Urban inland inundation dynamic simulation method and system responding to climate change in large scale

Abstract

The invention relates to the technical field of urban inland inundation risk simulation analysis, and discloses an urban inland inundation dynamic simulation method and system responding to climate change in a large scale, and the method comprises the following steps: S1, obtaining rainstorm information in social media through a crawler, and coupling an AI model and an ArcGIS to generate a large-scale space inundation map; s2, identifying waterlogging ponding time, obtaining corresponding rainfall capacity, and dividing inundation maps with different rainfall intensities; s3, analyzing urban waterlogging vulnerability influence factors and explicit and implicit features by using PDP and SHAP interpretable algorithms; s4, constructing an urban inland inundation disaster vulnerability simulation framework based on the long and short neural networks; s5, simulating future climate change; and S6, coupling future climate data and urban development scenarios to dynamically simulate the urban waterlogging vulnerability. According to the system, the deficiency of traditional observation data is made up by integrating social media multi-source data, the simulation precision is improved by means of an interpretable algorithm and a comprehensive weighting method, and accurate dynamic pre-judgment of current and future urban waterlogging risks under a large scale is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of urban waterlogging risk simulation analysis, in particular to a large-scale urban waterlogging dynamic simulation method and system responding to climate change. BACKGROUND

[0002] With the intensification of global climate change, extreme precipitation events occur frequently, coupled with the expansion of impervious surface area, the reconstruction of water system network, and the change of underlying surface properties in the process of rapid urbanization, urban waterlogging disasters have become an important problem threatening the safe operation of urban infrastructure, the safety of residents' life and property, and the stability of ecological environment. Precise and large-scale urban waterlogging dynamic simulation and risk assessment have become the core demand for improving the ability of urban disaster prevention and mitigation and optimizing urban planning layout.

[0003] Currently, urban waterlogging simulation technology mainly develops around three aspects of data collection, model construction and risk analysis. In the aspect of data collection, it mainly relies on traditional fixed observation sites such as meteorological stations and hydrological stations to obtain basic data such as rainfall and water level. Some technologies combine digital elevation model (DEM) and geographic information system technology to realize preliminary estimation of waterlogging inundation area. In the aspect of model construction, it gradually evolves from early hydrological and hydrodynamic models to machine learning models to improve simulation efficiency. In the aspect of risk analysis, it constructs a vulnerability assessment system by selecting rainfall intensity, terrain slope and other indicators to assist in waterlogging risk judgment.

[0004] However, the existing technology still has significant limitations and gradually exposes many technical problems: first, the data source has limitations. Traditional fixed observation sites are sparse and have limited coverage, which cannot meet the monitoring needs of large-scale areas, especially local waterlogging points in high-density cities, resulting in poor spatial continuity of simulation data and inability to accurately capture small-scale and sudden waterlogging events. At the same time, the existing technology does not effectively integrate new data sources such as storm-related texts, videos and other new data sources published on social media platforms, further limiting the coverage and timeliness of data, making it difficult to support the calibration and verification of waterlogging models with sufficient and continuous data. Second, the depth of waterlogging vulnerability analysis is insufficient. The existing technology mainly focuses on explicit influencing factors such as rainfall and terrain, and lacks coupling analysis of implicit factors such as population distribution, economic density, impervious surface coverage, soil type and vegetation coverage in the process of urbanization. Moreover, there is a lack of interpretable algorithms to analyze the influence mechanism of each factor on different levels of waterlogging disasters and the interaction between factors, resulting in insufficient scientificity and accuracy of vulnerability assessment results, which makes it difficult to support precise waterlogging risk positioning. These problems jointly restrict the effectiveness of existing urban waterlogging simulation technology in large-scale, high-precision and dynamic application scenarios, making it difficult to meet the needs of waterlogging disaster prevention and mitigation in the face of current climate change and urbanization challenges. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides a large-scale urban waterlogging dynamic simulation method and system responding to climate change, which solves the problems raised in the above background art.

[0006] To achieve the above object, the present application is implemented by the following technical solutions: a large-scale urban waterlogging dynamic simulation method responding to climate change, comprising the following steps:

[0007] S1, obtain a city space-time inundation map: obtain rainstorm related information published by social media such as Douyin, Toutiao, Weibo and Xiaohongshu through crawling, collect geographic name information and picture or video data in the information, determine the coordinates of waterlogging points based on the geographic name information, obtain the water depth of the waterlogging points by using an artificial intelligence model combined with the picture or video data, couple a digital elevation model with ArcGIS to generate the inundation range of the waterlogging points, and fuse the inundation ranges of multiple waterlogging points to generate a large-scale space inundation map;

[0008] S2, obtain real-time rainfall intensity: identify the time when the waterlogging point occurs waterlogging, obtain the rainfall of the waterlogging point corresponding to the time, couple the rainfall data with the space inundation map generated in S1, and divide the inundation map corresponding to different rainfall amounts;

[0009] S3, analyze the influence characteristics of urban waterlogging vulnerability: analyze the urban waterlogging vulnerability influence factors, analyze the urban waterlogging vulnerability formation process, and identify the dominant and recessive influence characteristics of urban waterlogging vulnerability;

[0010] S4, construct a high-density urban waterlogging vulnerability dynamic simulation framework based on an artificial intelligence model: construct an urban waterlogging risk vulnerability simulation index system, obtain and process the data required for simulation, construct an urban waterlogging risk vulnerability weight system, and construct an urban waterlogging vulnerability simulation model based on a convolution long short neural network;

[0011] S5, simulate future climate change: obtain CMIP6 climate prediction data, perform downscaling preprocessing on the CMIP6 prediction data, and couple future climate data to carry out urban waterlogging vulnerability simulation, wherein the CMIP6 data is derived from the NEX-GDDP-CMIP6 database, the grid data provided by the CNRM climate model developed by the French National Meteorological Research Center is used, the prediction time period is 2030-2050, and the meteorological elements include daily precipitation, daily temperature maximum and daily temperature average;

[0012] S6, dynamically simulate urban waterlogging risk by coupling future climate and urbanization data: determine a future urban development scenario, construct a high-density urban waterlogging vulnerability simulation model, and carry out high-density urban waterlogging vulnerability dynamic simulation.

[0013] By the technical solution, the S1 integrates multi-source data of social media and acquires large-scale space-time flooding map through geographic space technology, which makes up for the defect of insufficient spatial continuity of traditional observation data; the S2 establishes the correlation between rainfall intensity and flooding range, and clearly defines the distribution characteristics of urban flooding under different rainfall conditions; the S3 analyzes the influence mechanism and explicit and implicit characteristics of urban flooding vulnerability, and provides scientific factor basis for simulation; the S4 constructs a high-density urban flooding vulnerability simulation framework based on artificial intelligence, and improves the systematicness and accuracy of simulation; the S5 realizes response and adaptation to future climate change through CMIP6 data processing and downscaling preprocessing; and the S6 couples future climate and urbanization data to carry out dynamic simulation, and finally realizes accurate characterization and dynamic prediction of current and future large-scale and high-density urban flooding risk, and provides comprehensive technical support for urban flooding disaster prevention and mitigation strategy making and urban planning optimization.

[0014] Preferably, in the S1, the coupling of the digital elevation model and ArcGIS to generate the flooding point flooding range specifically includes the following sub-steps:

[0015] S11, acquire the raster cell data of the digital elevation model, wherein the digital elevation model is in a raster format and is from a geographic space data cloud;

[0016] S12, establish a rectangular grid based on the raster data of the digital elevation model;

[0017] S13, construct a polygon grid based on the rectangular grid;

[0018] S14, acquire the flooding point flooding range in combination with the waterlogging depth of the flooding point and the polygon grid.

[0019] By the technical solution, the DEM data in a raster format is first acquired, and then the rectangular grid and the polygon grid are gradually constructed, so that the abstract terrain information is converted into a structured space unit suitable for flooding range calculation, and finally the specific flooding range of a single flooding point under the constraint of the actual terrain is accurately calculated in combination with the determined waterlogging depth of the flooding point, which provides accurate and reliable single flooding point space boundary data support for generating a large-scale space flooding map by fusing the flooding ranges of multiple flooding points, effectively improves the scientificity and accuracy of the flooding range calculation, and avoids the estimation deviation of the flooding range caused by insufficient utilization of terrain information or insufficient adaptability of the space unit.

[0020] Preferably, in the S3, the influence factors of urban flooding vulnerability are analyzed by using the interpretable algorithms PDP and SHAP to analyze the influence of each factor on urban flooding disasters at different levels, and specifically include the following sub-steps:

[0021] S31, construct a factor analysis model by using an XGBoost algorithm;

[0022] S32, the influence of the characteristic factor on the urban waterlogging disaster is analyzed by using the SHAP interpreter;

[0023] S33, the interaction between the factors is analyzed by using the PDP.

[0024] Through the above technical solutions, a robust urban waterlogging vulnerability factor analysis model is constructed by using the XGBoost algorithm, which provides a reliable data processing and modeling basis for subsequent factor influence analysis; then, the independent influence mechanism of each characteristic factor on different levels of urban waterlogging disaster is clearly analyzed by using the SHAP interpreter, which breaks the limitation of traditional model black box and improves the explainability of factor influence analysis; finally, the interaction relationship between the factors such as cooperation or restriction is analyzed by using the PDP, so as to comprehensively and accurately reveal the action law of the urban waterlogging vulnerability influencing factor, which provides key support for subsequent construction of waterlogging risk vulnerability simulation index system and improvement of the scientificity of waterlogging simulation and risk assessment.

[0025] Preferably, the S4 specifically comprises the following sub-steps:

[0026] S41, construction of urban waterlogging risk vulnerability simulation index system;

[0027] S42, data acquisition and data processing;

[0028] S43, construction of urban waterlogging risk vulnerability weight system;

[0029] S44, construction of urban waterlogging vulnerability simulation model based on convolutional long short neural network.

[0030] Through the above technical solutions, the urban waterlogging risk vulnerability simulation index system is constructed by S41, the core dimensions and evaluation elements required for simulation are clearly defined, and the scientific scope of waterlogging vulnerability simulation is determined; through S42, data acquisition and processing are carried out, the integrity, effectiveness and consistency of the data required for simulation are ensured, and high-quality data basis is provided for subsequent modeling; through S43, the urban waterlogging risk vulnerability weight system is constructed, the influence degree of different indexes on vulnerability is balanced, the simulation result is avoided to be dominated by a single index, and the objectivity of simulation is improved; through S44, the simulation model based on convolutional long short neural network is constructed, relying on the efficient capture ability of the network to space-time characteristics, adapting to the dynamic change characteristics of high-density urban waterlogging, and finally jointly building a systematic and accurate high-density urban waterlogging vulnerability dynamic simulation framework, which provides key technical support for subsequent coupling of climate and urbanization data to carry out waterlogging risk simulation.

[0031] Preferably, in the S5, the downscaling pretreatment of the CMIP6 prediction data specifically comprises the following sub-steps:

[0032] S51. Interpolate the multi-year monthly average of the observed data to the CMIP6 model resolution to obtain the deviation between the observed field and the simulated field;

[0033] S52. Interpolate the bias field data to maintain the same resolution as the original observation data, and sum the data with the original observation data.

[0034] S53. When the resolution of the observation data is higher than 0.25°, the CMIP6 climate prediction grid data is interpolated to the same resolution as the observation data to reflect the spatial heterogeneity of urban precipitation forecasts.

[0035] Through the above technical solution, the deviation between the observation field and the CMIP6 simulation field is calculated in S511 to identify data system errors. In S512, the deviation field is fused with the original observation data to achieve data calibration. In S513, the resolution of CMIP6 data is adjusted for high-resolution observation data scenarios to reflect the spatial heterogeneity of urban precipitation. Ultimately, the problem of mismatch between the original resolution of CMIP6 data and the needs of urban-scale waterlogging simulation is solved, the spatial accuracy and precision of future climate data are improved, and high-quality data support is provided for subsequent precise urban waterlogging vulnerability simulation coupled with future climate data.

[0036] Preferably, in step S5, before conducting urban flood vulnerability simulation by coupling future climate data, an extreme precipitation calculation step is also included: extracting samples with daily precipitation ≥0.2mm from the CMIP6 estimated data for 2030-2050, arranging the samples in descending order of precipitation, selecting the daily precipitation corresponding to the 85th percentile as the extreme precipitation threshold, and identifying precipitation events exceeding this threshold as extreme precipitation events.

[0037] By using the above technical solutions, precipitation samples that meet specific conditions are selected from future climate forecast data and extreme precipitation thresholds are determined, thus accurately identifying extreme precipitation events that have a significant driving effect on urban flooding and avoiding the inclusion of non-extreme precipitation in the flooding vulnerability simulation process.

[0038] Preferably, a comprehensive linear weighting method is used to help avoid these shortcomings. In this method, subjective weights are represented by the analytic hierarchy process (AHP) and objective weights by the entropy weighting method. The calculation method for the linear weighting can be determined by the following formula:

[0039]

[0040] in and The first The comprehensive weight, subjective weight, and objective weight of each factor. 10 5. That is, the equal importance of subjective weight and objective weight;

[0041] The consistency of the pair-wise comparison matrix is determined by establishing a judgment matrix and determining a weight vector based on the judgment matrix, and a consistency test formula is as follows:

[0042]

[0043] Size.

[0044] By the above technical solution, the comprehensive linear weighting method of representing subjective weight by analytic hierarchy process and representing objective weight by entropy weight method can effectively avoid the defects of single subjective weighting (easily affected by experience judgment deviation) or single objective weighting (easily disturbed by data noise), balance subjective experience and objective data characteristics to determine the factor comprehensive weight; at the same time, the logical rationality of the pair-wise comparison matrix in the analytic hierarchy process is judged by consistency test to avoid the contradiction of subjective judgment, and finally a scientific and reliable urban waterlogging risk vulnerability weight system is constructed, which provides reasonable factor weight support for the accurate calculation of the subsequent waterlogging vulnerability simulation model, and improves the scientificity and reliability of the simulation results.

[0045] A large-scale urban waterlogging dynamic simulation system responding to climate change, comprising:

[0046] The data acquisition module is used to acquire heavy rain related information of Douyin, Toutiao, Weibo and Xiaohongshu social media through the crawler unit, extract geographic name information and determine the waterlogging point coordinates through the geographic analysis unit, obtain the water depth by using the artificial intelligence model combined with the picture or video data through the depth recognition unit, generate the inundation range by coupling the digital elevation model and ArcGIS through the range generation unit, and generate the large-scale spatial inundation map through the fusion unit;

[0047] The rainfall intensity acquisition module determines the waterlogging point water accumulation time through the time recognition unit, acquires the rainfall of the corresponding time through the rainfall acquisition unit, and couples the rainfall data and the spatial inundation map and divides the inundation map of different rainfall through the coupling division unit;

[0048] The vulnerability analysis module uses the XGBoost algorithm to build a factor analysis model through the factor analysis unit, analyzes the influence of single factor and the mutual influence between factors through the SHAP interpreter and PDP;

[0049] The simulation framework construction module constructs the waterlogging risk vulnerability simulation index system through the index system unit, pre-processes the simulation data through the data processing unit, constructs the weight system by using the AHP-entropy weight method through the weight system unit, constructs the vulnerability simulation model based on the convolution long short neural network through the model construction unit, and evaluates the model accuracy by using the mean square error and the root mean square error through the accuracy evaluation unit;

[0050] Future climate simulation module: Obtain the 2030-2050 grid data of the CNRM climate model in the NEX-GDDP-CMIP6 database through the climate data unit, perform downscaling processing on the CMIP6 data through the downscaling unit, calculate the extreme precipitation threshold through the extreme precipitation unit, and couple the future climate data to carry out simulation of waterlogging vulnerability through the coupling simulation unit;

[0051] Coupling simulation module: determine the future development scenario through the scenario unit, construct a high-density urban waterlogging vulnerability simulation model through the high-density model unit, and carry out dynamic simulation of waterlogging vulnerability through the dynamic simulation unit.

[0052] Through the above technical solutions, the data acquisition module integrates social media storm-related information and geographic spatial technology to generate a large-scale spatial inundation map, making up for the limitations of traditional observation data; the rainfall intensity acquisition module establishes the association between rainfall and the inundation map to divide the inundation characteristics under different rainfall scenarios; the vulnerability analysis module analyzes the role of waterlogging influencing factors and the mutual relationship between the factors; the simulation framework construction module builds a scientific index system, processes simulation data, constructs a reasonable weight system, and establishes a high-precision simulation model; the future climate simulation module processes future climate data to adapt to the simulation needs of urban scale; the coupling simulation module determines the future development scenario in combination with urbanization data and carries out dynamic simulation of waterlogging vulnerability, and the modules work together to ultimately achieve accurate characterization and dynamic prediction of large-scale urban current and future waterlogging risks, providing comprehensive and reliable technical support for urban disaster prevention and mitigation strategy formulation and urban planning optimization.

[0053] The present application provides a large-scale urban waterlogging dynamic simulation method and system responding to climate change, which has the following advantages:

[0054] 1. The present application effectively solves the technical problems of few measured data, fixed source and spatial discontinuity of urban waterlogging in the prior art, which leads to invalid rating and verification of waterlogging model and reduces the effectiveness of flood control measures. By using a crawler to obtain storm-related information on social media such as Douyin and Toutiao, combining a digital elevation model with ArcGIS to generate an inundation range of waterlogging points, and fusing multi-source data to form a large-scale spatial inundation map, the SHAP interpreter and PDP interpretable algorithm are used to analyze the influencing factors of urban waterlogging vulnerability, and the influence of each factor on different levels of waterlogging disasters and the interaction between the factors are determined, which significantly improves the spatial continuity and accuracy of urban waterlogging simulation, and provides scientific and reliable data support and technical basis for accurately identifying waterlogging risk vulnerability influencing indicators and carrying out waterlogging disaster control.

[0055] 2.The application can effectively respond to the influence of climate change and urbanization on urban waterlogging, couple CMIP6 future climate prediction data and urbanization related data, rely on convolution long short neural network to build a high-density urban waterlogging vulnerability dynamic simulation framework, and use a comprehensive linear weighting method combining the analytic hierarchy process and the entropy weight method to build a weight system, avoiding the defects of a single weighting method, and realizing dynamic simulation and prediction of future urban waterlogging risk in a large-scale range. This method can not only accurately simulate the current waterlogging situation, but also prospectively predict the waterlogging vulnerability changes under future climate and urbanization scenarios, providing forward-looking decision support for urban planning formulation and disaster prevention and mitigation strategy optimization, and significantly improving the comprehensive ability of cities to cope with waterlogging disasters. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 The flowchart of the application. DETAILED DESCRIPTION

[0057] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the specification of the application. Obviously, the described embodiments are only a part of the embodiments of the application, not all. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0058] Please refer to the drawings in the specification of the application Figure 1 The embodiments of the application provide a large-scale urban waterlogging dynamic simulation method and system responding to climate change, comprising the following steps:

[0059] S1, obtaining a city space-time inundation map: through a crawler, obtain rainstorm related information published by social media such as Douyin, Toutiao, Weibo and Xiaohongshu, collect geographic name information and picture or video data in the information, determine the coordinates of waterlogging points based on the geographic name information, use an artificial intelligence model to obtain the water depth of the waterlogging points combined with the picture or video data, couple a digital elevation model with ArcGIS to generate the inundation range of the waterlogging points, and fuse the inundation ranges of multiple waterlogging points to generate a large-scale space inundation map;

[0060] In S1, coupling the digital elevation model with ArcGIS to generate the inundation range of the waterlogging points specifically comprises the following sub-steps:

[0061] S11, obtaining the raster cell data of the digital elevation model, the digital elevation model is in raster format, and is from the geographic spatial data cloud;

[0062] S12, establishing a rectangular grid based on the raster data of the digital elevation model;

[0063] S13, constructing a polygon grid based on the rectangular grid;

[0064] S14, combine the water depth of the waterlogging point with the polygon grid to obtain the waterlogging range of the waterlogging point.

[0065] Specifically, the principle of S1 obtaining the urban space inundation map is that the user-generated content on social media platforms has the characteristics of wide range and timeliness. The dispersed heavy rain related information on platforms such as Douyin and Toutiao is collected through crawler technology to make up for the shortcomings of traditional fixed observation methods in monitoring small-scale and sudden waterlogging points. The geographic name information is extracted from the information to realize the spatial positioning of the waterlogging point through geographic analysis and determine the specific coordinates. The picture or video data is processed by using artificial intelligence models to identify and quantify the water depth of the waterlogging point from the visual image by using computer vision and other AI technologies, which provides key parameters for subsequent inundation range calculation. Coupling the terrain elevation data contained in the digital elevation model (DEM) with the spatial analysis function of ArcGIS can calculate the water spreading range of a single waterlogging point under the constraint of terrain conditions according to the difference in terrain and water depth. Finally, the local point information is integrated into a large-scale space inundation map covering a wide area by fusing the inundation ranges of multiple scattered waterlogging points, realizing the spatial presentation of waterlogging from "point" to "surface".

[0066] S2, real-time rainfall intensity acquisition: identify the time when the waterlogging point occurs waterlogging, acquire the rainfall of the waterlogging point at that time, couple the rainfall data with the space inundation map generated in S1 to divide the inundation map corresponding to different rainfall;

[0067] Specifically, the principle of S2 real-time rainfall intensity acquisition is to establish the spatio-temporal correlation between waterlogging and rainfall driving factors. By identifying the specific time when the waterlogging point occurs waterlogging, the rainfall data corresponding to the time node is matched to ensure the consistency of rainfall data and waterlogging inundation data in the time dimension. Then, the rainfall data is coupled with the space inundation map generated in S1 because rainfall is the core driving factor affecting the inundation range and depth of waterlogging. Through this coupling, the correlation between different rainfall intensities and corresponding inundation characteristics can be established, and the inundation map under different rainfall conditions can be divided, laying a foundation for subsequent analysis of the influence of rainfall intensity on urban waterlogging and identification of waterlogging risk distribution under different rainfall scenarios.

[0068] S3, analysis of urban waterlogging vulnerability influence characteristics: analyze the urban waterlogging vulnerability influence factors, analyze the formation process of urban waterlogging vulnerability, and identify the dominant and recessive influence characteristics of urban waterlogging vulnerability;

[0069] In S3, the interpretable algorithm PDP and SHAP are used to analyze the influence of each factor on different levels of urban waterlogging disasters. The specific steps include:

[0070] S31, constructing a factor analysis model using an XGBoost algorithm;

[0071] S32, analyzing the influence of characteristic factors on urban waterlogging disasters using a SHAP interpreter;

[0072] S33, explaining the interaction between factors using a PDP.

[0073] S4, constructing a high-density urban waterlogging vulnerability dynamic simulation framework based on an artificial intelligence model: constructing an urban waterlogging risk vulnerability simulation index system, obtaining and processing data required for simulation, constructing an urban waterlogging risk vulnerability weight system, and constructing an urban waterlogging vulnerability simulation model based on a convolutional long short neural network;

[0074] S4 specifically includes the following sub-steps:

[0075] S41, constructing an urban waterlogging risk vulnerability simulation index system;

[0076] S42, data acquisition and data processing;

[0077] S43, constructing an urban waterlogging risk vulnerability weight system;

[0078] Specifically, the principle of urban waterlogging vulnerability impact feature analysis is to systematically sort out and select various factors that may affect urban waterlogging vulnerability (including rainfall, terrain, and urbanization-related population, underlying surface properties, etc.) from the core logic of waterlogging disaster formation and development, and to identify key driving factors. Then, by analyzing the interaction of these factors under different spatio-temporal conditions (such as the synergistic effect of rainfall intensity and impervious surface coverage), the complete formation process of waterlogging vulnerability from "factor accumulation" to "risk manifestation" is restored. Finally, by distinguishing between explicit and implicit impact features, the composition mechanism of waterlogging vulnerability is comprehensively mastered, providing scientific features for the construction of subsequent simulation frameworks. The principle of S4, constructing a high-density urban waterlogging vulnerability dynamic simulation framework based on an artificial intelligence model, is to first convert the identified explicit and implicit impact features into quantifiable and calculable waterlogging risk vulnerability simulation indicators, forming a structured index system, ensuring that the simulation dimensions are comprehensive and focused on core elements, based on the feature analysis in S3. Then, by obtaining data that meets the requirements of the index system and preprocessing it (such as unifying the format and eliminating noise), high-quality input is provided for the model. Subsequently, a weight system is constructed to balance the importance of different indicators on vulnerability and avoid the dominance of a single indicator in the simulation results. Finally, a convolutional long short neural network (ConvLSTM) is selected to construct the simulation model, which has the ability to extract spatial features (adapt to the spatial heterogeneity of complex underlying surfaces in high-density cities) and time dynamic modeling, achieving dynamic and accurate simulation of high-density urban waterlogging vulnerability, forming a complete technical framework from index construction to model output.

[0079] In S43, a comprehensive linear weighting method is used to help avoid these defects, in which the analytic hierarchy process represents the subjective weight and the entropy weight method represents the objective weight, and the calculation method of linear weighting can be determined by the following formula:

[0080]

[0081] wherein and are the comprehensive weight, subjective weight and objective weight of the i-th factor, 10 5i.e., the subjective weight and the objective weight are equally important;

[0082] By establishing a judgment matrix, and determining the weight vector based thereon, the consistency index is used to judge the consistency of the pair-wise comparison matrix, and the consistency test formula is as follows:

[0083]

[0084] size.

[0085] The weight is determined by the amount of data information, the greater the weight, the lower the weight, which is an objective weighting method, and the calculation steps of the entropy weight method are as follows:

[0086] The factor normalization formula is as follows:

[0087]

[0088] wherein is the number of factors; is the evaluation object; represents the attribute value of the i-th factor of the j-th object; and are the maximum value and the minimum value in respectively; represents the normalized value of

[0089] The calculation formula of the entropy value of the i-th factor is as follows:

[0090] In the formula, represents the proportion of the j-th evaluation object in the i-th factor. The calculation formula is as follows:

[0091] The entropy weight of the i-th factor​​​​ The calculation formula is as follows:

[0092] Wherein m represents the number of factors.

[0093] S44, construct a city waterlogging vulnerability simulation model based on a convolution long short neural network.

[0094] S5, future climate change simulation: obtain CMIP6 climate prediction data, perform downscaling preprocessing on the CMIP6 prediction data, and perform city waterlogging vulnerability simulation coupled with future climate data, wherein the CMIP6 data is from the NEX-GDDP-CMIP6 database, the grid data provided by the CNRM climate model developed by the French National Meteorological Research Center is used, the prediction time period is 2030-2050, and the meteorological elements include daily precipitation, daily temperature maximum and daily temperature average;

[0095] In S5, the downscaling preprocessing of the CMIP6 prediction data specifically includes the following sub-steps:

[0096] S51, interpolate the multi-year monthly average values of the observation data to the CMIP6 model resolution, and obtain the bias between the observation field and the simulation field;

[0097] S52, keep the bias field data consistent with the original observation data resolution by interpolation, and sum it with the original observation data;

[0098] S53, when the observation data resolution is higher than 0.25°, interpolate the CMIP6 climate prediction grid data to the resolution consistent with the observation data to reflect the spatial heterogeneity of the urban precipitation forecast.

[0099] ​Specifically, the core principle of S5 future climate change simulation is to accurately capture the impact of future climate scenarios on urban waterlogging vulnerability. First, obtain CMIP6 climate data with global climate prediction capability (this type of data is the result of climate model simulation of future climate conditions). However, since the original CMIP6 data is mostly low-resolution grid data at the global or regional scale, its spatial accuracy cannot directly meet the needs of urban-scale waterlogging simulation in terms of local regional details (such as climate differences in different city blocks and different underlying surfaces). Therefore, it is necessary to improve the adaptability of the data to the urban simulation scale through downscaling preprocessing. The core logic of downscaling preprocessing is to reduce the deviation between climate simulation data and actual urban climate characteristics through bias correction and resolution matching: S51 first converts the average state of long-term actual observation data to the resolution of CMIP6 original data to calculate the systematic deviation between CMIP6 simulated climate field and actual observation climate field, providing a benchmark for subsequent data calibration; S52 then adjusts the calculated deviation field to the resolution of the original observation data and fuses it with the original observation data to correct the systematic error of CMIP6 data and ensure data accuracy; S53 for scenes with higher resolution observation data, directly interpolate the CMIP6 climate prediction grid data to the accuracy consistent with the high-resolution observation data, effectively reflecting the spatial heterogeneity of urban internal rainfall distribution (such as rainfall differences in different locations of the city), and finally making the preprocessed future climate data accurately match the scale and accuracy requirements of urban waterlogging simulation. Coupling it with the urban waterlogging vulnerability simulation process can effectively simulate and analyze the impact of future climate change on urban waterlogging risk.

[0100] In S5, before coupling future climate data to simulate urban waterlogging vulnerability, an extreme precipitation calculation step is included: extract the daily precipitation ≥0.2mm sample from the CMIP6 prediction data for 2030-2050, arrange the samples in descending order of precipitation, select the daily precipitation corresponding to the 85th percentile as the extreme precipitation threshold, and identify the precipitation event exceeding the threshold as an extreme precipitation event.

[0101] S6, coupling future climate and urbanization data to dynamically simulate urban waterlogging risk: determine future urban development scenarios, build a high-density urban waterlogging vulnerability simulation model, and conduct dynamic simulation of high-density urban waterlogging vulnerability.

[0102] Specifically, S6 is the core principle of coupling future climate and urbanization data to dynamically simulate urban waterlogging risk. It is based on urban waterlogging risk caused by climate change and urbanization process. By integrating two types of key data, the future waterlogging risk is dynamically deduced. The future urban development scenario is determined. The reason for including data such as water system network, road network, vegetation coverage, soil type and imperviousness is that these data directly depict the core characteristics of the future urban physical environment: water system and road network determine the urban drainage capacity and runoff path, vegetation coverage and soil type affect the rainwater infiltration rate, and impervious surface is directly related to the amount of surface runoff. These data together constitute the underlying surface basic conditions for future cities to cope with rainfall, ensuring that the set development scenario can truly reflect the impact of urbanization on the formation of urban waterlogging. On this basis, a high-density urban waterlogging vulnerability simulation model is built. The essence is to deeply couple the preprocessed future climate data with the above urbanization characteristic data, breaking through the simulation limitations of relying on single climate data or urban data, and accurately capturing the linkage mechanism of meteorological conditions-underlying surface response-waterlogging formation in high-density cities. Finally, dynamic simulation of high-density urban waterlogging vulnerability is carried out by giving the model dynamic deduction capability, i.e. considering the progressive changes of future climate at different time stages and the phased characteristics of urbanization in the simulation process, rather than a fixed single scenario, so as to dynamically track the distribution and intensity changes of urban waterlogging risk at different development nodes in the future, and provide scientific support for predicting future high-risk areas of waterlogging and developing disaster prevention and mitigation strategies adapted to different development stages.

[0103] A large-scale urban waterlogging dynamic simulation system responding to climate change, comprising:

[0104] Data acquisition module: used for acquiring heavy rain related information of social media such as Douyin, Toutiao, Weibo and Xiaohongshu through a crawler unit, extracting geographic name information and determining waterlogging point coordinates through a geographic analysis unit, obtaining water depth through a deep recognition unit using an artificial intelligence model combined with picture or video data, generating a flooded area through a range generation unit coupled with a digital elevation model and ArcGIS, and generating a large-scale spatial flooded map through a fusion unit;

[0105] Rainfall intensity acquisition module: determining waterlogging point waterlogging time through a time recognition unit, obtaining rainfall at the corresponding time through a rainfall acquisition unit, and coupling and dividing rainfall data and spatial flooded map through a coupling division unit to divide the flooded map into different rainfall flooded maps;

[0106] Vulnerability analysis module: constructing a factor analysis model through a factor analysis unit using an XGBoost algorithm, analyzing the influence of a single factor through a SHAP interpreter, and analyzing the mutual influence between factors through a PDP interpreter;

[0107] The simulation framework modeling module: the index system unit constructs the waterlogging risk vulnerability simulation index system, the data processing unit pre-processes the simulation data, the weight system unit adopts AHP-entropy weight method to construct the weight system by comprehensive linear weighting, the model construction unit constructs the vulnerability simulation model based on convolution long short neural network, and the precision evaluation unit adopts mean square error and root mean square error to evaluate the model precision;

[0108] The future climate simulation module: the climate data unit obtains the 2030-2050 grid data of CNRM climate model in NEX-GDDP-CMIP6 database, the downscaling unit carries out downscaling processing on CMIP6 data, the extreme precipitation unit calculates the extreme precipitation threshold, and the coupling simulation unit carries out waterlogging vulnerability simulation by coupling future climate data;

[0109] The coupling simulation module: the scenario unit determines the future development scenario, the high-density model unit constructs the high-density urban waterlogging vulnerability simulation model, and the dynamic simulation unit carries out dynamic simulation of waterlogging vulnerability.

[0110] In summary: the present application effectively solves the technical problems of few measured data, fixed source and spatial discontinuity of urban waterlogging in the prior art, which leads to invalidation of waterlogging model calibration and verification and reduction of effectiveness of flood control measures. By using a crawler to obtain heavy rain related information of social media such as Douyin and Toutiao, combining a digital elevation model and ArcGIS to generate an inundation range of waterlogging points, and fusing multi-source data to form a large-scale spatial inundation map, SHAP interpreter and PDP interpretable algorithm are used to analyze the influence factors of urban waterlogging vulnerability, the influence of each factor on different levels of waterlogging disasters and the interaction between factors are determined, and the spatial continuity and accuracy of urban waterlogging simulation are significantly improved, which provides scientific and reliable data support and technical basis for accurately identifying waterlogging risk vulnerability influence indicators and carrying out waterlogging disaster control.

[0111] Although the embodiments of the present application have been shown and described, it can be understood by those skilled in the art that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and spirits of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A large-scale response to climate change in urban waterlogging dynamic simulation method, characterized in that, Comprise the following steps: S1, urban space-time submerged map acquisition: through the crawler to obtain the rainstorm related information published by social media such as Douyin, today's headlines, microblog and Xiaohongshu, collect the geographic name information and picture or video data in the information, determine the inner waterlogging point coordinates based on the geographic name information, obtain the inner waterlogging point water depth by using the artificial intelligence model combined with the picture or video data, couple the digital elevation model with ArcGIS to generate the inner waterlogging point submerged range, and fuse multiple inner waterlogging point submerged ranges to generate a large-scale space submerged map; S2, real-time rainfall intensity acquisition: identify the time when the inner waterlogging point occurs waterlogging, obtain the rainfall of the inner waterlogging point corresponding to the time, couple the rainfall data with the space submerged map generated in S1, and divide the submerged map corresponding to different rainfall; S3, analysis of urban waterlogging vulnerability influence characteristics: analyze the urban waterlogging vulnerability influence factors, analyze the urban waterlogging vulnerability formation process, and identify the dominant and recessive influence characteristics of urban waterlogging vulnerability; S4, constructing a high-density urban waterlogging vulnerability dynamic simulation framework based on an artificial intelligence model: constructing an urban waterlogging risk vulnerability simulation index system, obtaining and processing the data required for simulation, constructing an urban waterlogging risk vulnerability weight system, and constructing an urban waterlogging vulnerability simulation model based on a convolution long-short neural network; S5, future climate change simulation: obtaining CMIP6 climate prediction data, pre-processing the CMIP6 prediction data, and coupling future climate data to simulate urban waterlogging vulnerability, wherein the CMIP6 data is from the NEX-GDDP-CMIP6 database, the grid data provided by the CNRM climate model developed by the French National Meteorological Research Center is used, the prediction time period is 2030-2050, and the meteorological elements include daily precipitation, daily temperature maximum and daily temperature average; S6, coupling future climate and urbanization data to dynamically simulate urban waterlogging risk: determining future urban development scenarios, constructing a high-density urban waterlogging vulnerability simulation model, and carrying out high-density urban waterlogging vulnerability dynamic simulation.

2. The method according to claim 1, wherein, In the S1, the coupling of the digital elevation model and ArcGIS to generate the inner waterlogging point submerged range specifically comprises the following sub-steps: S11, obtaining the grid cell data of the digital elevation model, the digital elevation model is in grid format, and is from the geographic spatial data cloud; S12, establishing a rectangular grid based on the grid data of the digital elevation model; S13, constructing a polygon grid based on the rectangular grid; S14, combining the inner waterlogging point water depth and the polygon grid to obtain the inner waterlogging point submerged range.

3. The method of claim 1, wherein, In the S3, the analysis of the urban waterlogging vulnerability influence factors uses the interpretable algorithms PDP and SHAP to analyze the influence of each factor on different levels of urban waterlogging disasters, and specifically comprises the following sub-steps: S31, constructing a factor analysis model using the XGBoost algorithm; S32, using the SHAP interpreter to analyze the influence of characteristic factors on urban waterlogging disasters; S33, using PDP to explain the mutual influence between factors.

4. The method of claim 1, wherein, The S4 specifically comprises the following sub-steps: S41, constructing an urban waterlogging risk vulnerability simulation index system; S42, data acquisition and data processing; S43, constructing a weight system of urban waterlogging risk vulnerability; S44, constructing a urban waterlogging vulnerability simulation model based on a convolution long short neural network.

5. The method of claim 1, wherein, In the S5, the downscaling preprocessing of the CMIP6 prediction data specifically includes the following sub-steps: S51, interpolating the multi-year monthly average values of the observation data to the resolution of the CMIP6 model to obtain the deviation of the observation field and the simulation field; S52, keeping the deviation field data consistent with the original observation data resolution by interpolation, and summing it with the original observation data; S53, when the observation data resolution is higher than 0.25°, interpolating the CMIP6 climate prediction grid data to the resolution consistent with the observation data to reflect the spatial heterogeneity of urban precipitation prediction.

6. The method of claim 1, wherein, In the S5, before coupling the future climate data to carry out urban waterlogging vulnerability simulation, an extreme precipitation calculation step is further included: extracting the daily precipitation ≥0.2mm samples in the CMIP6 prediction data from 2030 to 2050, arranging the samples in descending order of precipitation, selecting the daily precipitation corresponding to the 85th percentile as the extreme precipitation threshold, and identifying the precipitation events exceeding the threshold as extreme precipitation events.

7. The method of claim 4, wherein the method is characterized by, In the S43, a comprehensive linear weighting method is used to help avoid these defects, in which the analytic hierarchy process represents the subjective weight and the entropy weight method represents the objective weight, and the calculation method of linear weighting can be determined by the following formula: wherein and are the subjective weight and the objective weight of the i-th factor, respectively, 10 .5 the equal importance of the subjective weight and the objective weight.​ By establishing a judgment matrix, and determining the weight vector based on it, the consistency index is used to judge the consistency of the pair comparison matrix, and the consistency test formula is as follows: size.

8. A large-scale urban waterlogging dynamic simulation system responding to climate change, according to the large-scale urban waterlogging dynamic simulation method responding to climate change of any one of claims 1-7, characterized in that, It includes: The data acquisition module is used to acquire heavy rain related information of Douyin, Toutiao, Weibo and Xiaohongshu social media through the crawler unit, extract geographic name information and determine the waterlogging point coordinates through the geographic analysis unit, obtain the water depth through the deep recognition unit using the artificial intelligence model combined with picture or video data, generate the submerged range through the range generation unit coupled with the digital elevation model and ArcGIS, and generate the large-scale spatial submerged map through the fusion unit; The rainfall intensity acquisition module determines the waterlogging point waterlogging time through the time recognition unit, acquires the rainfall of the corresponding time through the rainfall acquisition unit, and couples the rainfall data with the spatial submerged map through the coupling division unit and divides the submerged map of different rainfall. The vulnerability analysis module uses the XGBoost algorithm to build a factor analysis model through the factor analysis unit, and analyzes the influence of single factor and the mutual influence between factors through the SHAP interpreter and PDP. The simulation framework construction module constructs an urban waterlogging risk vulnerability simulation index system through the index system unit, pre-processes the simulation data through the data processing unit, constructs a weight system through the weight system unit using the AHP-entropy weight method comprehensive linear weighting, constructs a vulnerability simulation model based on a convolution long short neural network through the model construction unit, and evaluates the model accuracy through the precision evaluation unit using the mean square error and the root mean square error. Future climate simulation module: Obtain the 2030-2050 grid data of CNRM climate model in NEX-GDDP-CMIP6 database through climate data unit, perform downscaling processing on CMIP6 data through downscaling unit, calculate extreme precipitation threshold through extreme precipitation unit, and carry out simulation of waterlogging vulnerability by coupling future climate data through coupling simulation unit; Coupling simulation module: Determine future development scenarios through scenario unit, build high-density urban waterlogging vulnerability simulation model through deep learning model unit, and carry out dynamic simulation of waterlogging vulnerability through dynamic simulation unit.

Images