A dual model coupled flood forecasting method and system

Abstract

This invention provides a dual-model coupled flood forecasting method and system, relating to hydrological forecasting, including: calculating the basin-wide average rainfall using the Thiessen polygon method based on real-time rainfall and meteorological grid forecast data; calculating Pa value, runoff generation, and confluence based on the basin-wide average rainfall using the API model to obtain the channel outlet cross-section flow and the first forecast water level; converting the channel outlet cross-section flow data calculated by the API model into the boundary conditions of the hydraulic model based on a "coupled time step synchronization mechanism" and a "boundary condition spatial interpolation algorithm"; establishing a hydraulic model of the channel to be forecasted, solving the hydraulic model based on the boundary conditions to obtain the second forecast water level; and after obtaining measured water level data, using an ensemble Kalman filter algorithm to jointly invert and correct the Pa value of the API model and the roughness n value of the hydraulic model to update the model parameters in real time. The dual-model coupling method improves the accuracy of flood water level forecasting.

Description

Technical Field

[0001] This invention relates to the field of hydrological forecasting technology, and more specifically, to a flood forecasting method and system with dual-model coupling. Background Technology

[0002] In hilly areas, floods in small and medium-sized watersheds rise and fall rapidly, with short flood lead times, making forecasting difficult. Existing technologies typically employ the following methods: "black box" regression models or empirical models: these establish statistical relationships between rainfall and runoff based on historical data, lacking physical mechanisms, having poor extrapolation capabilities, and resulting in large forecast errors over long lead times.

[0003] Single hydrological models (such as the unit hydrograph method and the Xin'anjiang model) only simulate the runoff generation and confluence processes of the watershed, failing to describe the spatial morphology of flood evolution within the river channel and providing information on water level and discharge processes at various cross-sections along the river. Pure hydrodynamic models, based on the Saint-Venant equations, discretize river cross-sections and input upstream and downstream boundary conditions to solve for the spatiotemporal variations of water level and discharge. These models can accurately describe the evolution of floods within the river channel, but they are highly dependent on boundary conditions (requiring accurate measured water level and discharge data) and cannot handle watershed runoff generation (requiring a runoff generation model; current coupling methods are typically manual input and lack automation).

[0004] A simple hydrodynamic model is the closest existing technology to this invention. Its core governing equations are the Saint-Venant equations. By discretizing river cross-sections and inputting the upstream inflow boundary (usually the measured flow rate or design flood hydrograph) and the downstream water level boundary, it solves for the changes in water level and flow rate at each cross-section over time. The main drawbacks include:

[0005] (1) Strong boundary dependence: The upstream inflow boundary requires measured flow data, but hydrological stations in high mountain and canyon areas are sparse, making it difficult to obtain reliable boundary conditions.

[0006] (2) Unable to handle runoff: The hydrodynamic model itself does not include a watershed runoff calculation module, and an external runoff model (such as API, SCS, etc.) is required. Moreover, the existing coupling methods are mostly manual file transfer, with low automation and difficulty in real-time forecasting.

[0007] (3) Difficulty in parameter calibration: When there is a lack of measured data for hydraulic parameters such as roughness, calibration relies on experience, and the accuracy of prediction is limited.

[0008] Therefore, improving the accuracy and applicability of flood forecasts has become a technical problem that needs to be solved. Summary of the Invention

[0009] The present invention aims to solve at least one of the technical problems existing in the prior art or related art, and discloses a flood forecasting method and system with dual model coupling, which can effectively improve the accuracy of flood forecasting.

[0010] The first aspect of this invention discloses a dual-model coupled flood forecasting method, comprising: data preprocessing: receiving real-time rainfall data and meteorological grid forecast data; calculating the basin-wide average rainfall using the Thiessen polygon method based on the real-time rainfall data and meteorological grid forecast data; API model calculation: calculating Pa value, runoff generation, and confluence based on the basin-wide average rainfall using the API model to obtain the discharge at the river outlet section, and calculating the forecast water level based on the discharge at the river outlet section, denoted as the first forecast water level; model coupling and boundary generation: converting the discharge data at the river outlet section calculated by the API model into the boundary conditions of the hydraulic model based on a "coupling time step synchronization mechanism" and a "boundary condition spatial interpolation algorithm"; wherein, the coupling time step synchronization mechanism specifically includes: defining the runoff generation calculation step of the API model and the evolution of the hydraulic model. The integer multiple relationship between step sizes ensures that the runoff calculation results of the API model are injected at the starting point of each evolution step size. The boundary condition spatial interpolation algorithm specifically includes: when the spatial location of the outlet section of the API model is inconsistent with that of the upstream section of the hydraulic model, a linear interpolation method based on the channel distance is used to distribute the flow of the channel outlet section of the API model to multiple upstream sections of the hydraulic model; hydraulic evolution calculation: based on real-time rainfall data and meteorological grid forecast data, a hydraulic model of the channel to be predicted is established, and the hydraulic model is solved based on boundary conditions to obtain the predicted water level of the channel section, which is denoted as the second predicted water level; joint inversion correction: after obtaining the measured water level data, the ensemble Kalman filter algorithm is used to perform joint inversion correction on the Pa value of the API model and the roughness n value of the hydraulic model, and the model parameters are updated in real time to reduce the forecast error.

[0011] The flood forecasting method using a dual-model coupling disclosed in this invention preferably further includes: adjusting the contribution ratio of the two models according to the flood forecasting stage: during the rising water period, the API model has a weight of 0.7 and the hydraulic model has a weight of 0.3; during the peak period, the API model has a weight of 0.5 and the hydraulic model has a weight of 0.5; during the receding water period, the API model has a weight of 0.3 and the hydraulic model has a weight of 0.7; triggering a weight switching operation by calculating the rate of change of the flow data at the river outlet section in real time; and merging the first forecast water level and the second forecast water level according to the weights to obtain the final forecast water level.

[0012] According to the flood forecasting method with dual-model coupling disclosed in this invention, preferably, the hydraulic model is a one-dimensional unsteady flow model, the governing equations are the Saint-Venant equations, and the solution is performed using the Preissmann implicit difference scheme.

[0013] According to the flood forecasting method with dual-model coupling disclosed in this invention, preferably, the joint inversion correction step specifically includes:

[0014] (1) Definition of state variables: The Pa value of the API model and the roughness n value of the hydraulic model are combined into a joint state vector. Where X is the joint state vector with a dimension of 2×1; Pa is the anterior rainfall (mm), ranging from 0 to Im (Im is the average water storage capacity of the basin); n is the Manning roughness coefficient (dimensionless), ranging from 0.02 to 0.15, depending on the roughness of the river channel wall; [·]^T represents matrix transpose; the joint state vector couples the runoff parameters of the API model with the river channel parameters of the hydraulic model to achieve joint inversion of parameters from both models.

[0015] (2) Observation operator: Establish a nonlinear mapping H=h(X) between the joint state vector and the measured water level H_obs, which is achieved through forward simulation of the coupled model;

[0016] (3) Set initialization: Generate multiple set members, and the state variable of each member is randomly perturbed within the prior range;

[0017] (4) Forecasting step: Each set member runs the coupled model independently to obtain the forecast water level set. ;

[0018] (5) Update step: When the measured water level H_obs is obtained, calculate the Kalman gain K and update the state vector set: Where X_a^i is the set of analyzed (updated) state vectors, i represents the set member index (i=1,2,...,N, N is usually 50~100); X_f^i is the set of predicted state vectors; K is the Kalman gain matrix (2×1); H_obs is the measured water level (m), provided by hydrological station or satellite altimetry data; H_f^i is the predicted water level, obtained by forward simulation of the coupled model; the symbol ^ represents the concept of a set, rather than an exponent or power.

[0019] (6) Convergence criterion: When the standard deviation of the set of state vectors changes by less than 5% for three consecutive iterations, convergence is determined, and the updated Pa and n are output;

[0020] (7) Real-time update: Repeat steps (4)-(6) for each new set of measured water level data received to achieve online parameter correction.

[0021] According to the dual-model coupled flood forecasting method disclosed in this invention, preferably, the real-time rainfall data specifically includes: real-time rainfall data provided by ground stations, real-time rainfall data provided by radar, and real-time rainfall data provided by satellite.

[0022] According to the dual-model coupled flood forecasting method disclosed in this invention, preferably, the data preprocessing step specifically includes:

[0023] For areas with sparse ground stations, the optimal interpolation method is used to fuse GPM / TRMM satellite precipitation data, radar quantitative precipitation estimation, and ground station data to obtain real-time rainfall data.

[0024] According to the dual-model coupled flood forecasting method disclosed in this invention, preferably, the joint inversion correction step specifically includes: extracting the river slope S and vegetation index NDVI based on DEM (Digital Elevation Model) data, and establishing an empirical relationship between roughness coefficient n and S and NDVI:

[0025] ;

[0026] Where a, b, c, and d are calibration coefficients used for the roughness initialization of hydraulic models for rivers without data.

[0027] According to the flood forecasting method with dual-model coupling disclosed in this invention, the joint inversion correction step preferably includes: using satellite altimetry data to calculate the water level time series of a river section without ground stations, and using ensemble Kalman filtering to invert and correct the water level output of the hydraulic model at that river section.

[0028] A second aspect of the present invention discloses a dual-model coupled flood forecasting system, comprising: a memory for storing program instructions; and a processor for calling the program instructions stored in the memory to implement the dual-model coupled flood forecasting method as described in any of the above technical solutions.

[0029] The beneficial effects of the present invention include at least the following:

[0030] 1. Significantly improved forecast accuracy: Compared with a simple hydrodynamic model, this invention reduces the relative error of peak flow from 15-20% in traditional methods to 5-8% through dual-model coupling and joint inversion correction, shortens the peak occurrence time error from 2-3 hours to 0.5-1 hour, and increases the Nash efficiency coefficient (NSE) from 0.65-0.75 to 0.85-0.92.

[0031] 2. High degree of automation: It achieves full automation from rainfall data access, areal rainfall calculation, runoff evolution, inversion correction to result output, without manual intervention, solving the pain point of traditional coupled models that require manual file transfer. The automated processing time is reduced from more than 30 minutes in traditional methods to less than 5 minutes.

[0032] 3. Strong applicability: For areas with no data in high mountains and canyons, the model effectively solves the problem of data scarcity and expands the application scope of the model by using technologies such as multi-source rainfall fusion, topography-roughness relationship model and remote sensing inversion of water level.

[0033] 4. Real-time correction capability: The joint inversion correction based on EnKF can update model parameters online and absorb measured information in real time, significantly improving the accuracy of subsequent forecast lead time. Attached Figure Description

[0034] Figure 1 A flowchart illustrating a dual-model coupled flood forecasting method according to an embodiment of the present invention is shown.

[0035] Figure 2 A schematic block diagram of a dual-model coupled flood forecasting system according to an embodiment of the present invention is shown. Detailed Implementation

[0036] To better understand the above-described objects, features, and advantages of the present invention, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Many specific details are set forth in the following description to provide a thorough understanding of the invention; however, the invention may be practiced in other ways different from those described herein, and therefore, the invention is not limited to the specific embodiments disclosed below.

[0037] like Figure 1 As shown, an embodiment of the present invention discloses a dual-model coupled flood forecasting method, comprising:

[0038] Step 101, Data Preprocessing: Receive real-time rainfall data and meteorological grid forecast data; calculate the average rainfall over the watershed using the Thiessen polygon method based on the real-time rainfall data and meteorological grid forecast data.

[0039] Step 102, API model calculation: Based on the average rainfall of the watershed, the Pa value, runoff generation and confluence are calculated using the API model to obtain the discharge at the river outlet section. The predicted water level is calculated based on the discharge at the river outlet section and is recorded as the first predicted water level.

[0040] In this step, the calculation of the initial impact rainfall (Pa) includes: continuous rolling calculation using the daily average rainfall data of the watershed, as shown in the following formula:

[0041] Pa t+1 =K(Pa t +P t )

[0042] Among them, Pa t+1P represents the Pa value on day t+1, K represents the decay coefficient (range 0.8~0.95), and P represents the decay coefficient. t Let Pa be the daily rainfall on day t. t This represents the Pa value on day t.

[0043] The specific calculation of the production flow includes: ;

[0044] Where R represents the production flow rate, This represents the average water storage capacity of the basin;

[0045] The flow calculation process includes: using the Sherman unit hydrograph method, the net rainfall depth R(t) is converted into the outlet cross-sectional flow rate Q(t). The unit hydrograph u(τ) is defined as the flow rate formed by a unit net rainfall (10 mm) at the outlet cross-section, which has time-shift invariance. The formula is as follows:

[0046]

[0047] in, The unit time interval, This indicates that the unit linear function is delayed in time. Units, Indicates the time period Net rain depth generated by internal accumulation.

[0048] Step 103, Model Coupling and Boundary Generation: Based on the "coupling time step synchronization mechanism" and the "boundary condition spatial interpolation algorithm", the river outlet section flow data calculated by the API model is transformed into the boundary conditions of the hydraulic model. The coupling time step synchronization mechanism specifically includes: defining an integer multiple relationship between the flow generation calculation step of the API model and the evolution step step of the hydraulic model, so that the flow generation calculation results of the API model are injected at the starting point of each evolution step step. The boundary condition spatial interpolation algorithm specifically includes: when the spatial location of the outlet section of the API model is inconsistent with that of the upstream section of the hydraulic model, a linear interpolation method based on the river distance is used to distribute the flow of the river outlet section of the API model to multiple upstream sections of the hydraulic model.

[0049] In this step, taking the July 2025 flood of the Quxi River as an example, the detailed calculation process includes:

[0050] Rainfall input: From 08:00 to 20:00 on July 15, 2025, the average rainfall in the basin was 85 mm;

[0051] API model calculation: Pa initial value 45mm, K=0.9, calculated flow rate R=42mm, and obtained outlet cross-sectional flow rate Q_API(t) by unit line convolution;

[0052] Coupling time step synchronization: API model step size Δt_API = 1 hour, hydraulic model step size Δt_H = 5 minutes, integer multiple relationship N = 12. Check for new API results at each Δt_H starting point, with a time synchronization error ≤ 30 seconds;

[0053] Boundary condition spatial interpolation: API outlet flow rate Q_out = 420 m³ / s. The hydraulic model has three upstream cross-sections at distances L1 = 2.1 km, L2 = 3.5 km, and L3 = 4.8 km, distributed inversely proportional to distance. Thus, Q1 = 84.8 m³ / s, Q2 = 141.3 m³ / s, and Q3 = 193.9 m³ / s;

[0054] Hydraulic model calculation: Solve the Saint-Venant equations to obtain the water level H(t) and flow rate Q(t) at each cross section;

[0055] Dynamic weight adjustment: During the rising water period, the API model weight α=0.7 and the hydraulic model weight β=0.3 are combined to obtain the final forecast result.

[0056] The specific calculation process of the linear interpolation method based on river channel distance includes:

[0057] Let the API outlet flow rate be Q_out, and the hydraulic model have n upstream cross-sections. Let L_i be the channel distance from the i-th cross-section to the API outlet. Then, the formula for calculating the flow rate Q_i allocated to the i-th cross-section is:

[0058] ;

[0059] In the formula, ∑L_j is the sum of the distances to all upstream cross-sections, ensuring mass conservation:

[0060] ;

[0061] Where Q_out = 420 m³ / s, the distances between the three upstream cross sections are L1 = 2.1 km, L2 = 3.5 km, and L3 = 4.8 km, then ∑L_j = 10.4 km.

[0062] The calculations yield: Q1 = 420 × (2.1 / 10.4) = 84.8 m³ / s, Q2 = 420 × (3.5 / 10.4) = 141.3 m³ / s, Q3 = 420 × (4.8 / 10.4) = 193.9 m³ / s.

[0063] Verification: 84.8 + 141.3 + 193.9 = 420 m³ / s, which satisfies the law of conservation of mass.

[0064] Specific methods for traffic allocation:

[0065] The closer the cross-section is to the API outlet, the smaller the allocated flow rate (because the water flow is not yet fully developed); the farther the cross-section is, the larger the allocated flow rate. The allocation formula is as follows:

[0066] Q_i = Q_out × (L_i / ∑L_j);

[0067] Where L_i is the river channel distance from the i-th cross section to the API outlet, and ∑L_j is the sum of the distances of all upstream cross sections. This allocation method ensures mass conservation and conforms to the physical laws of flood wave propagation along the river channel.

[0068] Step 104, Hydraulic evolution calculation: Based on real-time rainfall data and meteorological grid forecast data, establish a hydraulic model of the river channel to be predicted, solve the hydraulic model based on boundary conditions, and obtain the predicted water level of the river channel section, which is denoted as the second predicted water level;

[0069] In this step, the flow process calculated by the API model is used as the boundary condition. Combined with the river cross-section parameters, the calculation is performed from the downstream cross-section to the upstream cross-section to obtain the water level of each intermediate cross-section, thereby obtaining the forecast water level data of the key flood control areas.

[0070] Step 105, Joint Inversion Correction: After obtaining the measured water level data, the Pa value of the API model and the roughness n value of the hydraulic model are jointly inverted and corrected using the ensemble Kalman filter algorithm to update the model parameters in real time and reduce forecast errors.

[0071] According to the above embodiments, the method further includes: adjusting the contribution ratio of the two models according to the flood forecast stage: during the rising water period, the API model has a weight of 0.7 and the hydraulic model has a weight of 0.3; during the peak period, the API model has a weight of α=0.5 and the hydraulic model has a weight of β=0.5; during the receding water period, the API model has a weight of 0.3 and the hydraulic model has a weight of 0.7; triggering the weight switching operation by calculating the rate of change of the flow data at the river outlet section in real time; and merging the first forecast water level and the second forecast water level according to the weight to obtain the final forecast result (water level).

[0072] According to the above embodiment, the hydraulic model is further a one-dimensional unsteady flow model, the governing equations are the Saint-Venant equations, and the Preissmann implicit difference scheme is used for solving. Specifically, the specific method for constructing the hydraulic model includes:

[0073] Topographic data preparation: Collect river channel DEM data (resolution ≤30m), and extract the river channel centerline and cross-sectional geometric parameters (bottom width, slope coefficient, Manning roughness n).

[0074] Grid division: A calculation section is set every 500 meters along the center line of the river channel to form a one-dimensional grid. The spacing between sections is adjusted according to the river slope, and the grid is denser in steep slope sections.

[0075] Boundary condition settings: The upstream boundary is the flow process Q(t), provided by the API model, and the downstream boundary is the water level-flow curve HQ, calibrated by measured data;

[0076] Initial conditions: It is assumed that the entire river channel is in a steady flow state at the initial moment, and the water level and flow rate are determined by the design low water level, or the initial conditions of the dry riverbed are adopted.

[0077] Solver configuration: The Saint-Venant equations are discretized using the Preissmann implicit difference scheme, and the nonlinear algebraic equations are solved using the Newton-Raphson iterative method. The time step Δt = 5 minutes, and the convergence accuracy ε = 10. -5 .

[0078] According to the above embodiments, the joint inversion correction step further includes:

[0079] (1) Definition of state variables: The Pa value of the API model and the roughness n value of the hydraulic model are combined into a joint state vector. Where X is the joint state vector with a dimension of 2×1; Pa is the anterior rainfall (mm), ranging from 0 to Im (Im is the average water storage capacity of the basin); n is the Manning roughness coefficient (dimensionless), ranging from 0.02 to 0.15, depending on the roughness of the river channel wall; [·]^T represents matrix transpose; the joint state vector couples the runoff parameters of the API model with the river channel parameters of the hydraulic model to achieve joint inversion of parameters from both models.

[0080] (2) Observation operator: Establish a nonlinear mapping H=h(X) between the joint state vector and the measured water level H_obs, which is achieved through forward simulation of the coupled model;

[0081] (3) Set initialization: Generate multiple set members, and the state variable of each member is randomly perturbed within the prior range;

[0082] (4) Forecasting step: Each set member runs the coupled model independently to obtain the forecast water level set. ;

[0083] (5) Update step: When the measured water level H_obs is obtained, calculate the Kalman gain K and update the state vector set: Where X_a^i is the set of analyzed (updated) state vectors, i represents the set member index (i=1,2,...,N, N is usually 50~100); X_f^i is the set of predicted state vectors; K is the Kalman gain matrix (2×1); H_obs is the measured water level (m), provided by hydrological station or satellite altimetry data; H_f^i is the predicted water level, obtained by forward simulation of the coupled model; the symbol ^ represents the concept of a set, rather than an exponent or power.

[0084] The formula for calculating the Kalman gain K is:

[0085] ;

[0086] Among them, P - Let H be the forecast error covariance matrix (2×2), H be the observation operator Jacobian matrix (1×2), and R be the observation noise covariance (scalar).

[0087] The calculation steps include: calculating the forecast error covariance P. - =F·P·F 1 +Q (F is the state transition matrix, Q is the process noise covariance); calculate the observed predicted value H(X). - ); Calculate the Kalman gain K; Update the state vector Update error covariance P = (IK·H)·P - The physical meaning of K is: balancing the weights of predicted and observed values. A large K indicates greater confidence in the observations when the observation accuracy is high (small R); a large K indicates greater confidence in the observations when the model accuracy is high (small P). - When K is small, the model is more trusted.

[0088] (6) Convergence criterion: When the standard deviation of the set of state vectors changes by less than 5% for three consecutive iterations, convergence is determined, and the updated Pa and n are output;

[0089] (7) Real-time update: Repeat steps (4)-(6) for each new set of measured water level data received to achieve online parameter correction.

[0090] According to the above embodiments, the real-time rainfall data further includes: real-time rainfall data provided by ground stations, real-time rainfall data provided by radar, and real-time rainfall data provided by satellite.

[0091] According to the above embodiments, the data preprocessing steps specifically include: for areas with sparse ground stations, using the optimal interpolation method to fuse GPM / TRMM satellite precipitation data, radar quantitative precipitation estimation, and ground station data as real-time rainfall data.

[0092] According to the above embodiments, the joint inversion correction step further includes:

[0093] Based on DEM data, the river slope S and vegetation index NDVI are extracted, and an empirical relationship between roughness n and S and NDVI is established:

[0094] ;

[0095] Where a, b, c, and d are calibration coefficients used for the roughness initialization of hydraulic models for rivers without data.

[0096] According to the above embodiments, the joint inversion correction step further includes:

[0097] Using satellite altimetry data, the water level time series of a river section without ground stations is calculated, and the water level output of the hydraulic model at this river section is corrected by ensemble Kalman filtering inversion.

[0098] like Figure 2 As shown, according to another embodiment of the present invention, a dual-model coupled flood forecasting system 200 is also disclosed, comprising: a memory 201 for storing program instructions; and a processor 202 for calling the program instructions stored in the memory to implement the dual-model coupled flood forecasting method as described in the above embodiment.

[0099] All or part of the steps in the various methods of the above embodiments can be implemented by a program controlling the relevant hardware. The program can be stored in a readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other readable medium that can be used to carry or store data.

[0100] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A flood forecasting method coupled with dual models, characterized in that, include: Data preprocessing: Receiving real-time rainfall data and meteorological grid forecast data; Based on the real-time rainfall data and the meteorological grid forecast data, the Thiessen polygon method was used to calculate the average rainfall over the watershed. API model calculation: Based on the average rainfall of the watershed, Pa value, runoff generation and confluence calculation are performed based on the API model to obtain the discharge at the river outlet section. The predicted water level is calculated based on the discharge at the river outlet section and is denoted as the first predicted water level; where Pa means the amount of rainfall that affected the water in the previous period. Model Coupling and Boundary Generation: Based on the "coupling time step synchronization mechanism" and the "boundary condition spatial interpolation algorithm," the river outlet section flow data calculated by the API model is transformed into the boundary conditions of the hydraulic model. Specifically, the coupling time step synchronization mechanism includes defining an integer multiple relationship between the runoff calculation step size of the API model and the evolution step size of the hydraulic model, ensuring that the runoff calculation results of the API model are injected at the starting point of each evolution step. The boundary condition spatial interpolation algorithm specifically includes: when the spatial location of the outlet section of the API model is inconsistent with that of the upstream section of the hydraulic model, a linear interpolation method based on river distance is used to distribute the river outlet section flow of the API model to multiple upstream sections of the hydraulic model. Hydraulic evolution calculation: Based on the real-time rainfall data and meteorological grid forecast data, a hydraulic model of the river channel to be predicted is established. The hydraulic model is solved based on the boundary conditions to obtain the predicted water level of the river channel section, which is denoted as the second predicted water level. Joint inversion correction: After obtaining the measured water level data, the Pa value of the API model and the roughness n value of the hydraulic model are jointly inverted and corrected using the ensemble Kalman filter algorithm, and the model parameters are updated in real time to reduce the prediction error.

2. The dual model coupled flood forecasting method of claim 1, wherein, Also includes: The contribution ratios of the two models are adjusted according to the flood forecast stage: during the rising water period, the API model has a weight of 0.7 and the hydraulic model has a weight of 0.3; during the peak period, the API model has a weight of 0.5 and the hydraulic model has a weight of 0.5; during the receding water period, the API model has a weight of 0.3 and the hydraulic model has a weight of 0.

7. The weight switching operation is triggered by calculating the rate of change of the flow data at the river outlet section in real time. The first forecast water level and the second forecast water level are then combined according to their weights to obtain the final forecast water level. 3.The dual-model coupled flood forecasting method of claim 1, wherein, The hydraulic model is a one-dimensional unsteady flow model, and the governing equations are the Saint-Venant equations, which are solved using the Preissmann implicit difference scheme.

4. The flood forecasting method with dual-model coupling according to claim 1, characterized in that, The joint inversion correction steps specifically include: (1) Definition of state variables: The Pa value of the API model and the roughness n value of the hydraulic model are combined into a joint state vector X; (2) Observation operator: Establish a nonlinear mapping H=h(X) between the joint state vector and the measured water level H_obs, which is achieved through forward simulation of the coupled model; (3) Set initialization: Generate multiple set members, and the state variable of each member is randomly perturbed within the prior range; (4) Forecasting step: Each set member runs the coupled model independently to obtain the forecast water level set. ; (5) Update step: When the measured water level H_obs is obtained, calculate the Kalman gain K and update the state vector set; (6) Convergence criterion: When the standard deviation of the set of state vectors changes by less than 5% for three consecutive iterations, convergence is determined, and the updated Pa and n are output; (7) Real-time update: Repeat steps (4)-(6) for each new set of measured water level data received to achieve online parameter correction.

5. The flood forecasting method with dual-model coupling according to claim 1, characterized in that, The real-time rainfall data specifically includes: real-time rainfall data provided by ground stations, real-time rainfall data provided by radar, and real-time rainfall data provided by satellite.

6. The flood forecasting method with dual-model coupling according to claim 1, characterized in that, The data preprocessing steps specifically include: For areas with sparse ground stations, the optimal interpolation method is used to fuse GPM / TRMM satellite precipitation data, radar quantitative precipitation estimation, and ground station data as the real-time rainfall data.

7. The flood forecasting method with dual-model coupling according to claim 4, characterized in that, Also includes: Based on DEM data, river channel slope S and vegetation index NDVI are extracted, and an empirical relationship between roughness coefficient n and S and NDVI is established: ； Where a, b, c, and d are calibration coefficients used for the roughness initialization of hydraulic models for rivers without data.

8. The flood forecasting method with dual-model coupling according to claim 4, characterized in that, Also includes: Using satellite altimetry data, the water level time series of a river section without ground stations is calculated, and the water level output of the hydraulic model at this river section is corrected by ensemble Kalman filtering inversion.

9. A dual-model coupled flood forecasting system, characterized in that, include: Memory, used to store program instructions; A processor is configured to invoke the program instructions stored in the memory to implement the dual-model coupled flood forecasting method as described in any one of claims 1 to 7.

Images