Urban inland inundation dynamic early warning method fusing multi-source data and high-performance numerical model

Abstract

The invention discloses an urban inland inundation dynamic early warning method fusing multi-source data and a high-performance numerical model, and belongs to the technical field of urban flood control and disaster reduction and disaster early warning, and the method comprises the following steps: 1, obtaining and preprocessing multi-source data; 2, constructing a high-performance urban flood coupling model; step 3, carrying out assimilation correction on the ensemble Kalman filter; 4, performing rolling prediction and dynamic updating; and step 5, risk identification and early warning release. According to the method, the advantages of multi-source data can be fully fused, and the problems of inaccurate rain condition and unclear waterlogging condition in rapid early warning of urban waterlogging are solved; through multi-source data fusion and high-performance numerical model real-time assimilation correction, minute-level rolling simulation and dynamic correction are achieved, the regional refined ponding risk can be forecasted in advance, the problems that traditional early warning is low in precision, updating lags behind, and the model lacks the self-updating capacity are solved, and the practical value is remarkable.

Description

Technical Field

[0001] This invention belongs to the field of urban flood control, disaster reduction and early warning technology, and in particular relates to a dynamic early warning method for urban waterlogging that integrates multi-source data and high-performance numerical models. Background Technology

[0002] With the acceleration of global climate change and urbanization, urban flooding caused by extreme heavy rainfall is becoming increasingly frequent, severely impacting urban transportation and resident safety. Timely and accurate flood warnings are crucial for preventing casualties and property damage. However, existing urban flood warning methods have many limitations: On the one hand, traditional urban stormwater models often employ fixed parameters and offline simulation methods, making it difficult for the models to adjust to real-time conditions, resulting in low early warning accuracy. Model accuracy largely depends on parameter accuracy, but many existing solutions do not provide parameter correction mechanisms, and the long-term fixation of model parameters reduces simulation accuracy. As urban topography, underlying surfaces, and drainage networks continuously change, models using static parameters have poor applicability and increased prediction errors. Therefore, existing urban flooding simulations are often not accurate enough to reflect actual waterlogging conditions in a timely and accurate manner. On the other hand, traditional early warning systems suffer from response lag, failing to reflect the rapid evolution of urban flooding in a timely manner. Many early warnings rely on rainfall threshold triggers or alarms from a small number of monitoring points, resulting in limited spatial accuracy and an inability to refine to the street or block level. Conventional forecast update cycles are long (e.g., hourly or longer), which cannot meet the needs of rapid, minute-by-minute development of urban flooding, causing early warning information to lag behind the flood evolution process and reducing disaster prevention efficiency. Furthermore, some early urban flooding early warning systems lack dynamic updates and assimilation of the model, failing to fully utilize real-time monitoring data to correct the model state, resulting in significant deviations between early warning results and actual conditions.

[0003] To address these issues, the industry has begun exploring the use of multi-source monitoring data to enhance early warning capabilities in recent years. However, purely data-driven methods still have shortcomings in terms of physical consistency and generalization, and further improvements are needed by introducing physical models and data assimilation techniques.

[0004] In summary, existing urban flooding early warning methods still have significant shortcomings in terms of accuracy, timeliness, and adaptive model updates. There is an urgent need for a new method that integrates multi-source data, high-performance numerical models, and real-time assimilation correction to improve the precision and reliability of early warnings. Summary of the Invention

[0005] The purpose of this invention is to provide a dynamic early warning method for urban flooding that integrates multi-source data and high-performance numerical models, so as to solve the above-mentioned technical problems.

[0006] To achieve the above objectives, the present invention provides the following technical solution: This invention discloses a dynamic early warning method for urban flooding that integrates multi-source data and a high-performance numerical model. The method includes the following steps: Step 1: Acquisition and preprocessing of multi-source data: Acquire multi-source data of the study area, including short-term rainfall forecast data, real-time rainfall monitoring data, real-time surface water monitoring data, basic geographic data, and drainage network data; and preprocess the acquired multi-source data, including formatting, quality control, and spatiotemporal registration. Step 2: Construction of a high-performance urban flood coupling model: Construct a one-dimensional river channel model, a two-dimensional surface model, and an underground drainage network model. The underground drainage network model includes a runoff generation and runoff collection submodule and a one-dimensional network runoff collection submodule. Based on the one-dimensional river channel model, the two-dimensional surface model, and the underground drainage network model, establish an urban flood coupling model. Step 3, Ensemble Kalman Filter Assimilation Correction: During the operation of the urban flood coupling model, the ensemble Kalman filter algorithm is introduced to compare the real-time acquired observation data with the simulation results in each assimilation period, and to jointly update the model state variables and key parameters. Step 4, Rolling Forecast and Dynamic Update: A minute-level rolling forecast mechanism is adopted to continuously monitor real-time rainfall. When the cumulative rainfall exceeds a set threshold, a forecast is automatically triggered. When the rolling forecast is started, the real-time rainfall is used to drive the urban flood coupling model to simulate the current waterlogging state. Then, the measured waterlogging data at the current time is used to perform assimilation correction on the model. Finally, the short-term rainfall forecast is used to drive the model to simulate the evolution of waterlogging in the future. The measured rainfall and short-term rainfall forecast data are updated every fixed time step, and the process of "measured rainfall simulation - measured waterlogging correction - short-term rainfall forecast" is repeated. The latest correction state is used as the initial field for the next round of forecasting with a hot start file. This iterative process continues until the forecast rainfall ends and there is no real-time rainfall. Step 5, Risk Identification and Early Warning Issuance: After each rolling simulation is completed, the flood points are automatically identified and risk levels are classified based on the simulation results, i.e., the water accumulation evolution results for future periods. Then, the number of flood points at each risk level is counted according to administrative divisions, and the early warning level is determined according to the triggering rules. Different color early warning signals are used to correspond to different early warning levels.

[0007] Furthermore, the short-term rainfall forecast data mentioned in step 1 uses measured and corrected data provided by existing meteorological operational systems or third-party meteorological data services; the real-time rainfall monitoring data uses radar-rain gauge fusion rainfall products provided by existing meteorological operational systems or third-party meteorological data services; the real-time ground water accumulation monitoring data is acquired by water accumulation sensors, water level gauges, and video monitoring devices deployed at urban low-lying areas, underpasses, and tunnel entrances; the basic geographic data includes DEM data, river system and cross-sectional data, and land use data; DEM data is used to characterize urban micro-topography, with a resolution requirement of better than 5 m; river system and cross-sectional data is used to define external water boundaries and, in conjunction with the operating parameters of hydraulic structures, reflects the hydraulic exchange and backwater risk between the city and external rivers; land use data is used to distinguish different underlying surface types and assign corresponding roughness, infiltration rate, and runoff generation and confluence characteristics to each area; the drainage network data includes pipe diameter, pipe length, material, slope, burial depth, roughness, manhole elevation, and node connection relationships, and provides the operating parameters of hydraulic structures.

[0008] Furthermore, the one-dimensional river channel model mentioned in step 2 is a one-dimensional hydrodynamic model of the river channel with the river cross-section as the main body, while also considering hydraulic structures, used to describe the river water level and backwater effect; the two-dimensional surface model simulates the evolution of surface runoff and water accumulation based on DEM data, divides the road areas of focus and those where surface water accumulation may occur into grids, and sets roughness coefficient, building area coefficient, and infiltration rate parameters; the underground drainage network model includes a runoff generation and runoff collection submodule and a one-dimensional network runoff collection submodule, which provides upstream inflow process lines for the network drainage nodes by calculating the runoff generation and runoff collection process based on the sub-catchment area, and then includes control facilities. The one-dimensional integrated pipe network system is hydraulically simulated to calculate its free surface flow and pressure flow states. The specific process of establishing the urban flood coupling model is as follows: the underground drainage pipe network model is connected to the two-dimensional surface model through the elevation relationship between the rainwater wells and the corresponding ground, realizing the water flow exchange between the overflow of the rainwater wells and the surface drainage; the river outlet of the underground drainage pipe network model is connected to the one-dimensional river channel model to realize the water flow exchange between the pipe network system and the river channel; the two-way coupling between the two banks of the river channel and the two-dimensional surface model is realized through lateral connection, and the real boundary conditions are constructed in combination with the operation rules of hydraulic structures, thereby constructing the urban flood coupling model.

[0009] Furthermore, the model state variables mentioned in step 3 include water level; the key parameters include roughness and infiltration rate.

[0010] Furthermore, the automatic identification of waterlogging points in step 5 is specifically formulated by referring to the drainage standards and experience of Beijing, Shenzhen and Shanghai, taking a water depth of 15 cm as the basic threshold for identifying waterlogging points, and further combining the duration, flow velocity and water accumulation range to increase the criteria for judging waterlogging points. The risk level classification is specifically as follows: based on different thresholds, it is divided into general risk, medium risk, relatively high risk, and high risk. The method of determining the warning level according to the triggering rules, and using different colored warning signals to correspond to different warning levels, is as follows: blue, yellow, orange, and red warning signals are used to correspond to different warning levels; a blue warning is triggered when there are two or more general-risk waterlogging points or one medium-risk waterlogging point in a street; a yellow warning is triggered when there are two or more medium-risk waterlogging points or one high-risk waterlogging point; an orange warning is triggered when there are two or more high-risk waterlogging points; and a red warning is triggered when there are two or more high-risk waterlogging points.

[0011] The beneficial effects of this invention are as follows: The method described in this invention can fully integrate the advantages of meteorological departments' real-time and forecasted rainfall results and water resources departments' water accumulation monitoring data, solving the problems of "inaccurate rainfall" and "unclear water accumulation" in rapid early warning of urban flooding. Compared with existing technologies, this invention achieves minute-level rolling simulation and dynamic correction through multi-source data fusion and real-time assimilation and correction using high-performance numerical models. It can provide early warning of refined regional water accumulation risks, overcoming the problems of low accuracy, delayed updates, and lack of self-updating capabilities in traditional early warning systems, and has significant practical value.

[0012] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of the rolling forecast and early warning process in Example 1; Figure 2 This is a schematic diagram of the risk level of waterlogging points in the study area at a certain moment in Example 1; Figure 3 This is a schematic diagram illustrating the street-level early warning levels for a given area at a specific moment. Detailed Implementation

[0014] This invention discloses a dynamic early warning method for urban flooding that integrates multi-source data and a high-performance numerical model. The method includes the following steps: Step 1: Multi-source data acquisition and preprocessing: Acquire multi-source data for the study area, including short-term rainfall forecast data, real-time rainfall monitoring data, real-time surface water monitoring data, basic geographic data (including DEM, river system and cross-section data, land use data, etc.), and drainage network data. Preprocess the acquired multi-source data, including formatting, quality control, and spatiotemporal registration.

[0015] The short-term rainfall forecast data uses measured and corrected short-term rainfall forecast data provided by existing meteorological operational systems or third-party meteorological data services as the driving input for future periods in subsequent models. Real-time rainfall monitoring data uses radar-rain gauge fusion rainfall products provided by existing meteorological operational systems or third-party meteorological data services. Its accuracy and spatial continuity are superior to single-rain gauge observations, and it can be used as equivalent real-time rainfall input for the current period in subsequent models. Real-time groundwater monitoring data is acquired by water accumulation sensors, water level gauges, and video surveillance devices deployed at urban low-lying areas, underpasses, and tunnel entrances.

[0016] Basic geographic data and drainage network data are the core foundation for constructing refined urban flood models, and their accuracy directly determines the spatial reliability of the simulation results. Specifically, DEM data is used to characterize urban micro-topography, requiring a resolution better than 5 m. River system and cross-sectional data are used to define external water boundaries and, in conjunction with the operating parameters of hydraulic structures such as gates and pumping stations, reflect the hydraulic exchange and backwater risk between the city and external rivers. Land use data is used to distinguish different underlying surface types such as roads, green spaces, and buildings, and assigns corresponding roughness, infiltration rate, and runoff generation and confluence characteristics to each area. Drainage network data must include pipe diameter, length, material, slope, burial depth, roughness, manhole elevation, and node connection relationships, and provide operating parameters for hydraulic structures such as pumping stations, gates, and overflow outlets. All of the above data must undergo topology consistency checks to ensure a closed and connected network structure to support subsequent calculations of the urban flood coupling model.

[0017] Step 2: Construction of a high-performance urban flood coupling model: Construct a one-dimensional river channel model, a two-dimensional surface model, and an underground drainage network model (including a runoff generation and runoff generation submodule and a one-dimensional network runoff submodule), and establish a high-performance urban flood coupling model.

[0018] The one-dimensional river channel model is a hydrodynamic model based on the river cross-section, while also considering hydraulic structures such as sluice gates, weirs, and dams, used to describe river water levels and backwater effects. The two-dimensional surface model simulates surface runoff and water accumulation evolution based on a high-resolution DEM. It divides areas of focus and potential surface water accumulation, such as roads, into grids and sets parameters such as roughness, building area coefficient, and infiltration rate. The underground drainage network model integrates a runoff generation and runoff submodule and a one-dimensional network runoff submodule. It provides upstream inflow process lines for drainage nodes (rainwater wells, rainwater grates) by calculating runoff generation and runoff processes based on sub-catchments. This allows for hydraulic simulation of a one-dimensional integrated network system (including rainwater, sewage, combined sewer, and hydraulic structures such as sluice gates, flap gates, and pumping stations) containing complex control facilities, calculating its free surface flow and pressure flow states. By connecting the underground drainage network model to the corresponding ground elevation relationship between storm drains and the ground surface, the model achieves water exchange between storm drain overflows and surface runoff. Similarly, the model connects the river outlets of the underground drainage network model to the one-dimensional river channel model, enabling water exchange between the network and the river. Lateral connections achieve bidirectional coupling between the riverbanks and the two-dimensional ground surface model. Furthermore, by incorporating the operational rules of hydraulic structures such as gates and pumping stations, realistic boundary conditions are constructed, thereby building a coupled urban flooding model. This coupled urban flooding model is solved in parallel on a high-performance computing platform, enabling high-precision and rapid flood simulation at the street-level.

[0019] Step 3: Ensemble Kalman Filter (EnKF) Assimilation Correction: During the operation of the urban flood coupling model, the Ensemble Kalman Filter (EnKF) algorithm is introduced. Real-time observation data from water level sensors and water level gauges are compared with the simulation results in each assimilation cycle. This jointly updates model state variables (such as water level) and key parameters (such as roughness and infiltration rate). By constructing multiple ensemble members and iteratively executing the "prediction-comparison-update" process, model errors are corrected in real time, preventing the accumulation of biases over time and ensuring that the simulation results always closely match actual observations. This mechanism effectively improves the accuracy and reliability of forecasts, providing dynamic optimization initial conditions for subsequent rolling forecasts.

[0020] The ensemble Kalman filter algorithm needs to be executed in conjunction with an urban flooding coupled model and deployed uniformly on a GPU parallel architecture, forming an integrated high-performance computing process of "simulation-correction". Specifically, the flooding model achieves synchronous prediction by advancing multiple ensemble members in parallel. Then, the core assimilation operations, such as state error covariance calculation, Kalman gain calculation, and state update, are transformed into batch linear algebra calculations of small-scale matrices, and accelerated by calling GPU high-performance computing libraries. Through a multi-level parallel strategy combining ensemble parallelism, grid parallelism, and matrix batch processing, model computation and assimilation updates are executed continuously on the same GPU architecture, significantly improving assimilation efficiency and model correction speed, ensuring real-time correction of model state and avoiding error accumulation in minute-level rolling simulations.

[0021] Step 4, Rolling Forecast and Dynamic Update: A minute-level rolling forecast mechanism is adopted to continuously monitor real-time rainfall. The update is triggered when the cumulative rainfall exceeds a set threshold (i.e., 1-hour rainfall ≥ 100 mm). A mm, A The value is selected based on the actual management needs of the city, generally 15), and then the forecast is automatically triggered. When starting the rolling forecast, firstly, the current waterlogging state is simulated using a high-performance urban flood coupling model driven by real-time rainfall; then, the model is assimilated and corrected using the measured surface waterlogging data at the current moment; finally, the model is simulated to simulate the evolution of waterlogging in the future period using short-term rainfall forecasts. The measured rainfall and short-term rainfall forecast data are updated every fixed time step (e.g., 10-30 minutes, set according to actual needs), and the process of "measured rainfall simulation - measured waterlogging correction - short-term rainfall forecast prediction" is cyclical. The latest correction state is used as the initial field for the next round of forecasting with a hot start file. This iterative process continues until the forecast rainfall ends and there is no real-time rainfall, thereby realizing dynamic tracking and early forecasting of the urban flooding process, which significantly improves real-time performance and accuracy compared to conventional hourly or longer-cycle update methods.

[0022] Step 5: Risk Identification and Early Warning Issuance: After each rolling simulation, the system automatically identifies waterlogging points and classifies them into risk levels based on the simulation results (i.e., the future water accumulation evolution). These risk levels are categorized as general risk, moderate risk, relatively high risk, and high risk based on different thresholds. Then, the number of waterlogging points at each risk level is counted according to administrative divisions (e.g., streets or blocks), and the warning level is determined according to triggering rules. These rules can be flexibly adjusted based on the flood control standards or management needs of different cities, thereby achieving refined risk assessment and early warning triggering at the regional level. Four-color warning signals (blue, yellow, orange, and red) are used to correspond to different warning levels. Specifically: a blue warning is triggered when two or more general risk (Level IV) waterlogging points or one moderate risk (Level III) waterlogging point appear in a street; a yellow warning is triggered when two or more moderate risk or one relatively high risk (Level II) waterlogging point appears; an orange warning is triggered when two or more relatively high risk or one high risk (Level I) waterlogging point appears; and a red warning is triggered when two or more high risk waterlogging points exist.

[0023] When a certain area triggers an early warning condition, an early warning information is automatically generated and issued. This information includes the warning level (blue, yellow, orange, red), the start time and expected duration of the warning, the affected area, a list of key risk points, the maximum estimated water depth and arrival time, and recommendations for public travel, traffic control, underground space closure, and drainage scheduling. This risk identification and four-color early warning mechanism allows for a refined assessment of urban flooding, providing a basis for emergency response decisions.

[0024] The criteria for identifying waterlogging points were developed by referencing drainage standards and experiences from cities such as Beijing, Shenzhen, and Shanghai. Generally, a water depth of 15 cm is considered sufficient to significantly impact traffic and pedestrians, serving as a basic threshold for waterlogging point identification. Further criteria can be added, incorporating factors such as duration, flow velocity, and water accumulation area, and the risk level can be categorized. These criteria and risk classifications can be adjusted and optimized based on the drainage system design capacity, road conditions, and historical disaster data of different cities to ensure the scientific validity and applicability of the early warning results.

[0025] Example 1 This embodiment is an application example of the above method.

[0026] This embodiment discloses a method for dynamic early warning of urban flooding that integrates multi-source data and a high-performance numerical model, such as... Figure 1 As shown, it includes the following steps: Step 1: Multi-source data acquisition and preprocessing: This example uses the central urban area of ​​a megacity as an example, accessing publicly released precipitation analysis products and short-term precipitation forecast products from existing meteorological operational systems. The precipitation analysis product is a grid-based real-time rainfall field generated based on radar-rain gauge fusion technology, which retains the quantitative accuracy of ground observations and reflects the spatial characteristics of radar inversion, and can be directly used as the rainfall driving input for the model in the current period. The short-term precipitation forecast product is a grid-based forecast rainfall field based on measured correction and radar extrapolation, which can be directly used as the rainfall driving input for the model's simulation period of 0 to 3 hours in the future. After preprocessing such as coordinate transformation and cropping, each grid point (500-1000m) is assumed to be a virtual rain gauge station, corresponding to multiple grid areas (average 10-15m) in the urban flood model. Simultaneously, more than 80 automatic road waterlogging monitoring points have been deployed in the central urban area, focusing on underpasses, tunnel entrances, and historically flood-prone areas, and waterlogging depth data is uploaded at a minute-by-minute frequency. This invention can directly access this type of IoT monitoring data, and after outlier removal, it can be used for dynamic correction using ensemble Kalman filtering.

[0027] Secondly, basic urban geography (topography, river cross-sections, land use, etc.) and drainage network data are collected, and preprocessing work such as coordinate transformation, cross-section formatting, drainage network topology checking and generalization is carried out so that the relevant data can be used for the construction of urban flood coupling models.

[0028] Step 2: Construction of a high-performance urban flood coupling model: In this example, the high-performance urban flood coupling model is constructed using IFMS Urban, a flood analysis software independently developed by the China Institute of Water Resources and Hydropower Research. The model mainly includes an underground drainage network model, a one-dimensional river channel model, and a two-dimensional surface model.

[0029] The drainage network model includes a flow generation and runoff submodule and a one-dimensional network runoff submodule. The flow generation and runoff submodule uses the Thiessen polygon method to draw sub-catchments based on the node locations and flow generation / runoff calculation range within the drainage network system. A total of 146,669 sub-catchments are divided, and parameters such as impermeability and infiltration rate are set. The flow generation and runoff are calculated using the nonlinear reservoir method or other conventional methods, and then the flow calculated by the flow generation and runoff model is mapped to specific drainage nodes (rainwater wells, rainwater grates). The one-dimensional network runoff submodule includes a comprehensive network system encompassing rainwater network systems, sewage network systems, combined sewer systems, and hydraulic structures such as sluice gates, flap gates, and pumping stations.

[0030] One-dimensional river model: The river model is a one-dimensional hydrodynamic model of the river with the river cross section as the main body, while also considering hydraulic structures such as river gates, weirs and dams. It covers 48 major flood control and drainage rivers and 98 hydraulic structures in the central urban area.

[0031] Two-dimensional surface model: To reduce the computational load of the two-dimensional model, the two-dimensional surface model only divides the grid in key areas of interest and areas where surface water accumulation may occur, such as roads. A total of 330,000 unstructured grids are divided, with an average size of 10-15 meters, and parameters such as roughness, building area coefficient, and infiltration rate are set.

[0032] Constructing a coupled model: By connecting the underground drainage network model with the two-dimensional surface model through the elevation relationship between the rainwater wells and the corresponding ground, the water flow exchange between rainwater well overflow and surface runoff is realized; the river outlet of the underground drainage network system is connected with the one-dimensional river channel model, realizing the water flow exchange between the network system and the river channel; the two-way coupling between the riverbanks and the two-dimensional surface model is realized through lateral connections, and the actual boundary conditions are constructed by combining the operation rules of hydraulic structures such as gates and pumping stations, thereby constructing an urban flood coupling model.

[0033] Step 3: Ensemble Kalman Filter (EnKF) Assimilation Correction: Within each assimilation cycle (in this example, the assimilation frequency is consistent with the warning update frequency), firstly, multiple ensemble members are constructed based on initial conditions, rainfall uncertainties, and parameter perturbations; secondly, these ensemble members are used to advance the simulation to the observation time based on the latest measured rainfall data, thereby obtaining a prediction set for the current water accumulation state; then, the observed values ​​are compared with the prediction values ​​of each ensemble, and the gain matrix is ​​calculated according to the EnKF algorithm to update the state variables and related model parameters of each ensemble member; finally, the updated ensembles are averaged to obtain the corrected model state. Through the above assimilation, the water level field of the model is consistent with the sensor measurements, and the key parameters are also adaptively adjusted with environmental changes. By continuously executing the "prediction-comparison-update" cycle, the model can adaptively adjust its internal state and parameters during rainfall, avoiding the accumulation of errors over time, maintaining high consistency and high accuracy, and providing a dynamically optimized initial field for subsequent rolling simulations.

[0034] The entire model (including "simulation-correction") is deployed on a GPU parallel computing platform, enabling multi-threaded parallelism, grid partitioning parallelism, and matrix operation parallelism. This allows for the simulation of urban flooding for the next 60 minutes to be completed within 1 minute, meeting the computational requirements for minute-level rolling forecasts.

[0035] Step 4, Rolling Forecast and Dynamic Update: Taking a rainstorm event as an example, rolling forecasts are initiated after the hourly rainfall exceeds 15mm at 5:00 AM. The fixed time step for rolling forecasts is set to 10 minutes. At 5:00 AM, the model first simulates the current waterlogging situation using measured rainfall data from the start of rainfall to the current moment, performs assimilation correction using data from one existing waterlogging observation point, and then predicts the future process using short-term rainfall. After the simulation is completed, the corrected model state at the current moment (5:00 AM) is written to the hot-start file as the initial condition for the next round. Observed rainfall and forecast data are updated every 10 minutes (e.g., 5:10, 5:20, etc.), and the cycle of "measured rainfall simulation - measured waterlogging correction - short-term rainfall prediction" is repeated until the rainfall ends and there is no more real-time rainfall. This mechanism ensures continuous updates to the simulation state and minimizes error accumulation, dynamically tracking and forecasting the development of urban flooding.

[0036] Step 5, Risk Identification and Early Warning Issuance: After completing each round of rolling simulation and obtaining the water accumulation evolution results for the next 0 to 3 hours, identify the waterlogging points and classify their risk levels, such as... Figure 2 As shown in the example, a water depth of 15cm or more in the grid is defined as an area prone to flooding. Water depths of less than 27cm are defined as general risk (Level IV), 27–40cm as moderate risk (Level III), 40–60cm as high risk (Level II), and more than 60cm as high risk (Level I).

[0037] The number of flood-prone areas at each risk level is counted according to administrative divisions (such as streets or blocks), and the warning level is determined according to triggering rules. These rules can be flexibly adjusted based on the flood control standards or management needs of different cities, thereby achieving refined risk assessment and warning triggering at the regional level. In this example, the count is based on streets, such as... Figure 3 As shown, a blue alert is triggered when there are two or more general risk (Level IV) flooding points or one medium risk (Level III) flooding point in a street; a yellow alert is triggered when there are two or more medium risk or one relatively high risk (Level II) flooding point; an orange alert is triggered when there are two or more relatively high risk or one high risk (Level I) flooding point; and a red alert is triggered when there are two or more high risk flooding points.

[0038] Once the area meets the warning trigger conditions, a corresponding urban flooding warning will be generated. The warning information includes, but is not limited to: the warning level (blue, yellow, orange, red), the start time and expected duration, the potentially affected area, a list of key risk points, the maximum estimated water depth and arrival time, and recommendations for public travel, traffic control, underground space closure, and drainage scheduling. The generated warning information can be simultaneously released through the command platform and public channels, achieving precise, pre-disaster level warnings.

[0039] Finally, it should be noted that the above description is only used to illustrate the technical solution of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred arrangement, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solution of the present invention without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for dynamic early warning of urban flooding that integrates multi-source data and a high-performance numerical model, characterized in that, The method includes the following steps: Step 1: Acquisition and preprocessing of multi-source data: Acquire multi-source data of the study area, including short-term rainfall forecast data, real-time rainfall monitoring data, real-time surface water monitoring data, basic geographic data, and drainage network data; and preprocess the acquired multi-source data, including formatting, quality control, and spatiotemporal registration. Step 2: Construction of a high-performance urban flood coupling model: Construct a one-dimensional river channel model, a two-dimensional surface model, and an underground drainage network model. The underground drainage network model includes a runoff generation and runoff collection submodule and a one-dimensional network runoff collection submodule. Based on the one-dimensional river channel model, the two-dimensional surface model, and the underground drainage network model, establish an urban flood coupling model. Step 3, Ensemble Kalman Filter Assimilation Correction: During the operation of the urban flood coupling model, the ensemble Kalman filter algorithm is introduced to compare the real-time acquired observation data with the simulation results in each assimilation period, and to jointly update the model state variables and key parameters. Step 4, Rolling Forecast and Dynamic Update: A minute-level rolling forecast mechanism is adopted to continuously monitor real-time rainfall. When the cumulative rainfall exceeds a set threshold, a forecast is automatically triggered. When the rolling forecast is started, the real-time rainfall is used to drive the urban flood coupling model to simulate the current waterlogging state. Then, the measured waterlogging data at the current time is used to perform assimilation correction on the model. Finally, the short-term rainfall forecast is used to drive the model to simulate the evolution of waterlogging in the future. The measured rainfall and short-term rainfall forecast data are updated every fixed time step, and the process of "measured rainfall simulation - measured waterlogging correction - short-term rainfall forecast" is repeated. The latest correction state is used as the initial field for the next round of forecasting with a hot start file. This iterative process continues until the forecast rainfall ends and there is no real-time rainfall. Step 5, Risk Identification and Early Warning Issuance: After each rolling simulation is completed, the flood points are automatically identified and risk levels are classified based on the simulation results, i.e., the water accumulation evolution results for future periods. Then, the number of flood points at each risk level is counted according to administrative divisions, and the early warning level is determined according to the triggering rules. Different color early warning signals are used to correspond to different early warning levels.

2. The urban flooding dynamic early warning method according to claim 1, which integrates multi-source data and a high-performance numerical model, is characterized in that... The short-term rainfall forecast data mentioned in step 1 uses measured and corrected data provided by existing meteorological operational systems or third-party meteorological data services; the real-time rainfall monitoring data uses radar-rain gauge fusion rainfall products provided by existing meteorological operational systems or third-party meteorological data services; the real-time ground water accumulation monitoring data is acquired by water accumulation sensors, water level gauges, and video monitoring devices deployed at urban low-lying areas, underpasses, and tunnel entrances; the basic geographic data includes DEM data, river system and cross-sectional data, and land use data; DEM data is used to characterize urban micro-topography, with a resolution requirement of better than 5 m; river system and cross-sectional data is used to define external water boundaries and, combined with the operating parameters of hydraulic structures, reflects the hydraulic exchange and backwater risk between the city and external rivers; land use data is used to distinguish different underlying surface types and assign corresponding roughness, infiltration rate, and runoff generation and confluence characteristics to each area; the drainage network data includes pipe diameter, pipe length, material, slope, burial depth, roughness, manhole elevation, and node connection relationships, and provides the operating parameters of hydraulic structures.

3. The urban flooding dynamic early warning method according to claim 1, which integrates multi-source data and a high-performance numerical model, is characterized in that... The one-dimensional river channel model mentioned in step 2 is a one-dimensional hydrodynamic model of the river channel with the river cross-section as the main body and considering hydraulic structures, used to describe the river water level and backwater effect; the two-dimensional surface model simulates the evolution of surface runoff and water accumulation based on DEM data, divides the road areas of focus and those where surface water accumulation may occur into grids, and sets roughness coefficient, building area coefficient, and infiltration rate parameters; the underground drainage network model includes a runoff generation and runoff collection submodule and a one-dimensional network runoff collection submodule, which provides upstream inflow process lines for the network drainage nodes by calculating the runoff generation and runoff collection process based on the sub-catchment area, and then performs hydraulic simulation on the one-dimensional integrated network system including control facilities, calculating its free surface flow and pressure flow states; The specific process of establishing the urban flood coupling model is as follows: By connecting the underground drainage network model with the two-dimensional surface model through the elevation relationship between the rainwater wells and the corresponding ground, the water flow exchange between the overflow of the rainwater wells and the surface drainage is realized; the river outlet of the underground drainage network model is connected with the one-dimensional river channel model to realize the water flow exchange between the network system and the river channel; the two-way coupling between the riverbanks and the two-dimensional surface model is realized through lateral connection, and the real boundary conditions are constructed in combination with the operation rules of hydraulic structures, thereby constructing the urban flood coupling model.

4. The urban flooding dynamic early warning method according to claim 1, which integrates multi-source data and a high-performance numerical model, is characterized in that... The model state variables mentioned in step 3 include water level; the key parameters include roughness and infiltration rate.

5. The urban flooding dynamic early warning method according to claim 1, which integrates multi-source data and a high-performance numerical model, is characterized in that... The automatic identification of waterlogging points in step 5 is specifically formulated by referring to the drainage standards and experience of Beijing, Shenzhen and Shanghai, and setting a water depth of 15 cm as the basic threshold for identifying waterlogging points. Furthermore, the criteria for determining waterlogging points are increased by combining the duration, flow velocity and water accumulation range. The risk level classification is specifically as follows: based on different thresholds, it is divided into general risk, medium risk, relatively high risk, and high risk. The method of determining the warning level according to the triggering rules and using different colored warning signals to correspond to different warning levels is as follows: blue, yellow, orange, and red warning signals are used to correspond to different warning levels; a blue warning is triggered when two or more general risk waterlogging points or one moderate risk waterlogging point appear in a street. A yellow alert is triggered when there are two or more medium-risk or one high-risk flooding point; an orange alert is triggered when there are two or more high-risk or one high-risk flooding point; and a red alert is triggered when there are two or more high-risk flooding points.

Images