CN115391712A - Urban flood risk prediction method - Google Patents

Abstract

The invention discloses a method for predicting urban flood risk, comprising the following steps: based on the historical radar echo map of the city to be predicted, the radar echo map is preprocessed and reclassified and calculated by GIS, and the rainfall forecast of the city to be predicted is constructed model; build a SWMM-based flood risk prediction model for the city to be predicted; couple the rainfall prediction model flood risk model to obtain a flood risk prediction model; preprocess the rainfall prediction value output by the rainfall prediction model as the flood risk model The input value of rainfall; input the real-time radar echo map of the city to be predicted into the flood risk prediction model to obtain the analysis result of the flood risk of the city to be predicted. The invention has fast prediction and identification speed and high accuracy of results. The position of the main rainfall area can be predicted through the flood risk prediction model, the rainfall estimation accuracy of the main rainfall section is high, and the flood risk is analyzed according to the rainfall.

Description

A Method for Urban Flood Risk Prediction

technical field

The invention relates to the technical field of water conservancy engineering, in particular to a method for predicting urban flood risk.

Background technique

The rapid development of urbanization has significantly increased the impermeable area of the underlying surface of the city. Local sudden rainstorms often lead to serious floods in the city, threatening people's lives and causing huge property losses. city development. At this stage, there are problems such as low warning accuracy, short forecast period, and unsound mechanism in the release of waterlogging early warnings, which directly restrict the accuracy and timeliness of waterlogging early warnings. Therefore, carrying out research on key technologies for rapid urban flood risk identification is conducive to building a set of comprehensive dispatching indicators (covering forecasted rainfall, predicted flow, measured rainfall, and measured flow) and dispatching principles to improve the scientific dispatching level of urban water systems.

Contents of the invention

Purpose of the invention: The present invention aims to provide a rapid and accurate urban flood risk prediction method.

Technical solution: A method for predicting urban flood risk of the present invention comprises the following steps:

(1) Based on the historical radar echo map of the city to be predicted, the radar echo map is preprocessed and reclassified by GIS, and the rainfall prediction model of the city to be predicted is constructed;

(2) Collect the basic data of the city to be predicted, and preprocess the collected basic data;

(3) According to the preprocessed basic data in step (2), build a SWMM-based flood risk prediction model for the city to be predicted; generalize the node pipe network of the city and divide and calculate the sub-catchments, Construct the basin runoff model and the pipe network confluence model, and use ArcGIS software to simplify the local river data, import the simplified river data into SWMM, check and check the node pipe network logic and topological relationship, and build the river confluence sub-model ; Extract depressions and overflow points, and calculate the sub-catchment area corresponding to each depression; use the shortest path extraction tool to extract the confluence path between the overflow points of the connected depression; The obtained depression sub-catchment area is generalized into the surface water catchment area, thereby constructing the surface depression confluence model; coupling the river channel confluence sub-model with the surface depression confluence model to obtain the flood risk model of the city to be predicted;

(4) Coupling the flood risk model of the rainfall forecast model to obtain the flood risk forecast model; the rainfall forecast value output by the rainfall forecast model obtained in step (1) is preprocessed as the flood risk model in step (3) Rainfall input value;

(5) Input the real-time radar echo map of the city to be predicted into the flood risk prediction model to obtain the analysis results of the flood risk of the city to be predicted.

Further, in step (1), the method of constructing the rainfall prediction model of the city to be predicted is as follows: based on the historical radar echo images of multiple rainfall events, the radar echo images are preprocessed and reclassified by GIS, and their It is divided into high reflectance area, medium reflectance area and low reflectance area; the characteristic value of each cloud type and the identification law of rainfall type are extracted; according to the spectral characteristics of each reflectance in the radar echo map, the rainfall type identification index system is constructed.

Among them, the range of high reflectivity area is 35-70dBZ, the medium reflectivity area is 25-35dBZ, and the low reflectivity area is 0-25dBZ; the constructed rainfall type identification index system is as follows:

(1) (area of high reflectivity area + area of medium reflectivity area)/total area of echo map ≤ 0.1, stratiform cloud rainfall;

(2) (area of high reflectance area + area of medium reflectance area)/total area of echo map ≥ 0.3, convective cloud rainfall;

(3) 0.1<(high albedo area + medium albedo area)/total echo map area<0.3 and brightness gradient<0.15, mixed cloud rainfall dominated by stratiform clouds;

(4) 0.1<(high albedo area + medium albedo area)/total echo map area<0.3 and brightness gradient ≥0.15, mixed cloud rainfall dominated by convective clouds.

Step (1) also includes forecasting, inversion and precision analysis of rainfall, summarizing cloud motion vector calculation using pyramid optical flow method through radar echo chart; Falling area accuracy assessment and magnitude accuracy assessment.

Further, in step (2), the urban DEM data, river channel data, rainfall data, and underlying surface type data are collected and preprocessed, and the underlying surface of the watershed is automatically interpreted and extracted by software. Interpretation analysis is carried out by using types, so as to further calculate the impervious rate of the underlying surface of the watershed.

Further, in step (3), the rainfall runoff of each sub-catchment area is calculated separately in the watershed runoff model. The formula for the total production flow is expressed as:

R＝R ₁ +R ₂ +R ₃

In the formula, R is the total production flow, in mm; R ₁ is the production flow in the permeable area, in mm; R ₂ is the surface production flow in the impervious area with depressions, in mm; R ₃ is the impervious storage without depressions The surface yield of the area, in mm.

Furthermore, the SWMM model uses the nonlinear reservoir method to calculate the confluence of the permeable area and the impermeable area through the water balance equation and the Manning equation.

Further, in the construction of the pipe network confluence model in step (2), the dynamic wave method is used to simulate the pipe network confluence in the SWMM model.

Furthermore, in step (3), the parameters of the subcatchment include the impermeability rate, which interprets and analyzes the land use type of the underlying surface of the urban watershed to be predicted through the interpretation and analysis of the preprocessed remote sensing image data, and according to the Use the type to calculate the impermeability of the underlying surface of the watershed.

Furthermore, in step (3), the parameters of the sub-catchment also include the Manning coefficient of the slope value, the depression water storage depth, and the infiltration rate, wherein the slope value is extracted after processing the DEM data; by substituting it into the SWMM model and The final values of Manning coefficient, depression storage depth and infiltration rate are determined by repeated parameter adjustment.

Further, in step (3), the method of extracting depressions is as follows: set different filling thresholds for high-precision DEM data through the filling tool in hydrological analysis in ArcGIS and perform two calculations to obtain the first calculation results and The result of the second calculation is then subtracted from the DEM layer with a fill threshold of 0.1m obtained in the first calculation result with the Minus tool to extract the easy product In the low-lying area of water, the volume of the depression is finally calculated, and the formula for the volume of the depression is expressed as:

V _wa =S _wa ×h _mean

In the formula, V _depression is the volume of the depression, the unit is m ³ ; S _depression is the area of the depression, the unit is m ² ; h _mean is the average depth of all pixels within the extracted depression range, the unit is m.

Furthermore, in step (3), the extraction method of the subcatchment corresponding to the depression is as follows: firstly find the lowest point of the depression, and then combine the flow direction of the water flow to determine all the streams in the upstream area of the lowest point in the entire large watershed. A raster passing through this nadir is used to complete the subcatchment extraction.

Furthermore, in step (3), the extraction method of the depression overflow point is: first calculate the flow direction of the water flow in the DEM data, and then calculate the size of the confluence accumulation according to the water flow direction, so as to obtain the depression overflow point point position.

Furthermore, step (3) includes parameter testing of the established flood risk prediction model, and input the design rainfall data once in 20 years into the model for simulation. According to the comparison between the simulation results and the design flow process line, if two The curves differ greatly, and the parameters are readjusted for simulation until the simulated flow hydrograph is closer to the design flow hydrograph and the Nash coefficient of the two sets of data is close to 1.

Beneficial effect: compared with the prior art, the present invention has the following significant advantages:

(1) Objectively, quickly and accurately judge the type of cloud layer, and the accuracy of the rainfall prediction model is high;

(2) The speed of prediction and recognition is fast, and the accuracy of the results is high. The location of the main rainfall areas can be predicted through the flood risk prediction model, and the rainfall estimation accuracy of the main rainfall segments is high, and the flood risk is analyzed based on the rainfall.

Description of drawings

Figure 1 is a map of land use distribution in the Qingyang River Basin;

Figure 2 shows the distribution of water permeability and impermeability in the Qingyang River Basin;

Figure 3 is a diagram of the design rainstorm process once in 10 years under the condition of T=1h;

Figure 4 is the design rainstorm process diagram for the 10a and 20a once in the case of T=24h;

Figure 5 is a generalized map of the pipeline network in the Qingyang River Basin;

Figure 6 is a generalized map of the node pipe network channel in the study area;

Figure 7 is an enlarged view of the outlet of the watershed at A in Figure 6;

Figure 8 is a schematic diagram of the model network structure, (a) DEM unit grid, (b) depression overflow point and sub-catchment area, (c), flow path of water flow (d), profile of rainwater flow on the surface;

Figure 9 is a distribution map of depressions, overflow points and confluence paths in the Qingyang River Basin;

Figure 10 is a slope map of the catchment area of each depression in the Qingyang River Basin;

Figure 11 is the impermeability map of the Qingyang River Basin;

Figure 12 is the generalized model of the depression system in SWMM;

Figure 13 shows the flow process calculated by the instantaneous unit line method and the flow process simulated by the SWMM model;

Figure 14 is a diagram of the once-in-a-time rainfall simulation of the rapid flood risk identification model and the refined flood model 10a;

Figure 15 is the radar echo map of the rainfall event at 21:42 on August 12, 2020

Figure 16 is the radar echo map of the rainfall event at 21:48 on August 12, 2020;

Figure 17 is the radar echo map of the rainfall event at 21:54 on August 12, 2020;

Figure 18 is the "2020.08.12" rainfall forecast map (1h total rainfall);

Figure 19 is the "2020.8.12" distribution map of Beijing's predicted rainfall and measured rainfall (top: the first 6 minutes, bottom: the last 6 minutes);

Figure 20 shows the corresponding situation between the spatial distribution of predicted water accumulation points and the actual water accumulation points in August 12, 2020;

Figure 21 shows the simulated waterlogging situation of the T1 session rainfall;

Figure 22 is the simulated water accumulation situation of T2 rainfall;

Figure 23 is the simulated water accumulation situation of T3 rainfall;

Figure 24 shows the simulated water accumulation situation of the T4 rainfall event.

Detailed ways

The technical solution of the present invention will be further described below in conjunction with the accompanying drawings.

Example 1

This embodiment provides a method for predicting urban flood risk, which specifically includes the following steps:

S1 constructs the rainfall prediction model of the city to be predicted:

The rainfall prediction model is used to automatically identify the type of cloud layer, among which is to judge the type of cloud layer objectively, quickly and accurately. Based on the radar echo images of multiple rainfall events, the radar echo images are preprocessed and reclassified by GIS. It is divided into high reflectivity area (35-70dBZ), medium reflectivity area (25-35dBZ) and low reflectivity area (0-25dBZ), and extracts the characteristic values of four cloud types, and summarizes the rainfall type Identify patterns to provide fast and accurate rainfall type input for near-term rainfall nowcasting. According to the spectral characteristics of each reflectivity of the radar echo map, the rainfall type identification index system is constructed as follows:

(1) (area of high reflectivity area + area of medium reflectivity area)/total area of echo map ≤ 0.1, stratiform cloud rainfall;

(2) (area of high reflectance area + area of medium reflectance area)/total area of echo map ≥ 0.3, convective cloud rainfall;

(3) 0.1<(high albedo area + medium albedo area)/total echo map area<0.3 and brightness gradient<0.15, mixed cloud rainfall dominated by stratiform clouds;

(4) 0.1<(high albedo area + medium albedo area)/total echo map area<0.3 and brightness gradient ≥0.15, mixed cloud rainfall dominated by convective clouds.

Prediction, inversion, and precision analysis of rainfall are carried out on the rainfall prediction model constructed above, so as to evaluate the accuracy of the rainfall falling area and magnitude of the model prediction.

Prediction, inversion and accuracy analysis of rainfall are as follows: through the radar echo chart, it is concluded that the cloud motion vector calculation is carried out by using the pyramid optical flow method: the radar rainfall nowcasting based on the pyramid optical flow method is mainly composed of cloud motion vector calculation and rainfall The extrapolation and interpolation forecast consists of two parts:

Combining the LK optical flow method with the image sequence pyramid to construct the pyramid optical flow method, the linear equation of the LK optical flow method is as follows:

I _x (q) u+I _y (q) v=-I _t (q), q∈Ω

In the formula, I _x , I _y , I _t are the partial derivatives of the rainfall corresponding to point q to x, y, t respectively;

When calculating the optical flow velocity, firstly calculate the weighted square sum of a certain specific window function in the neighborhood Ω of the previous pixel point q, and then calculate the minimum value, and then use the least square method to use the Gaussian function to calculate the The weight proportion of the point is distributed to other points in the field Ω, and the calculation formula is as follows:

In the formula, * is convolution; K _ρ is the Gaussian kernel of standard deviation ρ;

Then use the relationship between echo reflectivity and rainfall intensity to forecast and invert the rainfall in the urban flood season. The principle of weather radar inversion of rainfall information is to invert the relationship between reflectivity factor Z and rainfall intensity R push calculation. Select a number of different rainfall events, and use the summed up cloud type discrimination threshold to determine the type of each rainfall event.

Z=A×R ^b

In the formula, Z is the radar reflection factor; R is the rainfall intensity; A and b are the inversion parameters. The values of parameters A and b of different types of clouds refer to previous research results.

Finally, combined with the measured rainfall data, the predicted rainfall data is evaluated for the accuracy of the falling area and the magnitude of the accuracy.

S2 basic data collection and preprocessing:

Collect and preprocess urban DEM data, river channel data, rainfall data, and underlying surface type data, and automatically interpret and extract the underlying surface of the watershed through software and manual visual interpretation, and analyze the land use type of the underlying surface of the watershed Interpretation analysis is carried out to further calculate the impervious rate of the underlying surface of the watershed.

S3 builds a SWMM-based flood risk prediction model for cities to be predicted:

Generalize the city's node pipe network, divide and calculate the sub-catchment area, build a watershed runoff model and a pipe network convergence model, and then simplify the local river data through ArcGIS software; import the processed river data into SWMM , and check and check the node pipe network logic and topological relationship, and construct the river confluence sub-model. Extract depressions and overflow points, and calculate the sub-catchment area corresponding to each depression; use the shortest path extraction tool to extract the confluence path between the overflow points of the connected depression; The sub-catchment area of the depression is generalized into the surface catchment area, so as to construct the surface depression confluence model; the flood risk model of the city to be predicted is obtained by coupling the channel confluence sub-model and the surface depression confluence model.

S301 river confluence sub-model:

The SWMM model belongs to the distributed watershed hydrological model, and the basic unit of rainfall and runoff calculation in this model is each sub-catchment; considering the surface characteristics of different underlying surfaces, the runoff calculation divides each sub-catchment in the watershed into different sub-catchments. There are two types of permeable areas and permeable areas, and the impermeable areas are divided into impermeable areas with depressions and impermeable areas without depressions.

When the rainfall flows through the impermeable area with depressions, the net rain is formed after filling the depressions on the surface, and the evaporation and production flow of depressions should be removed during calculation. The calculation formula is:

R ₂ =PED

In the formula, _R2 is the surface runoff in the impermeable area with subsidence storage, in mm; P is the rainfall, in mm; E is the evaporation, in mm; D is the surface subsidence storage, in mm.

The impact of evaporation should be considered in the impermeable area without depression storage. When calculating the production flow, the rainfall should be subtracted from the evaporation loss. The calculation formula of production flow is:

R ₃ =PE

In the formula, _R3 is the surface yield of the impermeable area without depression storage, in mm; P is the rainfall, in mm; E is the evaporation, in mm.

When the rainfall produces runoff in the permeable area, the surface subsidence storage and evaporation loss should be subtracted, and the infiltration of the soil should also be subtracted. The formula for calculating the runoff is:

R ₁ =(if)×Δt

In the formula, _R1 is the flow rate of the permeable area, in mm; i is the rainfall intensity, in mm/h; f is the infiltration intensity, in mm/h; Δt is the time interval, in h.

The calculation of infiltration in the SWMM model covers three models, namely: Horton model (Horton model), Green-Ampt model (Green-Ampt model) and runoff curve number model (SCS model), in which Horton The calculation of the model takes into account the decrease of soil infiltration rate with the increase of time. In addition, the model has fewer parameters and is easy to operate. It has strong applicability to small watershed research areas and is more suitable for urban areas.

The Horton infiltration formula is:

f _t =f _c +(f ₀ -f _c )×e ^-kt

In the formula, f _t is the soil infiltration rate at time t, in mm/h; f _c is the stable infiltration rate of soil, in mm/h; f ₀ is the initial infiltration rate of soil, in mm/h h; k is the infiltration attenuation coefficient; t is the soil infiltration time, the unit is h.

Each subcatchment is composed of three types of surfaces: permeable area, impervious area with depressions and impervious areas with depressions. The loss of rainfall through these three types of surfaces is not the same, so it is necessary to separate each type The surface runoff of the basin is calculated separately, and then added together to obtain the total surface runoff of the watershed, the formula is expressed as:

R＝R ₁ +R ₂ +R ₃

In the formula, R is the total production flow, in mm; R ₁ is the production flow in the permeable area, in mm; R ₂ is the surface production flow in the impervious area with depressions, in mm; R ₃ is the impervious storage without depressions The surface yield of the area, in mm.

The SWMM model adopts the nonlinear reservoir method, through the water balance equation and the Manning equation, to calculate the confluence of the permeable area and the impermeable area:

The water balance equation is:

In the formula, V is the surface water catchment, in m ³ ; d is the water depth of the sub-catchment, in m; t is the time, in s; A is the area of the sub-catchment, in m ² ; Rain intensity, the unit is mm/s; Q is the outlet flow of the sub-catchment, the unit is m ³ /s.

The Manning equation is:

In the formula, Q is the outflow of the sub-catchment, in m ³ /s; W is the characteristic width of the sub-catchment, in m; S is the average slope of the sub-catchment; N is the Manning coefficient.

Combine the water balance equation and the Manning equation, and use the finite difference method to solve:

In the formula, t is the time step, and the unit is s; d ₁ is the water depth value at the beginning of time t, and the unit is m; d ₂ is the water depth value at the end of time t, and the unit is m.

Use Horton's formula to find the value of i, then solve the above formula with iterative method to obtain the value of d ₂ , and then substitute it into Manning's equation to get the outflow Q ₂ at the end of the period.

The SWMM model covers three pipe network confluence methods, including: constant flow method, dynamic wave method and kinematic wave method. The dynamic wave method can simulate more complex water flow conditions, and can better solve the countercurrent situation in the drainage network, and has good universality for underground drainage networks of different shapes and sizes. Generally speaking, the drainage pipe network in urban areas is relatively complex, so this paper uses the dynamic wave method for simulation in the calculation.

The principle of the dynamic wave method is to solve the complete Saint-Venant equations for confluence calculation. Saint-Venant's equations contain three equations, which are pipe continuity equation, pipe momentum equation and node control equation.

Pipeline continuity equation:

In the formula, Q is the flow rate in the pipe network in m ³ /s; S is the cross-sectional area of the pipe network in m ² ; t is time in s; l is the distance in m.

Pipe momentum equation:

In the formula, H is the water depth in the pipeline, in m; g is the acceleration of gravity, generally 9.8m/s ² ; S _f is the resistance coefficient in the pipeline.

Node governing equation:

In the formula, Q _t is the flow in and out of the node, the unit is m ³ /s; Ask is the bottom area of the node, the unit is m.

S302 Construct and obtain the surface depression confluence model

First, with the help of the fill tool in hydrological analysis in ArcGIS, set different filling thresholds for high-precision DEM data (plane resolution of 2m, vertical height of 0.1m) to perform two calculations, and then use the Minus tool to use the second step Subtract the DEM layer of all the depressions filled in the first step and the DEM layer with a filling threshold of 0.1m obtained in the first step to extract the low-lying areas that are prone to water accumulation, and finally calculate the volume of the depressions and perform preliminary deletion processing . The formula for calculating the volume of a swale is as follows:

V _wa =S _wa ×h _mean

In the formula, V depression is the volume of the depression, the unit is m ³ ; S depression is the area of the depression, the unit is m ² ; h _mean is the average depth of all pixels within the extracted depression range, the unit is m.

Afterwards, the overflow points of the water lifting area and depressions are extracted. First find the outlet point, which is the lowest point of the depression's catchment area. Combined with the flow direction of the water flow, all grids flowing through the water outlet in the upstream area of the water outlet point are determined within the entire large watershed to complete the extraction of the water lifting area.

The overflow point of the depression needs to calculate the flow direction of the water flow in the DEM data first, and then calculate the cumulative amount of the confluence according to the direction of the water flow, so as to obtain the position of the overflow point of the depression.

The sink path for the swale is then generated. After the depressions and overflow points are extracted, the confluence path between the overflow points of the connected depressions can be obtained by extracting the shortest path extraction tool. With the help of the cost path tool in GIS spatial analysis, the minimum cost path from the start point to the end point is obtained through the calculation of the shortest path function.

Finally, using the SWMM model, the overflow point is generalized as a storage tank (the bottom area of the storage tank is the area of the depression; the height of the storage tank is the average depth of the depression), and the confluence path of the depression is generalized as a pipe channel. The extracted depression sub-catchments are generalized into surface water catchment areas, so as to construct the basic model of "overflow point-confluence path-sub-catchment area". At the same time, with the help of the scale tool in ArcGIS, the depth of the depression is calculated as the simulated ponding depth.

Rational Analysis of S303 Flood Risk Model

Firstly, the river flow process is verified by local hydrological data. Afterwards, the waterlogging depth was verified by using the parameters in the post-production and confluence model of the river flow verification, and the maximum 1-hour rainfall data in 10 years was input into the flood risk rapid identification model for simulation, and the model simulated water depth results were obtained. The water risk point account is compared with the simulation situation of the refined flood model to verify the reliability of the model simulation.

S4 rainfall prediction model coupled with flood risk model to get flood risk prediction model

The rainfall prediction value output by the rainfall prediction model obtained in step S1 is preprocessed as the rainfall input value of the flood risk model in step S3; the rainfall prediction model is coupled with the flood risk model to obtain a flood risk prediction model.

S5 inputs the real-time radar echo map of the city to be predicted into the flood risk prediction model to obtain the analysis result of the flood risk of the city to be predicted.

Example 2

In this example, the method in Example 1 is applied to the Qingyang River Basin in Beijing to construct a watershed-scale urban flood risk rapid identification model to predict urban waterlogging.

S1 constructs the rainfall prediction model of the city to be predicted:

The construction method is the same as that in Example 1, in which four rainfall events on July 22, July 29, 2019, and July 31, 2020, and August 12, 2020 in the Beijing area were selected for the construction of the rainfall data of 48 scenes of radar reflectivity The forecasting model carries out rainfall forecasting, inversion and precision analysis.

The radar echo image data in this embodiment comes from China Weather Network. The radar is located in the Yizhuang area of Beijing. The time resolution of the data is 6 minutes, the effective radius is 230 km, and the spatial resolution is 1 km×1°. The data details are shown in Table 1. The measured rainfall data in this paper come from the Beijing Meteorological Department, involving 460 rainfall stations. The time resolution of the measured rainfall data is 5 minutes, and the time scale of rainfall is 1 hour.

Table 1 Brief table of radar echo data

Table 2 Comparison table of the situation after two different ways of processing the missing data

In the case that the original generation radar missing image is missing, it is replaced by an adjacent image image. Table 2 analyzes the situation of replacing the latter image with the previous image and replacing the former image with the latter image when the middle image is missing when the radar echo image is used to predict rainfall in four rainfall events. After a comprehensive comparison of the deviation BIAS, root mean square error RMSE, correlation coefficient r, and relative error calculation results under the two processing conditions, the precision of replacing the former scene image with the latter scene image in T1-T3 scenes is higher, so When there are missing images in the forecast, the latter method is selected for processing.

Near-term rainfall forecast: Based on the meteorological radar echo map and the pyramid optical flow method, four rainfall events in the flood seasons of 2019 and 2020 were forecasted in the near-term, with a prediction period of 1 hour and a time resolution of 6 minutes. According to the extracted radar echo image area with reflectivity greater than or equal to 25dBZ and the predicted data area with inverted rainfall intensity greater than or equal to 1.54mm/h, the accuracy analysis of the falling area is carried out. The average accuracy of the four rainfall area predictions is 60.0%. Extraction The magnitude accuracy analysis is carried out on the measured results and inversion results of accurate prediction sites in the region, and the average accuracy of magnitude prediction is 62.79%.

S2 basic data collection and preprocessing:

Through the collection and arrangement of various types of data and data preprocessing, the basic data are obtained as shown in Table 3.

Table 3 Basic data of Qingyang River Basin model construction

(1) Image data preprocessing

Based on the Rapid Eye image data without cloud cover, the present invention interprets and analyzes the land use type of the underlying surface of the Qingyang River Basin through automatic interpretation and extraction of software and manual visual interpretation, thereby further calculating the underlying surface of the basin of impermeability. The underlying surface distribution of the Qingyang River Basin was extracted through image preprocessing, software automatic extraction, and manual correction.

(2) Calculation and extraction of underlying surface type

According to the attribute characteristics of different ground objects in the image, Yikang software was used to extract and calculate, and the Qingyang River Basin was divided into 16 types of underlying surfaces such as construction land, roads, and water systems.

Table 4 Statistical statistics of the area ratio of various land features in the Qingyang River Basin

(3) Statistics on classification results of underlying surface of Qingyang River

The underlying surface types of the Qingyang River Basin were extracted using the above method (Table 4). Among them, the impermeable area is 17.96km ² , accounting for 59.27%; the permeable area is 12.34km ² , accounting for 40.73%. Based on remote sensing image extraction, the distribution of various surface features in the Qingyang River Basin is shown in Figure 1, and the distribution of permeable and impermeable water is shown in Figure 2.

Since the data accuracy of the short-duration rainfall forecast is higher than that of the long-duration rainfall forecast, according to the actual situation of flood control emergency management, this paper uses the 1-hour short-duration design rainstorm that occurs once in 10 years when simulating and analyzing the waterlogging situation; while the river catchment The time confluence process is more complex, the confluence area is large and the confluence time is long, so the 24-h design long-duration design rainstorm that occurs once in 20 years is used in the simulation analysis of river flow.

(4) Design storm calculation

When calculating the 1h short-duration design rainstorm process, the rainstorm volume calculation adopts the rainstorm intensity formula in the "Calculation Standard for Stormwater System Planning and Design of Urban Rainwater System" (DB11/T969-2016), the design rain type selects the Chicago rain type, and then refers to the "Water Supply and Drainage Design According to the relevant content in the Handbook, the time course allocation of the rainfall process is carried out. Since this study only needs to use the 10-year once-in-1-hour data, the short-duration design rainstorm only calculates the 10-year once-in-a-year value.

The Beijing rainstorm intensity formula is calculated using the annual maximum method. According to the administrative region as the division unit, the western region is divided into the heavy rain zone I, and the eastern region is divided into the heavy rain zone II. The calculations correspond to two different rainstorm intensity formulas. The study area of this paper (Qingyang River Basin) belongs to Zone II, and its design storm intensity formula is as follows:

适用范围为：5min≤t≤1440min

The scope of application is: 5min≤t≤1440min

In the formula, q is the design rainstorm intensity in L(s·hm ² ); P is the return period of the design rainstorm in a unit; t is the design rainstorm duration in min.

After substituting the design storm calculation formula, the time history distribution calculation of the rainfall process is carried out according to the method in the "Water Supply and Drainage Design Manual", and according to the long-term actual observation data, the value of the rain peak coefficient r is taken as 0.167, and the formula is as follows:

In the formula, i is the rainfall intensity, the unit is mm/min; t _b represents the rainfall duration before the rainfall peak, the unit is min; t _a represents the rainfall duration after the rainfall peak, the unit is min; r is the rain peak coefficient.

When calculating the 24h long-duration design rainstorm process, it is calculated according to the "Beijing Hydrology Manual - Rainstorm Atlas", and the design rain type is selected as the two-shower rain type. -2016) in the 24h every 5min rain pattern distribution table for rainfall process distribution. Since this study needs to use the 10-year and 20-year design rainstorm data with a maximum of 24 hours, two sets of data were calculated for the long-duration design rainstorm. The short-duration and long-duration design rainstorm process lines are shown in Figure 3 and Figure 4, respectively.

S3 builds a SWMM-based flood risk prediction model for cities to be predicted:

S301 River Confluence Submodel

(1) Generalization of node pipe network

Construct the sub-model of the pipe network based on the underground drainage pipe network data and terrain elevation data of the Qingyang River Basin, check whether the upstream and downstream connections are correct according to the connection of the rainwater pipelines, and check the shape, elevation and bottom elevation of the upstream and downstream pipes , the height of the interface from the bottom of the pipe, etc.; check the logical relationship between the node pipe network connection, and the slope and other related information. The generalization of the underground drainage network should follow the principle of reducing complexity and simplifying, that is, deleting the pipeline paths in the study area that have nothing to do with the drainage function or have little impact on it, leaving only the most important part of the rainwater network, and checking the elevation and Matching of coordinate positions. Underground drainage pipe network data often have errors such as reverse slope, disconnected pipes, or isolated points. Therefore, it is necessary to carefully check the connection elevation of the upstream and downstream of the pipes to ensure that the underground drainage pipes, inspection wells and inspection wells The connection logic relationship is correct. After the model is constructed, it is obtained that there are 1921 nodes in the Qingyang River Basin, 1913 sections of the drainage pipe network, and a total length of 74.25km. The generalization of the constructed pipe network is shown in Figure 5.

(2) Division of sub-catchments

The division of sub-catchments needs to refer to the distribution of pipe networks in the study area, topography, underlying surface and other data. The division of sub-catchments considers the time conditions of the study area, and is divided according to the topographic features of the study area, the layout of underground pipe networks, and the principle that the water in the sub-catchments is discharged to the nearest river. The invention combines two methods of Thiessen polygon method and manual interpretation to divide the watershed into sub catchment areas. The first step is to manually divide the generalized inspection well catchment area based on information such as underground drainage pipelines in the watershed, the channel of the Qingyang River, the slope of the study area, and the proportion of impervious water, combined with remote sensing images. In the second step, the molecular catchment is further automated through the Thiessen polygon method based on the manually demarcated catchment. After the above steps, it can be obtained that the study area is divided into 1745 sub-catchments, with an area of 0.002-3.99ha, and the average area of sub-catchments is 0.38ha. The division results of nodes, pipe network, river course and sub catchment area are shown in Figure 6.

The GIS software simplifies the channel data of the Qingyang River; imports the processed channel data into SWMM, and checks and checks the logic and topological relationship of the node pipe network to build a sub-model of the channel confluence. The upper boundary of the channel is the starting point of Qingyang River, and the outlet boundary is where Qingyang River enters the mouth of Qinghe River.

S302 Construct and obtain the surface depression confluence model

(1) Depression extraction

With the help of the filling tool in hydrological analysis in ArcGIS, two calculations were performed with different filling thresholds for high-precision DEM data (plane resolution 2m, vertical height 0.1m). For the first time, use the filling tool to set the filling threshold to 0.1m, that is, to fill the depressions with a depth less than 0.1m as the error of the system itself, and keep the depressions with a depth greater than 0.1m. For the second time, use the same filling tool to fill again. At this time, without setting any filling threshold, fill all the depressions in the study area, and output the DEM data without depressions. Then, use the Minus tool to subtract all the DEM layers filled in the second step from the DEM layer with a filling threshold of 0.1m obtained in the first step to extract the low-lying areas that are prone to water accumulation, and finally calculate The volume of the swale.

Because the cell depth is the layer obtained by subtracting two different threshold fillings, so when calculating the sum of cell depths, it is necessary to first calculate the difference of each pixel, and then calculate the sum of all pixels. The cell area size is equal to the area size of a single cell multiplied by the total number of cells in the swale. In this study, because the planar resolution of the selected high-precision DEM data is 2m, the size of a single grid cell is 2m×2m during calculation.

After the depressions are extracted through the above-mentioned filling process, the depressions need to be preliminarily screened, that is, to remove areas with too small an area and water bodies such as lakes and rivers.

(2) Depression water catchment extraction

When extracting a depression catchment, the first step is to find the outlet point, which is the lowest point of the depression catchment. In the second step, combined with the flow direction of the water flow, determine all the grids that flow through the outlet in the upstream area of the water outlet within the entire large watershed. Pay attention to searching every corner of the watershed when searching to avoid missing data.

First use the catch overflow point tool in the hydrological analysis to search for the grid points with high accumulation of confluence within a certain range, then use the watershed tool to extract the catchment sub-basin corresponding to each depression, and calculate the sub-watershed corresponding to each depression catchment area.

(3) Extraction of depression overflow points

A swale overflow point is the point at which rainwater overflows from the edge of a swale. When the potential water accumulation area in the area (i.e. the depression area) and the corresponding cumulative flow of each depression are calculated, the position of the overflow point of the depression can be extracted by calculation, that is, the point where the maximum cumulative flow of water in a depression is located Location. When calculating the overflow point of the depression, it is necessary to calculate the flow direction of the water flow in the DEM data first, and then calculate the size of the cumulative flow according to the flow direction. The water flow direction of each grid in GIS is the direction when the water flows out of the grid, and the eight grids around the central grid are coded, and the numbers they represent are assigned to obtain the value of the water flow direction.

(4) Generation of confluence paths in depressions

During a rainfall process, the rainwater falls to the surface and collects in the low-lying places first. When the depression is filled, the water in the depression will overflow from the edge. This point is called the overflow point of the depression, as shown in Figure 8 (a), (b) shown. Through calculation, comparison and screening, find the point with the largest confluence accumulation in the depression area, which is the overflow point. When the depression is full, the water overflows from the overflow point, leaves the depression, and continues to flow to the next depression with a lower terrain. The entire process of continuous filling and flow of rainwater is generalized into a network, as shown in Figure 8 (c ), (d), and finally go through the whole research area. This method can quickly and accurately extract the location of the depression area (potential water accumulation area), and is easy to operate.

After the depressions and overflow points are extracted, the confluence path between the overflow points of the connected depressions can be obtained by extracting the shortest path extraction tool. With the help of the cost path tool in GIS spatial analysis, the minimum cost path from the start point to the end point is obtained through the calculation of the shortest path function. This method can find the fastest and most efficient route between two points, and is a method frequently used in optimal route selection. Its working principle is to obtain a pixel-wide grid network through the direction information of the overflow point in each flow direction grid, which can be converted into a downstream direction-determined curve through this tool.

The distribution of depressions, overflow points and confluence paths in the Qingyang River Basin extracted through the above steps is shown in Figure 9. Because this study mainly studies the areas that have an impact on people's lives within the city, focusing on roads and construction areas, the depressions with an area of less than 0.3ha, water areas, and green areas in the research area were removed, and Qingyang was finally screened. There are 46 effective depressions in the river area, with a total area of 0.32km ² ; there are 64 confluence paths between depressions, with a total length of 20.18km.

Afterwards, generalize the depression confluence network system in SWMM, and generalize the overflow point into a storage tank in the SWMM model (wherein, the bottom area of the storage tank is the area of the depression; the height of the storage tank is the average depth of the depression), and the The confluence path in the depression is generalized into pipes and canals, and the extracted depression sub-catchment is generalized into the surface catchment area, so as to construct the basic model of "overflow point-confluence path-sub-catchment area".

S303 Subcatchment Parameter Calculation

(1) Calculation of the slope of the subcatchment

The slope of the subcatchment will affect the speed of water flow, which will have a great impact on the model slope confluence. The change of slope will change the flow rate and confluence time of the slope confluence, as well as the flow hydrograph of the outlet section of the watershed, thereby reducing the Accuracy and realism of model simulations. The high-precision DEM data used in this study comes from the results of the first water affairs census in Beijing, and the precision of the elevation data is 2m. Load the high-precision DEM data of the Qingyang River Basin into the GIS software, use the slope tool in the surface analysis to extract the slope, and obtain the ground slope value in the entire Qingyang River Basin. Then, the average slope value of each sub-catchment is extracted with the help of the zoning statistics tool in the regional analysis function, and then the slope value of each sub-catchment is matched with each sub-catchment to obtain the The slope value after division, and then generate the average slope map of the sub-catchments in the study area shown in Figure 10. The average slope value of the smallest sub-catchment is 1.16%, and the maximum slope value is 20.77%.

(2) Calculation of impermeability in sub-catchments

The impermeability of the subcatchment is calculated according to the land use type of the underlying surface in the study area. In this study, the impermeable rate was calculated based on the data of different land use types on the underlying surface of the Qingyang River Basin. The types of land used for the underlying surface of the Qingyang River Basin include: roads, vegetation on both sides of roads, cultivated land, public green spaces, park green spaces, river courses, lakes, construction land, pits and ponds, bare soil, ordinary houses, pools, parking lots, and unused areas There are 16 kinds of roof greening and community greening, which are divided into two categories: permeable and impermeable. The value of impervious is 1, and the value of permeable is 0. Each sub-catchment area is counted by the partition statistics tool in GIS The average impervious rate of each sub-catchment, and then the average impermeable rate value of each sub-catchment area is corresponding to the serial number of each sub-catchment area, and the average impermeable rate value after dividing according to the sub-catchment area is obtained, and finally the graph is generated 11 shows the proportion of impermeability ratio of the sub-catchments in the study area. After calculation, the average impervious rate of the smallest sub-catchment is 23.79%, and the largest average impermeable rate is 100%.

(3) Determination of relevant parameters of sub-catchments

Other relevant parameter values of the subcatchment, such as the Manning coefficient of the permeable area and the impermeable area, the depth of the depression, the maximum and minimum infiltration rates, etc., are generally not obtained by direct calculation, but by reference Based on previous research conclusions, or select the initial value from the SWMM model manual for simulation, and then determine the final value through repeated parameter adjustment. The value of the overland flow parameter will affect the accuracy of the model calculation results. In combination with the operability of the actual area, the following formula is selected in the present invention to calculate the overland flow width of the sub-catchment area:

Width＝k×Sqrt(Area)(0.2＜k＜5)

Among them, Width is the width of the sub-catchment, the unit is m; Area is the area of the sub-catchment, the unit is m2. The value of K is 2. The generalized model in SWMM is shown in Figure 12.

S304 Model parameter test

Input the design rainfall data once in 20 years into the model for simulation, test the model parameters, and compare the simulation results with the design flow process line. If the two curves have a large difference, readjust the parameters for simulation until the simulated flow rate is obtained. The hydrograph is close to the design flow hydrograph and the Nash coefficient of the two sets of data is close to 1. According to Guo Qiqi’s research, in the SWMM model, the Manning coefficient of the impermeable zone and the permeable zone has a greater impact on the peak flow rate, which is a sensitive parameter; the maximum infiltration rate has little effect on the total flow rate, and is an insensitive parameter. Refer to this rule to calibrate the parameters. After inputting the initial selection values of the parameters in the model user manual in SWMM, input the rainfall events to simulate the model. Then, the simulated discharge hydrograph at the outlet of the river channel and the calculated hydrological manual For comparison, the flow process line calculated in the hydrological manual is used as a reference, and the simulated flow process line is as close as possible to the reference flow process line through repeated fine-tuning near the initial parameter value, and the final parameter of the model after calibration See Table 5 for the values.

Table 5 Selection table of relevant parameters of sub-catchments

Finally, calculate the depth of the depression in ArcGIS. After inputting the field rainfall data through SWMM software, the depression confluence network system can simulate the actual water storage volume V ₂ in the storage tank, and compare the simulated actual water storage volume V ₂ with the originally extracted depression. Compared with the total volume V ₁ , the water storage volume ratio V ₂ /V ₁ of each depression after the rainfall can be obtained, and then use the proportional tool in ArcGIS to reduce the volume of the depression according to the ratio V ₂ /V ₁ , and then extract The average depth value of each depression, this average depth value is the simulated ponding depth of the rainfall on the ground for this session.

Rational Analysis of S305 Flood Risk Model

(1) River flow process verification

Due to the lack of measured flow data in the study area, under the design rainfall once in 20 years, the flow hydrograph at the channel outlet section calculated by the instantaneous unit line method in the hydrological manual was compared with the flow hydrograph obtained by the SWMM model simulation. . When using the hydrological manual method to calculate the flow process line at the river outlet section, the design flow process is calculated using the instantaneous unit line, and the infiltration loss in each period is deducted by the runoff coefficient method. Refer to the Beijing Hydrological Manual (Second Volume Flood Chapter) 》, according to the area and related attributes of the study area, the confluence parameter n is 1.5, and k is 1.2. After parameter calibration, under the condition that the design rainfall is once in 20 years, the flow hydrograph obtained from the SWMM model simulation and the hydrograph calculated from the hydrological manual are shown in Figure 13. Among them, the flood peak flow at the outlet of the basin calculated once in 20 years is 79m ³ /s, and the flood peak flow simulated by the constructed flood risk rapid identification model is 63.6m ³ /s, with a relative deviation of 19.50% and a Nash coefficient of 0.86. It can be seen from Figure 13 that the peak time and the shape of the flow hydrographs calculated by the two methods are basically the same, and the flood peak flow of the hydrological manual method is larger than the simulated flow of the model. It can be seen that the simulation results of the model are reasonable and reliable, and this parameter can be used in the construction of the model of the sink confluence system. (2) Verification of waterlogging depth

First, verify the waterlogging depth, use the river flow to verify the parameters in the confluence model, and input the maximum 1-hour rainfall data in 10 years into the flood risk rapid identification model for simulation, and obtain the model simulated water depth results, according to "Qinghe Basin Compared with the simulation situation of the refined flood model, the deviation is shown in Table 6, and the simulation situation of rainfall and waterlogging is shown in Figure 14.

Table 6 Error table of water depth simulated by this model and simulated water depth by refined flood model after calibration of parameters

It can be seen from Table 6 that in the 10-year maximum 1-hour rainfall event, the simulated value of the W9 waterlogging point is 0.051m different from the waterlogging depth value simulated by the refined flood model, and the error rate is 13.25%; the W33 waterlogging point The difference between the simulated value and the water depth value simulated by the refined flood model is 0.026m, and the error rate is -12.68%; the difference between the simulated value of W44 water point and the water depth value simulated by the refined flood model is 0.002m, and the error rate is 1.12 %. It can be seen from the calculation results that the simulation results are reasonable, and the constructed urban flood risk rapid identification model can better simulate the waterlogging situation in the depressions in the study area.

S4 rainfall prediction model coupled with flood risk model to get flood risk prediction model

The rainfall prediction value output by the rainfall prediction model obtained in step S1 is preprocessed as the rainfall input value of the flood risk model in step S3; the rainfall prediction model is coupled with the flood risk model to obtain a flood risk prediction model.

S5 inputs the real-time radar echo map of the city to be predicted into the flood risk prediction model to obtain the analysis result of the flood risk of the city to be predicted.

(1) Research data

The input data in the near-term rainfall forecast comes from the radar echo map downloaded from the China Weather Radar Network, and the three-scene radar at 21:42, 21:48, and 21:54 of the rainfall on August 12, 2020 was downloaded during the rainfall forecast analysis Figure, as shown in Figure 15-Figure 17. The accumulated water data of the "2020.08.12" rainfall in the Qinghe River Basin comes from field surveys to collect accumulated water conditions. The measured rainfall data comes from the Beijing Meteorological Department, and the time resolution of the measured rainfall data is 5 minutes.

"2020.08.12" rain forecast analysis

Affected by the northwest vortex, the East China Sea and the Bay of Bengal airflow, the rainfall on August 12, 2020 was the strongest since Beijing entered the flood season. According to statistics, the city's average rainfall is 69mm, and the rainstorm center is distributed in strips from southwest to northeast. Changping, Huairou, and urban areas have relatively heavy rainfall. The urban heavy rain center is 244mm in Shijingshan, and the suburban heavy rain center is 237mm in Huairou. Most of the channels of the North Canal and a few channels such as the Baihe River, the Yanqi River, and the Juma River experienced significant flooding.

Through the near-term rainfall forecast platform, based on the pyramid optical flow method and cloud image extrapolation method, input the identified cloud type for calculation, and obtain the rainfall forecast product for the next hour. The rainfall interpolation map of the 1h total rainfall is shown in Figure 18, " Figure 19 shows the distribution of the predicted falling area and the measured rainfall falling area of the 8.12"1h near-term rainfall forecast product in the first 6 minutes and the last 6 minutes.

From the comparison of the predicted rainfall data and the measured rainfall data, it can be seen that the predicted results are basically consistent with the actual rainfall, and the predicted results in urban areas are slightly smaller. The comparison between the predicted situation and the actual situation is shown in Table 7.

Table 7 Contrast between forecasted situation and actual situation

Due to the fact that the actual survey of water accumulation points in Qingyang River is limited, the location distribution of predicted rainfall and actual water accumulation points in the Qinghe River Basin (the large watershed of the Qingyang River Basin in the study area) was analyzed. In the Qinghe River Basin, LiDAR technology was used to build a refined elevation model (planar resolution 2m, vertical height 0.1m). After terrain analysis, 170 low-lying risk areas of waterlogging were identified, and the predicted rainfall and actual water accumulation points were superimposed to obtain Figure 20. It can be seen from the figure that the actual location of the water accumulation point and the location of the heavy rainfall forecast fall area have a high degree of agreement.

(2) Analysis of forecasted rainfall, waterlogging and waterlogging in four fields

Human activities and natural factors jointly affect the runoff of urban watersheds. The rainfall-runoff relationship of urban flood is affected by the level of urban development and is closely related to the urban impervious area. The drainage capacity of underground drainage pipes in a certain design recurrence period is not only related to its design storm volume, but also closely related to its selected flood peak runoff coefficient. In general, the design drainage capacity of rainwater pipes (referring to the ability to discharge net rain) is determined. Referring to the content in the "Beijing Hydrology Handbook (Second Volume Flood)", the current flood peak runoff coefficient of rainwater pipes in urban areas of Beijing Generally, 0.55 is used. Considering the influence of the underground drainage pipe network this time, it is necessary to preprocess the rainfall data first.

Input the four rainfall data obtained in the near-term rainfall forecast into the built waterlogging risk rapid identification model to simulate and calculate the water accumulation depth. Through the four rainfall simulation calculations, it can be obtained that the minimum water accumulation depth of the T1 rainfall is 0.01m , the maximum value is 0.36m, and the average water depth is 0.28m; the minimum water depth of the T2 rainstorm is 0.02m, the maximum value is 0.34m, and the average water depth is 0.14m; the T3 rainwater depth is the smallest The value is 0.03m, the maximum is 0.79m, and the average water depth is 0.29m; the minimum water depth of T4 rainfall is 0.04m, the maximum is 0.79m, and the average water depth is 0.33m. The water accumulation conditions of the four rainfall events are shown in Table 8, and the schematic diagrams of the simulated water accumulation conditions are shown in Figures 21 to 24.

Table 8 Simulated accumulation of water in the four predicted rainfall events

Based on the constructed urban flood risk rapid identification model, the forecast situation of "8.12" rainfall in Beijing in 2020 was analyzed from the perspective of rainfall prediction. The flood risk rapid identification model can predict the location of the main rainfall area, and the rainfall estimation accuracy of the main rainfall segment is high, and the flood risk can be analyzed according to the rainfall.

Claims (10)

1. A city flood risk prediction method is characterized by comprising the following steps:

(1) Based on the historical radar echo map of the city to be predicted, preprocessing the radar echo map through a GIS (geographic information system), reclassifying and calculating, and constructing a rainfall prediction model of the city to be predicted;

(2) Collecting basic data of a city to be predicted, and preprocessing the collected basic data;

(3) Constructing a flood risk prediction model of the city to be predicted based on the SWMM according to the basic data preprocessed in the step (2);

generalizing a node pipe network of a city, dividing a sub-catchment area, calculating data, constructing a basin runoff generating model and a pipe network confluence model, simplifying local river channel data through ArcGIS software, introducing the simplified river channel data into an SWMM, checking and checking the logic and topological relation of the node pipe network, and constructing a river channel confluence sub-model;

extracting the depression and the overflow points, and calculating the area of the sub-catchment area corresponding to each depression; a confluence path communicated with all overflow points of the depression is obtained through extraction of a shortest path extraction tool; the depression confluence path is generalized into a pipe channel, and the extracted depression sub-catchment areas are generalized into surface catchment areas, so that a surface depression confluence model is constructed and obtained;

coupling the river channel confluence sub-model and the surface depression confluence model to obtain an flood risk model of the city to be predicted;

(4) Coupling the rainfall prediction model and the flood risk model to obtain a flood risk prediction model; preprocessing a rainfall prediction value output by the rainfall prediction model obtained in the step (1) to be used as a rainfall input value of the flooding risk model in the step (3);

(5) And inputting the real-time radar echo map of the city to be predicted into the flood risk prediction model to obtain a flood risk analysis result of the city to be predicted.

2. The urban flood risk prediction method according to claim 1, wherein in the step (1), the method for constructing the rainfall prediction model of the city to be predicted comprises: based on the historical radar echo map of rainfall of multiple times, preprocessing and reclassifying the radar echo map through a GIS (geographic information system), and dividing the radar echo map into a high-reflectivity area, a medium-reflectivity area and a low-reflectivity area; extracting characteristic values of all cloud layer types and identification rules of rainfall types; and constructing a rainfall type identification index system according to the spectral characteristics of each reflectivity of the radar echo map.

3. The method of urban flood risk prediction according to claim 2, characterized in that the high-reflectivity zone ranges from 35-70dBZ, the medium-reflectivity zone ranges from 25-35dBZ, and the low-reflectivity zone ranges from 0-25dBZ; the constructed rainfall type identification index system comprises the following steps:

(1) (area of high reflectivity region + area of medium reflectivity region)/total area of echo map is less than or equal to 0.1, and layer cloud rainfall occurs;

(2) (area of high reflectivity region + area of medium reflectivity region)/total area of echo map is not less than 0.3, and convection cloud rainfall is carried out;

(3) 0.1< (area of high reflectivity region + area of middle reflectivity region)/total area of echo diagram <0.3 and brightness gradient <0.15, mixed cloud rainfall mainly comprising layered cloud;

(4) 0.1< (area of high reflectivity region + area of middle reflectivity region)/total area of echo diagram <0.3 and brightness gradient ≥ 0.15, and mixed cloud rainfall mainly comprising convection cloud.

4. The urban flood risk prediction method according to claim 1, wherein in the step (3), rainfall runoff of each sub-catchment area is calculated in the runoff yield model, the sub-catchment areas comprise a water-permeable area, a water-storing and water-impermeable area and a water-storing and water-impermeable area, and the total runoff yield of the surface of the watershed is expressed by a formula:

R＝R ₁ +R ₂ +R ₃

in the formula, R is total output flow rate and unit mm; r is ₁ The unit is mm, which is the output flow of the permeable area; r ₂ The unit is the ground surface production flow of the water-impermeable area with hollow storage; r is ₃ The surface flow rate of the water-tight area without hollow storage is in mm.

5. The urban flood risk prediction method according to claim 4, wherein in the step (3), the parameters of the sub-catchment areas comprise water impermeability rates, the urban river basin underlying surface land utilization type to be predicted is analyzed through interpretation of the preprocessed remote sensing image data, and the water impermeability rates of the river basin underlying surface are calculated according to the underlying surface land utilization type.

6. The urban flood risk prediction method according to claim 5, wherein in the step (3), the parameters of the sub-catchment areas further comprise a slope value Manning coefficient, a hollow water storage depth and a infiltration rate, wherein the slope value is extracted after the DEM data processing; and (3) determining the final values of the Mannich coefficient, the depression water storage depth and the infiltration rate by substituting the SWMM model and repeatedly adjusting parameters.

7. The urban flood risk prediction method according to claim 1, characterized in that in step (3), the method for extracting the depression is: different filling thresholds are set for high-precision DEM data through a hole filling tool in hydrological analysis in ArcGIS, calculation is carried out twice, a first calculation result and a second calculation result are obtained respectively, then a Minus tool is used for subtracting all hole filling DEM layers in the second calculation result from the DEM layers with the filling threshold of 0.1m obtained in the first calculation result, namely, a hole area easy to accumulate water is extracted, and finally the hole volume is calculated, wherein the hole volume formula is expressed as:

V _{hollow-shaped} ＝S _{Hollow-shaped} ×h _mean

In the formula, V _{Hollow-shaped} Is the volume of the depression in m ³ ；S _{Hollow-shaped} Is the area of the depression in m ² ；h _mean Is the average value of the depths of all the pixels in the range of the extracted depression and has a unit of m.

8. The urban flood risk prediction method according to claim 7, wherein in the step (3), the sub-catchment areas corresponding to the depressions are extracted by: firstly, the lowest point of the depression is found, and then all grids flowing through the lowest point in the upstream area of the lowest point are determined in the whole large watershed range by combining the flowing direction of the water flow so as to complete the extraction of the sub catchment area.

9. The urban flood risk prediction method according to claim 7, wherein in step (3), the method for extracting overflow points comprises: the flowing direction of water flow in DEM data is calculated, and then the confluence cumulant of the water flow is calculated according to the flowing direction of the water flow, so that the position of a depression overflow point is obtained.

10. The method of claim 1, wherein the step (3) includes performing parameter testing on the constructed flood risk prediction model, inputting 20-year-round design rainfall data into the model for simulation, comparing the simulation result with the design flow process line, and if the difference between the two curves is large, readjusting the parameters for simulation until the simulated flow process line and the design flow process line are close and the nash coefficients of the two sets of data are close to 1.

Images