Numerical simulation method for synergistic effect of urban heat island and high-temperature heat wave

Abstract

The invention relates to the technical field of data processing, and discloses a numerical simulation method for a synergistic effect of an urban heat island and high-temperature heat waves. The method comprises the following steps: collecting urban building form data to construct a data set, and combining remote sensing data to carry out local climate zoning; extracting a high-temperature heat wave process and performing grading by utilizing machine learning; analyzing the urban heat island intensity time sequence change based on the remote sensing data; constructing an urban canopy mode coupling hydrology and an artificial heat source; and inputting different levels of heat wave data to simulate a synergistic effect. The technical problem that the urban heat island effect and high-temperature heat wave synergistic effect mechanism cannot be accurately simulated in the prior art is solved, and the numerical simulation method for quantitatively evaluating the urban heat island intensity amplification coefficient under the condition of different grades of high-temperature heat waves is established.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a numerical simulation method for the synergistic effect of urban heat island and high-temperature heat wave. Background Technology

[0002] The urban heat island effect, as a significant urban climate phenomenon, has long been a hot topic in urban climatology research due to its formation mechanism and evolution patterns. Current technologies utilize meteorological observation data analysis, satellite remote sensing monitoring, and numerical simulations to study the urban heat island effect, establishing various urban canopy models to simulate the thermal processes of the urban boundary layer. These models include single-layer canopy models, multi-layer canopy models, and building energy consumption models, which can reflect the impact of the urban underlying surface on local climate to a certain extent and can be used to assess the impact of urban planning schemes on the thermal environment. Meanwhile, meteorological techniques for identifying, classifying, and forecasting high-temperature heat waves have also matured, establishing high-temperature heat wave definition standards based on temperature thresholds and duration.

[0003] However, existing technologies have significant shortcomings in researching the synergistic effects of the urban heat island effect and high-temperature heat waves. First, current urban canopy models primarily focus on simulating the urban thermal environment under normal weather conditions, lacking specific research on the changing patterns of the urban heat island effect under extreme high-temperature weather conditions, making it difficult to accurately reflect the amplification mechanisms of the urban heat island effect during high-temperature heat waves. Second, existing high-temperature heat wave classification methods are mainly based on regional meteorological characteristics, failing to fully consider the influence of urbanization level and urban morphology on the intensity classification of high-temperature heat waves, resulting in classification results lacking relevance for urban applications. Furthermore, existing technologies lack effective methods for coupling the spatiotemporal distribution characteristics of the urban heat island effect with the characteristics of high-temperature heat wave events, making it impossible to quantify the intensity of their interaction. Summary of the Invention

[0004] This application provides a numerical simulation method for the synergistic effect of urban heat island and high-temperature heat wave, which solves the technical problem that existing technologies cannot accurately simulate the synergistic mechanism of urban heat island effect and high-temperature heat wave. A numerical simulation method is established to quantitatively evaluate the intensity amplification factor of urban heat island under different levels of high-temperature heat wave conditions.

[0005] This application provides a numerical simulation method for the synergistic effect of urban heat island and high-temperature heat wave, the numerical simulation method for the synergistic effect of urban heat island and high-temperature heat wave includes: Collect data on urban building height, building density, building coverage, and street-canyon ratio to construct an urban building dataset. Combine this with satellite remote sensing interpretation data to perform local climate zoning of the city, thus obtaining urban local climate zoning data. We extracted heat wave events from multiple cities and used machine learning methods to classify the heat wave events into different levels, thus obtaining heat wave level classification data. Based on satellite remote sensing data, the intensity of urban heat islands in different eras was analyzed and processed to obtain time-series data on the changes in urban heat island intensity. A city canopy model was constructed by coupling a city hydrological parameterization scheme and anthropogenic thermal scheme, and the city canopy model was parameterized to obtain a city canopy numerical model. The urban heat island effect under different levels of high temperature heat wave classification data is simulated by inputting the urban canopy numerical model to obtain data on the synergistic effect of urban heat island and high temperature heat wave.

[0006] The technical solution provided in this application constructs an urban building dataset by collecting data on urban building height, building density, building coverage, and street-canyon ratio. This dataset, combined with satellite remote sensing interpretation data, is used to determine the technical characteristics of local climate zoning. This solves the technical problem that existing technologies cannot accurately characterize urban spatial heterogeneity, laying a spatial foundation for differentiated simulation of the synergistic effects of urban heat islands and high-temperature heat waves. By extracting the technical characteristics of multiple urban high-temperature heat wave processes and using machine learning methods for hierarchical classification, the application overcomes the shortcomings of existing technologies, such as a single high-temperature heat wave classification standard and a lack of city-specificity, establishing a hierarchical system suitable for studying the synergistic effects of urban heat islands and high-temperature heat waves. The technical characteristics of urban heat island intensity over different eras, analyzed based on satellite remote sensing data, fill the gap in the quantitative analysis of the historical evolution of urban heat islands, providing data support for the time-scale analysis of synergistic effects. The technical characteristics of constructing an urban canopy model that couples urban hydrological parameterization schemes and anthropogenic thermal schemes overcome the limitations of existing urban canopy models in extreme weather conditions, achieving a refined description of the complex interaction processes between the urban underlying surface and the atmosphere. The technical feature of inputting high-temperature heat wave classification data into the urban canopy numerical model to simulate the urban heat island effect under different levels of conditions has achieved, for the first time, dynamic coupling simulation of high-temperature heat waves and urban heat island effect, revealing the physical mechanism and numerical characteristics of their synergistic effect.

[0007] In the specific application area of ​​numerical simulation of the synergistic effect of urban heat island and high-temperature heat waves, the K-means clustering algorithm uses building height data, building density data, impervious surface ratio data, and green space coverage data as feature vectors for cluster analysis, achieving objective zoning of urban space and avoiding the arbitrariness of subjective division, thus ensuring the scientific nature of the boundary condition settings for different zones in subsequent numerical simulations. The support vector machine classification algorithm, employing radial basis functions, kernel functions, and optimized parameter configurations, constructs a nonlinear classification boundary for high-temperature heat wave levels in a high-dimensional feature space, significantly improving the accuracy and stability of extreme weather event classification and providing reliable boundary condition inputs for synergistic effect simulation. The single-window algorithm, in surface temperature inversion processing, accurately corrects for atmospheric influences and surface emissivity, ensuring the accuracy and consistency of temporal variation data of urban heat island intensity, laying a data quality foundation for the quantitative assessment of synergistic effects. The urban canopy numerical model, by integrating urban hydrological parameterization schemes and anthropogenic heat source parameterization schemes, comprehensively considers the complex physical processes of the urban underlying surface in numerical simulations, enabling the simulation results to accurately reflect the real changes in urban microclimate under high-temperature heat wave conditions. Attached Figure Description

[0008] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is a schematic diagram of an embodiment of the numerical simulation method for the synergistic effect of urban heat island and high-temperature heat wave in this application. Figure 2 This is a schematic diagram showing the amplification factor of urban heat island intensity under different levels of high-temperature heat waves in the embodiments of this application. Detailed Implementation

[0010] This application provides a numerical simulation method for the synergistic effect of urban heat islands and high-temperature heat waves. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data used can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0011] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the numerical simulation method for the synergistic effect of urban heat island and high-temperature heat wave in this application includes: Step S101: Collect data on urban building height, building density, building coverage and street-canyon ratio, construct an urban building dataset, combine satellite remote sensing interpretation data, and perform local climate zoning of the city to obtain urban local climate zoning data. Step S102: Extract high temperature heat wave processes from multiple cities and use machine learning methods to classify the high temperature heat wave processes into levels, thereby obtaining high temperature heat wave level classification data. Step S103: Analyze the urban heat island intensity of different eras based on satellite remote sensing data to obtain time series variation data of urban heat island intensity; Step S104: Construct an urban canopy model that couples the urban hydrological parameterization scheme and the anthropogenic thermal scheme, and parameterize the urban canopy model to obtain the urban canopy numerical model. Step S105: Input the high temperature heat wave level classification data into the urban canopy numerical model to simulate the urban heat island effect under different levels of high temperature heat wave processes, and obtain the synergistic effect data of urban heat island and high temperature heat wave.

[0012] It is understood that the executing entity of this application can be a numerical simulation system for the synergistic effect of urban heat island and high-temperature heat wave, or it can be a terminal or a server; the specific implementation is not limited here. This application's embodiment uses a server as an example for illustration.

[0013] Specifically, collecting urban building morphology data requires obtaining accurate spatial information through various technical means. Building height data is acquired using lidar scanning technology. LiDAR emits laser pulses to the building surface and measures the return time to calculate three-dimensional coordinate information. Building density data refers to the number of buildings per unit area, calculated by statistically analyzing the ratio of the total number of buildings in a given area to the area of ​​that area. Building coverage data represents the ratio of the building's footprint to the total land area, and street-canyon ratio data refers to the ratio of building height to street width. Satellite remote sensing interpretation data processing employs a supervised classification algorithm. First, geometric and radiometric corrections are performed on the multispectral satellite imagery for preprocessing. Then, training sample areas are selected, spectral features of each land use type are extracted, and a classifier is built for pixel classification. The K-means clustering algorithm uses building height, building density, impervious surface ratio, and green space coverage as feature vectors for cluster analysis. The algorithm iteratively calculates to minimize the distance between data points within the same cluster and maximize the distance between different clusters, ultimately dividing the city into several regions with similar climatic characteristics.

[0014] The extraction of heat wave processes is based on historical meteorological observation data. A heat wave event is defined as a weather process in which the daily maximum temperature continuously exceeds a certain threshold for a specified number of days. The duration parameter records the number of consecutive days of the heat wave event, the maximum temperature parameter records the peak temperature during the period, the average temperature parameter calculates the average temperature during the entire heat wave period, and the cumulative heat parameter sums the daily temperatures exceeding the baseline. The support vector machine (SVM) classification algorithm uses radial basis functions as the kernel function and classifies heat wave events into four levels by finding the optimal separating hyperplane. The algorithm first maps the feature parameter data to a high-dimensional feature space, then finds the hyperplane that maximizes the classification margin, and solves the optimization problem using the Lagrange multiplier method to obtain the support vectors and classification boundaries. During training, the algorithm adjusts the penalty parameter C and the kernel function parameter γ. C controls the degree of penalty for misclassified samples, and γ determines the width of the radial basis function. The optimal parameter combination is found using a grid search method.

[0015] The urban heat island intensity analysis employs a single-window algorithm to retrieve land surface temperature from thermal infrared satellite data. The single-window algorithm first performs atmospheric correction on the received thermal infrared radiance values ​​to remove the effects of atmospheric absorption and scattering. Then, it calculates the true land surface temperature based on land surface emissivity and atmospheric transmittance parameters. The urban heat island intensity index is calculated by the difference between urban and surrounding rural land surface temperatures, with the rural area serving as a background reference temperature. The average temperature of the rural area is determined through statistical analysis. Time series analysis uses a trend decomposition method to decompose the heat island intensity time series into a trend term, a seasonal term, and a stochastic term. The trend term reflects the long-term variation pattern of the heat island intensity, while the seasonal term reflects the intra-annual periodic variation characteristics.

[0016] The construction of the urban canopy model requires the establishment of a three-dimensional mesh system to discretize the computational domain. The mesh system employs a structured grid, with horizontal boundaries determined according to local urban climate zoning data, and vertical extension from the surface to the top of the boundary layer. The urban hydrological parameterization scheme includes three modules: an impervious surface runoff module calculates stormwater runoff from hardened surfaces using the runoff coefficient method, where runoff equals precipitation multiplied by the runoff coefficient; a green space evapotranspiration module calculates vegetation transpiration and soil evaporation based on the Penman equation, considering meteorological factors such as net radiation, temperature, humidity, and wind speed; and a stormwater drainage module simulates the water flow process of the urban drainage system. The anthropogenic heat source parameterization scheme is divided into three categories: building air conditioning heat emissions are calculated based on building area and air conditioning load; vehicle heat emissions are estimated based on road network density and traffic volume; and industrial heat emissions are determined based on industrial area distribution and energy consumption statistics. Parameter sensitivity analysis employs a one-factor-at-a-time approach, gradually changing the model parameters, observing the magnitude of changes in the output results, and identifying the key parameters that have the greatest impact on the simulation results.

[0017] Data on different levels of high-temperature heat waves were input into the urban canopy numerical model as boundary conditions, including the spatiotemporal distributions of temperature, humidity, wind speed, and radiation flux. The numerical simulation employed the finite difference method to solve the governing equations, including the momentum equation, heat equation, and water vapor equation. The momentum equation describes the spatiotemporal evolution of the wind field, the heat equation calculates the temperature field distribution, and the water vapor equation simulates humidity field changes. The model's time step was set to the second level, and the spatial step to the meter level, with calculations advanced using an explicit time integration scheme. Urban microclimate parameters included temperature at a height of 2 meters above the ground, wind speed at a height of 10 meters, and relative humidity, extracted from grid points using interpolation methods. Synergistic effect analysis was used to compare changes in urban heat island intensity under different levels of high-temperature heat waves, calculating the heat island intensity amplification factor, which represents the degree of enhancement of the urban heat island effect during high-temperature heat waves relative to normal weather conditions. Statistical analysis employed analysis of variance to test the significance of differences in the impact of different levels of high-temperature heat waves on heat island intensity, and the F-test was used to determine whether the differences between groups reached statistical significance.

[0018] In one specific embodiment, a city building dataset is constructed, and combined with satellite remote sensing interpretation data, local climate zoning of the city is performed to obtain city local climate zoning data, including: Three-dimensional lidar scanning of urban buildings yields data on building height, building density, building coverage, and street-canyon ratio. Supervised classification processing of multispectral satellite remote sensing images yields data on impervious surface ratio, green coverage rate, and water body distribution. Building height data, building density data, impervious surface ratio data, and green coverage rate data are used as feature vectors and input into the K-means clustering algorithm for clustering to obtain urban local climate zoning data; Spatial connectivity analysis was performed on the urban local climate zoning data to obtain the corrected urban local climate zoning data.

[0019] Specifically, the lidar system emits laser pulses onto the building surface and receives the reflected signals. Distance is calculated by measuring the round-trip time of the laser beam, and the 3D point cloud data of the building is obtained by combining this with the laser beam scanning angle. Building height data is calculated by extracting the difference between the coordinates of the highest point of each building in the point cloud and the ground reference elevation. Building density data refers to the number of buildings per unit area, calculated by dividing the number of buildings identified within the scanned area by the area of ​​that area. Building coverage data represents the ratio of the building footprint to the total area; it requires extracting the building boundary outlines from the point cloud data, calculating and summing the footprint of each building, and then dividing by the total area of ​​the study area. Street-to-street ratio data represents the ratio of the average building height to the average street width. Street width is obtained by measuring the horizontal distance between adjacent buildings, and building height is the average height of buildings on both sides of the street.

[0020] Supervised classification of multispectral satellite remote sensing imagery first requires preprocessing of the original imagery, including geometric and radiometric correction. Geometric correction eliminates geometric distortions from satellite imaging by converting image coordinates to geographic coordinates using ground control points. Radiometric correction eliminates the effects of atmospheric scattering and absorption on the spectra of ground features by converting digitally quantized values ​​into surface reflectance. The supervised classification algorithm requires establishing a training sample set, selecting representative areas on the imagery as standard samples for various land cover types, including impervious surfaces, green spaces, and water bodies. The algorithm extracts the spectral features of the training samples, calculates the mean and variance of each band, and establishes a classifier model. The maximum likelihood classifier, based on Bayes' theorem, calculates the posterior probability of each pixel belonging to each land cover type and assigns the pixel to the category with the highest probability. Impervious surface proportion data is calculated by statistically analyzing the proportion of impervious surface pixels to the total number of pixels in the classification results; green space coverage data is obtained by statistically analyzing the proportion of vegetation pixels to the total number of pixels; and water body distribution data is formed by extracting the spatial location information of water body category pixels.

[0021] The K-means clustering algorithm uses building height, building density, impervious surface ratio, and green space coverage as four-dimensional feature vectors to perform urban local climate zoning. The algorithm first standardizes the data to eliminate the influence of different data units, converting each variable's value into a standardized value with a mean of zero and a standard deviation of one. The algorithm randomly initializes cluster centers, calculates the distance from each data point to each cluster center using Euclidean distance, and assigns the data point to the nearest cluster center. The algorithm recalculates the average value of all data points within each cluster as the new cluster center, repeating this assignment and update process until the cluster centers no longer change or the change is less than a set threshold. The number of clusters is determined using the elbow rule, calculating the sum of squares within the population under different numbers of clusters, and selecting the inflection point where the decrease in the sum of squares significantly decreases as the optimal number of clusters. Urban local climate zoning data is formed by labeling each spatial grid cell according to the cluster label to which its feature vector belongs.

[0022] Spatial connectivity analysis addresses the spatial discontinuities in K-means clustering results. The algorithm employs a region growing method, starting from the seed point of each cluster and expanding outwards to neighboring cells of the same type, forming spatially continuous partitions. Connectivity is determined based on the eight-neighbor principle, which considers the connectivity between each cell and its eight neighboring cells. The algorithm identifies isolated small-area partitions and merges them into the largest neighboring partition. The merging rule is based on the minimum distance principle, calculating the distance from the center point of the isolated partition to the center points of all neighboring partitions and selecting the partition with the smallest distance for merging. The correction process also includes boundary smoothing, using morphological filtering to eliminate jagged protrusions at partition boundaries and employing opening and closing operations to make the partition boundaries smoother and more regular.

[0023] In one specific embodiment, machine learning methods are used to classify high-temperature heat wave processes into levels, resulting in high-temperature heat wave level classification data, including: High temperature event identification and processing are performed on historical meteorological observation data to extract continuous high temperature weather processes and obtain high temperature heat wave event data; Based on high-temperature heat wave event data, the duration parameter, maximum temperature parameter, average temperature parameter, and cumulative heat parameter are calculated to obtain high-temperature heat wave characteristic parameter data; The high-temperature heat wave feature parameter data are input into the support vector machine classification algorithm for training to obtain a high-temperature heat wave level classification model. Based on the high-temperature heat wave classification model, the high-temperature heat wave process is classified into four levels to obtain data on mild, moderate, severe, and extreme high-temperature heat waves.

[0024] Specifically, clear high-temperature thresholds and duration standards are set to screen for heatwave events. Data processing first involves quality control of multi-year historical meteorological observation data, including missing value imputation, outlier detection, and data consistency checks. Missing value imputation uses linear interpolation to calculate the temperature value at the missing time based on adjacent temperature values. Outlier detection identifies obviously erroneous observations by setting reasonable upper and lower temperature limits. High-temperature event identification is based on a relative threshold method, calculating the 90th percentile of the highest daily temperature for the same historical period as the high-temperature threshold. A heatwave event is identified when the highest daily temperature exceeds this threshold for three consecutive days or more. The algorithm traverses the entire time series, checking whether the temperature exceeds the threshold daily, recording the start and end dates of consecutive threshold exceedances to form the time period information of the heatwave event. During data processing, the independence of high-temperature events needs to be considered. When the interval between two high-temperature periods is less than two days, they are merged into a single heatwave event to avoid incorrectly segmenting a continuous high-temperature process due to a brief cooling period.

[0025] When calculating feature parameters based on high-temperature heat wave event data, the duration parameter is obtained by counting the number of days from the start to the end of each heat wave event. The maximum temperature parameter is extracted by finding the maximum daily maximum temperature during the high-temperature period. The average temperature parameter is calculated by taking the arithmetic mean of all daily maximum temperatures during the entire high-temperature period. The calculation process for the cumulative heat parameter is more complex. It requires first determining a baseline temperature, usually using the multi-year average of historical daily maximum temperatures for the same period as the baseline. Then, the difference between the daily maximum temperature and the baseline temperature is calculated, and the positive differences are summed to obtain the cumulative heat parameter. During the calculation, attention must be paid to unit conversion: the temperature difference is in degrees Celsius, while the cumulative heat parameter is in degrees-days. Data processing also includes standardization of the feature parameters. Since the four parameters have different numerical ranges and dimensions, each parameter needs to be converted to a standard normal distribution, i.e., subtracting the mean and dividing by the standard deviation, to ensure that each parameter has the same weight in subsequent machine learning algorithms.

[0026] The training process of the Support Vector Machine (SVM) classification algorithm requires first dividing the dataset into training and test sets. Stratified sampling is used to ensure that the proportion of each level of heat wave in the training and test sets remains consistent. The algorithm uses radial basis functions (RBFs) as kernel functions, which map data from the original feature space to a higher-dimensional feature space, making linearly inseparable data in the original space linearly separable in the higher-dimensional space. The training process finds the optimal separating hyperplane by solving a quadratic programming problem. The objective function includes maximizing the classification margin and minimizing the classification error. Algorithm parameters include a penalty parameter C and a kernel function parameter γ. The value of C controls the degree of penalty for misclassified samples; a larger C value indicates a lower tolerance for misclassification. The value of γ determines the width of the radial basis function; a larger γ value indicates a more drastic change in the function. Parameter optimization uses a grid search method, exhaustively exploring all parameter combinations within a preset parameter range. Cross-validation is used to evaluate the classification performance of each parameter combination, and the parameter combination with the highest validation accuracy is selected as the optimal parameter.

[0027] The four-level classification process, based on a trained Support Vector Machine (SVM) model, categorizes high-temperature heat waves into four levels: mild, moderate, severe, and extreme, according to their severity. Classification decisions are made using the SVM decision function, which calculates the distance from the sample to be classified to the separating hyperplane of each class; the class with the largest distance is the predicted class. Mild heat waves correspond to short-duration, relatively low-temperature events; moderate heat waves have moderate duration and temperature intensity; severe heat waves are characterized by long duration or high temperature intensity; and extreme heat waves are characterized by very long duration and extremely high temperature intensity. The classification results output includes a level label and a confidence score for each heat wave event, with the confidence score reflecting the reliability of the classification result.

[0028] In one specific embodiment, the intensity of the urban heat island intensities in different eras is analyzed based on satellite remote sensing data to obtain time-series variation data of urban heat island intensity, including: Atmospheric and geometric corrections were performed on multi-temporal thermal infrared satellite remote sensing data to obtain surface brightness temperature data. Urban surface temperature data is obtained by using a single-window algorithm to retrieve surface temperature data based on surface brightness temperature data. Statistical analysis was performed on surface temperature data in rural areas surrounding the city to obtain background temperature reference values. The urban heat island intensity index data is obtained by calculating the urban surface temperature data and background temperature reference values. Time series analysis was performed on multi-year urban heat island intensity index data to obtain time series variation data of urban heat island intensity.

[0029] Specifically, atmospheric correction of multi-temporal thermal infrared satellite remote sensing data requires eliminating the absorption and scattering effects of water vapor, aerosols, and other gas molecules in the atmosphere on thermal infrared radiation. The atmospheric correction algorithm employs a radiative transfer model, calculating atmospheric transmittance and uplink radiation based on atmospheric profile data, which includes temperature, humidity, and pressure information at different altitudes. The algorithm first uses meteorological reanalysis data to obtain atmospheric state parameters at the satellite's transit time, then calculates atmospheric transmittance using the MODTRAN radiative transfer model. This parameter represents the proportion of thermal infrared radiation emitted from the Earth's surface that passes through the atmosphere and reaches the satellite sensor. Geometric correction eliminates geometric distortions during satellite imaging, including positional shifts caused by Earth's curvature, terrain undulations, and changes in satellite attitude. The correction process uses a rational function model, establishing a mathematical relationship between pixel coordinates and geographic coordinates through ground control points. These ground control points are selected from landmark buildings or road intersections with clear and precise geographical features. Geometric correction also includes terrain correction, using a digital elevation model to eliminate the impact of terrain undulations on pixel positioning, ensuring consistent spatial registration accuracy across different temporal images. Surface brightness temperature data is obtained by converting the corrected digital quantization values ​​into brightness temperature values. The conversion process is based on Planck's law and takes into account the sensor's spectral response function and calibration parameters.

[0030] When performing land surface temperature retrieval using the single-window algorithm, a quantitative relationship between land surface temperature and satellite-observed brightness temperature is established based on the radiative transfer equation in the thermal infrared band. The core of the algorithm is to eliminate the influence of atmosphere and land surface emissivity, which represents the ratio of the land surface's ability to emit thermal infrared radiation to that of a blackbody. The algorithm first determines the land surface emissivity based on land cover type. Different land surface types have different emissivities; urban buildings typically have higher emissivity, while vegetation and water bodies have relatively lower emissivity. Atmospheric transmittance is estimated through water vapor content, extracted from atmospheric profile data or estimated using a split-window algorithm from the brightness temperature differences across multiple thermal infrared bands. The calculation process of the single-window algorithm includes: correcting the satellite-observed brightness temperature using land surface emissivity and atmospheric transmittance, eliminating the influence of upward and downward atmospheric radiation, and finally obtaining the true land surface temperature. The algorithm also needs to consider the effective average atmospheric temperature, which is calculated by weighted integration of the atmospheric temperature profile. The weighting function reflects the contribution of different atmospheric layers to land surface temperature retrieval.

[0031] Statistical analysis of surface temperature data in rural areas surrounding cities requires first determining the spatial extent of the rural reference area. This reference area is selected from farmland or naturally vegetated areas at a certain distance from the city boundary and unaffected by urbanization. Spatial selection is based on land use type data, excluding construction land and industrial land, retaining agricultural land and forest land as the rural background area. Temperature statistical analysis employs various statistical measures to describe the distribution characteristics of rural surface temperature, including the arithmetic mean, median, mode, and standard deviation. The arithmetic mean is calculated by summing the surface temperatures of all rural area pixels and dividing by the total number of pixels. The median represents the value in the middle after arranging all temperature values ​​in ascending order, and the mode represents the most frequently occurring temperature value. Determining the background temperature reference value requires considering the stability of the temperature distribution. By comparing the coefficient of variation of rural area temperatures at different time points, the statistical measure with the smaller coefficient of variation is selected as the background temperature reference value. The coefficient of variation is equal to the standard deviation divided by the mean, reflecting the relative dispersion of the temperature distribution.

[0032] The calculation of urban heat island intensity is based on the difference analysis between urban surface temperature data and background temperature reference values. The heat island intensity index represents the temperature increase of urban areas relative to surrounding rural areas. The calculation process is performed pixel-by-pixel. For each pixel within the urban area, the surface temperature value is subtracted from the corresponding background temperature reference value to obtain the heat island intensity value for that pixel. Urban areas are identified based on land use type data, classifying artificial surfaces such as construction land, industrial land, and commercial land as urban areas. The spatial distribution of heat island intensity index data reflects the differences in thermal environment among different areas within the city. Commercial centers and industrial areas typically have higher heat island intensity values, while parks, green spaces, and areas surrounding water bodies have relatively lower values. Data processing also includes the classification of heat island intensity levels, setting different levels based on the distribution range of heat island intensity values ​​to establish a grading standard for subsequent synergistic effect analysis.

[0033] Time series analysis was used to perform trend decomposition and periodic analysis on multi-year urban heat island intensity index data to identify long-term trends and seasonal patterns of heat island intensity. Trend decomposition employed a moving average method, calculating the average heat island intensity over several consecutive years to eliminate interannual fluctuations and extract long-term trend components. Seasonal analysis, based on monthly average heat island intensity data, calculated the multi-year average for each month to identify seasonal variation patterns. The time series analysis also included outlier detection. By setting a normal range for heat island intensity variation, it identified anomalous years or months that significantly deviated from the normal range. Outlier detection was based on the three-standard-deviation criterion, marking values ​​exceeding the average plus or minus three standard deviations as outliers. The analysis results of the time series data on urban heat island intensity revealed the evolution of the heat island effect over different decades, establishing quantitative indicators for assessing the impact of urbanization on the thermal environment.

[0034] In one specific embodiment, an urban canopy model is constructed by coupling an urban hydrological parameterization scheme and anthropogenic thermal scheme, and the urban canopy model is parameterized to obtain an urban canopy numerical model, including: A three-dimensional grid system for urban canopy models was constructed based on local urban climate zoning data, with a horizontal resolution of 100 meters and a vertical resolution of 10 meters, to obtain the model grid data. The parameters of impervious surface runoff, green space evapotranspiration, and stormwater drainage are integrated into the urban hydrological parameterization scheme to obtain urban hydrological parameterization scheme data. The parameterization modules for building air conditioning heat exhaust, transportation vehicle heat exhaust, and industrial heat exhaust are integrated into the anthropogenic heat source parameterization scheme to obtain anthropogenic heat source parameterization scheme data. The urban hydrological parameterization scheme data and the anthropogenic heat source parameterization scheme data are coupled into the urban canopy model to obtain the urban canopy numerical model; Parameter sensitivity analysis was performed on the urban canopy numerical model to obtain optimized parameter configuration data.

[0035] Specifically, when constructing a three-dimensional grid system for the urban canopy model, the establishment of the grid system requires discretizing the urban space into regular computational units. The grid construction process determines the computational domain boundary, which is based on the spatial extent of the city's local climate zoning data, ensuring coverage of all zoning types. Horizontal grid division adopts an equidistant grid scheme, dividing the urban area into square grid cells with a horizontal resolution of 100 meters, each grid cell representing a computational node. Vertical grid division extends from the Earth's surface to the top of the atmospheric boundary layer, typically set at a height of 1500 meters, divided into 150 vertical layers with a vertical resolution of 10 meters. Grid layering adopts an equidistant scheme, with each vertical layer having a thickness of 10 meters, and grid nodes located at the center of each layer. The model grid data includes the three-dimensional coordinate information of each grid cell, the grid cell volume, the connection relationships between adjacent grid cells, and boundary condition settings. The grid data structure adopts a structured grid format, facilitating spatial difference and integration operations in numerical computation. Grid quality checks include verification of the orthogonality and smoothness of grid cells, ensuring that the grid geometric quality meets the accuracy requirements of numerical computation.

[0036] The integration process of the urban hydrological parameterization scheme requires merging three independent parameterization modules into a unified hydrological cycle calculation framework. The impervious surface runoff parameterization module calculates rainwater runoff from hardened surfaces based on the runoff coefficient method. The runoff coefficient is determined according to different impervious surface materials: 0.9 for asphalt pavement, 0.8 for concrete pavement, and 0.95 for building roofs. Runoff calculation uses a time-stepping method, calculating the runoff volume of each grid cell step by step. The runoff volume equals the precipitation intensity multiplied by the impervious surface area and then multiplied by the corresponding runoff coefficient. The green space evapotranspiration parameterization module calculates vegetation transpiration and soil evaporation based on the Penman-Montis formula. The calculation process requires input meteorological elements such as net radiation, air temperature, relative humidity, wind speed, and soil moisture content. Evapotranspiration calculation is divided into two steps: potential evapotranspiration and actual evapotranspiration. Potential evapotranspiration represents the maximum evapotranspiration under conditions of sufficient water supply, while actual evapotranspiration is reduced by the potential evapotranspiration based on soil moisture content. The stormwater drainage parameterization module simulates the water flow process of an urban drainage network. The network's drainage capacity is determined based on pipe diameter and slope. The drainage calculation uses the Manning formula and considers the impact of pipe friction resistance on water flow velocity. The modules are loosely coupled, with each module maintaining independent computational logic and exchanging information through a data interface.

[0037] The parameterization scheme for anthropogenic heat sources transforms the heat emission characteristics of three types of anthropogenic heat sources into gridded heat flux distribution data. The building air conditioning heat emission parameterization module calculates heat emissions based on building type, building area, and air conditioning load. The air conditioning heat emission intensity is set at 15 watts per square meter for residential buildings, 25 watts per square meter for commercial buildings, and 20 watts per square meter for office buildings. Heat emission calculations consider temporal variations, with daytime emissions higher than nighttime emissions and summer emissions higher than winter emissions; this variation is reflected by setting a time correction coefficient. The transportation heat emission parameterization module calculates vehicle heat emissions based on road network density and traffic flow. The heat emission intensity is determined according to vehicle type and operating status: 3 kilowatts per vehicle for small cars, 8 kilowatts per vehicle for large trucks, and 5 kilowatts per vehicle for buses. The industrial heat emission parameterization module calculates industrial heat emissions based on industrial area distribution and unit energy consumption intensity. Different industrial types show significant differences in heat emission intensity: 50 watts per square meter for the steel industry, 30 watts per square meter for the chemical industry, and 10 watts per square meter for the electronics industry. The module outputs data in a uniform gridded heat flux distribution format, with the unit being watts per square meter, and the data update frequency is 1 hour.

[0038] The coupling process of the urban canopy model integrates hydrological parameterization scheme data and anthropogenic heat source parameterization scheme data into a unified numerical model framework. The coupling employs an operator splitting method, decomposing the complex coupling process into multiple sub-processes for separate solving. The hydrological process first calculates the surface water balance, including precipitation, runoff, evapotranspiration, and soil moisture changes; this water balance calculation affects surface temperature and humidity distribution. The anthropogenic heat source process superimposes the heat fluxes of various heat sources into the energy balance equation, participating as an additional heat source term in the temperature field calculation. The coupling calculation uses an explicit time integration scheme with a time step of 60 seconds, sequentially performing hydrological calculations, heat source calculations, and temperature field updates within each time step. Data transfer utilizes a memory-shared approach to avoid frequent file read / write operations; the various parameterization schemes exchange calculation results through a shared data structure. The coupled model contains a complete physical process description, covering fundamental physical laws such as radiative transfer, turbulent transport, latent heat of phase change, and surface energy balance.

[0039] The parameter sensitivity analysis employed a one-factor-at-a-time approach to assess the impact of model parameters on the calculation results. The analysis included key physical parameters and empirical parameters. Physical parameters included surface albedo, roughness length, and soil thermal conductivity; empirical parameters included turbulent mixing length and impedance coefficient. The analysis process involved changing the values ​​of each parameter one by one, keeping other parameters constant, running the model, calculating, and comparing the magnitude of changes in the output results. The parameter variation range was set to ±20% of the standard value, with five variation levels set for each parameter for sensitivity testing. Sensitivity evaluation used a normalized sensitivity index, calculated as the output change rate divided by the input change rate; a larger absolute value of the index indicates higher parameter sensitivity. The analysis results were sorted according to the sensitivity index to identify the key parameters with the greatest impact on the simulation results. Based on the sensitivity analysis results, parameter configuration optimization was performed on highly sensitive parameters. The least squares method was used to fit the observed data to determine the optimal parameter values.

[0040] In one specific embodiment, high-temperature heat wave level classification data is input into an urban canopy numerical model to simulate the urban heat island effect under different levels of high-temperature heat waves, including: Data on mild, moderate, severe, and extreme high-temperature heat waves were used as boundary conditions and input into the urban canopy numerical model for numerical simulation to obtain urban meteorological field data under different levels of high-temperature heat waves. Based on urban meteorological field data, the urban near-surface temperature field, urban near-surface wind speed field, and urban near-surface relative humidity field are calculated to obtain urban microclimate parameter data. Synergistic effect analysis was performed based on urban microclimate parameter data and temporal variation data of urban heat island intensity to obtain heat island intensity amplification coefficient data. Statistical analysis and processing of the heat island intensity amplification factor data yielded data on the synergistic effect of urban heat island and high-temperature heat wave.

[0041] Specifically, when inputting data from different levels of high-temperature heat waves as boundary conditions into the urban canopy numerical model, the boundary conditions include the spatiotemporal distribution of temperature, humidity, wind speed, and radiation flux. When using data from mild high-temperature heat waves as boundary conditions, the upper boundary temperature is set to the characteristic temperature value of that heat wave level, the lateral boundaries use periodic boundary conditions, and the lower boundary is set to surface temperature. The boundary condition data format includes time series of meteorological elements for each boundary grid point, with a temporal resolution of 1 hour and a spatial resolution consistent with the model grid. Numerical simulation processing employs the finite difference method to solve the governing equations, which include the momentum equation, heat equation, water vapor equation, and continuity equation. The momentum equation describes the spatiotemporal evolution of the wind speed field, considering the combined effects of pressure gradient force, Coriolis force, and turbulent friction. The heat equation calculates the temperature field distribution, including physical processes such as advection transport, radiative heating, turbulent transport, and latent heat of phase change. The water vapor equation simulates humidity field changes, considering water vapor advection, diffusion, and phase change processes. Numerical integration uses an explicit Euler scheme with a time step of 30 seconds, satisfying the Courant condition to ensure numerical stability. The simulation process proceeds step by step, calculating the spatial distribution of each physical quantity in turn within each time step, ultimately obtaining complete urban meteorological field data, including the spatiotemporal distribution of three-dimensional temperature field, three-dimensional wind speed field, and three-dimensional humidity field.

[0042] The calculation of urban microclimate parameters is based on the post-processing analysis of urban meteorological field data, requiring the extraction of near-surface physical quantity distributions from the three-dimensional meteorological field. The urban near-surface temperature field is obtained by extracting temperature values ​​at a height of 2 meters from the three-dimensional temperature field using interpolation. The interpolation algorithm employs linear interpolation, calculating the temperature at 2 meters based on the temperature values ​​and height differences between adjacent vertical layers. The urban near-surface wind speed field extracts wind speed vectors at a height of 10 meters, including the east-west and north-south components of the horizontal wind speed. The wind speed magnitude is calculated by taking the square root of the sum of the squares of the two components, and the wind direction is determined by calculating the angle between the two components using the arctangent function. The calculation of the urban near-surface relative humidity field requires first extracting specific humidity and temperature at a height of 2 meters, then calculating the saturated vapor pressure using the Clausius-Clapeyron equation. Relative humidity is equal to the actual vapor pressure divided by the saturated vapor pressure and then multiplied by 100%. The data extraction process employs a combination of spatial and temporal interpolation methods. Spatial interpolation eliminates grid discretization errors, while temporal interpolation obtains a continuous time series. Microclimate parameter data are statistically analyzed according to local climate zones, and the spatial average, maximum, minimum and standard deviation of each physical quantity in each zone are calculated to form zone statistical data.

[0043] Synergistic effect analysis compares urban microclimate parameter data with temporal variation data of urban heat island intensity to identify the degree of impact of high-temperature heat waves on the urban heat island effect. The analysis first establishes a temporal correspondence by matching urban microclimate parameter data under different levels of high-temperature heat waves with temporal variation data of urban heat island intensity for the corresponding time periods. The heat island intensity amplification factor is calculated using a ratio method, dividing the urban heat island intensity during a high-temperature heat wave by the intensity under normal weather conditions to obtain the relative change in heat island intensity. An amplification factor greater than 1 indicates that the high-temperature heat wave enhances the urban heat island effect, while an amplification factor less than 1 indicates that the high-temperature heat wave weakens the urban heat island effect. The calculation process is performed separately for different levels of high-temperature heat waves; the amplification factor for mild high-temperature heat waves reflects a slight enhancement effect, while the amplification factor for extreme high-temperature heat waves reflects a significant enhancement effect. Spatial analysis calculates the amplification factor for different local climate zones. Significant differences in the amplification factors for commercial centers, residential areas, and industrial areas reflect the moderating effect of different underlying surface types on the synergistic effect. Time analysis calculates the diurnal and seasonal variations of the magnification factor. The magnification factor is usually higher during the day than at night, and higher in summer than in other seasons.

[0044] Statistical analysis was performed on the heat island intensity amplification factor data using descriptive and inferential statistical analysis to quantify the statistical characteristics and significance level of the synergistic effect. Descriptive statistics calculated the mean, median, standard deviation, skewness, and kurtosis of the amplification factor. The mean reflects the average level of the amplification factor, the standard deviation reflects the dispersion of the amplification factor, the skewness reflects the symmetry of the distribution, and the kurtosis reflects the sharpness of the distribution. Inferential statistics used analysis of variance (ANOVA) to test the differences in the impact of different levels of high-temperature heat waves on the heat island intensity amplification factor. ANOVA decomposed the total variance into between-group and within-group variance, and the significance of the between-group differences was tested using the F-statistic. Paired t-tests were used to compare the significance of the differences in urban heat island intensity under high-temperature heat wave conditions and normal weather conditions, testing the null hypothesis that there is no significant difference in heat island intensity under the two conditions. Correlation analysis calculated the correlation coefficients between the amplification factor and the characteristic parameters of high-temperature heat waves, including the correlation with duration, maximum temperature, average temperature, and cumulative heat. Regression analysis was used to establish a quantitative relationship model between the amplification factor and the characteristic parameters of high-temperature heat waves. The multiple linear regression method was adopted, with the characteristic parameters of high-temperature heat waves as independent variables and the amplification factor as dependent variable. The regression coefficients were estimated by the least squares method.

[0045] Figure 2This diagram illustrates the amplification factor of urban heat island intensity under different levels of high-temperature heat waves in this application embodiment. By inputting high-temperature heat wave data of different levels as boundary conditions into an urban canopy numerical model for synergistic effect analysis, the distribution of heat island intensity amplification factors for each functional area was obtained. The commercial center area exhibited the highest amplification factor under all levels of high-temperature heat waves, gradually increasing from 1.2 for mild heat waves to 2.3 for extreme heat waves, indicating that the commercial center area is most sensitive to high-temperature heat waves. The amplification factor for residential areas was relatively low, varying between 1.1 and 1.9, while the amplification factor for industrial areas fell between the two. This result indicates that with the increase of high-temperature heat wave intensity, the urban heat island effect is significantly amplified, and the degree of amplification varies significantly among different functional areas.

[0046] In one specific embodiment, building height data, building density data, impervious surface ratio data, and green space coverage data are used as feature vectors and input into a K-means clustering algorithm for clustering processing to obtain urban local climate zoning data, including: Standardize the data on building height, building density, impermeable surface ratio, and green coverage to obtain standardized feature vector data. Based on standardized feature vector data, the elbow rule is used to determine the optimal number of clusters, thus obtaining the optimal number of clusters. The standardized feature vector data is input into the K-means clustering algorithm, the number of clusters is set to the optimal number of clusters, the number of iterations is 100, and the clustering calculation is performed to obtain the clustering result data. Based on the clustering results, the urban spatial grid is partitioned and labeled to obtain local urban climate zoning data.

[0047] Specifically, when standardizing data on building height, building density, impervious surface ratio, and green space coverage, the Z-score standardization method is used to eliminate the influence of different variable dimensions and numerical ranges. The standardization calculation process first calculates the mean and standard deviation of each variable. The mean of building height data is calculated by summing the building height values ​​of all grid cells and dividing by the total number of grid cells. The standard deviation is obtained by taking the square root of the sum of the squares of the differences between the building height values ​​of each grid cell and the mean. The standardization formula subtracts the mean of the variable from each original data value and divides it by the standard deviation of the variable, resulting in a standardized value with a mean of zero and a standard deviation of one. The standardization process for building density data is the same as for building height data: first, the mean and standard deviation of the building density of all grid cells are calculated, and then the standardized value is calculated for each grid cell individually. The standardization of impervious surface ratio and green space coverage data uses the same method, resulting in variables with the same numerical range and distribution characteristics after standardization. Standardized feature vector data combines the four standardized variable values ​​of each grid cell into a four-dimensional vector. The vector format is such that each row represents a grid cell and each column represents the standardized value of a variable.

[0048] The elbow rule determines the optimal number of clusters based on the trend analysis of the sum of squares within clusters as the number of clusters changes. The elbow rule calculation process sets the candidate number of clusters to a range of 2 to 15. For each candidate number of clusters, the K-means algorithm is run to calculate the sum of squares within each cluster. The sum of squares within a cluster represents the sum of the squared distances from all data points to the center of their respective clusters; a smaller value indicates better cluster density. The algorithm calculates the sum of squares within clusters for different numbers of clusters and plots the relationship curve between the number of clusters and the sum of squares within clusters. Elbow identification is based on the detection of inflection points in the curve slope. The decrease in the sum of squares within clusters corresponding to two adjacent numbers is calculated. The cluster number where the decrease in the sum of squares within clusters significantly decreases is the elbow point. The elbow point indicates that the marginal utility of increasing the number of clusters to improve cluster quality begins to diminish; the number of clusters corresponding to the elbow point is selected as the optimal number of clusters. The algorithm also calculates the silhouette coefficient as an auxiliary criterion for the elbow rule. The silhouette coefficient measures the similarity of a data point to other points in the same class and the difference between data points and other classes. The number of clusters corresponding to the maximum silhouette coefficient is used as a verification of the optimal number of clusters.

[0049] The K-means clustering algorithm employs an iterative optimization method to find the optimal cluster centers and data point allocation scheme. During initialization, several data points from the optimal clusters are randomly selected as initial cluster centers. The K-means++ method prioritizes data points that are geographically distant from each other to avoid over-concentration and convergence to local optima. Euclidean distance is used to calculate the Euclidean distance from the four-dimensional normalized feature vector of each data point to each cluster center, assigning the data point to the nearest cluster center. For cluster center updates, the mean of the four-dimensional feature vectors of all data points within each cluster is calculated as the new cluster center. The update process calculates the mean coordinates of all four dimensions. The iteration process repeats the data point allocation and cluster center update steps, with a maximum of 100 iterations. The algorithm converges when the change in cluster center position is less than a set threshold or when the maximum number of iterations is reached. Convergence is determined based on the change in cluster center position between two consecutive iterations, measured by the Euclidean distance between the old and new cluster center positions. The clustering results data include the cluster label for each grid cell and the final cluster center coordinates. The cluster label uses an integer number to indicate the cluster category to which each grid cell belongs.

[0050] The zoning labeling process divides the urban spatial grid into regions with similar climatic characteristics based on the clustering results. The labeling process is based on the clustering labels of grid cells, merging adjacent grid cells with the same clustering label into the same climate zone. Adjacency determination adopts the eight-neighbor principle, where each grid cell has an adjacency relationship with its eight surrounding grid cells. The region merging algorithm uses a connected component analysis method, starting from any unlabeled grid cell, searching for all connected grid cells with the same clustering label, and labeling these grid cells as the same zone. The search process uses a breadth-first search algorithm, maintaining a queue of grid cells to be visited, visiting each grid cell in the queue one by one, and checking the clustering labels of its eight neighboring grid cells. Zone numbers are assigned sequentially according to the order of zone identification, with the first identified zone numbered 1, the second identified zone numbered 2, and so on. The urban local climate zone data includes the zone number of each grid cell and the spatial boundary coordinates of each zone. The zone boundaries are formed by extracting the geographic coordinates of the edge grid cells within the zone to create polygonal boundaries. The data format adopts a vector data structure, with each partition as an independent polygon feature. The attribute table includes partition number, partition area, average building height within the partition, average building density, average impermeable surface ratio, and average green coverage.

[0051] In one specific embodiment, the high-temperature heat wave feature parameter data is input into a support vector machine classification algorithm for training to obtain a high-temperature heat wave level classification model, including: The high-temperature heat wave characteristic parameter data were randomly divided in a 7:3 ratio to obtain training set data and test set data; The model was trained using a support vector machine algorithm with radial basis function as kernel function based on the training set data. The penalty parameter C was set to 100 and the kernel function parameter γ was set to 0.01 to obtain the initial classification model. The test set data is input into the initial classification model to verify the classification accuracy, and the model performance evaluation data is obtained. Based on the model performance evaluation data, the penalty parameter C and the kernel function parameter γ were optimized using the grid search method to obtain a high-temperature heat wave classification model.

[0052] Specifically, when randomly partitioning the high-temperature heat wave characteristic parameter data in a 7:3 ratio, the data partitioning needs to ensure that the sample distribution of the training set and the test set remains consistent across all levels of high-temperature heat waves. The random partitioning algorithm first indexes all high-temperature heat wave events, then uses a pseudo-random number generator to produce a random index sequence, and allocates data samples to the training and test sets according to this sequence. The partitioning process employs a stratified sampling strategy, randomly sampling from the data at a 7:3 ratio across the four levels of high-temperature heat waves (mild, moderate, severe, and extreme) to ensure that the proportion of samples from each level in the training and test sets remains consistent with the original dataset. The training set contains the four characteristic parameter values ​​and corresponding level labels for each high-temperature heat wave event, while the test set has the same data format as the training set but is not used in the model training process. The data partitioning results are validated using cross-validation, calculating the distribution differences between the training and test sets to ensure the representativeness and randomness of the partitioning results. The partitioned datasets are then standardized by calculating the mean and standard deviation of each characteristic parameter in the training set and converting the characteristic parameter values ​​in both the training and test sets into standardized values.

[0053] The support vector machine (SVM) algorithm model training process based on the structural risk minimization principle in statistical learning theory uses radial basis function (RBF) kernels. The RBF kernel is a Gaussian kernel, which calculates the similarity between two sample points; the similarity decreases exponentially with increasing distance between the sample points. The penalty parameter C is set to 100, indicating a strong penalty constraint on misclassified samples. A larger C value means the algorithm is less tolerant of training errors and tends to find a hyperplane that completely separates the training samples. The kernel parameter γ is set to 0.01 to control the width of the RBF; a smaller γ value indicates a larger kernel width, resulting in a wider influence of a single support vector on the decision boundary. The algorithm training process finds the optimal separating hyperplane by solving a convex quadratic programming problem. The optimization objective function includes maximizing the classification margin and minimizing the classification error. The Lagrange multiplier method transforms the constrained optimization problem into a dual problem, and the Lagrange multipliers are iteratively solved using a sequential minimum optimization algorithm. The training process identifies support vectors, which are key sample points located near the classification boundary and determine the final location of the classification hyperplane. The initial classification model includes a set of support vectors, the corresponding Lagrange multipliers, and a classification threshold parameter.

[0054] The classification accuracy validation process inputs test set data into the initial classification model to calculate the consistency between the predicted results and the true labels. The validation process calculates the decision function value for each test set sample, derived using support vectors, Lagrange multipliers, and kernel functions. For multi-class classification, a one-to-one strategy is employed, pairwise combining samples from the four levels of high-temperature heat waves to construct six binary classifiers. Each binary classifier determines which of two categories a sample belongs to. The final classification result is determined through a voting mechanism, selecting the category with the most votes as the prediction result. Model performance evaluation data includes metrics such as classification accuracy, precision, recall, and F1 score. Classification accuracy is calculated as the proportion of correctly classified samples out of the total number of samples; precision is calculated as the proportion of correctly predicted samples in each category out of the total number of predicted samples in that category; and recall is calculated as the proportion of correctly predicted samples in each category out of the total number of true samples in that category. The F1 score is calculated as the harmonic mean of precision and recall, comprehensively evaluating the model's classification performance across all categories. The confusion matrix records the distribution of prediction results for each category of samples. The diagonal elements of the matrix represent the number of correctly classified samples, and the off-diagonal elements represent the number of misclassified samples.

[0055] The grid search method for parameter optimization seeks the optimal combination of penalty parameter C and kernel function parameter γ through exhaustive search. The parameter search space is set with C values ​​ranging from 1 to 1000, using 10 log-spaced points, and γ values ​​ranging from 0.001 to 1, using 10 log-spaced points, forming a search grid of 100 parameter combinations. The optimization process trains a support vector machine model for each parameter combination and evaluates model performance using cross-validation. A 5-fold cross-validation method is used, randomly dividing the training set into 5 subsets, using 4 subsets alternately to train the model, and using the remaining subset to validate performance. The cross-validation score is calculated as the average accuracy of the 5 validation results, and the parameter combination with the highest average accuracy is selected as the optimal parameters. Parameter optimization also considers the model's generalization ability, identifying overfitting by calculating the difference between training accuracy and validation accuracy. Overfitting is characterized by a significantly higher training accuracy than validation accuracy. After determining the optimal parameters, the final high-temperature heatwave classification model is retrained using the entire training set data. The model includes the optimized parameter configuration, support vector set, and classification decision function.

[0056] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A numerical simulation method for the synergistic effect of urban heat island and high-temperature heat wave, characterized in that the method... The application comprises the following steps: Collecting urban building height, building density, building coverage rate and street canyon ratio data, constructing an urban building dataset, combining satellite remote sensing interpretation data, and performing local climate zoning on the city to obtain urban local climate zoning data; Extracting multiple urban high-temperature heat wave processes and using machine learning methods to grade and classify the high-temperature heat wave processes to obtain high-temperature heat wave grade classification data; Analyzing and processing the urban heat island intensity of different years based on satellite remote sensing data to obtain urban heat island intensity time series change data; Constructing a city canopy model coupled with an urban hydrological parameterization scheme and a human heat scheme and parameterizing the city canopy model to obtain a city canopy numerical model; Inputting the high-temperature heat wave grade classification data into the city canopy numerical model to simulate the urban heat island effect under different grade high-temperature heat wave processes and obtain urban heat island and high-temperature heat wave synergistic effect data.

2. The method of claim 1, wherein, Constructing an urban building dataset, combining satellite remote sensing interpretation data, and performing local climate zoning on the city to obtain urban local climate zoning data, including: Performing three-dimensional laser radar scanning on urban buildings to obtain building height data, building density data, building coverage rate data and street canyon ratio data; Performing supervised classification processing on multispectral satellite remote sensing images to obtain impervious surface ratio data, green coverage rate data and water body distribution data; Inputting the building height data, building density data, impervious surface ratio data and green coverage rate data as feature vectors into a K-means clustering algorithm for clustering processing to obtain urban local climate zoning data; Performing spatial connectivity analysis processing on the urban local climate zoning data to obtain corrected urban local climate zoning data. 3.The method of claim 1, wherein, Using machine learning methods to grade and classify high-temperature heat wave processes to obtain high-temperature heat wave grade classification data, including: Performing high-temperature event identification processing on historical meteorological observation data to extract continuous high-temperature weather processes to obtain high-temperature heat wave event data; Calculating duration parameters, maximum temperature parameters, average temperature parameters and cumulative heat parameters based on high-temperature heat wave event data to obtain high-temperature heat wave characteristic parameter data; Inputting the high-temperature heat wave characteristic parameter data into a support vector machine classification algorithm for training processing to obtain a high-temperature heat wave grade classification model; Performing four-grade classification processing on high-temperature heat wave processes based on the high-temperature heat wave grade classification model to obtain mild high-temperature heat wave data, moderate high-temperature heat wave data, severe high-temperature heat wave data and extreme high-temperature heat wave data.

4. The method of claim 1, wherein, Analyzing and processing the urban heat island intensity of different years based on satellite remote sensing data to obtain urban heat island intensity time series change data, including: Performing atmospheric correction and geometric correction processing on multi-temporal thermal infrared satellite remote sensing data to obtain land surface brightness temperature data; Performing land surface temperature inversion processing based on the land surface brightness temperature data using a single window algorithm to obtain urban land surface temperature data; Performing statistical analysis processing on the land surface temperature data of the surrounding rural areas of the city to obtain a background temperature reference value; Performing heat island intensity calculation processing based on the urban land surface temperature data and the background temperature reference value to obtain urban heat island intensity index data; Time series analysis is performed on multi-year urban heat island intensity index data to obtain urban heat island intensity time series data. 5.The method of claim 1, wherein, A city canopy model coupled with an urban hydrological parameterization scheme and an anthropogenic heat scheme is constructed, and parameterization settings are made for the city canopy model to obtain a city canopy numerical model, including: A three-dimensional grid system of the city canopy model is constructed based on city local climate zoning data, with a horizontal resolution of 100 meters and a vertical resolution of 10 meters, to obtain model grid data; An impervious surface runoff parameterization module, a green space evapotranspiration parameterization module, and a rainwater pipe network discharge parameterization module are integrated into the urban hydrological parameterization scheme to obtain urban hydrological parameterization scheme data; A building air conditioner heat discharge parameterization module, a vehicle heat discharge parameterization module, and an industrial heat discharge parameterization module are integrated into the anthropogenic heat source parameterization scheme to obtain anthropogenic heat source parameterization scheme data; The urban hydrological parameterization scheme data and the anthropogenic heat source parameterization scheme data are coupled to the city canopy model to obtain a city canopy numerical model; Parameter sensitivity analysis is performed on the city canopy numerical model to obtain optimized parameter configuration data. 6.The method of claim 1, wherein, High-temperature heat wave classification data are input into the city canopy numerical model to simulate the urban heat island effect under different levels of high-temperature heat wave processes, including: Mild, moderate, severe, and extreme high-temperature heat wave data are input into the city canopy numerical model as boundary conditions for numerical simulation to obtain city meteorological field data under different levels of high-temperature heat wave conditions; Based on the city meteorological field data, city near-surface temperature field, city near-surface wind speed field, and city near-surface relative humidity field are calculated to obtain city microclimate parameter data; Synergistic effect analysis is performed based on the city microclimate parameter data and the urban heat island intensity time series data to obtain heat island intensity amplification coefficient data; Statistical analysis is performed on the heat island intensity amplification coefficient data to obtain city heat island and high-temperature heat wave synergistic effect data.

7. The method of claim 2, wherein, The building height data, building density data, impervious surface ratio data, and green coverage rate data are input into the K-means clustering algorithm as feature vectors for clustering processing to obtain city local climate zoning data, including: Standardization processing is performed on the building height data, building density data, impervious surface ratio data, and green coverage rate data to obtain standardized feature vector data; Based on the standardized feature vector data, the elbow rule is used to determine the optimal number of clusters to obtain the optimal number of clusters; The standardized feature vector data are input into the K-means clustering algorithm, with the number of clusters set to the optimal number of clusters and the number of iterations set to 100 for clustering calculation processing to obtain clustering result data; Based on the clustering result data, city spatial grid partitioning and labeling processing are performed to obtain city local climate zoning data. 8.The method of claim 3, wherein, High-temperature heat wave feature parameter data are input into the support vector machine classification algorithm for training processing to obtain a high-temperature heat wave classification model, including: The high-temperature heat wave feature parameter data are randomly divided into training set data and test set data in a 7:3 ratio; The support vector machine algorithm with a radial basis function as a kernel function is used for model training and processing based on the training set data, a penalty parameter C is set to 100, a kernel function parameter gamma is set to 0.01, and an initial classification model is obtained; The test set data is input into the initial classification model for classification accuracy verification processing, and model performance evaluation data is obtained; Based on the model performance evaluation data, the penalty parameter C and the kernel function parameter gamma are optimized by using the grid search method, and a high-temperature heat wave grade classification model is obtained.

Images