A method for calculating critical rainfall of important facilities based on digital surface model

Abstract

The invention discloses a critical rainfall calculation method for important facilities based on a digital surface model, comprising S1: collecting high-precision point cloud mapping data, current land use classification data and other data of the city where the important facilities are located, and collecting flood depth data of key locations; S2: correcting various collected data; S3: fusing various corrected data; S4: dynamically dividing sub-catchment areas into three levels based on the fused terrain data; S5: constructing a multi-city stormwater runoff model based on Bayesian averaging; S6: inputting rainfall data under different return periods into the multi-city stormwater runoff model based on Bayesian averaging constructed in step S5, predicting flooding results under rainfall conditions with different return periods, and calculating the critical rainfall for warning of heavy rain disasters for each important facility through the flooding results; the invention overcomes the simulation error caused by the structural uncertainty of a single city stormwater runoff model, and improves the accuracy of the heavy rain disaster warning.

Description

Technical Field

[0001] The invention relates to the technical field of urban flood control and drainage, and in particular to a method for calculating critical rainfall for early warning of important facilities based on a high-precision digital surface model. Background Art

[0002] In recent years, affected by climate change and human activities, many cities at home and abroad are facing varying degrees of waterlogging risks. Urban key facilities generally refer to urban protection objects such as hospitals and high-speed rail stations that affect people's livelihood. Urban waterlogging will occur when the critical rainfall is exceeded. Accurate calculation of critical rainfall for early warning of urban key facilities is helpful for early response and prevention and reducing the negative impact of heavy rain on urban social economy.

[0003] Urban stormwater runoff model is the main tool for evaluating and calculating the critical rainfall for urban rainstorm disasters. It mainly determines the critical rainfall threshold by simulating the waterlogging disaster situation under different rainfall events. Common urban stormwater runoff models include SWMM stormwater flood management model, STORM data flow model, MOUSE urban stormwater runoff model, etc. Accurate calculation of urban warning critical rainfall depends on the accurate description and extraction of the main characteristic information of the current catchment area (such as the catchment area range, topography, land type attributes and underground pipe network, etc.), and on the robustness and accuracy of the simulation of urban stormwater runoff model.

[0004] On the one hand, historical studies usually believe that the catchment area of ​​a certain outlet is fixed under different rainfall conditions. Conventional static catchment area drawing methods include: (1) manual delineation; (2) catchment area division method based on drainage pipe network; (3) catchment area division method based on terrain data. However, in recent years, with the continuous development and progress of measurement technology, domestic and foreign scholars have gradually begun to introduce high-precision point cloud data and apply it to the accurate acquisition of urban catchment area characteristics. Some studies have found that the catchment area under different rainfall characteristics is different. If the dynamic changes of the catchment area boundary are not considered when simulating urban stormwater runoff, it may lead to systematic deviations in the simulated peak flow.

[0005] On the other hand, most of the existing literature on the calculation of critical rainfall for urban warning is limited to the use of a single urban stormwater runoff model. The use of a single urban stormwater runoff model affects the robustness of the simulation results due to the uncertainty caused by the structural defects of the model itself. Summary of the invention

[0006] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a method for calculating the critical rainfall for early warning of important facilities based on a high-precision digital surface model, specifically a method for calculating the critical rainfall for important urban facilities, which simultaneously considers the dynamic change characteristics of the catchment area boundaries under different rainfall conditions coupled with multi-city stormwater runoff models.

[0007] To achieve the above objectives, the technical solution of the present invention is:

[0008] A method for calculating critical rainfall for warning of important facilities based on a high-precision digital surface model comprises the following steps:

[0009] S1: Data collection: collect high-precision point cloud mapping data of cities where important facilities are located, current land use classification data, topographic data of buildings, roads and canals, drainage network data, historical rainfall data with different return periods and flooding depth results under corresponding rainstorm events;

[0010] S2: Correct the various data collected in step S1;

[0011] S3: Fusing various data corrected in step S2, specifically adding the corrected land use classification data, building, road and river channel terrain data, and drainage network data to the corrected point cloud mapping data, and correcting the boundary data of the fused data. Correcting the boundary data refers to correcting the coordinates and elevation of the boundary data;

[0012] S4: Based on the point cloud mapping data, land use classification data, and terrain data of buildings, roads, and canals that have been fused in step S3, the sub-catchment areas are dynamically divided into three levels, specifically, the coordinates of the boundary points of the catchment areas are determined on the fused point cloud mapping data;

[0013] S5: constructing a multi-city stormwater runoff model based on Bayesian averaging, wherein the multi-city stormwater runoff model based on Bayesian averaging is used for the calculation of runoff generation and runoff and waterlogging simulation of each sub-catchment area obtained in step 4 of S4;

[0014] S6: Input rainfall data under different return periods into the multi-city stormwater runoff model based on Bayesian averaging constructed in step S5, repeat the calculation of watershed runoff production and simulation calculation of waterlogging depth in step S5, and predict the flooding results under rainfall conditions with different return periods through the multi-city stormwater runoff model based on Bayesian averaging. The flooding results include flooding depth and flooding duration. When the flooding depth and flooding duration predicted based on the input rainfall conditions are the flooding depth and flooding duration that will cause waterlogging in the existing specifications, the rainfall involved in the input rainfall conditions is the critical rainfall for warning of heavy rain disasters for important facilities.

[0015] Furthermore, the step S1 specifically includes the following steps:

[0016] S1.1: Use LiDAR technology to collect high-precision point cloud mapping data of cities where important facilities are located, and use LiGeoreference solution software or LiDAR360 software to analyze the data collected by LiDAR technology to obtain high-precision point cloud mapping data;

[0017] S1.2: Obtain the current land use classification data through the latest land use classification results of the important setting area;

[0018] S1.3: Obtain the overall outline and elevation data of the building perimeter in the area where the important facilities are located, obtain the boundary data and elevation data of municipal roads and community roads; obtain the cross-sectional and longitudinal morphology and direction of the river channel, the boundary position of the river channel, and the elevation data of the levee crest. The boundary data or boundary position is reflected by coordinate data;

[0019] S1.4: Obtain the location, elevation, burial depth, pipe diameter, material, and direction data of the drainage network, as well as the damage location, damage level, siltation type and level data of the existing drainage pipelines;

[0020] S1.5: Obtain historical data containing at least one complete rainfall event record, as well as the measured flooding duration and flooding depth near the location of important facilities during that rainfall event.

[0021] Furthermore, the step S2 specifically includes the following steps:

[0022] S2.1: The original point cloud mapping data is corrected by using the Gaussian filtering method. The noise of the original point cloud mapping data is removed by the Gaussian filtering method. The Gaussian filtering function H is defined as:

[0023]

[0024] In the formula, the normal distribution function f(x) is expressed as:

[0025]

[0026] In formulas (1) and (2), e is the base of the natural logarithm and is taken as 2.71828; x is the scanning window width; σ is the standard deviation of x, and σ is the variable value calculated by the normal function as x changes;

[0027] The distance value D of the current filtering scanning point is obtained by formula (1) and formula (2): fil The relationship between each data point, i.e., the Gaussian filtering result, is shown in formula (3), and the scanning window width x is shown in formula (4):

[0028] D fil =H×D(x i ) (3)

[0029] x=lin space(-d,d,2·d+1) (4)

[0030] In the formula, x i is the position of data point i scanned by the radar after excluding the maximum value position and invalid point position data in the window; D(x i ) is the distance value corresponding to each data point; d is the radius of the window;

[0031] S2.2: Verify drainage network data through pipeline closed-circuit television detection CCTV, pipeline periscope detection QV, sonar detection, and manual observation methods, including verification of drainage network longitudinal section elevation and direction, connection status, and pipe diameter size data.

[0032] S2.3: Verify the rationality of the current land use classification data set through field survey of land use classification data;

[0033] S2.4: The rationality of the topographic data of buildings, roads and canals shall be verified by field measurement. The elevations of buildings, roads, canals above the water surface and the top of the dikes shall be verified by total station and digital photography; the underwater cross-sectional data of canals shall be measured and verified by total station combined with flower pole and steel ruler or underwater robot;

[0034] S2.5: The measured maximum inundation depth and inundation duration of rainstorm floods at the locations of important facilities are analyzed and verified using water level monitoring methods such as multi-point water level sensors. The required rainfall data for the same event are obtained from nearby meteorological stations in the area.

[0035] Furthermore, the step S3 specifically includes the following steps:

[0036] S3.1: Based on the area where important facilities are located, add the terrain data of buildings, roads, and canals to the point cloud mapping data, and correct the added and fused terrain data. Specifically, add the horizontal and vertical coordinates and elevation data of underground pipe networks and above-ground buildings to the point cloud data. Correcting the fused terrain data means adjusting the coordinates of the fused boundary data according to the actual situation.

[0037] S3.2: When the urban drainage pipe network and rainwater grate data are added to the point cloud mapping data, the elevation data of the urban drainage pipe network and rainwater grate are corrected so that the ground elevation data of the urban drainage pipe network and rainwater grate are lower than the ground elevation data of the location of the urban drainage pipe network and rainwater grate;

[0038] S3.3: Fusion of the land classification data with other data, including adding land classification data, pipe network coordinate data, and measured local terrain elevation data to point cloud mapping;

[0039] S3.4: After fusing the data collected through different channels, correct the boundaries of the fused data, including correcting the boundary coordinate data of adjacent buildings through field surveys and detection verification, so that the boundary data of adjacent buildings can be connected.

[0040] Furthermore, the step S4 specifically includes the following steps:

[0041] S4.1: Take the main urban river with actual drainage function as the first-level sub-catchment area, and mark the boundary coordinates of the first-level sub-catchment area on the fused point cloud mapping data;

[0042] S4.2: Based on the division of the primary catchment area, the main and secondary roads with drainage pipes on both sides of the main urban river and the main canals with drainage functions are regarded as secondary sub-catchments, and the boundary coordinates of the secondary sub-catchments are marked on the fused point cloud mapping data;

[0043] S4.3: Based on the demarcation boundaries of the primary and secondary sub-catchments, the Hydrology tool in ArcToolBox in ArcGIS software is used to demarcate the tertiary sub-catchments SC based on elevation and flow direction; and the boundary coordinate data of the tertiary sub-catchments are marked in the fused point cloud mapping data using ArcGIS software;

[0044] S4.3.1: If the vertical pipe of a single building is connected to a drainage pipe, use ArcGIS software to adjust the watershed division to ensure that the rainwater of the building is connected to a nearby watershed area. Specifically, the coordinate data of the nearby watershed area is added to the point cloud mapping data of the single building;

[0045] S4.3.2: If no drainage pipes are laid near the building riser, the rainwater from a single building will be collected in the direction of the largest slope of the terrain adjacent to the riser and converge to the lowest point; specifically, the coordinate data of the third-level sub-catchment area with the largest slope of the adjacent terrain will be added to the point cloud mapping data of the single building.

[0046] Furthermore, the step S5 specifically includes the following steps:

[0047] S5.1: Construct an urban stormwater runoff model based on the Bayesian model average. Let y be the multi-model simulation variable, i.e., simulated runoff. The posterior probability distribution p(y|m k ,Q obs ,ξ) is:

[0048]

[0049] In the formula, Q obs =[y1,y2,...,y T] is the measured runoff process required for the calibration model, the model input is ξ, T is the measured sequence length, M = [m1, ..., m k ,...,m K ] represents the model space composed of k selected urban stormwater runoff models, and K is the total number of selected urban stormwater runoff models. k (y|m k ,Q obs ,ξ) is the given observation data Q obs , input data ξ and model simulation value m k The posterior distribution of the simulated variable y under the condition of k (y|m k ,Q obs ,ξ) is the posterior probability distribution of a single model k under given input conditions;

[0050] S5.2: Let ω k is the model weight coefficient, then The mean of the posterior probability distribution of the Bayesian average simulated variable E(y|M,Q obs ,ξ) and variance D(y|M,Q obs ,ξ) are shown in formulas (6) and (7) respectively:

[0051]

[0052]

[0053] In the formula, σ k 2 For a given observation data Q obs and the variance of the simulated variables under the conditions of the model input data ξ.

[0054] S5.3: Select the maximum NSE Nash efficiency coefficient as the objective function, calculate and solve the simulated outflow process line of each sub-catchment based on the measured rainfall and other data, and determine the Bayesian average probability distribution parameter ω based on the sub-catchment runoff data. k and σ k 2 As well as the internal parameters of the rainwater runoff model of each city, RMSE root mean square error and CC linear correlation coefficient are used as evaluation indicators for evaluating the simulation results and the measured outflow process:

[0055]

[0056]

[0057]

[0058] Where: Q sim,irepresents the simulated flow of the subcatchment in the i-th period, Q obs,i represents the measured flow rate of the subcatchment in the i-th period, Q obs represents the historical measured flow sequence of the subcatchment area; Q sim represents the simulated flow series of the subcatchment;

[0059] S5.4: The simulated outflow process of each sub-catchment obtained by solving is input into Mike flood software together with the corrected and fused point cloud mapping data, current land use classification data, topographic data of buildings, roads and canals, catchment area boundary data, and pipe network vector data. The Mike flood software is used to simulate and analyze the waterlogging inundation situation of each point on the road network in the area where important facilities are located, and compared and analyzed with the waterlogging results measured in historical data and the model parameter settings are adjusted; S5.4 is specifically divided into the following steps:

[0060] S5.4.1: When the outlet flow of each subcatchment area within the discharge standard of the canal system is connected to the drainage system through pipes and stormwater grates, the flow, confluence and overflow of water bodies in the drainage system shall be calculated by solving the one-dimensional hydrodynamic equations;

[0061] S5.4.2: When the rainfall is too heavy and the required drainage volume exceeds the discharge standard of the canal system, the surface runoff process of the road that exceeds the discharge standard of the canal system is calculated by solving the two-dimensional hydrodynamic equation, simulating the maximum flooding depth and flooding duration of each important facility point, and drawing the boundary coordinate data of the catchment area under the rainfall conditions of this event;

[0062] S5.5 Compare the two-dimensional road confluence path and simulated flow direction obtained by simulation with the boundaries of each sub-catchment area in step 4, and verify the rationality of the catchment area division boundary. If the two-dimensional water flow direction does not correspond to the catchment area boundary, adjust the catchment area boundary coordinate data, re-calibrate the parameters of each part, and use the model parameters that are closest to the measured waterlogging results when the waterlogging conditions at each point on the road network in the area where important facilities are located are simulated and analyzed using Mike flood software. The optimal model parameters are used.

[0063] Furthermore, the step S6 specifically includes the following steps:

[0064] S6.1: Based on the Bayesian weight parameters and model parameter results after simulation and calibration in step S5 and the watershed division range, input different rainfall conditions, recalculate and predict the runoff generation and confluence process of each sub-watershed under different return period rainfall conditions, and use Mike flood software to calculate the two-dimensional road slope flow process. After simulating and calculating the confluence flow direction and confluence process, compare and analyze the consistency of the simulated water flow direction with the watershed division boundary;

[0065] S6.2: Dynamically adjust the boundary coordinate data of the third-level sub-catchment area according to the actual runoff generation and runoff process and the two-dimensional water flow of the road;

[0066] S6.3: Based on sub-steps 6.1 and 6.2, the flooding depth and duration of each key location, and the actual confluence path simulated under each rainfall event, dynamically adjust the coordinates of the catchment area boundary on the fused point cloud mapping data, re-enter the rainfall conditions, and determine the flooding depth and duration of each point under each simulated condition based on the Bayesian average multi-city stormwater runoff model.

[0067] S6.4: Based on the actual elevation of each point, as well as the allowable flooding depth and duration, evaluate and determine the critical rainfall for rainstorm disaster warning at each point. When the actual rainfall in a certain period reaches the critical rainfall, a waterlogging warning for important urban facilities will be issued.

[0068] The beneficial effects of the present invention are:

[0069] Compared with the prior art, the present invention has the following beneficial effects:

[0070] 1. The existing evaluation and calculation of critical rainfall for urban rainstorm disasters often adopts a single SWMM storm flood management model, STORM data flow model or MOUSE urban storm runoff model. The present invention proposes a multi-city storm runoff model based on Bayesian average. The multi-city storm runoff model based on Bayesian average couples multiple city storm runoff models, overcomes the simulation error caused by the structural uncertainty of a single city storm runoff model, and reduces the error of catchment area runoff simulation by coupling the multi-model simulation method and applying it to the urban dynamic storm runoff simulation, thereby reducing the simulation forecast error and making the urban waterlogging warning structure more accurate.

[0071] 2. When simulating urban stormwater runoff, if the dynamic changes of the catchment area boundaries are not taken into account, it may lead to systematic deviations in the simulated peak flow. The present invention proposes a division method based on high-precision point cloud data and taking into account the dynamic changes of the catchment area boundaries under different rainfall characteristics, thereby improving the accuracy of rainstorm disaster warning.

[0072] 3. The present invention is based on a high-precision digital surface model and combines multiple data sources and methods. It has strong operability, can accurately calculate the critical rainfall threshold of important facility areas, and realize the early warning function. This method has broad application prospects in the protection of important urban facilities and urban flood control work under the background of frequent extreme climate conditions in the future. It makes full use of various cutting-edge data to improve the accuracy and timeliness of flood control warnings, ensure the safe operation of important facilities and the sustainable development of cities. The present invention is scientific and reasonable and fits the actual engineering. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1It is a specific flow chart of the method of the present invention.

[0074] Figure 2 It is a schematic diagram of the Gaussian filtering principle of the present invention.

[0075] Figure 2 (a) is a schematic diagram of the Gaussian filter template of the present invention.

[0076] Figure 2 (b) is a schematic diagram of the Gaussian distribution probability density function of the present invention.

[0077] Figure 3 It is a schematic diagram of the step-by-step division of the catchment area of ​​the present invention.

[0078] Figure 4 It is a schematic diagram of the change of the catchment area under different precipitation conditions of the present invention.

[0079] Figure 4 (a) is a schematic diagram of the present invention in which the catchment area of ​​outlet C is only the catchment area of ​​B2 when the actual rainfall does not exceed the critical rainfall condition of the low-lying area.

[0080] Figure 4 (b) is a schematic diagram of the catchment area of ​​outlet C when the actual rainfall P exceeds the critical rainfall condition of the low-lying area in the present invention, which is the sum of the catchment areas of B1 and B2.

[0081] Figure 5 It is a schematic diagram of determining water depth inundation and early warning critical rainfall according to the present invention.

[0082] Figure 5 (a) is a schematic diagram of the maximum flooding depth and rainfall return period of the present invention.

[0083] Figure 5 (b) is a schematic diagram of the maximum flooding depth and rainfall of the present invention. DETAILED DESCRIPTION

[0084] In order to make the purpose, technical solutions and advantages of the invention clearer, the present invention is further described below with reference to the accompanying drawings.

[0085] The present invention first collects and obtains multiple data sources such as high-precision point cloud data, current land use classification data, drainage network data, etc., and then uses multiple methods to correct data errors. For example, the Gaussian filtering method is used to remove the noise of the original point cloud data. Then, based on the corrected point cloud data, the important terrain elements and drainage facility data such as coupled buildings, roads, underground pipelines, and current land use classification data are superimposed. Then, by integrating the three-level catchment area dynamic division method of rivers and pipes, natural terrain, land types and flow directions, the fine division of the catchment area is achieved; on the basis of the fine division of the catchment area, a comprehensive simulation method of multi-city stormwater runoff model based on Bayesian model averaging is further constructed, and the SCEM-UA algorithm is used to solve and deduce the weight coefficients and internal parameters of different models, and the outflow process lines of each catchment area are calculated; then, the Mikeflood software is used to simulate the one-dimensional pipeline flow and the two-dimensional road network slope flow, and the results such as the flooding depth and the flow direction are simulated. The rationality of the watershed division is verified by reverse checking the direction of water flow, and the boundary coordinates of the watershed are dynamically adjusted on the fused point cloud mapping data to dynamically divide the boundaries of the watershed. Finally, the weight coefficients and internal parameters of the model obtained by simulation calibration are combined with the dynamic adjustment of the watershed boundary to predict the flooding depth of each target point under different rainfall conditions, and the critical rainfall threshold of each point is determined. For details, see Figure 1 .

[0086] In the present invention, historical meteorological data, including flooding data, flooding duration data, etc., can be obtained through meteorological stations, and the model can be trained with the historical data; then a certain rainfall data is collected, and the flooding situation of the certain rainfall is predicted with the trained model.

[0087] The technical solution of the present invention is further specifically described below through embodiments and in combination with the accompanying drawings: Step one, collecting high-precision point cloud mapping data of the city, collecting current land use classification data, important terrain data such as buildings, roads and canals, drainage network data, collecting historical rainfall data under different recurrence periods and flooding results under corresponding rainstorm events, the flooding results including flooding depth and flooding duration, etc.

[0088] The step 1 further comprises the following sub-steps:

[0089] (1.1) The high-precision point cloud mapping data required to be collected is obtained through laser radar and other technologies, combined with dedicated solution software such as LiGeoreference solution software and LiDAR360 software to analyze the data. The accuracy of the mapping results should not exceed 2m.

[0090] In step 1.1, a range is demarcated around the key facilities, and high-precision point cloud data within the demarcated range is collected. The demarcated range is slightly larger than the range of the catchment area determined by the key facilities. The high-precision point cloud data is the surface data within the demarcated range, that is, the high-precision terrain data within the demarcated range. The high-precision terrain data includes point cloud data of surface terrain such as house elevation data and river surface data; the accuracy of surveying and mapping results should not exceed 2m, which means that the edge width of a single grid of point cloud data does not exceed 2m.

[0091] The point cloud data used has high accuracy and fast processing efficiency, and is currently widely used in surveying and mapping, agriculture and other fields. There are currently relatively mature technical means for the acquisition, solution and modeling of point cloud data.

[0092] (1.2) The required current land use classification data shall comply with the relevant requirements of the “Current Land Use Classification” (GB / T21010-2017). The conventional land use data source within the field can be used: ESRI’s global 10-meter resolution land cover data or the land type map layer data in the latest land survey data of the project area.

[0093] When there are multiple data sources for the land use data used, the update time and data accuracy of these data sets should be compared and analyzed, and the data source with the latest sampling time and the highest accuracy should be selected. There are relatively mature and common classification methods for land use data. In general, the project area planning bureau and other relevant departments have relatively high-accuracy current land use classification data, which should be given priority.

[0094] (1.3) The required building terrain data shall include the overall outline and elevation data of the building perimeter; the required road terrain data shall include the boundaries and elevation data of municipal roads and community roads; the required canal data shall include the cross-sectional and longitudinal shapes and directions of the canal, the location of the river boundary, and the elevation of the embankment top. The required important terrain data such as buildings, roads, and canals shall be obtained through actual measurement.

[0095] These local terrain data are mainly used to make up for the deficiencies and omissions of point cloud data. The sampling and acquisition of these data and the analysis and extraction using ArcGIS and other software are routine means for the owner.

[0096] (1.4) The drainage network data that needs to be collected should include the location, elevation, burial depth, pipe diameter, material, direction, etc. of the drainage network, as well as the damaged location, damage level, siltation type and level of the existing drainage pipelines. These data are obtained through geophysical survey.

[0097] The types of drainage network data used are all conventional data types in the industry and can be obtained using existing technical means such as pipeline closed-circuit television detection CCTV, pipeline periscope detection QV, etc. The data acquisition technology is mature.

[0098] (1.5) The measured rainstorm and flood processes to be collected should include at least one complete rainfall event record, specifically minute-level rainfall records and the measured flooding process line near the target location to be predicted under that event, including flooding duration and flooding depth changes.

[0099] Step 2: Use multiple methods to calibrate the various data collected in step 1.

[0100] The step 2 further includes the following sub-steps:

[0101] (2.1) Use Gaussian filtering to remove noise from raw point cloud data. Since raw point cloud data may be affected by noise during acquisition, such as randomly distributed noise points on the edge of an object, filtering the raw point cloud data is an important technical means to obtain high-precision, low-error point cloud data.

[0102] As attached Figure 2 As shown in the figure, a Gaussian filter function is introduced to remove the environmental noise in the original point data. The essence of Gaussian filtering is a linear smoothing filter with weights determined by normal distribution. It can effectively suppress noise and smooth data. Specifically, it is a process of weighted averaging each point cloud data scanned by the lidar and then performing convolution.

[0103] Define the Gaussian filter function H as:

[0104]

[0105] In the formula, the normal distribution function f(x) is expressed as

[0106]

[0107] In formulas (1) and (2), e is the base of the natural logarithm and is taken as 2.71828; x is the scanning window width; σ is the standard deviation of x, and σ is the variable value calculated by the normal function as x changes.

[0108] The distance value D of the current filtering scanning point is obtained by formulas (1) and (2): fil The relationship between each data point, i.e., the Gaussian filtering result, is shown in formula (3), and the scanning window width x is shown in formula (4):

[0109] D fil =H×D(x i )(3)

[0110] x=lin space(-d,d,2·d+1) (4)

[0111] In the formula, x i is the position of data point i scanned by the radar after excluding the maximum value position and invalid point position data in the window; D(x i ) is the distance value corresponding to each data point; d is the radius of the scanning window. The specific implementation of Gaussian filtering is to perform a convolution operation on the window center point and the left and right data scanned by the radar with a Gaussian template. In the present invention, a 1×3 Gaussian function is selected.

[0112] It should be noted that the use of Gaussian filtering to remove noise points in point cloud data is a commonly used technical means in the industry.

[0113] (2.2) Use pipeline closed-circuit television (CCTV), pipeline periscope (QV), sonar detection and manual observation to verify the rationality of drainage network data. Focus on verifying the elevation and direction of the drainage network longitudinal section, connection status, pipe diameter, etc.

[0114] CCTV, QV, sonar and manual observation methods are currently the most commonly used means of monitoring the condition of drainage pipes. Among them, the CCTV detection method is one of the most effective and advanced pipeline maintenance and troubleshooting methods, and is widely used in the field of municipal engineering.

[0115] (2.3) The required current land use classification data shall be verified by field surveys and other data sources to verify the rationality of the current land use classification. The land use classification shall strictly comply with the requirements of the "Current Land Use Classification" (GB / T 21010-2017).

[0116] (2.4) The necessary important terrain data such as buildings, roads and canals shall be verified by actual measurement to ensure the rationality of the data. The elevations of buildings, roads, canals above the water surface and the top of the dikes shall be verified by total station, GPS RTK measurement technology, digital photography and other methods. The underwater cross-sections of canals shall be measured and verified by total station combined with flower poles and steel rulers or underwater robots.

[0117] The use of total stations, GPS RTK, digital photography and other methods to survey and map terrain information such as ground elevation above water is a conventional technical means in the industry; similarly, the use of total stations combined with flower poles and steel rulers or underwater robots is also a commonly used and relatively mature technical means for underwater terrain surveying and mapping.

[0118] (2.5) The maximum inundation depth and inundation duration of the actual rainstorm flood at the required target points shall be analyzed and verified by using water level monitoring methods such as multi-point water level sensors. The required rainfall data of the same event shall be obtained from the nearby meteorological stations in the area, and the time step of rainfall data monitoring shall not exceed 5 minutes.

[0119] Step 3: Overlay, merge and correct various data. Based on the corrected point cloud data, the terrain data of important terrain elements such as buildings and roads, drainage facilities data such as drainage pipe networks, and current land use classification data are fused and superimposed after verification. After data fusion, the boundary data is corrected to ensure the rationality of the connection of various data results; the correction of boundary data refers to the correction of the coordinates and elevation data of each building boundary through verification methods such as manual survey.

[0120] The step 3 further comprises the following sub-steps:

[0121] (3.1) Based on the area scope, add terrain data such as buildings, roads, and canals on the basis of the corrected point cloud data, and correct the added terrain data. Focus on narrow-side water-blocking buildings such as solid walls and river embankments, special multi-layered features such as overpasses, viaducts, pedestrian overpasses, and sunken bridges. Correct the local terrain according to their actual characteristics to ensure the rationality of the water flow direction. The embankments and walls are added to the point cloud data in an actual measurement method. There are no embankments, and the water flows everywhere, with embankment constraints. Among them, the top elevation, plane position, and direction of solid walls and river embankments are drawn and corrected in the topographic map according to the actual measurement results; for features such as overpasses, viaducts, pedestrian overpasses, and sunken bridges, only their underlying elevation information is retained and superimposed on the topographic map.

[0122] Since high-precision point cloud data usually have large errors when surveying multi-layer terrain features such as overpasses, and narrow-edge features such as riverbanks or walls, the actual measurement method in step 2 is used to obtain information on these special terrain features and superimpose them into the point cloud data.

[0123] The point cloud data only has surface data. In the above step 3.1, specific data such as underground pipeline data, ground wall data, levee data, overpass data, house data, etc. are added to the point cloud data. Specifically, the horizontal and vertical coordinates and elevation data of the underground pipeline network, ground wall data, levee data, overpass data are added to the point cloud data.

[0124] The pipe network data specifically includes the coordinate data, depth, length and other basic pipe network data of the pipe network; the wall data specifically includes the coordinate data, height, length, width and other basic wall data of the wall; the embankment data specifically includes the coordinate data, height, length, width and other basic embankment data of the embankment; the house data specifically includes the coordinate data, height, length, width and other basic house data of the house.

[0125] Regarding the step 3.1, correcting the added terrain data refers to correcting the coordinate data of each building boundary by manual survey or other means.

[0126] (3.2) The ground elevation of the urban drainage network and rainwater grates should be fully integrated with the point cloud data and other ground feature elevation data to ensure the full connection between the drainage network data and the terrain data, and to ensure the rationality of the model simulation of the rainwater collection process. Focus on verifying the uniformity of the network elevation and the terrain data elevation system, and the connection between the rainwater grate ground elevation and the measured ground elevation. Usually, the grate ground elevation should be slightly lower than the ground elevation at that location.

[0127] In the present invention, when the coordinate data or ground elevation data of the urban drainage network and rainwater grates are added to the point cloud data of the ground, the coordinate data or ground elevation data of the urban drainage network and rainwater grates can be manually checked and corrected so that the ground elevation data of the urban drainage network and rainwater grates are slightly lower than the ground elevation data of the location.

[0128] (3.3) When integrating land classification data with other data, the focus should be on verifying the degree of overlap between land classification boundaries and measured terrain and object boundaries. When there are large discrepancies in data connection, adjustments should be made based on actual conditions. Since the accuracy of land classification data is usually slightly lower than that of terrain measurement results, the land classification boundaries in the corrected measured terrain data are generally used as the basis.

[0129] The fusion of land classification data and other data, land classification data according to the specifications include cultivated land, wetland, residential land, public management and public service land, etc.; the other data refers to all data except land classification, such as point cloud data, pipeline network data, measured local terrain data, etc. Land classification data, pipeline network coordinate data, measured local terrain elevation data can be added to point cloud data to realize the fusion of land classification data and other data.

[0130] (3.4) In order to solve the problems such as the inconsistency in the data fusion process, field surveys and detections are carried out to verify the boundaries of the above data from different sources to ensure that the topographic data, land classification data, pipe network drainage data and other data are accurately connected and fully integrated.

[0131] Due to the different ways of collecting different data, there may be problems with the connection between different data. For example, the boundaries of adjacent buildings may not be connected. In this case, the boundary coordinate data of adjacent buildings should be corrected through field surveys and detection verification so that the boundary data of adjacent buildings can be connected.

[0132] Regarding problems such as the lack of connection during the fusion of different data, another situation is that the ground elevation data of the rain grate should be lower than the ground elevation data of the location. In order to connect the boundary data of the rain grate with the boundary data of the ground where the rain grate is located, the boundary coordinate data of the rain grate and the boundary coordinate data of the ground where the rain grate is located can be corrected through field surveys and detection verification.

[0133] In the present invention, after completing the data fusion and correction in step three, the processing of each subsequent step is performed based on the fused data.

[0134] Step 4: Based on the natural terrain, river channels and pipe networks, and land type datasets integrated in step 3, the sub-catchment areas are dynamically divided into three levels.

[0135] It should be noted that the sub-catchment area in this step mainly refers to the area other than the hardened road and the impermeable hardened road surface connected to the road system, such as the residential area, green space, and pond. The sub-catchment area uses the urban stormwater runoff model to simulate the runoff generation and confluence. In the following step 5, the sub-catchment area uses the urban stormwater runoff model to simulate the runoff generation and confluence. The hardened system such as the road is simulated by the evolution of the two-dimensional hydrodynamic model. In step 6, the evolution of the two-dimensional hydrodynamic model is used to simulate the hardened system such as the road.

[0136] The step 4 further comprises the following steps:

[0137] (4.1) As attached Figure 3 As shown in the figure, the various types of data in step three are integrated. First, the first-level sub-catchment area is divided based on the macroscopic main river scale, and the boundary coordinates of the first-level sub-catchment area are marked on the fused point cloud mapping data. The first-level drainage system of the city includes the main river, large-scale drainage pumping stations along the river, etc. The first-level drainage system of the city generally undertakes the task of drainage in a large area, and the flood water is eventually discharged to the outer river. Therefore, the main river in the city with actual basin drainage function is initially used as the boundary of the first-level sub-catchment area to ensure that the first-level sub-catchment area division characteristics are consistent with the catchment characteristics of the main river in the city. The catchment characteristics refer to the direction of the river, the connection status with other main and tributary rivers, the river boundary, and the river basin boundary; the first-level sub-catchment area division characteristics are consistent with the catchment characteristics of the main river in the city, which means that the boundary of the first-level sub-catchment area is consistent with the facts, the flow direction of the main river in the fused data is consistent with the direction of the actual river, and the connection status of the main river with other main and tributary rivers is consistent with the facts.

[0138] Based on the various types of data fused in step 3 (the main urban river with actual drainage function is used as the first-level division boundary to ensure that the first-level catchment area division characteristics are consistent with the catchment characteristics of the main urban river.), firstly, the first-level catchment area division based on the macroscopic main river scale is constructed. Specifically, since the first-level drainage system of the city (usually composed of the main river, large-scale drainage pumping stations along the river, etc.) is generally responsible for the drainage task of a large area, the flood water is eventually discharged to the outer river. Therefore, the main urban river with actual drainage function is initially used as the first-level division boundary to ensure that the first-level sub-catchment area division characteristics are consistent with the catchment characteristics of the main urban river.

[0139] (4.2) Based on the division of the primary catchment area, the secondary catchment area is divided according to the main and secondary trunk roads with drainage pipes on both sides of the main road and the main channel with drainage function to ensure the rationality of the main drainage system.

[0140] On the basis of the first-level sub-catchment division, the urban second-level sub-catchment division is further refined based on the rainwater pipe network (or combined pipe network) and artificial trunk canal. Specifically, the division is based on the main and secondary trunk roads with drainage pipes on both sides and the trunk canals with drainage functions to ensure the rationality of the main drainage system.

[0141] (4.3) Based on the elevation data, current land use classification data, and building and road boundary data in the corrected and fused point cloud data in step 3, the third-level sub-catchment area is divided. Based on the boundary division of the first and second sub-catchment areas, the Hydrology tool in ArcToolBox in ArcGIS software is first used to divide the preliminary third-level catchment area based on elevation and flow direction. The Hydrology sub-catchment area division includes six steps: filling depressions, flow direction calculation, runoff accumulation calculation, control point insertion, setting extraction accuracy, and catchment area boundary extraction.

[0142] In the present invention, the boundaries of the three-level sub-catchment areas can be determined based on the terrain, and can be roads, drainage ditches, etc. It should be noted that using ArcGIS software to perform preliminary division of sub-catchment areas is one of the conventional means in the industry.

[0143] The preliminary three-level sub-catchment area is superimposed and integrated with the land use classification type data, building and road boundary data, with a focus on the building layer data to ensure that the same building does not connect to two catchments, that is, the catchment area boundary does not cross the building body. The following two situations are handled separately:

[0144] (4.3.1) If the building riser is connected to a drainage pipe, ArcGIS software is used to adjust the watershed division to ensure that rainwater from the building is connected to a nearby watershed area. The watershed area refers to the third-level watershed area. Using ArcGIS software to adjust the watershed division means manually adjusting the position, i.e., coordinates, of the boundary points of the third-level watershed area in the point cloud mapping data. For example, if the boundary line of the third-level watershed area passes through a building, ArcGIS software is used to manually modify the position of the points on the boundary line of the third-level watershed area in the fused point cloud mapping data. Specifically, the coordinates of the boundary line of the third-level watershed area are modified to prevent the boundary line of the third-level watershed area from passing through the building.

[0145] (4.3.2) If no drainage pipes are laid near the building riser, it is assumed that rainwater from a single building will flow into the direction of the largest slope of the terrain near the riser and converge to the lowest point; it is also necessary to ensure that the boundary of the catchment area does not divide the building. The boundary coordinates of the third-level sub-catchment area can be artificially modified so that the boundary of the third-level sub-catchment area bypasses the building, ensuring that a building can only be in one third-level sub-catchment area.

[0146] (4.4) It should be noted that the sub-catchment areas in this step mainly refer to areas other than hardened roads and impermeable hardened pavements connected to the road system, such as residential areas, green areas, and ponds. These sub-catchments use the urban stormwater runoff model to simulate runoff generation and confluence (see step five), and hardened systems such as roads are simulated using the two-dimensional hydrodynamic model evolution (see step six).

[0147] Step 5: Construct a comprehensive simulation method of multi-city stormwater runoff model based on Bayesian average. The multi-city stormwater runoff model based on Bayesian average is used to calculate the runoff of each sub-catchment obtained in step 4. The SCEM-UA algorithm (Shuffled Complex Evolution Metropolis Algorithm) and measured rainfall runoff data are used to simulate and calibrate the Bayesian weight coefficients of different models and the internal parameters of each model, and the runoff process of each catchment outlet is deduced. Then the runoff of each sub-catchment is preferentially connected to the rainwater pipe system and processed according to one-dimensional flow calculation; the excess runoff is solved according to the two-dimensional hydrodynamic model of the slope. The catchment outflow is connected to the pipe or the slope flow is simulated by Mikeflood software.

[0148] Each catchment area is calculated using a multi-city stormwater runoff model based on Bayesian averaging. The pipe network system uses one-dimensional flow calculations, and the runoff that exceeds the standard road surface is solved using two-dimensional hydrodynamics to simulate and calibrate the parameter domains of each variable. The SCEM-UA algorithm (Shuffled Complex Evolution Metropolis Algorithm) and measured rainfall runoff data are used to simulate and calibrate the Bayesian weight coefficients of different models and the internal parameters of each model. The runoff from each sub-catchment area is preferentially connected to the stormwater pipe system, and the flow in the pipe channel is processed according to one-dimensional flow calculations; the runoff that exceeds the channel discharge standard is solved according to the two-dimensional hydrodynamic model of the slope. Since rainfall that does not exceed the pipe channel discharge standard usually does not cause flooding and waterlogging, there is no waterlogging warning. The present invention mainly warns of rainfall events that exceed the pipe channel discharge standard and will cause waterlogging.

[0149] The step five further comprises the following steps:

[0150] (5.1) This step uses the Bayesian model averaging method to couple multiple urban stormwater runoff models. In essence, the posterior probability distribution of each model is used as the weight to perform weighted averaging on the predicted values ​​of a single model to obtain the Bayesian model average. Overall, the better the simulation effect, the higher the model weight.

[0151] Let y be the multi-model simulation variable (simulated runoff), Q obs =[y1,y2,...,y T ] is the measured runoff process required for the calibration model, the model input is ξ (rainfall process), T is the measured sequence length, M = [m1, ..., m k ,...,m K ] represents the model space composed of k selected urban stormwater runoff models, k = 1, 2, ..., K; K is the total number of selected urban stormwater runoff models. k (y|m k ,Q obs ,ξ) is the given observation data Q obs , input data ξ and model simulation value m k The posterior distribution of the simulated variable y under the condition of . According to the Bayesian total probability rule, the probability density function of the average simulated variable y (predicted runoff) of the Bayesian model is:

[0152]

[0153] In the formula, p k (y|m k ,Q obs ,ξ) is the posterior probability distribution of a single model k under given input conditions.

[0154] (5.2) Let ω k is the model weight coefficient, The mean of the posterior probability distribution of the Bayesian average simulated variable E(y|M,Q obs ,ξ) and variance D(y|M,Q obs ,ξ) are shown in formulas (6) and (7) respectively:

[0155]

[0156]

[0157] In the formula, σ k 2 For a given observation data Q obs and the variance of the simulated variable under the condition of the model input data ξ, the model weight coefficient ω k The acquisition method is shown in step 5.3 below.

[0158] (5.3) The objective function is to maximize NSE (Nash efficiency coefficient), and the simulated outflow process line of each sub-catchment is calculated based on the measured rainfall data. The Bayesian average probability distribution parameter ω is determined based on the sub-catchment runoff data. k and σ k 2 As well as the internal parameters of the stormwater runoff model of each city, RMSE (root mean square error) and CC (linear correlation coefficient) are used as evaluation indicators to evaluate the simulation results and the measured outflow process.

[0159] The objective function is to maximize the NSE (Nash efficiency coefficient), and the simulated outflow process lines of each sub-catchment are calculated based on the measured rainfall and other data. The SCEM-UA algorithm is used to calibrate the Bayesian average probability distribution parameter ω based on the sub-catchment runoff data. k and σ k 2 As well as the internal parameters of the stormwater runoff model of each city, RMSE (root mean square error) and CC (linear correlation coefficient) are used as the evaluation criteria for the simulation results and the measured outflow process.

[0160]

[0161]

[0162]

[0163] In this specific implementation method, the urban stormwater runoff model uses the SWMM model and the MOUSE model for simulation analysis, takes NSE as the objective function, adopts the SCEM-UA algorithm to solve the objective function, and calibrates the Bayesian weight coefficients of the two models and the internal parameters of the two models.

[0164] It should be noted that the SCEM-UA algorithm used to calibrate various unknown quantities is a conventional method in this field.

[0165] (5.4) The simulated outflow process of each sub-catchment area, the corrected DEM data, the catchment area boundary, and the pipe network vector data are input into the Mike flood software to simulate and analyze the waterlogging situation near each point of the project area road network, and compared with the measured waterlogging results to calibrate and adjust the various parameter settings.

[0166] The step 5.4 further comprises the following sub-steps:

[0167] (5.4.1) The outlet flow of each sub-catchment area that does not exceed the discharge standard of the pipe and canal system is connected to the drainage system through rainwater grates and pipes. The flow, confluence and overflow of water bodies in the drainage system are calculated by solving one-dimensional hydrodynamic equations.

[0168] (5.4.2) Rainwater exceeding the standard is mainly discharged through roads. Therefore, the two-dimensional hydrodynamic equation is used to solve and calculate the surface runoff process of roads that exceeds the discharge standard of the canal system, simulate and determine the maximum flooding depth and flooding duration of each target location, and draw the boundaries of the catchment area under the rainfall conditions of this event. In the modern society where the urbanization process is rapidly approaching, road pavement is an important channel for the discharge of rainwater exceeding the standard in cities and towns, and plays an important role in urban waterlogging and risk prevention. The size of the pavement drainage capacity has a significant impact on the efficiency of water discharge under high-recurrence rainfall and the risk of urban waterlogging.

[0169] (5.4.3) It should be noted that the use of Mike flood software to simulate two-dimensional flooding is a common practice in the industry. The operation process of Mikeflood software mainly includes: road boundaries, terrain, pipe network and watershed runoff input, grid division, model construction, two-dimensional water flow direction simulation, flooding depth result input, etc. For detailed operations, please refer to the software operation manual.

[0170] (5.5) Compare the two-dimensional road confluence path and water flow direction obtained by simulation with the boundaries of each sub-catchment area in step 4, and verify the rationality of the watershed division boundary. If the two-dimensional water flow direction reflected in the flow diagram does not correspond to the watershed boundary, for example, the water flow direction crosses the watershed boundary and other unreasonable problems, adjust the watershed boundary coordinate data on the fused point cloud mapping data, re-simulate each part and re-calibrate the parameters, including the Bayesian weight parameters and the learnable parameters of the Bayesian hydrological model itself. After repeatedly adjusting the parameters of the watershed boundary and optimizing the calculation, compare, test and analyze the simulation results with the measured results including the flooding depth and duration, calibrate the various parameters in the two-dimensional simulation, and use them for the next step of rainfall waterlogging simulation and risk assessment under different recurrence periods.

[0171] (5.6) It should be noted that since rainfall that does not exceed the canal discharge standard usually does not cause flooding and waterlogging, there is no waterlogging warning. The present invention mainly provides warnings for rainfall events that exceed the canal discharge standard and will cause waterlogging.

[0172] Step six: Based on the Bayesian weight parameters and internal parameter results of each model calibrated in the previous step, input rainfall data under different return periods, repeat the calculation of watershed runoff and waterlogging depth simulation and watershed boundary adjustment in step five, predict the flooding results under rainfall conditions with different return periods, and reversely infer the critical rainfall for rainstorm disaster warning at each point based on the allowable flooding depth at each point.

[0173] The step six further includes the following sub-steps:

[0174] (6.1) Based on the Bayesian weight parameters and model parameter results after simulation and calibration in the previous steps, as well as the scope of the catchment area, the rainfall conditions under different return periods are input, and the runoff generation and confluence process of each sub-catchment area under these different rainfall conditions is recalculated and predicted. Then, the Mike flood software is used to calculate the two-dimensional road slope flow and one-dimensional pipe flow process.

[0175] After simulating and calculating the flow direction and flow process, it is also necessary to compare and analyze the consistency of the simulated flow direction and the boundary of the watershed. That is, the two-dimensional overland flow does not cross the boundary of the watershed. If the two are inconsistent, the boundary of the sub-watershed is appropriately adjusted to ensure the rationality of the two-dimensional simulation prediction results.

[0176] (6.2) It should be noted that due to different rainfall events, the division and scope of the catchment area may change. Therefore, the catchment area boundaries under different recurrence periods are dynamically adjusted according to the actual runoff generation and flow rate of the two-dimensional road catchment to ensure the rationality of the division. Figure 4 As shown in the figure, due to the existence of local low-lying areas (such as concave areas), when the actual rainfall P does not exceed the critical rainfall condition P of the low-lying area, the inflow from the catchment area B1 upstream of the low-lying area gathers in the low-lying area and does not flow out; therefore, at this time, the catchment area range of outlet C is only the catchment area range of B2. However, when the actual rainfall P exceeds the critical rainfall condition P of the low-lying area, the low-lying area cannot receive all the inflow from the catchment area B1 upstream, and part of the excess flow flows downstream to the outlet C. At this time, the catchment area range of outlet C is the sum of the catchment areas B1+B2. Therefore, the catchment range of outlet C should be dynamically adjusted according to each rainfall event.

[0177] (6.3) According to sub-steps 6.1 and 6.2, the flooding depth and flooding duration near each target location are recalculated to determine the flooding depth and flooding duration of each point under each simulated condition after dynamically adjusting the catchment area boundary according to the actual confluence path simulated under each rainfall event.

[0178] (6.4) As attached Figure 5 As shown, based on the actual elevation of each point and information such as the allowable flooding depth, a comprehensive analysis and evaluation is conducted to determine the critical rainfall return period or disaster-causing rainfall for rainstorm disaster warning at each point.

[0179] Based on the actual elevation of each point, as well as the allowable flooding depth and duration, the critical rainfall for rainstorm disaster warning at each point is evaluated and determined.

[0180] The allowable flooding depth at each point can be determined comprehensively based on the demands of the ownership units of each point, the importance of the target location, the degree of social impact, and relevant regulations.

[0181] The present invention proposes a Bayesian average multi-city stormwater runoff model to calculate the water flow process of each catchment area, and calculates the flooding depth and flooding duration under different rainfall conditions based on the existing commercial software through a two-dimensional hydrodynamic model. According to the existing specifications, when the flooding depth reaches 15cm and the flooding duration reaches 30 minutes, urban waterlogging will occur. When different rainfall conditions are input, the flooding depth calculated based on different rainfall conditions reaches 15cm and the flooding duration reaches 30 minutes, then the rainfall involved in the rainfall conditions is the critical rainfall. In actual engineering, the critical rainfall threshold is a range value. If the critical rainfall is exceeded, it is considered that waterlogging will occur. During an actual rainfall process of a certain time, when the measured rainfall reaches the critical rainfall threshold, the city meteorological department or the important facility management department can issue an urban waterlogging warning.

[0182] Finally, it should be noted that the contents not described in detail in this specification belong to the prior art known to professional and technical personnel in this field. The above description is only the preferred embodiment of the present invention and is not used to limit the present invention. Although the present invention is described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions recorded in the aforementioned embodiments, or to replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for calculating critical rainfall of important facilities based on digital surface model, characterized in that: The following steps are involved: S1: Data collection: collect high-precision point cloud mapping data of cities where important facilities are located, current land use classification data, topographic data of buildings, roads and canals, drainage network data, historical rainfall data with different return periods and flooding depth results under corresponding rainstorm events; S2: Correct the various data collected in step S1; S3: Fusing the various data corrected in step S2, specifically adding the corrected land use classification data, building, road and river terrain data, and drainage network data to the corrected point cloud mapping data, and correcting the boundary data coordinates and elevation of the fused data; S4: Based on the integrated point cloud mapping data, land use classification data, and terrain data of buildings, roads, and canals in step S3, the sub-catchment areas are dynamically divided into three levels; S5: constructing a multi-city stormwater runoff model based on Bayesian average, wherein the multi-city stormwater runoff model based on Bayesian average is used for the calculation of runoff and waterlogging simulation of each sub-catchment area obtained in step 4 of S4; S6: Input rainfall data under different return periods into the multi-city stormwater runoff model based on Bayesian averaging constructed in step S5, repeat the calculation of runoff production and waterlogging depth simulation in step S5, and predict the flooding results under rainfall conditions with different return periods through the multi-city stormwater runoff model based on Bayesian averaging. The flooding results include flooding depth and flooding duration. When the flooding depth and flooding duration predicted based on the input rainfall conditions are the flooding depth and flooding duration that will cause waterlogging in the existing specifications, the rainfall involved in the input rainfall conditions is the critical rainfall for warning of heavy rain disasters for each important facility; Wherein: the step S5 specifically includes the following steps: S5.1: Construct an urban stormwater runoff model based on Bayesian model averaging. Let y be the multi-model simulation variable, i.e., simulated runoff. The posterior probability distribution p(y|m k ,Q obs ,ξ) is: In the formula, Q obs =[y1,y2,...,y T ] is the measured runoff process required for the calibration model, the model input is ξ, T is the measured sequence length, M = [m1, ..., m k ,...,m K ] represents the model space composed of k selected urban stormwater runoff models, K is the total number of selected urban stormwater runoff models; p k (y|m k ,Q obs ,ξ) is the given observation data Q obs , input data ξ and model simulation value m k The posterior distribution of the simulated variable y under the condition of k (y|m k ,Q obs ,ξ) is the posterior probability distribution of a single model k under given input conditions; S5.2: Let ω k is the model weight coefficient, then The mean of the posterior probability distribution of the Bayesian average simulated variable E(y|M,Q obs ,ξ) and variance D(y|M,Q obs ,ξ) are shown in formulas (6) and (7) respectively: In the formula, σ k 2 For a given observation data Q obs and the variance of the simulated variables under the condition of the model input data ξ; S5.3: Select the maximum NSE Nash efficiency coefficient as the objective function, calculate and solve the simulated outflow process line of each sub-catchment based on the measured rainfall data, and determine the Bayesian average probability distribution parameter ω based on the sub-catchment runoff data. k and σ k 2 As well as the internal parameters of the rainwater runoff model of each city, the RMSE root mean square error and CC linear correlation coefficient are used as evaluation indicators for evaluating the simulation results and the measured outflow process, among which: Where: Q sim,i represents the simulated flow of the subcatchment in the i-th period, Q obs,i represents the measured flow rate of the subcatchment in the i-th period, Q obs represents the historical measured flow sequence of the subcatchment area; Q sim represents the simulated flow series of the subcatchment; S5.4: The simulated outflow process of each sub-catchment area obtained by solving the problem is input into the Mike flood software together with the corrected and fused point cloud mapping data, the current land use classification data, the terrain data of buildings, roads and canals, the boundary data of the catchment area, and the vector data of the pipe network. The Mike flood software is used to simulate and analyze the waterlogging inundation situation of each point on the road network in the area where the important facilities are located, and the waterlogging results measured in the historical data are compared and analyzed, and the model parameter settings are adjusted; S5.5 Compare the two-dimensional road confluence path and simulated flow direction obtained by simulation with the boundaries of each sub-catchment area in step 4, and verify the rationality of the catchment area division boundary. If the two-dimensional water flow direction does not correspond to the catchment area boundary, adjust the catchment area boundary coordinate data, re-calibrate the parameters of each part, and use the model parameters used when the waterlogging and inundation conditions at each point on the road network in the area where important facilities are located obtained by Mike flood software simulation analysis are closest to the measured waterlogging results as the optimal model parameters.

2. The method for calculating critical rainfall of important facilities based on digital surface model according to claim 1 is characterized in that: The S1 step specifically includes the following steps: S1.1: Use LiDAR technology to collect high-precision point cloud mapping data of cities where important facilities are located, and use LiGeoreference solution software or LiDAR360 software to analyze the data collected by LiDAR technology to obtain high-precision point cloud mapping data; S1.2: Obtain the current land use classification data through the latest land use classification results of the area where the important facilities are located; S1.3: Obtain the overall outline and elevation data of the building perimeter in the area where the important facilities are located, obtain the boundary data and elevation data of municipal roads and community roads; obtain the cross-sectional and longitudinal morphology and direction of the river channel, the boundary position of the river channel, and the elevation data of the levee crest. The boundary data or boundary position is reflected by coordinate data; S1.4: Obtain the location, elevation, burial depth, pipe diameter, material, and direction data of the drainage network, as well as the damage location, damage level, siltation type and level data of the existing drainage pipelines; S1.5: Obtain through historical data a record of at least one complete rainfall event, and the measured flooding duration and flooding depth near the location of important facilities during the complete rainfall event.

3. The method for calculating critical rainfall of important facilities based on digital surface model according to claim 1 is characterized in that: The S2 step specifically includes the following steps: S2.1: The original point cloud mapping data is corrected by using the Gaussian filtering method. The noise of the original point cloud mapping data is removed by the Gaussian filtering method. The Gaussian filtering function H is defined as: In the formula, the normal distribution function f(x) is expressed as: In formulas (1) and (2), e is the base of the natural logarithm and is taken as 2.71828; x is the scanning window width; σ is the standard deviation of x, and σ is the variable value calculated by the normal function as x changes; The distance value D of the current filtering scanning point is obtained by formula (1) and formula (2): fil The relationship between each data point, i.e., the Gaussian filtering result, is shown in formula (3), and the scanning window width x is shown in formula (4): D fil =H×D(x i )(3) x=lin space(-d,d,2·d+1) (4) In the formula, x i is the position of data point i scanned by the radar after excluding the maximum value position and invalid point position data in the window; D(x i ) is the distance value corresponding to each data point; d is the radius of the window; S2.2: Verify the drainage network data through pipeline closed-circuit television detection CCTV, pipeline periscope detection QV, sonar detection, and manual observation methods, including verifying the drainage network longitudinal section elevation and direction, connection status, and pipe diameter size data; S2.3: Verify the rationality of the current land use classification data set through field survey of land use classification data; S2.4: The rationality of the topographic data of buildings, roads and canals shall be verified by field measurement. The elevations of buildings, roads, canals above the water surface and the top of the dikes shall be verified by total station and digital photography; the cross-sectional data of canals under water shall be measured and verified by total station combined with flower pole or steel ruler or underwater robot; S2.5: The maximum inundation depth and inundation duration of the measured rainstorm floods at the locations of important facilities are analyzed and verified by water level monitoring using water level sensors arranged at multiple points. The required rainfall data for the same event are obtained from nearby meteorological stations in the area.

4. The method for calculating critical rainfall of important facilities based on digital surface model according to claim 1 is characterized in that: The S3 step specifically includes the following steps: S3.1: Based on the area where important facilities are located, add the terrain data of buildings, roads, and canals to the point cloud mapping data, and correct the added and fused terrain data. Specifically, add the horizontal and vertical coordinates and elevation data of underground pipe networks and above-ground buildings to the point cloud data. Correcting the fused terrain data means adjusting the coordinates of the fused boundary data according to the actual situation. S3.2: When the urban drainage network and rainwater grate data are added to the point cloud mapping data, the elevation data of the urban drainage network and rainwater grate are corrected so that the ground elevation data of the urban drainage network and rainwater grate are lower than the ground elevation data of the location of the urban drainage network and rainwater grate; S3.3: Integration of land classification data with other data, including adding land classification data, pipe network coordinate data, and measured local terrain elevation data to point cloud mapping; S3.4: After fusing the data collected through different channels, correct the boundaries of the fused data, including correcting the boundary coordinate data of adjacent buildings through field surveys and detection verification, so that the boundary data of adjacent buildings can be connected.

5. The method for calculating critical rainfall of important facilities based on digital surface model according to claim 1 is characterized in that: The S4 step specifically includes the following steps: S4.1: Take the main urban river with actual drainage function as the first-level sub-catchment area, and mark the boundary coordinates of the first-level sub-catchment area on the fused point cloud mapping data; S4.2: Based on the division of the first-level sub-catchment area, the main and secondary roads with drainage pipes on both sides of the main urban river and the main canals with drainage functions are regarded as the second-level sub-catchment area, and the boundary coordinates of the second-level sub-catchment area are marked on the fused point cloud mapping data; S4.3: Based on the demarcation of the primary and secondary sub-catchments, the Hydrology tool in ArcToolBox in ArcGIS software is used to demarcate the tertiary sub-catchments based on elevation and flow direction. The boundary coordinate data of the tertiary sub-catchments are annotated in the fused point cloud mapping data using ArcGIS software; The S4.3 step includes the following two situations: S4.3.1: If a single building has a vertical pipe connected to a drainage channel, use ArcGIS software to adjust the boundary coordinates of the three-level catchment areas on the fused point cloud mapping data to ensure that rainwater from the building is connected to a nearby catchment area; S4.3.2: If no drainage pipes are laid near the building riser, the rainwater from a single building will be collected in the direction of the largest slope of the terrain adjacent to the riser and converge to the lowest point; specifically, the coordinate data of the third-level sub-catchment area with the largest slope of the adjacent terrain will be added to the point cloud mapping data of the single building.

6. The method for calculating critical rainfall of important facilities based on digital surface model according to claim 1 is characterized in that: The step S5.4 specifically includes the following steps: S5.4.1: When the outlet flow of each subcatchment area within the discharge standard of the canal system is connected to the drainage system through pipes and stormwater grates, the flow, confluence and overflow of water bodies in the drainage system shall be calculated by solving the one-dimensional hydrodynamic equations; S5.4.2: When the rainfall is too heavy, the drainage required by the canal system exceeds the discharge standard of the canal system. The surface runoff process of the road that exceeds the discharge standard of the canal system is calculated by solving the two-dimensional hydrodynamic equation, and the maximum flooding depth and flooding duration of each important facility are simulated and calculated. When the rainfall is too heavy, the drainage required by the canal system exceeds the discharge standard of the canal system, and the boundary coordinate data of the catchment area under rainfall conditions are drawn; S5.5 Compare the two-dimensional road confluence path and simulated flow direction obtained by simulation with the boundaries of each sub-catchment area in step 4, and verify the rationality of the catchment area division boundary. If the two-dimensional water flow direction does not correspond to the catchment area boundary, adjust the catchment area boundary coordinate data, re-calibrate the parameters of each part, and use the model parameters used when the waterlogging and inundation conditions at each point on the road network in the area where important facilities are located obtained by Mike flood software simulation analysis are closest to the measured waterlogging results as the optimal model parameters.

7. The method for calculating critical rainfall of important facilities based on digital surface model according to claim 1 is characterized in that: The step S6 specifically includes the following steps: S6.1: Based on the Bayesian weight parameters and model parameter results after simulation and calibration in step S5 and the watershed division range, input different rainfall conditions, recalculate and predict the runoff generation and confluence process of each sub-catchment under different return period rainfall conditions, and use Mike flood software to calculate the two-dimensional road slope flow process. After simulating and calculating the confluence flow direction and confluence process results, compare and analyze the consistency of the simulated water flow direction with the watershed division boundary; S6.2: Dynamically adjust the boundary coordinate data of the third-level sub-catchment area according to the actual runoff generation and runoff process and the two-dimensional water flow of the road; S6.3: According to sub-steps 6.1 and 6.2, the flooding depth and flooding duration of each key location, and dynamically adjust the boundary coordinates of the catchment area on the fused point cloud mapping data according to the actual confluence path simulated under each rainfall event, re-enter the rainfall conditions, and determine the flooding depth and flooding duration of each point under each simulated working condition based on the Bayesian average multi-city stormwater runoff model; S6.4: Based on the actual elevation of each point, as well as the allowable flooding depth and duration, evaluate and determine the critical rainfall for rainstorm disaster warning at each point. When the actual rainfall in a certain period reaches the critical rainfall, a waterlogging warning for important urban facilities will be issued.

Images