A machine learning driven urban land surface heat island intensity mechanism analysis method fusing local climate zone framework

Abstract

This application provides a machine learning-driven method for analyzing the intensity mechanism of urban surface heat islands, integrating a local climate zone framework. The method includes: acquiring geospatial data of an urban agglomeration region; dividing the urban agglomeration region into multiple LCZ type units based on the geospatial data under the LCZ1-LCZ10 / LCZA-LCZG classification system to obtain LCZ classification results; determining SUHII based on the geospatial data and LCZ classification results to obtain SUHII spatial distribution data; extracting two-dimensional and three-dimensional urban morphology factors based on multi-source geospatial data; training an XGBoost model based on the SUHII spatial distribution data and urban morphology factors to obtain an optimized XGBoost model; performing a two-scale analysis on the optimized XGBoost model using the SHAP analysis method to obtain two-scale analysis results; and generating thermal environment regulation strategies for the entire urban agglomeration and various LCZ units based on the two-scale analysis results.

Description

Technical Field

[0001] This application relates to the fields of urban thermal environment management and urban planning technology, and in particular to a machine learning-driven method for analyzing the intensity mechanism of urban surface heat islands that integrates a local climate zone framework. Background Technology

[0002] The driving mechanism of the urban heat island effect refers to the causal relationship between abnormally high urban surface temperatures and anthropogenic factors such as urban underlying surface structure, land use, and building layout. Accurately identifying the key urban morphological elements that dominate heat island formation and their mechanisms of action is fundamental to developing scientific thermal environment control strategies.

[0003] In related technologies, the analysis of the driving mechanism of the urban heat island effect is usually based on the inversion of land surface temperature from remote sensing data, and combined with statistical models or machine learning methods to analyze the impact of urban morphology indicators on the intensity of the heat island. Such methods are effective in reflecting the overall thermal environment characteristics of cities.

[0004] However, due to the significant differences in the underlying surface composition and spatial configuration of different areas within the city, it is difficult to fully capture the unique thermal response characteristics of each area when using a unified model to fit the entire region. This results in the limited applicability of the identified driving factors in local areas, which in turn affects the accurate implementation of thermal environment control strategies. Summary of the Invention

[0005] In view of the above, it is necessary to provide a machine learning-driven method for analyzing the intensity mechanism of urban surface heat islands that integrates a local climate zone framework. This method aims to solve the technical problem that existing urban heat island effect analysis techniques cannot accurately and systematically reveal the driving mechanism of surface urban heat island intensity in different local climate zone types at the urban agglomeration scale.

[0006] On the one hand, this application provides a machine learning-driven method for analyzing the urban surface heat island intensity mechanism by integrating a local climate zone framework. The method includes: acquiring multi-source geospatial data of a target urban agglomeration region; dividing the target urban agglomeration region into multiple local climate zone (LCZ) type units based on the multi-source geospatial data to obtain LCZ classification results; determining the surface urban heat island intensity (SUHII) of each LCZ type unit based on the multi-source geospatial data and the LCZ classification results to obtain SUHII spatial distribution data; and extracting multiple urban morphology factors based on the multi-source geospatial data, wherein the urban morphology factors are used to characterize the urban structure. The system is structured as follows: Based on the spatial distribution data of SUHII and the urban morphology factors, a training sample set is constructed. Using SUHII as the dependent variable and the multiple urban morphology factors as independent variables, a preset model is trained to obtain an optimized model. The optimized model is then subjected to dual-scale analysis to obtain the factor contribution of each urban morphology factor to SUHII, the influence law of each urban morphology factor on SUHII, and the interaction effect among the urban morphology factors at both the overall urban agglomeration scale and the LCZ unit scale. Based on the factor contribution, the influence law, and the interaction effect, thermal environment regulation strategies for the overall urban agglomeration and various LCZ units are generated.

[0007] In some embodiments of this application, the multi-source geospatial data includes land surface temperature, the LCZ type unit includes LCZD type unit, and the step of determining the surface urban heat island intensity (SUHII) of each LCZ type unit based on the multi-source geospatial data and the LCZ classification results to obtain the SUHII spatial distribution dataset includes: determining the average land surface temperature of the LCZD type unit based on the land surface temperature; and determining the surface urban heat island intensity (SUHII) of each LCZ type unit using the average land surface temperature of the LCZD type unit as a reference temperature.

[0008] In some embodiments of this application, the step of performing a dual-scale analysis on the optimization model to obtain the factor contribution of each urban morphology factor to SUHII at both the overall urban agglomeration scale and the LCZ unit scale, the influence law of each urban morphology factor on SUHII, and the interaction effect between the urban morphology factors includes: analyzing the factor contribution of each urban morphology factor to SUHII and the influence law of the urban morphology factors on SUHII at the overall urban agglomeration scale; analyzing the contribution of different urban morphology factors under the same LCZ type at the LCZ unit scale to determine the factor contribution set corresponding to each LCZ type unit; and determining the interaction effect between each urban morphology factor by constructing a SHAP interaction graph.

[0009] In some embodiments of this application, the preset model is an XGBoost model, and training the preset model to obtain an optimized model includes: using SUHII as the dependent variable and the multiple urban morphology factors as independent variables, training the XGBoost model based on the training sample set to obtain an XGBoost optimized model.

[0010] In some embodiments of this application, the analysis of the contribution of each urban morphology factor to SUHII at the overall scale of the urban agglomeration and the influence law of the urban morphology factor on SUHII includes: based on the training sample set, using the SHAP analysis method to analyze the XGBoost optimization model to obtain the SHAP value of each urban morphology factor to SUHII, wherein the SHAP value is the first factor contribution at the overall scale of the urban agglomeration; obtaining the nonlinear relationship between the SHAP value and the corresponding urban morphology factor, and determining the influence law of each urban morphology factor on SUHII based on the nonlinear relationship.

[0011] In some embodiments of this application, determining the interaction effects between various urban morphology factors by constructing a SHAP interaction graph includes: calculating the SHAP interaction value set corresponding to each LCZ type unit based on the XGBoost optimization model, where the SHAP interaction value is the interaction strength between pairs of interaction factors, and the pair of interaction factors consists of any two urban morphology factors; and constructing a SHAP interaction graph based on the SHAP interaction value set, using the main effect factor in the pair of interaction factors as the independent variable and the SHAP interaction value as the dependent variable.

[0012] In some embodiments of this application, the method includes: determining the dominant influencing factor at the overall scale of the urban agglomeration based on the SHAP value corresponding to each urban morphology factor; and determining the dominant influencing factor of each LCZ type unit based on the factor contribution set corresponding to each LCZ type unit.

[0013] In some embodiments of this application, the step of acquiring multi-source geospatial data of a target urban agglomeration region, and dividing the target urban agglomeration region into multiple local climate zone (LCZ) type units based on the multi-source geospatial data to obtain LCZ classification results includes: extracting urban feature data based on the multi-source geospatial data; acquiring a preset sample set, which includes urban feature sample data corresponding to all LCZ types; dividing the preset sample set into a training set and a test set; training a preset model based on the training set to obtain an initial classification model; inputting the urban feature data into the initial classification model to generate initial LCZ type labels; performing Gaussian filtering optimization on the initial LCZ type labels to form continuous LCZ type units; and performing accuracy verification on the continuous LCZ type units based on the test set. If the verification result does not meet the preset conditions, the initial classification model is iteratively optimized until the preset conditions are met, and the LCZ classification result is output.

[0014] In some embodiments of this application, the step of performing accuracy verification on the continuous LCZ type units based on the test set, and if the verification result does not meet the preset conditions, iteratively optimizing the initial classification model until the preset conditions are met, and outputting the LCZ classification result includes: comparing the continuous LCZ type units with the test set, constructing a confusion matrix and calculating the Kappa coefficient; if the Kappa coefficient is lower than a preset threshold, adjusting the parameters of the preset model and returning to the step: training the preset model based on the training set to obtain an initial classification model until the Kappa coefficient is lower than the preset threshold, and outputting the LCZ classification result.

[0015] In some embodiments of this application, the urban morphology factor includes two-dimensional feature factors and three-dimensional feature factors.

[0016] In the machine learning-driven urban surface heat island intensity mechanism analysis method based on the integrated local climate zone framework provided in this application embodiment, the urban agglomeration is divided into multiple LCZ type units based on multi-source geospatial data. Since LCZ classification itself is a standardized clustering of the urban underlying surface structure, it naturally distinguishes regions with different thermal environment response types, thereby enabling the division of cities into standard units with internal homogeneity and external heterogeneity. On this basis, the surface urban heat island intensity SUHII of each LCZ unit is calculated, and multiple morphological factors characterizing urban structural features are extracted by combining multi-source geospatial data to construct a machine learning optimization model with SUHII as the dependent variable. Furthermore, by performing dual-scale analysis of the optimization model at the overall scale of the urban agglomeration and the scale of the LCZ unit, the contribution of each factor at different scales, the nonlinear influence law, and the interaction effect between factors are obtained, thereby taking into account both macroscopic commonalities and local characteristics within a unified framework. Therefore, the generated thermal environment regulation strategy includes both an overall guidance applicable to the entire region and precise and operable differentiated measures for each type of LCZ unit, which significantly improves the accuracy and spatial specificity of the analysis of the driving mechanism of urban surface heat island intensity in local climate zones. It effectively solves the problem that existing urban heat island effect analysis techniques are insufficient to accurately and systematically reveal the driving mechanism of surface urban heat island intensity in different local climate zone types at the urban agglomeration scale. Attached Figure Description

[0017] Figure 1 This is a flowchart of a machine learning-driven method for analyzing the intensity mechanism of urban surface heat islands that integrates a local climate zone framework, according to an embodiment of this application.

[0018] Figure 2 This is a flowchart of an embodiment of the LCZ classification method provided in this application.

[0019] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0020] It should be noted that in this application, "at least one" means one or more, and "more than one" means two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and drawings of this application are used to distinguish similar objects, not to describe a specific order or sequence.

[0021] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a specific manner. Unless otherwise specified, the following embodiments and features described herein can be combined with each other.

[0022] In the field of urban heat island effect analysis, the traditional urban-rural dichotomy method, as an early analytical approach for urban heat island intensity, treats the city as a whole when calculating temperature differences with the suburbs. This completely ignores the significant heterogeneity of surface materials, spatial structure, and functional layout within the city, leading to an inability to accurately locate key areas of influence. Consequently, the analysis results offer very little guidance for optimizing the urban thermal environment. The core reason for this problem lies in its failure to consider the high degree of heterogeneity of the urban thermal environment under the influence of urbanization.

[0023] While analytical techniques for driving factors and urban heat islands have improved compared to earlier versions, fundamental shortcomings remain. First, although machine learning-driven analysis schemes cover multidimensional urban morphological factors, they fail to fully integrate the characteristics of different areas within the city for classification and analysis, neglecting the heterogeneity of factor influences under different underlying surface types.

[0024] To address the aforementioned technical issues, this application provides a machine learning-driven method for analyzing the urban surface heat island intensity mechanism, which integrates a local climate zone framework. Starting from the three core requirements of accuracy, systematicity, and interpretability, this method integrates the LCZ (Local Climate Zone) framework, multidimensional urban morphology indicators, and interpretable machine learning models to construct a dual-scale analysis system, thereby achieving in-depth analysis of the driving mechanism of SUHII (Surface Urban Heat Island Intensity).

[0025] The machine learning-driven urban surface heat island intensity mechanism analysis method based on the integrated local climate zone framework provided in this application can be applied to one or more electronic devices, including but not limited to servers, geographic information system workstations, cloud computing platforms, or localized computing units with spatial data analysis capabilities. This application does not limit the type of electronic device.

[0026] like Figure 1 The diagram shown is a flowchart of a machine learning-driven method for analyzing the intensity mechanism of urban surface heat islands, which integrates a local climate zone framework, according to an embodiment of this application. The order of the steps in this flowchart can be adjusted according to different needs, and some steps can be omitted.

[0027] S10, acquire multi-source geospatial data of the target urban agglomeration area, and based on the multi-source geospatial data, divide the target urban agglomeration area into multiple local climate zone (LCZ) type units to obtain LCZ classification results.

[0028] In some embodiments of this application, the electronic device acquires multi-source geospatial data of the target urban agglomeration area. The multi-source geospatial data may include remote sensing satellite image data, urban building vector data, open source map data, vegetation cover, water body distribution related data, terrain roughness data, land use type data, etc.

[0029] In some embodiments of this application, after acquiring multi-source geospatial data of the target urban agglomeration area, the electronic device performs unification and verification processing on the multi-source geospatial data to form a standardized dataset. Specifically, the electronic device converts different types of data into the same geographic coordinate system (such as WGS84 or CGCS2000) and resamples all raster data to a consistent spatial resolution. Simultaneously, the electronic device performs cross-comparison of data from different sources to identify and correct spatial misalignments, attribute conflicts, or outliers. After the above processing, a standardized multi-source dataset of the study area is generated.

[0030] like Figure 2 The image shows an embodiment of the LCZ classification method provided in this application. In some embodiments of this application, multi-source geospatial data of a target urban agglomeration region is obtained. Based on the multi-source geospatial data, the target urban agglomeration region is divided into multiple local climate zone LCZ type units to obtain LCZ classification results, including: S21. Based on the multi-source geospatial data, extract urban feature data.

[0031] In some embodiments of this application, the LCZ classification system is adopted, and based on the multi-source geospatial data, the target urban agglomeration region is divided into multiple local climate zone (LCZ) type units. The LCZ classification system is a classification system based on the concept of local climate, which divides regional climate into several local climate zones according to different surface types.

[0032] In one specific embodiment, based on the multi-source geospatial data, urban feature data is extracted, including: electronic devices extracting land cover type, building density, building height, and vegetation cover features for LCZ classification based on the multi-source geospatial data. Land cover type can be obtained through remote sensing image classification, for example, using maximum likelihood or random forest classifiers to classify remote sensing images into categories such as water bodies, vegetation, buildings, and bare soil. Building density is obtained by calculating the ratio of building coverage area to total land surface area. Building height can be extracted from LiDAR (Light Detection and Ranging) data or DSM (Digital Surface Model) data. Vegetation cover can be calculated using NDVI (Normalized Difference Vegetation Index).

[0033] S22, Obtain a preset sample set, which includes city feature sample data corresponding to all LCZ types; In some embodiments of this application, obtaining a preset sample set specifically includes: taking representative sample points covering all 17 LCZ types (LCZ1–LCZ10 and LCZA–LCZG) within the target urban agglomeration area; associating each sample point with a multi-dimensional urban feature vector and a corresponding LCZ type label, thereby constructing a balanced and spatially representative LCZ training sample set.

[0034] S23, the preset sample set is divided into a training set and a test set; S24, Train the preset model based on the training set to obtain an initial classification model; In some embodiments of this application, the preset model may be any one of random forest, XGBoost, support vector machine or convolutional neural network.

[0035] S25, Input the city feature data into the initial classification model to generate initial LCZ type labels; In some embodiments of this application, the urban feature data is input into an initial classification model to generate initial LCZ type labels. Specifically, the study area is divided into 17 LCZ types according to the local climate zone standard, including 10 built LCZ types (LCZ1–LCZ10) and 7 natural LCZ types (LCZA–LCZG). The built LCZ types include: compact high-rise (LCZ1), compact mid-rise (LCZ2), compact low-rise (LCZ3), open high-rise (LCZ4), open mid-rise (LCZ5), open low-rise (LCZ6), low-density low-rise (LCZ7), large low-rise (LCZ8), dispersed building area (LCZ9), and industrial area (LCZ10). The 7 natural LCZ types include: dense forest area (LCZA), sparse forest area (LCZB), shrub area (LCZC), low vegetation area (LCZD), rock or hardened bare area (LCZE), bare land area (LCZF), and water body (LCZG).

[0036] S26, Gaussian filtering optimization is performed on the initial LCZ type label to form continuous LCZ type units; In some embodiments of this application, after generating initial LCZ type labels, since the initial classification results are usually output in pixels, frequent category jumps or isolated patches may occur between adjacent pixels, resulting in discontinuous spatial distribution and fragmented boundaries, making it difficult to reflect the macroscopic morphological characteristics of the actual urban structure. Therefore, Gaussian filtering optimization is applied to the initial LCZ type labels to form continuous LCZ type units. Specifically, a local window centered on the current pixel is constructed, and the probability of each category occurring within the window is weighted and averaged. The weights are determined by a Gaussian function, with pixels closer to the center contributing more. After filtering, the LCZ type with the highest probability at each location is selected as the final label, thereby effectively suppressing salt-and-pepper noise, eliminating small isolated areas, and making similar LCZ units more spatially coherent and natural, better conforming to the regional division principles of local climate zones.

[0037] S27. Based on the test set, the accuracy of the continuous LCZ type units is verified. If the verification result does not meet the preset conditions, the initial classification model is iteratively optimized until the preset conditions are met, and the LCZ classification result is output.

[0038] In some embodiments of this application, step S27 includes: comparing the consecutive LCZ type units with the test set, constructing a confusion matrix and calculating the Kappa coefficient; if the Kappa coefficient is lower than a preset threshold, adjusting the parameters of the preset model and returning to step: training the preset model based on the training set to obtain an initial classification model until the Kappa coefficient is lower than the preset threshold, and outputting the LCZ classification result. In some embodiments of this application, comparing the consecutive LCZ type units with the test set, constructing a confusion matrix and calculating the Kappa coefficient includes: comparing the LCZ classification result optimized by Gaussian filtering with the true LCZ label of each sample point in the test set one by one; counting the number of correctly classified or misclassified categories in all samples to form a two-dimensional table, where rows represent true categories and columns represent predicted categories, and this table is the confusion matrix; based on this, further calculating the Kappa coefficient. The Kappa coefficient can reflect the degree of difference between the classification result and random guessing.

[0039] In some embodiments of this application, a visualization layer of LCZ classification results is generated, including an LCZ classification map, a classification result report, and an accuracy verification report. The LCZ classification map is output in a geographic information system format and includes the boundary and identifier of each LCZ type unit; the classification result report includes classification accuracy indicators, area proportions of each type of LCZ, etc.; the accuracy verification report includes a confusion matrix, Kappa coefficients, and the distribution of verification sample points.

[0040] In other embodiments of this application, the granularity of LCZ classification can be adjusted, for example, the LCZ types can be refined to 20 categories or simplified to 15 categories according to research needs, but the 17-category LCZ classification system has been widely recognized and can balance classification accuracy and practicality.

[0041] In other embodiments of this application, the LCZ classification can be combined with information such as urban functional zoning and administrative divisions to adjust spatial weights, so as to further improve the geographical rationality of the classification results.

[0042] In other embodiments of this application, the electronic device can spatially overlay the LCZ classification results with the urban heat island intensity (SUHII) data to generate a correspondence between LCZ units and SUHII, providing basic data for subsequent analysis of the heat island driving mechanism.

[0043] In this embodiment, refined LCZ classification of the target urban agglomeration region is achieved through the acquisition of multi-source geospatial data, feature extraction, LCZ classification, spatial aggregation, and accuracy verification. LCZ classification effectively captures the spatial heterogeneity within the city, laying the foundation for subsequent analysis of the heat island driving mechanism of different LCZ types of units.

[0044] S11. Based on the multi-source geospatial data and the LCZ classification results, determine the surface urban heat island intensity (SUHII) of each LCZ type unit to obtain SUHII spatial distribution data.

[0045] It should be noted that the urban heat island intensity refers to the temperature difference between the city and the suburbs. The urban heat island intensity can be obtained by subtracting the suburban temperature from the urban temperature.

[0046] In some embodiments of this application, the multi-source geospatial data includes land surface temperature, the LCZ type unit includes LCZD type unit, the average land surface temperature of the LCZD type unit is determined based on the land surface temperature, and the average land surface temperature of the LCZD type unit is used as a reference temperature to determine the surface urban heat island intensity (SUHII) of each LCZ type unit.

[0047] In some embodiments of this application, the land surface temperature can be obtained by processing the MOD11A2 remote sensing satellite product. Specifically, this includes the following steps: reading the raw digital values ​​from the MOD11A2 dataset and converting them to actual land surface temperature in Kelvin units according to the scaling factor provided in the product documentation; converting the temperature values ​​to Celsius; integrating multiple synthetic images of the same month within the study area and calculating the average temperature of each surface pixel for that month to form monthly-scale land surface temperature data; and filling in missing pixels caused by cloud cover or other reasons using methods such as temporal proximity interpolation or spatial neighborhood averaging to obtain a continuous and complete land surface temperature dataset.

[0048] In some embodiments of this application, the surface urban heat island intensity (SUHII) of each LCZ type unit is determined based on the reference temperature. Specifically, the surface urban heat island intensity (SUHII) of each block is obtained by calculating the difference between the average surface temperature of the LCZ unit and the average surface temperature of the LCZD type unit.

[0049] In other embodiments of this application, different methods for determining the reference temperature may be used. For example, the average surface temperature of the suburbs surrounding a city may be used as the reference temperature, or the average surface temperature of the natural area surrounding the city may be used as the reference temperature.

[0050] In some embodiments of this application, SUHII is classified into levels. Specifically, the electronic device calculates the overall mean and standard deviation of SUHII of all LCZ type units, and classifies them into six levels based on the overall mean and standard deviation: strong heat island, high heat island, medium-high heat island, medium heat island, medium-low heat island, and low heat island, which intuitively reflects the intensity of the heat island. For example, LCZ type units with SUHII>μ+1.5σ are classified as strong heat islands.

[0051] In some embodiments of this application, the SUHII classification results are spatially overlaid with the LCZ classification results to generate a SUHII level distribution map. Different color gradients represent the six levels of urban heat island intensity, making the spatial distribution characteristics of the urban heat island readily apparent.

[0052] In some embodiments of this application, a statistical analysis report on the SUHII level is generated, including the area proportion of each level of heat island region, the distribution characteristics of LCZ type, and typical regional examples.

[0053] In other embodiments of this application, the SUHII classification results can be directly used to generate thermal environment control strategies. For example, measures such as increasing vegetation cover and improving building reflectivity can be prioritized for areas with strong heat islands; areas with low heat islands can be preserved as urban cooling sources for protection.

[0054] In some embodiments of this application, the visualization results of the generated SUHII spatial distribution data may include SUHII heatmaps, SUHII contour maps, and SUHII statistical analysis charts. The SUHII heatmap uses color gradients to represent the magnitude of SUHII values, with redder colors indicating a stronger heat island effect; the SUHII contour maps show the continuous changes in SUHII values; and the SUHII statistical analysis charts display statistical indicators such as the SUHII mean and standard deviation for each LCZ type unit.

[0055] In the embodiments of this application, the intensity of the urban heat island on the surface of each LCZ type unit is quantified.

[0056] S12, extract multiple urban morphology factors based on the multi-source geospatial data, and the urban morphology factors are used to characterize urban structural features.

[0057] In some embodiments of this application, multiple urban morphology factors are extracted based on the multi-source geospatial data to characterize urban structural features.

[0058] In some embodiments of this application, urban morphology factors include two-dimensional feature factors and three-dimensional feature factors. The two-dimensional feature factors include building density, vegetation coverage index, water body index, landscape diversity index, etc., which characterize the two-dimensional structure of the city. The three-dimensional feature factors include building height, floor area ratio, sky openness, terrain roughness, etc., which characterize the three-dimensional structure of the city.

[0059] S13. Based on the SUHII spatial distribution data and the urban morphology factors, a training sample set is constructed. The SUHII is used as the dependent variable and the multiple urban morphology factors are used as independent variables to train the preset model and obtain the optimized model.

[0060] In some embodiments of this application, the preset model is an XGBoost model, and training the preset model to obtain an optimized model includes: using SUHII as the dependent variable and the multiple urban morphology factors as independent variables, training the XGBoost model based on the training sample set to obtain an XGBoost optimized model.

[0061] In some embodiments of this application, the construction of a training sample set based on the SUHII spatial distribution data and the urban morphology factors includes: constructing an independent sample for each LCZ type unit; for each LCZ type unit: extracting the surface urban heat island intensity (SUHII) value corresponding to the LCZ type unit as the target variable (i.e., dependent variable) of the sample; calculating and summarizing the statistical characteristics of each urban morphology factor in the LCZ type unit based on the multi-source geospatial data as the input features (i.e., independent variables) of the sample; and pairing the dependent variables and independent variables of all LCZ units to form a structured training sample set.

[0062] In some embodiments of this application, the XGBoost model is trained based on the training sample set to obtain an optimized XGBoost model, including: initializing the hyperparameters of the XGBoost model; dividing the training sample set into a training subset and a validation subset; iteratively fitting the model using the training subset; and evaluating the model performance using the validation subset after each round of training. During training, the internal decision rules are gradually optimized by minimizing the mean square error between the predicted SUHII and the actual SUHII. When the performance on the validation subset tends to stabilize or reaches a preset number of iterations, training is stopped, and the optimized XGBoost model is output. The optimized XGBoost model can quantify the influence of various urban morphology factors on SUHII and can be used to explain the driving mechanism of heat island intensity under different LCZ types.

[0063] In other embodiments of this application, the extracted urban morphology factors can be standardized to eliminate the impact of differences in the dimensions of different factors on model training. For example, the Z-score standardization method can be used to convert each factor value into a standardized value with a mean of 0 and a standard deviation of 1.

[0064] In this embodiment, by extracting multiple urban morphological factors characterizing urban structural features, a machine learning model with SUHII as the dependent variable is constructed, achieving quantitative modeling of the driving mechanism of urban heat island intensity. Through training the XGBoost model, the model can accurately capture the nonlinear relationship between urban morphological factors and SUHII.

[0065] S14. Perform a dual-scale analysis on the optimization model to obtain the contribution of each urban morphology factor to SUHII at the overall urban agglomeration scale and the LCZ unit scale, the influence law of each urban morphology factor on SUHII, and the interaction effect among the urban morphology factors.

[0066] In some embodiments of this application, the dual-scale analysis of the optimization model includes: analyzing the contribution of each urban morphology factor to SUHII at the overall urban agglomeration scale and the influence law of the urban morphology factors on SUHII; at the LCZ unit scale, analyzing the contribution of different urban morphology factors under the same LCZ type, and determining the factor contribution set corresponding to each LCZ type unit; and determining the interaction effect between each urban morphology factor by constructing a SHAP interaction graph.

[0067] In some embodiments of this application, the contribution of each urban morphology factor to SUHII at the overall scale of urban agglomeration and the influence law of the urban morphology factors on SUHII are analyzed, including: based on the training sample set, using the SHAP analysis method to analyze the XGBoost optimization model to obtain the SHAP value of each urban morphology factor to SUHII, where the SHAP value is the contribution of the first factor at the overall scale of urban agglomeration; obtaining the nonlinear relationship between the SHAP value and the corresponding urban morphology factor, and determining the influence law of each urban morphology factor on SUHII based on the nonlinear relationship. It should be noted that the SHAP analysis method is a feature contribution quantification method based on game theory. In the embodiments of this application, the SHAP analysis method is used to interpret the XGBoost optimization model. Specifically, the electronic device calculates the SHAP value of each urban morphology factor through the SHAP analysis tool of the XGBoost model (such as shap.TreeExplainer), and the SHAP value quantifies the contribution of the factor to SUHII. The larger the absolute value of the SHAP value, the more significant the influence of the factor on SUHII.

[0068] In some embodiments of this application, based on the training sample set, the SHAP analysis method is used to analyze the XGBoost optimization model to obtain the SHAP value of each urban morphology factor to SUHII. The SHAP value is the contribution of the first factor at the overall scale of the urban agglomeration. This includes: loading the XGBoost optimization model using a SHAP interpreter suitable for tree models, and interpreting each LCZ unit sample in the training sample set one by one; for each sample, calculating the offset contribution of each urban morphology factor relative to the baseline prediction value through the interpreter, i.e., the SHAP value of the factor; obtaining the absolute value of the SHAP value of the same factor in all samples, and taking the average value, using the average value as the comprehensive contribution of the factor at the overall scale of the urban agglomeration; the larger the average value, the stronger the overall influence of the urban morphology factor on SUHII.

[0069] In some embodiments of this application, the nonlinear relationship between SHAP values ​​and corresponding urban morphology factors is obtained, and the influence of each urban morphology factor on SUHII is determined based on the nonlinear relationship. This includes: pairing the original value of each urban morphology factor with its corresponding SHAP value and plotting a scatter plot; using a weighted regression method to fit a continuous relationship curve between the factor value and the SHAP value; and by analyzing the shape of the curve, such as monotonically increasing, monotonically decreasing, threshold effect, or inverted U-shape, determining the mode of action of the factor on SUHII.

[0070] In some embodiments of this application, at the LCZ unit scale, the contribution of different urban morphology factors under the same LCZ type is analyzed to determine the factor contribution set corresponding to each LCZ type unit, including: obtaining the SHAP value of each urban morphology factor in each LCZ type unit based on the SHAP interpreter; obtaining the absolute value of the SHAP value corresponding to each factor under the same LCZ type and taking the average value to obtain the contribution of each factor under the LCZ type; and obtaining the factor contribution set corresponding to all LCZ type units.

[0071] In some embodiments of this application, the interaction effects between various urban morphology factors are determined by constructing a SHAP interaction graph, including: calculating the SHAP interaction value set corresponding to each LCZ type unit based on the XGBoost optimization model, where the SHAP interaction value is the interaction strength between pairs of interaction factor pairs, each pair consisting of any two urban morphology factors; and constructing a SHAP interaction graph based on the SHAP interaction value set, using the main effect factor in each pair of interaction factor pairs as the independent variable and the SHAP interaction value as the dependent variable. In some embodiments of this application, after obtaining the SHAP values ​​corresponding to each urban morphology factor and the factor contribution set corresponding to each LCZ type unit, the method further includes: determining the dominant influencing factor at the overall scale of the urban agglomeration based on the SHAP values ​​corresponding to each urban morphology factor; and determining the dominant influencing factor of each LCZ type unit based on the factor contribution set corresponding to each LCZ type unit.

[0072] In some embodiments of this application, the dominant influencing factor at the overall scale of the urban agglomeration is determined based on the SHAP values ​​corresponding to each urban morphology factor. This includes: calculating the average value of the absolute values ​​of the SHAP values ​​of each urban morphology factor in all LCZ unit samples to obtain the average SHAP value of each factor at the overall scale of the urban agglomeration; sorting the factors from largest to smallest according to the average value; and selecting one or more factors with the highest ranking as the dominant influencing factor at the overall scale of the urban agglomeration. This dominant influencing factor represents the urban morphological feature with the strongest explanatory power for the intensity of the urban heat island on the surface across the entire region.

[0073] In some embodiments of this application, the dominant influencing factor of each LCZ type unit is determined based on the factor contribution set corresponding to each LCZ type unit. This includes: obtaining the corresponding factor contribution set for each LCZ type, where the factor contribution set contains the average SHAP value of each urban morphology factor under that type; sorting the factors in the factor contribution set according to their contribution magnitude; selecting the factor with the highest contribution as the dominant influencing factor of that LCZ type; and if multiple factors have similar contribution magnitudes, they can be listed as key influencing factors of that LCZ type simultaneously. This method can identify the main morphological elements driving the urban heat island effect in different local climate zones.

[0074] In other embodiments of this application, heatmaps can be used to visualize the contribution of each urban morphology factor to SUHII at the overall scale of the urban agglomeration, as well as the changes in the contribution of each urban morphology factor at the LCZ unit scale, so as to make it easier to intuitively compare the influence intensity of different factors and the differences in different LCZ types.

[0075] In other embodiments of this application, a report analyzing the driving mechanisms at both the overall urban agglomeration scale and the LCZ unit scale can be generated based on the results of the dual-scale analysis. This report may include a list of core influencing factors, nonlinear relationship curves, a comparison table of factor contributions, and conclusions on interaction effect analysis, providing systematic decision support for urban thermal environment governance.

[0076] In this embodiment, a deep analysis of the SUHII mechanism driven by urban morphology factors is achieved through dual-scale analysis at both the overall urban agglomeration scale and the LCZ unit scale. This dual-scale analysis not only grasps the common patterns of regional heat island driving but also identifies the specific mechanisms of local heat island driving, making the analysis of the heat island driving mechanism more comprehensive and in-depth.

[0077] S15. Based on the contribution of the factors, the influence patterns, and the interaction effects, generate thermal environment regulation strategies for the entire urban agglomeration and various LCZ units.

[0078] In some embodiments of this application, the core factors affecting the intensity of the heat island are determined based on the contribution ranking of various urban morphology factors at the overall scale of the urban agglomeration, and a general control strategy at the urban agglomeration level is formulated. At the same time, based on the contribution differences and interaction effects of various urban morphology factors at the LCZ unit scale, differentiated control strategies are customized for different LCZ type units.

[0079] In some embodiments of this application, control strategies are generated at the overall scale of urban agglomerations. For example, if building density has the highest contribution to the overall scale of the urban agglomeration and exhibits a positive correlation (i.e., increased building density leads to an increase in SUHII), a general strategy of "optimizing urban building density and avoiding excessive concentration" is generated; if vegetation cover has a significant negative correlation with SUHII, a general measure of "expanding urban green space coverage" is generated. The strategy content includes control objectives, implementation scope, expected reduction in heat island intensity, etc., forming a general outline for the control of the urban agglomeration's thermal environment.

[0080] In some embodiments of this application, differentiated control strategies are generated for various LCZ units. For example, for the LCZ1 (compact high-rise) unit, analysis shows that building height contributes the most and is positively correlated with SUHII, while building height and vegetation cover have a significant interaction effect (i.e., the cooling effect of vegetation cover is weakened under high building density), so a targeted strategy of "increasing vertical greening in compact high-rise areas and optimizing building layout to improve ventilation" is generated; for the LCZ11 (dense woodland) unit, vegetation cover contributes the most and is strongly negatively correlated with SUHII, so a strategy of "protecting existing woodland and expanding woodland cover area" is generated.

[0081] In some embodiments of this application, the generated control strategies are spatially overlaid with the LCZ classification results to generate a thermal environment control strategy layer. This layer is output in GIS format, with different colors or symbols used to represent the corresponding control strategies for different LCZ units. For example, compact high-rise areas display a "Increase Vertical Greening" indicator, while industrial areas display a "Optimize Building Reflectivity" indicator. The strategy layer can be directly imported into an urban planning system to guide the implementation of specific projects.

[0082] In other embodiments of this application, based on historical control data, the changes in SUHII after the implementation of different strategies are simulated to evaluate the feasibility of the strategies. For example, the input is "the vegetation coverage rate increases by 20% after implementing vertical greening", and the output is the predicted SUHII reduction value and spatial distribution.

[0083] In the machine learning-driven urban surface heat island intensity mechanism analysis method based on the integrated local climate zone framework provided in this application embodiment, the urban agglomeration is divided into multiple LCZ type units based on multi-source geospatial data. Since LCZ classification itself is a standardized clustering of the urban underlying surface structure, it naturally distinguishes regions with different thermal environment response types, thereby enabling the division of cities into standard units with internal homogeneity and external heterogeneity. On this basis, the surface urban heat island intensity SUHII of each LCZ unit is calculated, and multiple morphological factors characterizing urban structural features are extracted by combining multi-source geospatial data to construct a machine learning optimization model with SUHII as the dependent variable. Furthermore, by performing dual-scale analysis of the optimization model at the overall scale of the urban agglomeration and the scale of the LCZ unit, the contribution of each factor at different scales, the nonlinear influence law, and the interaction effect between factors are obtained, thereby taking into account both macroscopic commonalities and local characteristics within a unified framework. Therefore, the generated thermal environment regulation strategy includes both an overall guidance applicable to the entire region and precise and operable differentiated measures for each type of LCZ unit, which significantly improves the accuracy and spatial specificity of the analysis of the driving mechanism of urban surface heat island intensity in local climate zones. It effectively solves the problem that existing urban heat island effect analysis techniques are insufficient to accurately and systematically reveal the driving mechanism of surface urban heat island intensity in different local climate zone types at the urban agglomeration scale.

[0084] like Figure 3 The diagram shown is a structural schematic of an electronic device provided in one embodiment of this application. The electronic device 10 can be a server, a geographic information system workstation, a cloud computing platform, or a localized computing unit with spatial data analysis capabilities, etc. This application embodiment does not impose any restrictions on the specific type of the electronic device 10.

[0085] exist Figure 3The electronic device 10 may include a communication module 101, a memory 102, a processor 103, an input / output (I / O) interface 104, and a bus 105. The processor 103 is coupled to the communication module 101, the memory 102, and the input / output interface 104 via the bus 105.

[0086] Communication module 101 may include a wired communication module and / or a wireless communication module. The wired communication module may provide one or more wired communication solutions such as Universal Serial Bus (USB) and Controller Area Network (CAN). The wireless communication module may provide one or more wireless communication solutions such as Wireless Fidelity (Wi-Fi), Bluetooth (BT), mobile communication networks, frequency modulation (FM), near field communication (NFC), and infrared (IR).

[0087] Memory 102 may include one or more random access memory (RAM) and one or more non-volatile memory (NVM). The RAM can be directly read and written by the processor 103, and can be used to store executable programs (e.g., machine instructions) of other running programs, as well as user and application data. The RAM may include static random-access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), etc.

[0088] Non-volatile memory can also store executable programs and user and application data, and can be pre-loaded into random access memory for direct reading and writing by the processor 103. Non-volatile memory can include disk storage devices and flash memory.

[0089] Memory 102 is used to store one or more computer programs. The one or more computer programs are configured to be executed by processor 103. The one or more computer programs include multiple instructions that, when executed by processor 103, enable a machine learning-driven method for analyzing the intensity mechanism of urban surface heat islands, which integrates a local climate zone framework, and is executed on electronic device 10.

[0090] In other embodiments, such as Figure 3 The electronic device 10 shown also includes an external memory interface for connecting to an external memory to expand the storage capacity of the electronic device 10.

[0091] Processor 103 may include one or more processing units, such as application processors (APs), modem processors, graphics processing units (GPUs), image signal processors (ISPs), controllers, video codecs, digital signal processors (DSPs), and / or neural network processing units (NPUs). These different processing units may be independent devices or integrated into one or more processors.

[0092] The processor 103 provides computational and control capabilities, for example, to execute computer programs stored in the memory 102 to implement the aforementioned machine learning-driven method for analyzing the intensity mechanism of urban surface heat islands within a framework of local climate zones.

[0093] The input / output interface 104 is used to provide a channel for user input or output. For example, the input / output interface 104 can be used to connect various input / output devices, such as a mouse, keyboard, touch device, display screen, etc., so that users can enter information or visualize information.

[0094] Bus 105 is used at least to provide a channel for communication between communication modules 101, memory 102, processor 103, and input / output interface 104 in electronic device 10.

[0095] It is understood that the structures illustrated in the embodiments of this application do not constitute a specific limitation on the electronic device 10. In other embodiments of this application, the electronic device 10 may include more or fewer components than illustrated, or combine some components, or split some components, or have different component arrangements. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

[0096] This application also provides a computer-readable storage medium storing a computer program, which includes program instructions. When the program instructions are executed, the method implemented can refer to the methods in the above embodiments of this application.

[0097] The computer-readable storage medium can be the internal memory of the electronic device described in the above embodiments, such as the hard disk or memory of the electronic device. Alternatively, the computer-readable storage medium can be an external storage device of the electronic device, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., installed on the electronic device.

[0098] In some embodiments, a computer-readable storage medium may include a stored program area and a stored data area, wherein the stored program area may store an operating system, an application program required for at least one function, etc.; and the stored data area may store data created based on the use of the electronic device, etc.

[0099] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0100] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0101] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0102] Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of this application is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be embraced within this application. No appended diagram markings in the claims should be construed as limiting the scope of the claims.

[0103] Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices described in this application may also be implemented by a single unit or device through software or hardware. The terms "first," "second," etc., are used to indicate names and do not indicate any specific order.

[0104] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application and are not intended to limit it. Although this application has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of this application without departing from the spirit and scope of the technical solutions of this application.

Claims

1. A machine learning driven urban land surface heat island intensity mechanism analysis method fusing local climate zone framework, characterized in that, The method includes: To obtain multi-source geospatial data of a target urban agglomeration region, and based on the multi-source geospatial data, divide the target urban agglomeration region into multiple local climate zone (LCZ) type units to obtain LCZ classification results, the method includes: extracting urban feature data based on the multi-source geospatial data; obtaining a preset sample set, which includes urban feature sample data corresponding to all LCZ types; dividing the preset sample set into a training set and a test set; training a preset model based on the training set to obtain an initial classification model; inputting the urban feature data into the initial classification model to generate initial LCZ type labels; optimizing the initial LCZ type labels using Gaussian filtering to form continuous LCZ type units; and verifying the accuracy of the continuous LCZ type units based on the test set. If the verification result does not meet the preset conditions, iteratively optimizing the initial classification model until the preset conditions are met, and outputting the LCZ classification results. Based on the multi-source geospatial data and the LCZ classification results, the surface urban heat island intensity (SUHII) of each LCZ type unit is determined, and the spatial distribution data of SUHII is obtained. Multiple urban morphology factors are extracted based on the multi-source geospatial data, and these urban morphology factors are used to characterize urban structural features. A training sample set is constructed based on the SUHII spatial distribution data and the urban morphology factors. The preset model is trained with SUHII as the dependent variable and the multiple urban morphology factors as independent variables to obtain an optimized model. A dual-scale analysis was performed on the optimization model to obtain the contribution of each urban morphology factor to SUHII at the overall urban agglomeration scale and the LCZ unit scale, the influence law of the urban morphology factor on SUHII, and the interaction effect among the urban morphology factors. Based on the contribution of the factors, the influence patterns, and the interaction effects, thermal environment regulation strategies are generated for the urban agglomeration as a whole and various LCZ units.

2. The machine learning driven urban land surface heat island intensity mechanism analysis method fusing local climate zone framework as claimed in claim 1, wherein, The multi-source geospatial data includes land surface temperature, and the LCZ type units include LCZD type units. Based on the multi-source geospatial data and the LCZ classification results, the surface urban heat island intensity (SUHII) of each LCZ type unit is determined, resulting in a SUHII spatial distribution dataset including: Based on the surface temperature, the average surface temperature of the LCZD type unit is determined; Using the average surface temperature of the LCZD type units as the reference temperature, the surface urban heat island intensity (SUHII) of each LCZ type unit is determined.

3. The machine learning-driven method for analyzing the intensity mechanism of urban surface heat islands, which integrates a local climate zone framework as described in claim 1, is characterized in that... The optimization model is subjected to a dual-scale analysis to obtain the contribution of each urban morphology factor to SUHII at both the overall urban agglomeration scale and the LCZ unit scale, the influence pattern of the urban morphology factors on SUHII, and the interaction effects among the urban morphology factors, including: This study analyzes the contribution of various urban morphology factors to SUHII at the overall scale of urban agglomeration and the influence patterns of these factors on SUHII. At the LCZ unit scale, the contribution of different urban morphology factors under the same LCZ type is analyzed to determine the set of factor contribution for each LCZ type unit. By constructing a SHAP interaction graph, the interaction effects between various urban morphological factors are determined.

4. The machine learning-driven method for analyzing the intensity mechanism of urban surface heat islands, which integrates a local climate zone framework as described in claim 3, is characterized in that... The preset model is an XGBoost model, and the process of training the preset model to obtain an optimized model includes: Using SUHII as the dependent variable and the multiple urban morphology factors as independent variables, the XGBoost model is trained based on the training sample set to obtain the optimized XGBoost model.

5. The machine learning-driven method for analyzing the intensity mechanism of urban surface heat islands, which integrates a local climate zone framework as described in claim 4, is characterized in that... The analysis, at the overall scale of urban agglomerations, examines the contribution of various urban morphology factors to SUHII and the patterns of their influence on SUHII, including: Based on the training sample set, the SHAP analysis method is used to analyze the XGBoost optimization model to obtain the SHAP value of each urban morphology factor to SUHII. The SHAP value is the contribution of the first factor at the overall scale of the urban agglomeration. Obtain the nonlinear relationship between SHAP values ​​and corresponding urban morphology factors, and determine the influence of each urban morphology factor on SUHII based on the nonlinear relationship.

6. The machine learning-driven method for analyzing the intensity mechanism of urban surface heat islands, which integrates a local climate zone framework as described in claim 3, is characterized in that... The process of determining the interaction effects between various urban morphological factors by constructing a SHAP interaction graph includes: Based on the XGBoost optimization model, the SHAP interaction value set corresponding to each LCZ type unit is calculated. The SHAP interaction value is the interaction strength between interaction factor pairs, and the interaction factor pair is composed of any two urban morphology factors. Based on the set of SHAP interaction values, a SHAP interaction graph is constructed with the main effect factors in the interaction factor pairs as independent variables and the SHAP interaction values ​​as dependent variables.

7. The machine learning-driven method for analyzing the intensity mechanism of urban surface heat islands, which integrates a local climate zone framework as described in claim 5, is characterized in that... The method includes: Based on the SHAP values ​​corresponding to each urban morphology factor, the dominant influencing factors at the overall scale of the urban agglomeration are determined. Based on the set of factor contributions corresponding to each LCZ type unit, the dominant influencing factor of each LCZ type unit is determined.

8. The machine learning-driven method for analyzing the intensity mechanism of urban surface heat islands, which integrates a local climate zone framework as described in claim 1, is characterized in that... The accuracy verification of the continuous LCZ type units based on the test set is performed. If the verification result does not meet the preset conditions, the initial classification model is iteratively optimized until the preset conditions are met, and the LCZ classification result is output, including: The continuous LCZ type units are compared with the test set to construct a confusion matrix and calculate the Kappa coefficient; If the Kappa coefficient is lower than a preset threshold, the parameters of the preset model are adjusted and the process returns to step: train the preset model based on the training set to obtain an initial classification model until the Kappa coefficient is lower than the preset threshold, and then output the LCZ classification result.

9. The machine learning-driven method for analyzing the intensity mechanism of urban surface heat islands, which integrates a local climate zone framework as described in claim 1, is characterized in that... The urban morphology factors include two-dimensional feature factors and three-dimensional feature factors.

Images