A cross-domain data fusion early warning method for urban waterlogging

Abstract

The present application relates to the technical field of smart city and emergency management, and more particularly to a cross-domain data fusion early warning method for urban waterlogging. The method comprises the following steps: obtaining original multi-source heterogeneous data; spatially registering and rasterizing the original multi-source heterogeneous data to obtain a city grid attribute table; constructing a city bearing bottom plate according to the city grid attribute table to obtain a multi-dimensional static bearing capacity atlas; collecting city real-time state data flow, superimposing and analyzing influence factors on the multi-dimensional static bearing capacity atlas to obtain a comprehensive influence coefficient distribution map; extracting the reference drainage capacity value of each grid in the multi-dimensional static bearing capacity atlas, and combining the comprehensive influence coefficient distribution map to dynamically correct the drainage capacity to obtain an instantaneous drainage efficiency matrix. The present application solves the problem that the traditional method cannot dynamically quantify the influence of the city operation state on the drainage system, and ultimately realizes accurate tracing and cascade influence early warning of waterlogging risk.

Description

Technical Field

[0001] This invention relates to the field of smart city and emergency management technology, and in particular to a cross-domain data fusion early warning method for urban flooding. Background Technology

[0002] Existing urban flooding early warning methods generally adopt a static threshold model, which relies on a single or fixed rainfall threshold (such as "issue an early warning if rainfall exceeds 50 mm / hour"). This treats urban drainage capacity as a constant and ignores the significant impact of dynamic factors such as pipe network blockage, pump station operation status, and previous rainfall on actual drainage capacity, resulting in frequent false alarms and missed warnings.

[0003] Existing urban flooding early warning methods focus on the symptoms of water accumulation rather than the underlying mechanisms, failing to identify the causal chains and key triggers of flooding. They lack the ability to trace chain reactions such as "overflow at point A triggering widespread water accumulation in area B," leading to flood control resources often being invested in treating water accumulation rather than preventing it at its source, resulting in a cycle of "treating the symptoms but not the root cause."

[0004] The existing risk assessment system is limited to direct flooding losses, using water depth as a single risk indicator. It ignores the indirect impact of urban flooding on urban functional networks and cannot identify and assess the risks of functional isolation of key facilities caused by the fragmentation of transportation networks. This results in incomplete information dimensions for flood control decision-making and systemic blind spots in emergency response.

[0005] In summary, existing technologies suffer from systemic problems such as insufficient early warning accuracy, inadequate source identification capabilities, and a single dimension of risk assessment, which urgently need to be addressed. Summary of the Invention

[0006] Therefore, it is necessary to provide a cross-domain data fusion early warning method for urban flooding to solve at least one of the above-mentioned technical problems.

[0007] To achieve the above objectives, a cross-domain data fusion early warning method for urban flooding includes the following steps:

[0008] Step S1: Obtain raw multi-source heterogeneous data through the real-time data interfaces of various departments in the city; perform spatial registration and rasterization processing on the raw multi-source heterogeneous data to obtain the city raster attribute table; construct the city's carrying capacity base based on the city raster attribute table to obtain a multi-dimensional static carrying capacity map.

[0009] Step S2: Collect real-time urban status data stream, perform influence factor superposition analysis on the multidimensional static bearing capacity map to obtain the comprehensive influence coefficient distribution map; extract the baseline drainage capacity value of each grid in the multidimensional static bearing capacity map, and perform dynamic correction of drainage capacity in combination with the comprehensive influence coefficient distribution map to obtain the instantaneous drainage efficiency matrix;

[0010] Step S3: Obtain rainfall forecast data and construct effective runoff distribution data by combining instantaneous drainage efficiency matrix; use effective runoff distribution data to identify pipeline drainage overflow nodes in multidimensional static bearing capacity map to obtain overflow node feature set; perform time series extrapolation of surface water accumulation state on overflow node feature set to obtain spatiotemporal risk evolution sequence.

[0011] Step S4: Identify risk characteristics of the spatiotemporal risk evolution sequence to obtain a risk node time series feature table; perform source tracing analysis and initial bottleneck identification on the risk node time series feature table to obtain an initial bottleneck node set; identify indirect impact paths of the spatiotemporal risk evolution sequence to obtain a key impact path map; perform dual risk node identification and rating on the initial bottleneck node set and key impact path map to obtain graded early warning data; distribute multi-channel early warning information from the graded early warning data to form multi-channel early warning instructions.

[0012] This invention dynamically corrects urban drainage capacity by introducing real-time dynamic events and time gradient functions, generating an instantaneous drainage efficiency matrix that reflects real-time conditions such as pipe network blockage and traffic congestion. This completely solves the problem of insufficient early warning accuracy and frequent false alarms and missed alarms caused by existing technologies that rely solely on static rainfall thresholds. Through source tracing analysis and trigger chain identification methods, this invention can trace back from complex waterlogging phenomena and pinpoint the initial bottleneck nodes that trigger chain reactions, transforming early warning and emergency response from a passive "stopgap" approach to a proactive "root cause" approach, enabling precise allocation and pre-deployment of flood control resources. Furthermore, this invention assesses the disruptive effect of waterlogging on urban transportation networks and its impact on the accessibility of critical facilities (such as hospitals) through indirect impact path identification, expanding risk assessment from a single direct flood loss to a systemic dimension of urban functional network security, overcoming the significant deficiency of existing technologies' single-dimensional risk assessment. This significantly improves the initiative, accuracy, and scientific rigor of urban flooding early warning systems. Attached Figure Description

[0013] Figure 1 This is a flowchart illustrating the steps of a cross-domain data fusion early warning method for urban flooding. Detailed Implementation

[0014] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0015] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0016] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0017] To achieve the above objectives, please refer to Figure 1 This invention provides a cross-domain data fusion early warning method for urban flooding, comprising the following steps:

[0018] Step S1: Obtain raw multi-source heterogeneous data through the real-time data interfaces of various departments in the city; perform spatial registration and rasterization processing on the raw multi-source heterogeneous data to obtain the city raster attribute table; construct the city's carrying capacity base based on the city raster attribute table to obtain a multi-dimensional static carrying capacity map.

[0019] In one embodiment, raw, multi-source heterogeneous data, including digital elevation model, impermeable layer, drainage network, and soil type, are acquired through a data interface and unified into the CGCS2000 coordinate system. These data are then registered and rasterized into 50m x 50m units to form a basic spatial element layer set. Next, the vulnerability of the pipe network is quantified by scoring its pipe age and material, resulting in an urban raster attribute table. Simultaneously, the soil's pre-saturation is calculated using rainfall records from the past 72 hours. Finally, multiple attributes, including elevation, impermeability, network vulnerability score, and soil saturation, are overlaid onto a unified raster to construct a multi-dimensional static bearing capacity map that comprehensively reflects the city's static carrying capacity.

[0020] Step S2: Collect real-time urban status data stream, perform influence factor superposition analysis on the multidimensional static bearing capacity map to obtain the comprehensive influence coefficient distribution map; extract the baseline drainage capacity value of each grid in the multidimensional static bearing capacity map, and perform dynamic correction of drainage capacity in combination with the comprehensive influence coefficient distribution map to obtain the instantaneous drainage efficiency matrix;

[0021] In one embodiment, real-time urban status data streams, such as traffic congestion index, smart manhole cover blockage status, and pump station start / stop status, are collected via API interface at 5-minute intervals. These dynamic events are located to corresponding grid cells, and qualitative event states (such as "severe congestion" or "manhole cover blockage") are converted into quantitative surface runoff delay coefficients or inflow capacity reduction rates through preset quantification rules, forming a comprehensive impact coefficient distribution map. Simultaneously, a time gradient function is introduced to amplify the cumulative effect of persistent events (such as manhole cover blockage lasting 2 hours), and coupled situational analysis is performed based on the pipeline network topology. Finally, these dynamic correction factors are applied to the baseline drainage capacity values ​​extracted from the multidimensional static bearing capacity map to generate an instantaneous drainage efficiency matrix reflecting the current true drainage efficiency in every corner of the city.

[0022] Step S3: Obtain rainfall forecast data and construct effective runoff distribution data by combining instantaneous drainage efficiency matrix; use effective runoff distribution data to identify pipeline drainage overflow nodes in multidimensional static bearing capacity map to obtain overflow node feature set; perform time series extrapolation of surface water accumulation state on overflow node feature set to obtain spatiotemporal risk evolution sequence.

[0023] In one embodiment, radar rainfall forecasts for the next 6 hours are acquired and decomposed into 10-minute time series corresponding to 50m × 50m grids. At each time step, the effective runoff generated after deducting infiltration from rainfall is calculated. This runoff is collected on the surface based on elevation data. When it reaches the storm drain, it is compared with the real-time inflow capacity defined in the instantaneous drainage efficiency matrix to allocate the amount of water entering the pipe network and the amount of water remaining on the surface. The amount of water entering the pipe network is simulated using simplified hydraulic methods to identify manholes overflowing due to overload, forming a feature set of overflow nodes. The flow generated by these overflow nodes is combined with the amount of water remaining on the surface, and then the diffusion process on the surface is quickly simulated using a simplified hydrodynamic method based on depression filling and weir flow formulas. Finally, a spatiotemporal risk evolution sequence containing the entire process of the evolution of water accumulation range and depth over the next 6 hours is generated.

[0024] Step S4: Identify risk characteristics of the spatiotemporal risk evolution sequence to obtain a risk node time series feature table; perform source tracing analysis and initial bottleneck identification on the risk node time series feature table to obtain an initial bottleneck node set; identify indirect impact paths of the spatiotemporal risk evolution sequence to obtain a key impact path map; perform dual risk node identification and rating on the initial bottleneck node set and key impact path map to obtain graded early warning data; distribute graded early warning information through multiple channels to form multi-channel early warning instructions.

[0025] In one embodiment, risk identification is performed in two dimensions based on the spatiotemporal risk evolution sequence. First, a source tracing analysis is conducted by constructing a temporal causal relationship graph between overflow nodes to identify and quantify the "trigger chain," thereby tracing back to the "initial bottleneck node set" that triggers the chain overflow. Second, an indirect impact analysis is performed by assessing the predicted blocking effect of water accumulation on the urban traffic network and calculating the "accessibility attenuation rate" of key facilities such as hospitals and substations, thereby identifying the "critical impact path map" and "critical cutoff points" that can cause systemic traffic disruption. Finally, the "initial bottleneck nodes" (causes) and "critical cutoff points" (effects) are comprehensively scored and ranked in a dual-standard evaluation system, and combined with the urgency of their expected occurrence, hierarchical early warning data containing clear warning levels, locations, and time windows is generated. This data is then transformed into differentiated, executable multi-channel early warning instructions for different departments and the public for distribution.

[0026] Preferably, step S1 includes:

[0027] Acquire raw, multi-source heterogeneous data, including DEM data, impermeable layer distribution data, drainage network data, soil type data, and historical rainfall records;

[0028] Spatial feature rasterization and registration are performed on the original multi-source heterogeneous data to obtain a basic spatial feature layer set;

[0029] The vulnerability of the drainage network is quantified and assigned to the basic spatial element layer set using drainage network data, resulting in a city raster attribute table.

[0030] Collect cumulative rainfall data over the past 72 hours and combine it with raw multi-source heterogeneous data to calculate the current soil saturation status.

[0031] Multidimensional static attribute overlays are performed on the basic spatial element layer set, urban raster attribute table, and current soil saturation status data to form a multidimensional static bearing capacity map.

[0032] In one embodiment, firstly, data stored in a raw multi-source heterogeneous database is acquired through a preset data acquisition interface. This database includes: digital elevation model (DEM) data with a spatial resolution of 5 meters, in GeoTIFF format; surface impermeable layer distribution data obtained from remote sensing image interpretation, in raster format; urban drainage network data provided by the water resources department, in Shapefile format, including pipeline location, diameter, material, and construction date attributes; and urban soil type distribution maps and historical daily rainfall records for the past 365 days provided by the surveying and mapping department.

[0033] Next, spatial feature rasterization and registration are performed. All acquired vector and raster data are uniformly converted to the CGCS2000 geodetic coordinate system. Using the city's administrative boundaries as the scope, standard 50m × 50m raster units are defined. The digital elevation model, impermeable layer distribution data, and soil type distribution map are resampled or rasterized to generate elevation rasters, impermeability rasters, and soil type rasters with the same resolution and spatial range. These rasters together constitute the basic spatial feature layer set.

[0034] Next, the vulnerability of the pipe network is quantified and assigned. For each 50m x 50m grid cell, the vulnerability of the internal pipe network is calculated. This process is achieved by calculating a pipe network health status score, which is based on pipe age, historical maintenance frequency, and material. For example, pipes older than 20 years are scored 10 points, those between 10 and 20 years are scored 5 points; historical maintenance frequency more than once a year is scored 10 points; and concrete is scored 5 points. The scores of all pipe segments within the grid are weighted and averaged, and then combined with the pipe network density within the grid, to finally assign a quantified vulnerability score of 0 to 100 to each grid cell. This score is used as a new attribute column and merged with basic attributes such as elevation and impermeability to form an urban grid attribute table, where each row corresponds to a grid cell and each column represents an attribute.

[0035] Subsequently, the current soil saturation state data was calculated. Hourly rainfall data for the past 72 hours was extracted from historical rainfall records. The current soil moisture content was calculated using the anterior impact rainfall (API) method, which is performed iteratively using the following formula:

[0036] ;

[0037] in The amount of rainfall affecting the current moment. The amount of rainfall that affected the previous moment. This refers to the attenuation coefficient. The value is determined based on the soil type raster: 0.92 for clay, 0.87 for loam, 0.83 for sandy loam, and 0.85 for denser artificial fill. The calculated final antecedent rainfall is compared with the field capacity of the soil type to obtain a soil saturation value between 0 and 1, forming the current soil saturation status data.

[0038] Finally, multidimensional static attribute overlay is performed. The basic spatial feature layer set (including elevation, impermeability, etc.), the urban raster attribute table (including pipeline vulnerability scores), and the current soil saturation data are spatially connected according to a unified raster coordinate system. This operation generates a feature vector containing all static and quasi-static attributes for each raster cell. The resulting multidimensional static bearing capacity map is structurally represented as a multi-band raster file, where the values ​​of each cell (raster cell) in different bands represent a series of bearing capacity-related attributes such as elevation, impermeability, pipeline vulnerability, and soil saturation.

[0039] The city grid attribute table is shown in Table 1.

[0040] Table 1

[0041]

[0042] Preferably, step S2 involves collecting real-time urban status data streams and performing influence factor overlay analysis on the multidimensional static carrying capacity map, including:

[0043] Based on the multidimensional static bearing capacity map, multi-source dynamic data are collected in real time to form a real-time urban status data stream;

[0044] Dynamic spatial location of events is performed on the real-time status data stream of the city to obtain a spatial location event distribution map;

[0045] Based on the spatial location event distribution map, the impact factors are quantified and converted to obtain the urban event impact factor table;

[0046] By using the urban event impact factor table to conduct a multi-layer impact superposition analysis on the multi-dimensional static bearing capacity map, a comprehensive impact coefficient distribution map is obtained.

[0047] In one embodiment, firstly, multi-source dynamic data is collected in real time. Through a preset API interface, real-time traffic congestion index data covering main and secondary roads is obtained from the city's traffic command center with a refresh cycle of 5 minutes. Simultaneously, a smart manhole cover sensor network deployed in low-lying areas is accessed to obtain manhole cover water level and blockage status data; and the SCADA system of the water department's pumping stations is connected to obtain the start / stop status and real-time drainage flow of each pumping station. This data, carrying timestamps and location identifiers, collectively constitutes the city's real-time status data stream.

[0048] Then, dynamic event spatial localization is performed on the real-time urban status data stream. The geographic coordinates or device number in each event record are matched to the unique identifier of its corresponding 50m x 50m grid cell using a spatial index. For linear traffic congestion events, all grid cells within their affected area are marked. This operation generates a dynamically updated spatial location event distribution map, where each grid cell is labeled with the specific event type and status currently occurring.

[0049] Next, based on the spatial location event distribution map, the influencing factors are quantified and transformed. The logic behind this is to convert qualitative, multi-source urban operational events into unified quantitative indicators affecting drainage capacity. The specific steps are as follows:

[0050] 1. For traffic congestion events, the congestion level (divided into four levels: smooth, slow, congested, and severe congestion) issued by the traffic control center is converted into a surface runoff confluence time delay coefficient. For example, when the road corresponding to a certain grid is "severely congested", its delay coefficient is set to 1.3, which means that the time for surface water to flow into the storm drain of that grid is increased by 30% compared to the normal state.

[0051] 2. For manhole cover blockage events, the water level data reported by the smart manhole cover sensor is converted into an inflow capacity reduction rate. When the sensor water level exceeds 50% of the manhole opening height, it is determined to be a partial blockage, and the inflow capacity reduction rate of the corresponding rainwater inlet is set to 0.6, meaning that its drainage capacity is only 40% of the design capacity.

[0052] 3. For pump station operation events, the start / stop status fed back by the SCADA system is directly mapped to the drainage efficiency correction factor of the downstream pipe network. When a critical pump station is in a shutdown state, the drainage efficiency correction factor for all grids within its influence range is set to 0.1.

[0053] Using the above rules, all dynamic events are converted into specific numerical factors, forming a table of urban event impact factors.

[0054] Finally, a multi-layered impact analysis was conducted on the multi-dimensional static carrying capacity map using the urban event impact factor table. When a grid cell is simultaneously affected by multiple events, the stratified worst-case principle is adopted for superposition to reflect the comprehensive effect of risks. For example, a grid cell may have both a "severe congestion" event (affecting surface runoff) and a "partial manhole cover blockage" event (affecting pipeline inflow). First, the surface runoff lag coefficient is applied, and then multiplied by the pipeline inflow reduction rate to obtain a comprehensive impact coefficient, the calculation formula of which is:

[0055] ;

[0056] in The comprehensive impact coefficient, The surface event impact coefficient. This represents the impact coefficient of the network connection event. This calculation is applied to all grid cells to generate a comprehensive impact coefficient distribution map. Each grid value in this map represents the degree of reduction in the overall efficiency of the drainage system at that location due to real-time urban events.

[0057] Preferably, the quantitative transformation of influencing factors based on the spatial location event distribution map includes:

[0058] Based on the spatial location event distribution map, events are classified and their impacts are analyzed to obtain a mapping table of event influencing factors.

[0059] The historical cumulative effect was analyzed using the event influencing factor mapping table, resulting in a time cumulative influence coefficient table.

[0060] The baseline impact coefficient of urban waterlogging was calibrated based on the time cumulative impact coefficient table, and the quantitative data of the baseline impact coefficient were obtained.

[0061] Based on the baseline impact coefficient quantification data, a dynamic event hierarchy mapping is performed on the spatial location event distribution map to obtain the urban event impact factor table.

[0062] In one embodiment, firstly, events are classified and their impact analyzed based on a spatial location event distribution map. All dynamic events are divided into four categories: surface events (such as traffic congestion), network events (such as manhole cover blockage), pipeline events (such as pipe siltation), and discharge events (such as pump station shutdown). The core impact mechanism on the drainage system is defined for each type of event. For example, the impact mechanism of traffic congestion is defined as "slowing down the rate of surface runoff accumulation," and the impact mechanism of manhole cover blockage is defined as "reducing the inflow capacity of storm drains." These "event type-impact mechanism" correspondences are stored to form an event impact factor mapping table.

[0063] Next, a historical cumulative effect analysis is performed using an event influencing factor mapping table. The core of this analysis lies in quantifying the amplifying effect of event duration on the intensity of impact. Data on all recorded flooding events and their associated dynamic events from the past year are retrieved. Through analysis, functions are established to express the time-varying impact of different types of events. For example, the cumulative effect of the impact coefficient for manhole cover blockage events is calculated using the following formula:

[0064] ;

[0065] in, The duration is The final impact coefficient at that time This is the initial impact coefficient when the event occurs. This is the time accumulation factor for this type of event (e.g., for garbage congestion). (Set to 0.3) The reference time unit is set to 1 hour. Through fitting analysis of historical data, a time unit is determined for each event type. The values ​​are ultimately used to form a table of time-cumulative impact coefficients containing cumulative effect functions for different event types.

[0066] Then, the baseline impact coefficient of urban flooding is calibrated according to the time-cumulative impact coefficient table. At least 100 historical flooding cases are selected. For each case, the final water depth and extent, as well as the type and duration of the dynamic event causing the flooding, are known. Using the functions in the time-cumulative impact coefficient table, the "initial impact coefficient" F_0 required for various events (such as severe congestion or severe manhole cover blockage) to produce the known flooding consequences is calculated backwards without considering the time-cumulative effect. By statistically averaging the back-calculation results of a large number of cases, a standardized baseline impact coefficient lookup table is obtained; this table represents the quantitative data for the baseline impact coefficient.

[0067] Finally, a dynamic event hierarchy mapping is performed on the spatial location event distribution map based on the baseline impact coefficient quantification data. For any dynamic event obtained from the spatial location event distribution map at the current moment (e.g., the manhole cover blockage event with ID A01, which has lasted for 3 hours), the initial impact coefficient corresponding to "manhole cover blockage" is first retrieved from the baseline impact coefficient quantification data. (For example, 0.6). Then, using a function from the time-cumulative influence coefficient table, the duration is... Enter 3 hours into the formula to calculate the time-cumulative effect multiplier (e.g., 1.14). Multiply the initial impact coefficient by the time-cumulative multiplier (0.6 × 1.14 = 0.684) to obtain the final impact factor of the event at the current moment. Apply this calculation to all dynamic events to generate a table of city event impact factors, where each entry contains the event ID, location, type, and the final quantified impact factor after time-cumulative effect correction.

[0068] Preferably, step S2 involves extracting the baseline drainage capacity value of each grid cell in the multidimensional static bearing capacity map and performing dynamic correction of the drainage capacity in conjunction with the comprehensive influence coefficient distribution map, including:

[0069] Acquire and record the duration of historical drainage events to obtain an event duration table;

[0070] Time gradient analysis is performed based on the event duration table to obtain the timeliness impact coefficient;

[0071] Data on the urban drainage system is obtained through a real-time data interface. Based on the distribution maps of the timeliness impact coefficient and the comprehensive impact coefficient, the coupling status analysis of the urban drainage system is conducted to obtain the system coupling status index.

[0072] The system coupling situation index is used to correct the baseline drainage capacity value for dynamic drainage capacity, and a preliminary dynamic adjustment capacity value is obtained.

[0073] Physical constraints are imposed on the initial dynamic adjustment capacity value, and spatial consistency optimization is performed to obtain the dynamic adjustment drainage capacity value.

[0074] The dynamic adjustment drainage capacity value and the multidimensional static bearing capacity spectrum are spatiotemporally integrated to obtain the instantaneous drainage efficiency matrix.

[0075] In one embodiment, historical drainage event data is acquired and the event duration is recorded. A database containing records of all dynamic events (such as manhole cover blockages and traffic congestion) over the past 365 days is connected. For each event currently occurring, its start timestamp is recorded, and its duration up to the current time is calculated. For example, if a manhole cover blockage event started at 8:00 AM and the current time is 10:00 AM, its duration is recorded as 2 hours. These records constitute an event duration table.

[0076] The core logic of time gradient analysis based on event duration tables lies in the fact that the impact of an event is not constant but evolves over time. According to a pre-defined time gradient function library, the timeliness impact coefficient is calculated for each event. For example, for a manhole cover blockage event that has lasted for 2 hours, its corresponding exponential cumulative function is called, where the exponential cumulative function is:

[0077] ;

[0078] in For an event with a duration of The timeliness impact coefficient is calculated as follows: where t is the duration in hours.

[0079] ;

[0080] This means that the impact of the event is 14% stronger than at the initial moment. Applying this calculation to all events generates a timeliness impact coefficient.

[0081] A coupling situation analysis of the urban drainage system was conducted to assess its overall vulnerability. Real-time data on the topology of the urban drainage network and the real-time status of key nodes (such as large pumping stations and main pipe junctions) were obtained through a real-time data interface. Based on the comprehensive influence coefficient distribution map and the timeliness influence coefficient, multiple affected network nodes were identified. By analyzing the upstream and downstream relationships of these nodes in the network topology, the risk of "series failure" was identified. For example, when the inflow capacity of multiple upstream branches of a main pipe decreases by more than 50% due to an event, the main pipe is deemed to have coupling risk, and a system coupling situation index is calculated for its associated grid, with a value set at 0.8, indicating that the overall efficiency of the drainage system in that area is further reduced due to the cascading effect.

[0082] The baseline drainage capacity value (e.g., 100 cubic meters per hour) for each grid is extracted from the multidimensional static bearing capacity map. This baseline value is multiplied by the grid's comprehensive influence coefficient, timeliness influence coefficient, and system coupling situation index to obtain the preliminary dynamic adjustment capacity value. For example, if grid A has a baseline capacity of 100, a comprehensive influence coefficient of 0.7, a timeliness influence coefficient of 1.1, and a system coupling situation index of 0.9, then its preliminary dynamic adjustment capacity value is:

[0083] 100×(1-(1-0.7)×1.1)×0.9=60.3 cubic meters / hour;

[0084] Check whether the calculated capacity value exceeds the physical limits of the drainage facility corresponding to the grid (e.g., it cannot be lower than the minimum operating capacity or higher than the maximum design capacity). For values ​​that are out of range, correct them to the nearest limit value. At the same time, smooth out any unreasonable large differences in capacity between adjacent grids to ensure that the corrected drainage capacity is spatially continuous, ultimately obtaining a dynamically adjusted drainage capacity value.

[0085] Divide the dynamically adjusted drainage capacity value of each grid cell by its baseline drainage capacity value to obtain a percentage, i.e., instantaneous drainage efficiency. Integrate the instantaneous drainage efficiency values ​​of all grid cells into a matrix corresponding to the urban geospatial space, and append the current timestamp to this matrix. This matrix is ​​the instantaneous drainage efficiency matrix, which dynamically reflects the actual drainage capacity of each corner of the city relative to its design capacity at the current moment.

[0086] Preferably, step S3 includes:

[0087] Acquire rainfall forecast data from meteorological departments; use the instantaneous drainage efficiency matrix to perform spatiotemporal decomposition of rainfall forecast data from meteorological departments to obtain rasterized rainfall forecast sequences;

[0088] Inflow calculation and allocation are performed on the gridded rainfall forecast sequence to obtain effective runoff distribution data;

[0089] By using effective runoff distribution data to trace surface runoff paths in a multidimensional static bearing capacity map, a surface runoff path network is obtained.

[0090] By combining the surface runoff path network and the instantaneous drainage efficiency matrix, the inflow allocation of the pipe network is calculated, and the inflow allocation table of the pipe network nodes is obtained.

[0091] Based on the inflow allocation table of the pipeline node, the hydraulic transmission of the pipeline network is simulated to obtain the hydraulic state time series table of the pipeline network.

[0092] Overflow node identification and analysis were performed on the hydraulic state time series table of the pipeline network and the inflow distribution table of the pipeline network nodes to obtain the overflow node feature set;

[0093] The overflow node feature set is redistributed to the surface runoff path network for surface water formation and diffusion processing to obtain a dynamic distribution map of water accumulation.

[0094] The spatiotemporal risk sequence is integrated from the dynamic distribution map of water accumulation, the feature set of overflow nodes, and the time series table of hydraulic status of the pipe network to form a spatiotemporal risk evolution sequence.

[0095] In one embodiment, radar rainfall forecast data for the next 6 hours issued by the meteorological department is acquired. Then, the rainfall forecast data is decomposed in time and space. This process divides the rainfall forecast for the next 6 hours into 10-minute time steps and uses spatial interpolation methods to convert the radar echo intensity data into rainfall data that completely corresponds to a 50m × 50m grid in the city, ultimately forming a rasterized rainfall forecast sequence containing 36 time slices.

[0096] Inflow calculation and allocation are performed on the rasterized rainfall forecast sequence. For each raster cell in each time slice, the proportion of rainfall converted into surface runoff is calculated based on its impermeability attribute (e.g., 0.8) (e.g., 10 mm of rainfall results in 8 mm of runoff). This is then subtracted from the infiltration loss calculated based on soil saturation and soil type to obtain the final effective runoff generated by that raster within that time step. This calculation is applied to all raster cells to obtain effective runoff distribution data.

[0097] Based on the elevation data in the map, the D8 algorithm is used to determine the surface water flow direction of each grid cell to the lowest elevation among its eight neighbors. By iterating this process, the effective runoff generated by each grid cell is tracked to collect on the surface, ultimately forming a surface runoff path network describing the water flow from the source to the storm drain or depression.

[0098] For each grid cell located at a storm drain inlet, the total runoff collected from its surface is compared with the real-time inflow capacity of that grid cell in the instantaneous drainage efficiency matrix. If the runoff is less than the inflow capacity, all runoff enters the pipe network; if the runoff is greater than the inflow capacity, the excess is retained as surface water, and the portion that can enter the pipe network is recorded to form a pipe network node inflow allocation table.

[0099] Using the pipe network topology and a simplified motion wave method, the transport process of incoming water in the underground pipe network is simulated. This process calculates the flow rate of each pipe segment and the water level of each manhole at each time step, ultimately forming a time series table of the pipe network hydraulic state containing the predicted water level and flow rate of each pipe network node for the next 6 hours.

[0100] Check the hydraulic status time series table of the pipeline network to see if there are any cases where the water level of a manhole exceeds its head elevation. If so, mark the manhole as an overflow node. At the same time, calculate its overflow flow rate per minute (i.e., the flow rate exceeding the head), and record information such as the overflow start time and the expected peak time to form an overflow node feature set.

[0101] The overflow flow from each overflow node is treated as a new surface water source and added to the corresponding grid cell in the surface runoff path network. Combined with the previously unconnected surface water volume, a depression-filling algorithm is used to simulate the accumulation and diffusion process of this water on the surface, and the dynamic changes in water depth of each grid cell in the next 6 hours are calculated to form a dynamic water distribution map.

[0102] The spatiotemporal risk sequence is integrated from the dynamic distribution map of water accumulation, the feature set of overflow nodes, and the time series table of the hydraulic status of the pipe network. This process integrates all the results calculated by time step (including water accumulation depth, overflow nodes, pipe network water level, etc.) into a unified data structure. This structure consists of a series of time-sorted data frames, each of which is a complete snapshot of the urban flooding risk at a future moment, collectively forming a spatiotemporal risk evolution sequence.

[0103] Preferably, the process of redistributing the overflow node feature set to the surface runoff path network for surface water formation and diffusion treatment includes:

[0104] The overflow node feature set is used to locate and quantify the water source points, resulting in a water source point distribution map.

[0105] The water accumulation and depression filling calculation was performed using a multidimensional static bearing capacity map to obtain data on the depression filling status.

[0106] A simplified hydrodynamic diffusion simulation was performed based on the distribution map of water sources and the data on the filling status of depressions to obtain the temporal variation data of water volume in depressions.

[0107] By combining the temporal variation data of water volume in depressions and the multidimensional static bearing capacity map, the water depth is calculated in a gridded manner to form gridded water depth distribution data.

[0108] The data on water depth distribution in the raster is used to identify contiguous water accumulation areas and extract their boundaries. Then, the data is dynamically integrated over time to obtain a dynamic water accumulation distribution map.

[0109] In one embodiment, the overflow node feature set is used to locate and quantify the water source. The coordinates of each overflow node are mapped to a corresponding 50m × 50m grid cell. Simultaneously, the overflow flow rate (e.g., 0.1 cubic meters per second) of that node at each time step (e.g., 10 minutes) is used as the water source input for that grid cell. All overflow nodes and the grid cells containing surface water are aggregated to form a water source distribution map, which clearly marks the source location and volume of newly accumulated surface water within each time step.

[0110] Based on the high-precision digital elevation model in the map, all depressions formed by terrain closure are identified, and the water level-volume relationship curve is calculated for each depression. At each time step, water from the water source distribution map is preferentially injected into the corresponding depression. Based on the water volume and volume relationship curve, the water surface elevation of the depression after filling is calculated. This process generates depression filling status data, recording the current water storage and water surface elevation of each depression.

[0111] A simplified hydrodynamic diffusion simulation is performed based on the distribution map of water sources and the filling status data of depressions. The logic behind this simulation is to achieve rapid water diffusion through simplified calculations. Diffusion calculations are initiated when the water surface elevation of a depression exceeds the elevation of its lowest overflow point. It uses the weir flow formula to calculate the flow exchange between adjacent depressions based on the water level difference. The weir flow formula calculation process is as follows:

[0112] ;

[0113] in For traffic, For flow coefficient, The width of the overflow outlet. The water head difference is used to simulate the cascading diffusion of water from high-water-level depressions to low-water-level depressions by iterating this calculation at each time step, thus obtaining time-series data on the water volume variation in the depressions.

[0114] A rasterized calculation of water depth is performed by combining temporal variation data of depression water volume and multidimensional static bearing capacity maps. At each time step, the updated total water volume of each depression is redistributed to all the raster cells it covers, based on the topographic relief within the depression. All raster cells within a depression have the same water surface elevation, and the water depth of each cell is equal to that water surface elevation minus its own surface elevation. Through this calculation, discrete depression water storage information is transformed into continuous raster water depth distribution data.

[0115] The system identifies and extracts boundaries of contiguous water accumulation areas from raster-based water depth distribution data, and then dynamically integrates the data over time. A water depth threshold (e.g., 0.05 meters) is set. All adjacent raster cells with depths exceeding this threshold are identified as a contiguous water accumulation area. A boundary profile is generated for each contiguous area, and its total area, average depth, and other attributes are calculated. The water depth raster and contiguous area information generated at all time steps are organized chronologically to ultimately form a dynamic water accumulation distribution map that dynamically displays the entire process of water accumulation range and depth from formation to evolution.

[0116] Preferably, step S4, which involves source tracing analysis and initial bottleneck identification of the risk node time series feature table, includes:

[0117] Based on the first overflow time of the nodes in the risk node time sequence characteristic table, construct a time sequence overflow connection graph;

[0118] Overflow propagation path identification is performed on the time-series overflow connection graph to obtain the overflow propagation path tree;

[0119] Trigger chain identification and quantification are performed on the overflow propagation path tree to obtain a node trigger chain feature table;

[0120] Based on the node trigger chain feature table and the spatiotemporal risk evolution sequence, the water contribution is tracked to obtain the overflow contribution matrix;

[0121] The overflow source is located based on the overflow contribution matrix, and the source weight is calculated to form a node source weight table;

[0122] The initial bottleneck node set is determined based on the node source weight table and the overflow propagation path tree.

[0123] In one embodiment, a time-series overflow connection graph is constructed based on the timestamp of the first overflow of each overflow node in the risk node time-series characteristic table. This graph is a directed acyclic graph, where each overflow node is a graph node. If node A and node B are physically connected (through a pipeline or the ground surface), and A's overflow time is earlier than B's, then a directed edge from A to B is established between A and B, representing a potential overflow propagation relationship.

[0124] Next, overflow propagation paths are identified in the temporal overflow connection graph. Starting from all nodes with an in-degree of 0 (i.e., the earliest overflowing nodes without upstream overflow sources), a depth-first search algorithm is used to traverse all paths from these source nodes to other nodes. Each complete path represents the propagation process of an overflow event from the source to the downstream. All these paths together constitute the overflow propagation path tree.

[0125] Next, trigger chain identification and quantification are performed on the overflow propagation path tree. The core logic lies in identifying and quantifying the causal relationships between nodes. Analyzing the structure of the overflow propagation path tree, the set of all downstream nodes reachable from a source node is defined as a "trigger chain" triggered by that source node. The length (total number of nodes) and breadth (number of direct downstream branches) of the trigger chain for each source node are calculated. These quantified metrics are recorded to form a node trigger chain feature table.

[0126] Subsequently, water contribution is tracked based on the node trigger chain feature table and the spatiotemporal risk evolution sequence. For any downstream node in the trigger chain, if it has multiple upstream overflow sources, the overflow flow data of each upstream source is extracted from the spatiotemporal risk evolution sequence. Through simplified water balance calculation, the percentage contribution of each upstream source to the total overflow of the downstream node is assigned. For example, if the total overflow of downstream node C is 1 cubic meter per second, and its upstream A contributes 0.7 cubic meters per second and its upstream B contributes 0.3 cubic meters per second, then the contributions of A and B are 70% and 30%, respectively. The contribution relationships between all nodes are calculated and stored to form an overflow contribution matrix.

[0127] Next, the overflow sources are identified based on the overflow contribution matrix, and source weights are calculated. All nodes with an in-degree of 0 are identified as initial overflow sources. A comprehensive source weight is calculated for each source node, taking into account three factors: trigger chain length (higher value, higher weight), total water volume contribution (higher total water volume contribution to all downstream nodes, higher weight), and overflow time (earlier overflow, higher weight). The final weight score for each source is obtained through weighted summation, forming a node source weight table.

[0128] Finally, an initial bottleneck node set is determined based on the node source weight table and the overflow propagation path tree. A source weight threshold is set, for example, selecting the top 10% of nodes by weight among all source nodes. These nodes are initially identified as initial bottleneck nodes. Further checks are performed; if a node with a weight below the threshold is a necessary "passage" on the path of multiple high-weight nodes, then that node is also added to the initial bottleneck node list. The final set of nodes selected constitutes the initial bottleneck node set that has a significant impact on subsequent large-scale water accumulation.

[0129] Preferably, step S4, which involves identifying indirect influence paths of the spatiotemporal risk evolution sequence, includes:

[0130] Import spatial distribution data of key urban facilities and combine it with spatiotemporal risk evolution sequence to conduct risk correlation analysis of key facilities and obtain facility risk impact mapping table;

[0131] Obtain basic urban road data and construct an urban traffic network topology map;

[0132] Based on the urban transportation network topology map and facility risk impact mapping table, baseline accessibility calculation is performed to obtain the facility accessibility baseline table;

[0133] Predicted waterlogging distribution data are extracted from the spatiotemporal risk evolution sequence, and the waterlogging obstacle network is corrected on the urban traffic network topology map to generate a waterlogging obstacle traffic network.

[0134] Dynamic accessibility changes are calculated based on the waterlogged obstacle traffic network and facility accessibility benchmark table to obtain the facility accessibility change matrix;

[0135] An accessibility attribution analysis was performed on the facility accessibility change matrix to obtain a table of the impact contribution of water accumulation points.

[0136] Based on the impact contribution table of water accumulation points, critical cut-off points of the path are identified, and a list of critical cut-off points is obtained.

[0137] The critical impact path map is formed by integrating the list of critical cut-off points, the facility accessibility change matrix, and the urban transportation network topology.

[0138] In one embodiment, spatial distribution data of key urban facilities, including geographical coordinates of hospitals, substations, and subway stations, is imported. The coordinates of these facilities are then spatially overlaid with the water accumulation range in the spatiotemporal risk evolution sequence. If the coordinates of a facility fall within the predicted water accumulation range, it is determined that it is directly affected. This process generates a facility risk impact mapping table, recording whether each facility is directly affected.

[0139] Then, the city's basic road data is acquired and transformed into a city traffic network topology graph with intersections as nodes and road segments as edges. Each edge is assigned a weight based on its length and design speed.

[0140] Next, baseline accessibility calculations are performed based on the urban transportation network topology map and the facility risk impact mapping table. For any two critical facilities (e.g., hospital A and substation B), Dijkstra's algorithm is used to calculate the shortest travel time under conditions of no water accumulation. The pairwise shortest travel times between all critical facilities are calculated and stored to form a facility accessibility baseline table.

[0141] Subsequently, predicted waterlogging distribution data are extracted from the spatiotemporal risk evolution sequence, and the urban traffic network topology is corrected for waterlogging obstruction network. At each time step, each road segment is checked for waterlogging. If the water depth of a road segment exceeds 0.3 meters, that road segment is temporarily "removed" from the network, meaning its traffic weight is set to infinity. This process generates a dynamically changing waterlogging obstruction traffic network over time.

[0142] Subsequently, dynamic accessibility changes are calculated based on the waterlogged traffic network and facility accessibility benchmark table, with the core objective of quantifying the indirect impact of traffic disruptions. At each time step, the shortest travel times between all key facilities are recalculated using the waterlogged traffic network. These new travel times are compared to the benchmark travel times to calculate an accessibility decay rate. For example, if the benchmark time from A to B is 10 minutes, and it becomes 30 minutes after the flooding, the accessibility decay rate is 200%. The accessibility decay rates between all facilities at different times are recorded to form a facility accessibility change matrix.

[0143] Then, an accessibility attribution analysis was performed on the facility accessibility change matrix. For each pair of facilities with significant accessibility degradation, the new shortest path after flooding was analyzed. By comparing the differences between the old and new paths, the key flooded road segments that caused the path changes were located. The degree of accessibility recovery after removing the flooded road segment was calculated, thereby quantifying the contribution of the flooded point to accessibility degradation and forming a flood point impact contribution table.

[0144] Next, critical cut-off points were identified based on the impact contribution table of water accumulation points. Water accumulation points that simultaneously met the following two conditions were selected: 1. They ranked among the top in terms of their contribution to the accessibility attenuation of multiple pairs of critical facilities; 2. The detour path length caused by them increased by more than 5 times the original path. These water accumulation points were identified as "critical cut-off points" because they caused systemic disruptions to the transportation network with a relatively small area of ​​water accumulation. All identified points were compiled to form a list of critical cut-off points.

[0145] Finally, the critical cut-off point list, facility accessibility change matrix, and urban transportation network topology map were integrated into an impact path network. On the urban transportation network topology map, all critical cut-off points were highlighted, and the original critical paths blocked by these cut-off points, as well as the forced detour routes, were drawn using lines of different colors. This thematic map, integrating cut-off points, blocked paths, and affected facilities, is the critical impact path map, and its core information table is shown in Table 2.

[0146] Table 2

[0147]

[0148] Preferably, step S4, which involves dual risk node identification and rating of the initial bottleneck node set and the key impact path diagram, includes:

[0149] High-risk impact nodes are screened from the critical impact path map to obtain a set of high-risk impact nodes;

[0150] A dual-standard node comprehensive evaluation is conducted on the initial bottleneck node set and the high-risk impact node set to form a node comprehensive scoring table;

[0151] Based on the node comprehensive scoring table and the risk node time series characteristic table, early warning priority allocation and timeliness analysis are performed to generate graded early warning data.

[0152] In one embodiment, high-risk impact nodes are screened from the critical impact path map. The critical impact path map is analyzed, and waterlogged nodes marked as "critical cutoff points," as well as waterlogged nodes that directly submerge critical facilities (such as hospitals and substations), are jointly screened. These nodes are categorized as high-risk impact nodes because they severely impact the urban functional network.

[0153] Then, a dual-standard node comprehensive evaluation is performed on the initial bottleneck node set and the high-risk impact node set. The core logic lies in simultaneously assessing risk from two dimensions: "cause" (the source of the disaster) and "effect" (the node causing serious consequences). A two-dimensional evaluation matrix is ​​created, with the horizontal axis representing "causal weight" (derived from the source weight of the initial bottleneck node) and the vertical axis representing "impact rating" (derived from the cutoff effect index or direct impact level of the high-risk impact node). All initial bottleneck nodes and high-risk impact nodes are mapped onto this matrix. The comprehensive score S of each node is calculated using a weighted summation formula, where the calculation process of the weighted summation formula is as follows:

[0154] ;

[0155] in The causal weights of the nodes, Rate the impact of the node. and For the preset weighting coefficients (e.g.) =0.4, =0.6). The overall score of all nodes is recorded to form a node overall score table.

[0156] Finally, based on the node comprehensive score table and the risk node time series characteristic table, early warning priority allocation and timeliness analysis are performed. First, all nodes are sorted from highest to lowest according to their scores in the node comprehensive score table. Then, combined with the "expected start overflow time" of each node recorded in the risk node time series characteristic table, the "time window" of the early warning (i.e., the difference between the current time and the overflow time) is calculated. Based on the comprehensive score and the time window, an early warning level is assigned to each node. For example, nodes in the top 10% of the comprehensive score with a time window of less than 1 hour are assigned the highest level (red) early warning; nodes with higher scores but a time window of more than 3 hours are assigned a level 3 (yellow) early warning. This process ultimately generates hierarchical early warning data containing node ID, location, comprehensive score, early warning level, and time window. This data is the direct basis for generating differentiated early warning instructions subsequently.

[0157] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0158] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A cross-domain data fusion early warning method for urban flooding, characterized in that, Includes the following steps: Step S1: Obtain raw, multi-source, heterogeneous data through the real-time data interfaces of various departments in the city; Spatial registration and rasterization of the original multi-source heterogeneous data are performed to obtain a city raster attribute table; based on the city raster attribute table, a city carrying capacity base is constructed to obtain a multi-dimensional static carrying capacity map. Step S2: Collect real-time urban status data streams, perform influence factor superposition analysis on the multi-dimensional static carrying capacity map, and obtain a comprehensive influence coefficient distribution map; Acquire and record the duration of historical drainage events to obtain an event duration table; Time gradient analysis is performed based on the event duration table to obtain the timeliness impact coefficient; urban drainage system data is obtained through real-time data interface, and coupling situation analysis of urban drainage system is performed based on the distribution map of timeliness impact coefficient and comprehensive impact coefficient to obtain system coupling situation index; drainage dynamic capacity correction is performed on the benchmark drainage capacity value using the system coupling situation index to obtain preliminary dynamic adjustment capacity value. Physical constraints are imposed on the initial dynamic adjustment capacity value, and spatial consistency optimization is performed to obtain the dynamic adjustment drainage capacity value. The efficiency matrix of dynamic adjustment drainage capacity value and multidimensional static bearing capacity spectrum is spatiotemporally integrated to obtain instantaneous drainage efficiency matrix. Step S3: Obtain rainfall forecast data and construct effective runoff distribution data by combining instantaneous drainage efficiency matrix; use effective runoff distribution data to identify pipeline drainage overflow nodes in the multidimensional static bearing capacity map and obtain overflow node feature set; The spatiotemporal risk evolution sequence is obtained by performing a temporal deduction of the surface water accumulation state on the feature set of overflow nodes; Step S4: Identify risk features in the spatiotemporal risk evolution sequence to obtain a temporal feature table of risk nodes; Source tracing analysis and initial bottleneck identification are performed on the temporal characteristic table of risk nodes to obtain an initial bottleneck node set; indirect impact path identification is performed on the spatiotemporal risk evolution sequence to obtain a key impact path map; dual risk node identification and rating are performed on the initial bottleneck node set and the key impact path map to obtain graded early warning data; multi-channel early warning information is distributed from the graded early warning data to form multi-channel early warning instructions.

2. The cross-domain data fusion early warning method for urban flooding according to claim 1, characterized in that, Step S1 includes: Acquire raw, multi-source heterogeneous data, including DEM data, impermeable layer distribution data, drainage network data, soil type data, and historical rainfall records; Spatial feature rasterization and registration are performed on the original multi-source heterogeneous data to obtain a basic spatial feature layer set; The vulnerability of the drainage network is quantified and assigned to the basic spatial element layer set using drainage network data, resulting in a city raster attribute table. Collect cumulative rainfall data over the past 72 hours and combine it with raw multi-source heterogeneous data to calculate the current soil saturation status. Multidimensional static attribute overlays are performed on the basic spatial element layer set, urban raster attribute table, and current soil saturation status data to form a multidimensional static bearing capacity map.

3. The cross-domain data fusion early warning method for urban flooding according to claim 1, characterized in that, Step S2 involves collecting real-time urban status data streams and performing influencing factor overlay analysis on the multidimensional static bearing capacity map, including: Based on the multidimensional static bearing capacity map, multi-source dynamic data are collected in real time to form a real-time urban status data stream; Dynamic spatial location of events is performed on the real-time status data stream of the city to obtain a spatial location event distribution map; Based on the spatial location event distribution map, the impact factors are quantified and converted to obtain the urban event impact factor table; By using the urban event impact factor table to conduct a multi-layer impact superposition analysis on the multi-dimensional static bearing capacity map, a comprehensive impact coefficient distribution map is obtained.

4. The cross-domain data fusion early warning method for urban flooding according to claim 3, characterized in that, The quantitative transformation of influencing factors based on the spatial location event distribution map includes: Based on the spatial location event distribution map, events are classified and their impacts are analyzed to obtain a mapping table of event influencing factors. The historical cumulative effect was analyzed using the event influencing factor mapping table, resulting in a time cumulative influence coefficient table. The baseline impact coefficient of urban waterlogging was calibrated based on the time cumulative impact coefficient table, and the quantitative data of the baseline impact coefficient were obtained. Based on the baseline impact coefficient quantification data, a dynamic event hierarchy mapping is performed on the spatial location event distribution map to obtain the urban event impact factor table.

5. The cross-domain data fusion early warning method for urban flooding according to claim 1, characterized in that, Step S3 includes: Acquire rainfall forecast data from meteorological departments; use the instantaneous drainage efficiency matrix to perform spatiotemporal decomposition of rainfall forecast data from meteorological departments to obtain rasterized rainfall forecast sequences; Inflow calculation and allocation are performed on the gridded rainfall forecast sequence to obtain effective runoff distribution data; By using effective runoff distribution data to trace surface runoff paths in a multidimensional static bearing capacity map, a surface runoff path network is obtained. By combining the surface runoff path network and the instantaneous drainage efficiency matrix, the inflow allocation of the pipe network is calculated, and the inflow allocation table of the pipe network nodes is obtained. Based on the inflow allocation table of the pipeline node, the hydraulic transmission of the pipeline network is simulated to obtain the hydraulic state time series table of the pipeline network. Overflow node identification and analysis were performed on the hydraulic state time series table of the pipeline network and the inflow distribution table of the pipeline network nodes to obtain the overflow node feature set; The overflow node feature set is redistributed to the surface runoff path network for surface water formation and diffusion processing to obtain a dynamic distribution map of water accumulation. The spatiotemporal risk sequence is integrated from the dynamic distribution map of water accumulation, the feature set of overflow nodes, and the time series table of hydraulic status of the pipe network to form a spatiotemporal risk evolution sequence.

6. The cross-domain data fusion early warning method for urban flooding according to claim 5, characterized in that, The process of reassigning the overflow node feature set to the surface runoff path network for surface water formation and diffusion includes: The overflow node feature set is used to locate and quantify the water source points, resulting in a water source point distribution map. The water accumulation and depression filling calculation was performed using a multidimensional static bearing capacity map to obtain data on the depression filling status. A simplified hydrodynamic diffusion simulation was performed based on the distribution map of water sources and the data on the filling status of depressions to obtain the temporal variation data of water volume in depressions. By combining the temporal variation data of water volume in depressions and the multidimensional static bearing capacity map, the water depth is calculated in a gridded manner to form gridded water depth distribution data. The data on water depth distribution in the raster is used to identify contiguous water accumulation areas and extract their boundaries. Then, the data is dynamically integrated over time to obtain a dynamic water accumulation distribution map.

7. The cross-domain data fusion early warning method for urban flooding according to claim 1, characterized in that, Step S4 involves source tracing analysis and initial bottleneck identification of the risk node time series characteristic table, including: Based on the first overflow time of the nodes in the risk node time series characteristic table, construct a time series overflow connection graph; Overflow propagation path identification is performed on the time-series overflow connection graph to obtain the overflow propagation path tree; Trigger chain identification and quantification are performed on the overflow propagation path tree to obtain a node trigger chain feature table; Based on the node trigger chain feature table and the spatiotemporal risk evolution sequence, the water volume contribution is tracked to obtain the overflow contribution matrix; The overflow source is located based on the overflow contribution matrix, and the source weight is calculated to form a node source weight table; The initial bottleneck node set is determined based on the node source weight table and the overflow propagation path tree.

8. The cross-domain data fusion early warning method for urban flooding according to claim 1, characterized in that, Step S4, which involves identifying indirect impact paths of the spatiotemporal risk evolution sequence, includes: Import spatial distribution data of key urban facilities and combine it with spatiotemporal risk evolution sequence to conduct risk correlation analysis of key facilities and obtain facility risk impact mapping table; Obtain basic urban road data and construct an urban traffic network topology map; Based on the urban transportation network topology map and facility risk impact mapping table, baseline accessibility calculation is performed to obtain the facility accessibility baseline table; Predicted waterlogging distribution data are extracted from the spatiotemporal risk evolution sequence, and the waterlogging obstacle network is corrected on the urban traffic network topology map to generate a waterlogging obstacle traffic network. Dynamic accessibility changes are calculated based on the waterlogged obstacle traffic network and facility accessibility benchmark table to obtain the facility accessibility change matrix; An accessibility attribution analysis was performed on the facility accessibility change matrix to obtain a table of the impact contribution of water accumulation points. Based on the impact contribution table of water accumulation points, critical cut-off points of the path are identified, and a list of critical cut-off points is obtained. The critical impact path map is formed by integrating the list of critical cut-off points, the facility accessibility change matrix, and the urban transportation network topology.

9. The cross-domain data fusion early warning method for urban flooding according to claim 1, characterized in that, Step S4 involves dual risk node identification and rating for both the initial bottleneck node set and the key impact path graph, including: High-risk impact nodes are screened from the critical impact path map to obtain a set of high-risk impact nodes; A dual-standard node comprehensive evaluation is conducted on the initial bottleneck node set and the high-risk impact node set to form a node comprehensive scoring table; Based on the node comprehensive scoring table and the risk node time series characteristic table, early warning priority allocation and timeliness analysis are performed to generate graded early warning data.

Images