with high / low thermal risk.

[0119] The third identification unit is used to calculate the minimum cost path on the corresponding comprehensive resistance surface based on the minimum cumulative resistance model, and to identify the low-heat-risk corridor connecting the high-heat-risk corridor and the low-heat-risk surface.

[0120] In a preferred embodiment, the network construction module 200 further includes an obstacle point identification unit for identifying obstacle points on the high-heat-risk surface and the low-heat-risk surface based on the Barrier Mapper tool.

[0121] In a preferred embodiment, the model selection module 300 specifically includes:

[0122] The extraction unit is used to extract multi-dimensional landscape features of surfaces, corridors and obstacle points in high / low thermal risk networks using a fishing net tool. The multi-dimensional landscape features include two-dimensional landscape indicators and three-dimensional urban morphology indicators.

[0123] The training unit is used to construct multiple machine learning prediction models based on surfaces, corridors, and obstacle points in high / low thermal risk networks, and to perform hyperparameter tuning. Then, the corresponding preset parameter space is traversed through grid search combined with cross-validation to select the optimal hyperparameter combination of each machine learning model on a specific dataset. Based on preset evaluation indicators, the machine learning model with the highest accuracy is selected as the optimal prediction model. The input of the optimal prediction model is multidimensional landscape features, and the output is surface temperature.

[0124] Furthermore, in a preferred embodiment, the parsing module 400 specifically includes:

[0125] The second calculation unit uses the SHAP model to calculate the SHAP value of each landscape feature in the optimal prediction model, i.e., the feature contribution.

[0126] The summarization unit is used to summarize the average absolute value of all samples to show the global importance ranking and local interpretation of each landscape feature in different network components;

[0127] The parsing unit is used to establish a SHAP local dependency graph and, based on the SHAP local dependency graph, parse the key morphological thresholds of each landscape feature in different network components. The key morphological thresholds are the inflection points where the influence of a landscape feature on land surface temperature changes from positive to negative or from negative to positive.

[0128] Furthermore, in a preferred embodiment, the strategy generation module 500 is specifically used to formulate differentiated spatial morphology optimization strategies for the surfaces, corridors, and obstacle points of high-heat-risk networks and low-heat-risk networks based on the key morphology threshold and in combination with the role mechanism of different landscape features in physical thermodynamics and urban ecology, following the governance path of "spatial identification-factor diagnosis-threshold control".

[0129] This invention also provides a spatial morphology optimization device based on an urban thermal risk network, including a computer-readable storage medium and a processor. When the processor executes a computer program on the computer-readable storage medium, it implements the steps of the spatial morphology optimization method based on an urban thermal risk network described above.

[0130] Figure 8 This is a schematic diagram of the spatial morphology optimization device based on urban thermal risk network provided in an embodiment of the present invention, as shown below. Figure 8 As shown, the spatial morphology optimization device 8 based on an urban thermal risk network in this embodiment includes: a processor 80, a readable storage medium 81, and a computer program 82 stored in the readable storage medium 81 and executable on the processor 80. When the processor 80 executes the computer program 82, it implements the steps in the various method embodiments described above, for example... Figure 1The steps shown. Alternatively, when the processor 80 executes the computer program 82, it implements the functions of each module in the above-described device embodiments, for example... Figure 7 The functions of the module shown.

[0131] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0132] The present invention is not limited to the description in the specification and embodiments, and thus other advantages and modifications can be readily realized by those skilled in the art. Therefore, the present invention is not limited to the specific details, representative devices and illustrated examples shown and described herein without departing from the spirit and scope of the general concept as defined by the claims and their equivalents.

Claims

1. A spatial morphology optimization method based on urban thermal risk networks, characterized in that, Includes the following steps: Step 1: Acquire multi-source remote sensing data of the city, and preprocess and standardize the multi-source remote sensing data to construct a spatial basic dataset; Step 2: Based on the aforementioned spatial basic dataset, perform morphological spatial pattern analysis to identify risk surfaces, calculate the comprehensive resistance surface, and identify corridors and obstacle points based on the minimum cumulative resistance model and circuit theory, and construct independent high thermal risk networks and low-high thermal risk networks. Step 3: Extract multidimensional landscape features from each network component and train the optimal prediction model. The input of the optimal prediction model is the multidimensional landscape features, and the output is the surface temperature. The network components include the surface, corridors, and obstacle points of the high / low thermal risk network. Step 4: Use the SHAP model to calculate the feature contribution of each landscape feature in the optimal prediction model, and establish a SHAP local dependency graph to extract the key morphological thresholds of each landscape feature in different network components. Step 5: Generate spatial morphology optimization strategies for different network components in the high / low thermal risk network based on the key morphology threshold. Step 3, training the optimal prediction model, includes the following steps: Step 301: Use a fishing net tool to extract multi-dimensional landscape features of surfaces, corridors and obstacle points in high / low heat risk networks. The multi-dimensional landscape features include two-dimensional landscape indicators and three-dimensional urban morphology indicators. Step 302: Construct multiple machine learning prediction models based on the surface, corridors and obstacle points in the high / low thermal risk network, and perform hyperparameter tuning. Then, traverse the corresponding preset parameter space through grid search and cross-validation to select the optimal hyperparameter combination of each machine learning model on a specific dataset. Select the machine learning model with the highest accuracy as the optimal prediction model based on preset evaluation indicators. The input of the optimal prediction model is multidimensional landscape features and the output is surface temperature.

2. The spatial morphology optimization method based on urban thermal risk network according to claim 1, characterized in that, The multi-source remote sensing data includes surface temperature data, socioeconomic data, and environmental characteristic data. Step 1, constructing the spatial foundation dataset, includes the following steps: Step 101: Filter initial thermal infrared band images that meet preset conditions in Google Earth Engine, and fill in the missing areas in the thermal infrared band images using a preset interpolation strategy to obtain continuous surface temperature data. Step 102: Acquire socioeconomic data and environmental characteristic data, and use spatial registration, resampling and regional statistical methods to integrate the surface temperature data, socioeconomic data and environmental characteristic data into a fishing net grid to construct a spatial basic dataset.

3. The spatial morphology optimization method based on urban thermal risk network according to claim 2, characterized in that, The socioeconomic data includes road density data, thermal exposure data, and / or thermal vulnerability data; The environmental feature data includes at least one of the following: topographic data, local climate zone map, building morphology data, three-dimensional tree data, albedo data, land cover data, and spectral index.

4. The spatial morphology optimization method based on urban thermal risk network according to claim 1, characterized in that, Identifying surfaces of high / low thermal risk networks includes the following steps: Step 201: Construct two sets of spatial associations: thermal hazard-thermal exposure and thermal hazard-thermal vulnerability, and generate a local bivariate clustering graph; Step 202: Identify potential high-heat-risk areas and potential low-heat-risk areas using the local bivariate clustering graph. The potential high-heat-risk areas are the intersection of "high-heat hazard - high-heat exposure" and "high-heat hazard - high-heat vulnerability". The potential low-heat-risk areas are the intersection of "low-heat hazard - low-heat exposure" and "low-heat hazard - low-heat vulnerability". Step 203: Convert the potential high-heat-risk area and the potential low-heat-risk area into binary raster maps respectively. After extracting the core patches through morphological spatial pattern analysis, filter out the interfering patches through the area screening mechanism, and extract the preferred core patches through the possible connectivity index and / or the overall connectivity index to generate high-heat-risk surfaces and low-heat-risk surfaces.

5. The spatial morphology optimization method based on urban thermal risk network according to claim 4, characterized in that, Identifying corridors and obstacle points in high / low thermal risk networks includes the following steps: Step 204: Construct an evaluation framework based on "thermal hazard-thermal exposure-thermal vulnerability", and calculate the resistance values ​​of high / low thermal risk surfaces by weighted superposition to generate a comprehensive resistance surface for high / low thermal risk. Step 205: Calculate the minimum cost path on the corresponding comprehensive resistance surface based on the minimum cumulative resistance model, and identify the low-heat-risk corridors connecting the high-heat-risk corridors and the low-heat-risk surfaces. Step 206: Identify the barrier points on the high-heat-risk surface and the low-heat-risk surface using the BarrierMapper tool.

6. The spatial morphology optimization method based on urban thermal risk network according to claim 1, characterized in that, Step 4 extracts key morphological thresholds for each landscape feature across different network components, including the following steps: Step 401: Calculate the SHAP value of each landscape feature in the optimal prediction model using the SHAP model, i.e., the feature contribution. Step 402: Summarize the average absolute value of SHAP for all samples to show the global importance ranking and local interpretation of each landscape feature in different network components; Step 403: Establish a SHAP local dependency graph, and based on the SHAP local dependency graph, analyze the key morphological thresholds of each landscape feature in different network components. The key morphological thresholds are the inflection points where the influence of landscape features on land surface temperature changes from positive to negative or from negative to positive.

7. The spatial morphology optimization method based on urban thermal risk network according to claim 1, characterized in that, In step 5, a spatial morphology optimization strategy is generated. Specifically, based on the key morphology threshold and combined with the role mechanism of different landscape features in physical thermodynamics and urban ecology, a differentiated spatial morphology optimization strategy is formulated for the surfaces, corridors and obstacle points of high-heat-risk networks and low-heat-risk networks, following the governance path of "spatial identification-factor diagnosis-threshold control".

8. A spatial morphology optimization device based on an urban thermal risk network, based on the method described in any one of claims 1-7, characterized in that, It includes a data acquisition module, a network construction module, a model optimization module, a parsing module, and a policy generation module. The data acquisition module is used to acquire multi-source remote sensing data of the city, and to preprocess and standardize the multi-source remote sensing data to construct a spatial basic dataset. The network construction module is used to perform morphological spatial pattern analysis to identify risk surfaces based on the spatial basic dataset, calculate the comprehensive resistance surface, and identify corridors and obstacle points based on the minimum cumulative resistance model and circuit theory, and construct independent high thermal risk networks and low-high thermal risk networks. The model optimization module is used to extract multidimensional landscape features of each network component and train the optimal prediction model. The input of the optimal prediction model is multidimensional landscape features, and the output is surface temperature. The network components include the surface, corridors, and obstacle points of the high / low thermal risk network. The parsing module is used to calculate the feature contribution of each landscape feature in the optimal prediction model using the SHAP model, and to establish a SHAP local dependency graph to extract the key morphological thresholds of each landscape feature in different network components. The strategy generation module is used to generate spatial morphology optimization strategies for different network components in high / low thermal risk networks based on the key morphology threshold.

9. A spatial morphology optimization device based on an urban thermal risk network, comprising a computer-readable storage medium and a processor, characterized in that, When the processor executes the computer program on the computer-readable storage medium, it implements the steps of the spatial morphology optimization method based on urban thermal risk network as described in any one of claims 1-7.

Images