Method for improving flood inundation simulation precision based on fusion of VGI data and optimization of DEM

Abstract

The invention provides a method for improving flood inundation simulation precision based on fusion of VGI data and optimization of DEM, and the method comprises the steps: selecting HEC-RAS as a flood inundation simulation model, and inputting data including DEM data, river flow data, precipitation data and land coverage type data; according to a Manning coefficient range table provided by the land coverage database, setting corresponding Manning coefficients for different coverage types, and preliminarily setting the Manning coefficients as the median of each coverage type; the Manning coefficient setting is iteratively adjusted by comparing the difference between the simulation result and the actual water depth data; establishing a relation chain between the extracted water depth information and the submerged simulated water depth; height measurement data of a land observation satellite is combined with the COPDEM to estimate a residual error between the COPDEM at a water depth point position and a real elevation, and a conversion relation is established between extracted water depth and simulated water depth of a relatively optimized DEM according to an elevation difference value between the DEM and the COPDEM. Based on the technical scheme of the invention, the simulation precision is obviously improved, the VGI data is successfully combined with the optimized DEM, and the accurate correction of the Manning coefficient is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of disaster information simulation, and in particular to a method for improving flood and inundation simulation accuracy based on fusing VGI data and optimizing DEM. Background Art

[0002] Flood simulation requires high-precision terrain, hydrology, meteorology and other data as input. However, in practical applications, these data are often difficult to obtain completely or there are errors, which affects the accuracy of the simulation results. In particular, the accuracy and resolution of terrain data directly affect the accuracy of flood simulation. If the terrain data is inaccurate or the resolution is insufficient, the simulation results may deviate greatly from the actual situation. Existing flood inundation simulation models are often highly complex and involve multiple physical processes and parameters. This makes model construction and parameter setting difficult and requires professional knowledge and skills. Insufficient representation of urban surface microtopography is an important factor affecting the accuracy of flood inundation simulation. Compared with natural watersheds, cities have complex and changeable surface features, and traditional high-precision terrain data (such as 10m resolution or even 5m resolution) are not sufficient to fully characterize the local microtopographic characteristics of urban surface areas. These microtopographic features have an important impact on the surface runoff path, so the lack of terrain data will lead to deviations between the simulation results and the actual runoff process. Based on the extracted water information, precipitation information and optimized DEM, this application effectively integrates VGI data into flood inundation simulation, and calibrates the flood simulation results based on the optimized DEM, thereby improving the accuracy of inundation simulation. Summary of the invention

[0003] In response to the problems in the above-mentioned prior art, the present application proposes a method for improving the accuracy of flood and inundation simulation based on the fusion of VGI data and the optimization of DEM. HEC-RAS is selected as the flood and inundation simulation model, and the input data includes DEM data, river flow data, precipitation data, and land cover type data; according to the Manning coefficient range table provided by the land cover database, the corresponding Manning coefficient is set for different cover types, and is initially set to the median of each cover type; the Manning coefficient setting is iteratively adjusted by comparing the difference between the simulation results and the actual water depth data; a relationship chain between the extracted water depth information and the inundation simulation water depth is established; the altimetry data of the land observation satellite is combined with COPDEM to estimate the residual of COPDEM and the actual elevation at the water depth point, and then according to the elevation difference between DEM and COPDEM, a conversion relationship is established between the extracted water depth and the simulated water depth of the relatively optimized DEM.

[0004] In one embodiment, the relationship chain between the extracted water depth information and the submerged simulated water depth is: extracted water depth - COPDEM - simulated water depth, COPDEM is a rough representation of the extracted water depth, and the simulated water depth is based on COPDEM, and there is a progressive relationship between the three.

[0005] In one embodiment, the steps of extracting the water depth and establishing a conversion relationship with the simulated water depth of the relatively optimized DEM are as follows:

[0006] Taking the water depth point as the center, select the ATL03 photon event data within the predetermined range; then use the signal_conf_ph field in the ATL03 data to screen out the 4th level high-confidence photon events as elevation control points; the photon event points within the buffer radius are thinned according to the terrain, and 50-100 high-confidence photon event points are selected for each water depth point. The more complex the terrain, the more points are selected;

[0007] At the COPDEM corresponding to the water depth point, the pixel window is used to extract the elevation value of COPDEM, and the average elevation value of the window is calculated as the COPDEM elevation corresponding to the water depth point, which is recorded as A;

[0008] Use the ATL03 point elevation minus its corresponding COPDEM elevation to obtain the residual distribution between the ATL03 control point and COPDEM, recorded as B; use the Kriging interpolation method to interpolate the elevation difference B, and use window extraction to obtain the elevation residual C at the water depth point; add the residual C to the elevation A to obtain the corrected elevation at the water depth point, and add the residual C to the water depth to obtain the water depth relative to COPDEM; then subtract COPDEM from the optimized DEM to obtain the elevation difference between the optimized DEM and COPDEM; add the water depth relative to COPDEM to the elevation difference to obtain the water depth relative to the optimized DEM, and the conversion relationship of the submerged water depth is now established.

[0009] In one embodiment, the Manning coefficient correction process includes: setting an initial Manning coefficient, running a flooding simulation model, comparing the simulated water depth with the actual water depth, calculating the RMSE of the water depth difference, adjusting the Manning coefficient according to the water depth difference, and outputting the final Manning coefficient if the Manning coefficient reaches the convergence standard; if the Manning coefficient does not reach the convergence standard, returning to running the flooding simulation model.

[0010] The above technical features can be combined in various suitable ways or replaced by equivalent technical features as long as the purpose of the present invention can be achieved.

[0011] The method provided by the present invention for improving the accuracy of flood simulation based on fusing VGI data and optimizing DEM has at least the following beneficial effects compared with the prior art:

[0012] By iteratively comparing the difference between the simulated water depth and the measured water depth, the RMSE is used as the main indicator to dynamically adjust the Manning coefficient. After multiple iterations, the simulation results of the study area are closer to the actual flood situation, and the RMSE value of the verification result has dropped to 0.32 meters, which significantly improves the simulation accuracy. Through the above research, this application has constructed an effective flood inundation simulation process, successfully combined VGI data with optimized DEM, and achieved accurate correction of the Manning coefficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The present invention will be described in more detail below based on embodiments and with reference to the accompanying drawings, wherein:

[0014] Figure 1 A graph of river discharge is shown;

[0015] Figure 2 A time trend graph of precipitation is shown;

[0016] Figure 3 A comparison of the optimized DEM and FABDEM simulation details when the flood peak arrives is shown;

[0017] Figure 4 A schematic diagram of the water depth conversion process is shown;

[0018] Figure 5 A schematic diagram of the elevation residual interpolation is shown;

[0019] Figure 6 A flow chart of the Manning coefficient adjustment is shown. DETAILED DESCRIPTION

[0020] The present invention will be further described below in conjunction with the accompanying drawings.

[0021] The present invention provides a method for improving the accuracy of flood simulation based on fusing VGI data and optimizing DEM, and selects HEC-RAS as a flood simulation model. Compared with models such as LISFLOOD and FLO-2D, the HEC-RAS model is specially designed for river hydraulics, and has the advantages of strong integration capability, good visualization effect and easy operation in river flood simulation.

[0022] HEC-RAS supports one-dimensional (1D) and two-dimensional (2D) modes. The 1D mode is mainly used to simulate the movement of water flow in the river channel. It is based on the one-dimensional Saint-Venant equation, including the continuity equation and the momentum equation. The model assumes that the water flows longitudinally in the river channel and does not consider changes in the lateral direction. The 2D mode is used to simulate the movement of water flow in the river channel and its surrounding areas. It is based on the two-dimensional shallow water wave equation and considers the changes in the longitudinal and lateral directions of the water flow. It can more finely simulate the diffusion process of floods on the plane. Therefore, the study selected the HEC-RAS2D unsteady flow model as the inundation simulation model. The HEC-RAS2D unsteady flow model simulates the dynamic changes of floods on a two-dimensional plane by solving the two-dimensional shallow water wave equation. The equation includes the continuity equation and the momentum equation. The specific form is as follows:

[0023] The continuity equation is also known as the mass conservation equation. This equation describes the conservation of mass of water in a two-dimensional plane and ensures that at any given time, the amount of water entering and leaving a control volume is balanced. The equation is expressed as follows.

[0024]

[0025] Among them: h represents the water depth, u represents the flow speed of the water in the x direction, v represents the flow speed of the water in the y direction, and t represents time. Both x and y are geographic space coordinates.

[0026] The momentum equation describes the conservation of momentum of the water in the x-direction and the y-direction. This equation takes into account the effects of gravity, bed slope, and friction.

[0027] Momentum equation (x direction):

[0028]

[0029] Momentum equation (y direction):

[0030]

[0031] Where: h represents the water depth, u represents the flow velocity of the water in the x direction, v represents the flow velocity of the water in the y direction, g represents the acceleration of gravity, H represents the height of the water surface, and n represents the Manning's nValues ​​of the riverbed or waterway surface, that is, the surface roughness coefficient. R represents the hydraulic radius, t represents the time, and x and y are both geographic space coordinates. Since the two-dimensional shallow water wave equation is a nonlinear partial differential equation, analytical solutions are usually difficult to obtain, and numerical methods are usually used to solve it.

[0032] HEC-RAS2D numerically solves these equations using the finite volume method while handling complex terrain and boundary conditions to ensure conservation of mass and momentum. The basic steps of the solution process are as follows:

[0033] 1. Spatial discretization. HEC-RAS uses the finite volume method to spatially discretize the computational domain. The simulation area is divided into a series of grid cells, each of which represents a control volume or a computational point. Depending on the terrain and the required precision of the simulation, triangular or quadrilateral grids can be selected. This method can effectively handle complex terrain and irregular boundaries.

[0034] 2. Initial condition setting. Set the initial conditions for the simulation, including the initial water depth h, water flow velocities u and v, etc. These initial conditions usually come from historical data, observational data, or hypothetical scenarios. In addition, HEC-RAS supports two modes: "cold start" (starting from a dry riverbed) and "hot start" (starting from an existing water flow state).

[0035] 3. Boundary condition processing. The model also needs to set upstream and downstream boundary conditions. The upstream boundary conditions are usually water flow process data, and the downstream boundary conditions are usually water level height process data or water level-flow relationship data. HEC-RAS uses the characteristic line method to process boundary conditions to ensure the stability of the numerical solution.

[0036] 4. Equation discretization. Equation discretization is the process of converting continuity equations into algebraic equations so that they can be solved on a computer. For the continuity equation, discretization includes integrating the control volume, expressing changes in time and space as discrete steps, and calculating water depth and flow changes to ensure conservation of mass. For the momentum equation, discretization includes integrating the change in momentum while taking into account the effects of bed slope and friction to ensure conservation of momentum. This process ensures conservation of mass and momentum in the model under complex terrain and irregular boundary conditions.

[0037] 5. Time discretization. For the time term, HEC-RAS uses a dynamic time step adjustment strategy based on the Courant number to automatically optimize the calculation step size. The Courant number is a dimensionless number used to evaluate the relative distance that water flow propagates in a single time step under given spatial grid and flow velocity conditions. The definition formula of the Courant number is shown below.

[0038]

[0039] Where u represents the flow velocity (m / s), Δt represents the time step (s), and Δx represents the spatial grid size (m). The size of the Courant number directly affects the stability of the simulation; if the Courant number is too large, it may lead to unstable calculations, and the space spanned by the water flow in each time step may exceed the grid unit, thus failing to accurately capture the flow dynamics. A too small Courant number can ensure stability, but it will significantly increase the calculation time and resource consumption.

[0040] 6. Numerical solution algorithm. HEC-RAS uses an alternating direction implicit algorithm to solve the discretized system of equations. This algorithm decomposes the two-dimensional problem into two one-dimensional problems, alternately solving the equations in the x-direction and the y-direction, which greatly improves the computational efficiency. In each time step, HEC-RAS uses an iterative method to solve the nonlinear system of equations. The iteration process continues until the solution converges or the maximum number of iterations is reached.

[0041] 7. Parameterization of Manning coefficient. Manning roughness coefficient is reflected in the equation through friction slope. HEC-RAS can input land use type data and set different Manning coefficient values ​​for areas with different use types to improve simulation accuracy. In addition, during the simulation process, HEC-RAS uses the threshold method to process the process from dry to wet in the calculation domain to ensure the robustness of the calculation results.

[0042] The above steps can efficiently and stably solve the two-dimensional shallow water wave equation and simulate the flood diffusion process under complex terrain conditions. In practical applications, the accuracy of the model also depends on the quality of the input data, such as the accuracy of the DEM and the accuracy of the boundary conditions. Therefore, the optimized DEM data will be used to improve the accuracy of the terrain representation, and the water depth information extracted from the VGI data will be used to correct the model parameters and verify the simulation results, thereby further improving the accuracy of the simulation.

[0043] Considering that flood disasters often occur in an environment where professional data is scarce, this study uses publicly available data as the input data for the HEC-RAS2D model. The input data includes:

[0044] 1. DEM data. The DEM data used in the study is based on COPDEM, from which surface covers such as roads, viaducts, buildings, and trees are removed, and levees are extracted to produce optimized DEM data. DEM data is the key data for flooding simulation and the main factor affecting the simulation results. Therefore, this study will compare the flooding simulation effects of common open source DEM data, including ASTER, SRTM, COPDEM, AW3D, and FABDEM. These data can be downloaded from the GEE platform or its official website. Through comparative analysis, we can more clearly understand the simulation effects of various types of data and the simulation advantages of optimized DEM.

[0045] 2. River flow data. This data comes from the water flow information of the Mihe River and Danhe River released by the official Weibo of the W City Hydrological Center. It has been obtained through crawler technology and saved in the disaster information database. The time trend of flow is as follows Figure 1 shown.

[0046] 3. Precipitation data. This data also comes from the official Weibo of the W City Hydrological Center and has been obtained and saved in the database. The temporal trend of precipitation is as follows: Figure 2 shown.

[0047] 4. Land cover type data. This data comes from the 30-meter land use data released by the Chinese Academy of Sciences in 2023. The data is based on Landsat remote sensing images and is produced through manual visual interpretation. Different land cover types have different roughness levels and corresponding Manning coefficient ranges. When simulating flooding, you can first take any Manning coefficient within the range. In the subsequent model parameter correction section, the study will correct the Manning coefficient within the range based on the extracted water depth information.

[0048] The software version used for the flooding simulation is HEC-RAS 6.4.1, which is free and open source. The software only supports CPU operations, and the operation time depends on the CPU performance. The software operating environment is: Windows 11 operating system, CPU is AMD R75800H.

[0049] 1. Setting spatial discretization parameters. Create a new project file in HEC-RAS, set the workspace unit to international units (meters), and set the RAS Mapper drawing area to the WGS-84 geographic coordinate system. Import the optimized DEM into RASMapper, create a new 2D simulation area, and set the spatial grid size. The setting of the spatial grid will affect the results of the flooding analysis. If the spatial grid is too large, the simulation results will be insufficient in detail and may not accurately capture the impact of small-scale terrain; if the grid is too small, it will significantly increase the calculation time and resource consumption, and lose timeliness for large-scale simulations. Therefore, it is necessary to select an appropriate spatial grid size to ensure a balance between simulation accuracy and calculation efficiency. The study area is about 83 kilometers long from north to south and about 65 kilometers long from east to west. After testing, when the grid is set to 500 meters, the calculation takes 3 minutes, and when the grid is set to 100 meters, the calculation takes 65 minutes. When the grid continues to shrink, the computing resources will increase significantly, so the study sets the spatial grid to 100 meters, with a total of 255,754 computing units.

[0050] 2. Initial and boundary condition settings. After completing the spatial discretization, the river channels of the Mi River and the Dan River are vectorized and the corresponding boundary flow input points and outflow boundaries are added. According to the Shandong Provincial Annals of Natural Geography, the average annual runoff of the Mi River and the Dan River is 602 million cubic meters and 45.35 million cubic meters respectively. It can be inferred that the average annual flow of the Mi River and the Dan River is 19m 3 / s and 1.4m 3 / s, set it as the initial start-up flow of the 2D simulation. The outflow boundary is set to normal depth, which means that the hydraulic slope is used to estimate the water depth at the outlet, simulating the natural outflow of water at the end of the river. According to the study area, which is 83 kilometers long from north to south and has a drop of 41 meters, the slope can be estimated to be about 0.0002, which is applied to the flow input point and the outflow boundary. Finally, the time series flow data of the Mihe River and the Dan River are added in the 2D unsteady flow data editor.

[0051] 3. Set the Manning coefficient. Import the land cover data and set the corresponding Manning coefficient for different cover types according to the Manning coefficient range table provided by the land cover database. The initial setting is the median of each cover type. The Manning coefficient is shown in Table 4-1.

[0052] Table 4-1 Manning coefficients for different land cover types

[0053]

[0054] 4. Time step parameter setting. In flood simulation, the time step parameter will affect the simulation stability to a certain extent. In this model, the upper limit of the Courant number is set to 0.8 and the lower limit is set to 0.2, the number of steps to double before falling below the minimum Courant number is set to 3, the maximum number of times the basic time step is doubled, and the maximum number of times the basic time step is halved is set to 5, so as to maintain the stability of the numerical solution and reduce the calculation error. When the Courant number is lower than 0.2, if the Courant number continues to be lower than the threshold for 3 consecutive time steps, the time step will be increased by 1 time until the Courant number reaches or exceeds 0.2, and it will double up to 5 times. On the contrary, when the Courant number exceeds 0.8, the time step will be halved to prevent numerical instability during the calculation process, and it will be halved up to 5 times. By adaptively adjusting the time step, the model can effectively cope with the situation of rapid changes in flow velocity, especially when simulating short-time flood crest processes, it can ensure the accuracy and stability of the calculation. In addition, this method also reduces the errors that may be introduced when the time step is manually selected, and enhances the robustness of the simulation.

[0055] The quality of DEM directly affects the accuracy of flood simulation. High-quality DEM can more accurately express terrain features, thereby improving the accuracy of flood simulation. In order to evaluate the impact of different DEM data sources on flood inundation simulation results and the effect of optimizing DEM, this study used the HEC-RAS2D model to perform inundation simulation based on ASTER, SRTM, COPDEM, AW3D, FABDEM and optimized COPDEM. By comparing and analyzing these simulation results, we can better understand the performance of various DEM data in flood simulation, further emphasize the advantages of optimizing DEM, and provide a basis for subsequent model parameter correction and model verification using VGI data.

[0056] From the early morning of August 18 to the early morning of August 20, 2018, S City suffered the heaviest rainfall in the same period in recorded history. At the same time, the water levels of Tangshuiya, Yeyuan Reservoir, Songshan Reservoir and Heihushan Reservoir in the upper reaches of the Mi River also exceeded the safety warning line due to heavy rain. In order to avoid causing more serious reservoir dam collapse disasters, the W Flood Control and Drought Relief Command decided to start flood discharge at 8:00 on August 19 and end at 18:00 on August 21. This is the flood diffusion stage, a total of 58 hours. August 22 to 28 is the waterlogging stage, during which the flood gradually recedes. The study selected 12:00 on August 18 to 0:00 on August 22, 2018 as the flood diffusion simulation time period, and the calculation output time step is 30 minutes. The simulation of the waterlogging stage will be carried out in the model parameter correction part as a verification of the model correction effect.

[0057] According to the official Weibo of S City Radio and Television Station, the flood peak arrived in the urban area of ​​S City at 1:30 a.m. on the 20th. Considering the delay of the flood reaching S City, the study uses 12:00 on the 19th, 1:30 on the 20th, 12:00 on the 20th, 0:00 on the 21st, and 0:00 on the 22nd as the key simulation time points for the flood diffusion stage.

[0058] The performance of five commonly used DEMs, ASTER, SRTM, AW3D, COPDEM, and FABDEM, and the optimized DEM in flood spread simulation were compared. The colors from blue to red represent the water depth from shallow to deep. Since the water depth simulation results of different DEMs are quite different, the color depth only reflects the relative water depth and does not represent the absolute water depth value. This representation method is mainly used to intuitively display the spatial distribution characteristics of the flooded area and the trend of relative water depth changes, which is convenient for comparing the flooding conditions at different time points and in different areas. All models show that the flood range changes over time, but there are significant differences in the details of the flood range and the distribution of water depth. COPDEM, FABDEM and optimized DEM show a relatively detailed and dynamic flood evolution process, while the ASTER, AW3D and SRTM models show a relatively small flood range with no obvious changes.

[0059] There are differences in the flooding characteristics of different DEM models. The simulation results of ASTER show that the flooding range is small and scattered. Over time, the flooded area gradually expands, but the overall flooding range is still small. This performance may be due to the low resolution of ASTER DEM, which leads to inaccurate terrain details, thus affecting the accuracy of flood simulation. The flooding simulation results of SRTM DEM around rivers and low-lying areas are more continuous than ASTER DEM, and the flooding range will fluctuate with the change of time nodes, showing a certain flood diffusion trend, indicating that SRTM's terrain data has certain practicality in flood simulation. Both ASTER and SRTM models underestimate the flood range, mainly concentrated in the northern region.

[0060] Compared with SRTM DEM, AW3D DEM shows continuous flood distribution and better details. COPDEM shows a more detailed flood evolution process and can reflect the river network structure. Among the commonly used DEMs, FABDEM shows the most detailed and dynamic flood simulation results, clearly showing the process of flood diffusion along the river channel and reflecting the influence of complex terrain.

[0061] The optimized DEM removed the surface covers such as buildings, trees and elevated roads that affected the flooding simulation results, and reconstructed the elevation of the dam. The optimized DEM flooding simulation results showed a more coherent flooded area compared to FABDEM, making the simulation results closer to the actual situation.

[0062] From the perspective of temporal evolution characteristics, all models showed the scope of surface waterlogging caused by heavy rain at 12:00 on the 19th. The flood peak arrived at 01:30 on the 20th, and the river flood spread and overflowed. This process was most significant in the simulation results of COPDEM, FABDEM and optimized DEM, among which the optimized DEM showed the most consistent flooding simulation results. From 12:00 on the 20th to 0:00 on the 21st, the flood peak passed and the flood diffusion weakened. This process was more obvious in the simulation results of FABDEM and optimized DEM. By 0:00 on the 22nd, the flood diffusion weakened again. Overall, the flooding simulation results of the optimized DEM reached the best, and among the commonly used DEMs for flooding simulation, the simulation results of FABDEM were more reasonable.

[0063] The study is based on the local optimization of FAMDEM. In order to show a clearer contrast effect, the local details of the flooding when the flood peak arrives (1:30 on the 20th) will be compared with the flooding simulation effects of the two DEMs. The flooding simulation results are as follows: Figure 3 shown.

[0064] The figure shows the comparison results of the local details of the flood when the flood peak reached (1:30 on the 20th). The left picture is the original FABDEM, and the right picture is the optimized DEM. Five comparison areas are marked in the figure, marked with numbers 1 to 5. From the top two images, it can be seen that the optimized DEM is smoother in terrain details, eliminating abnormal elevations caused by artificial structures and trees, and is closer to the real terrain. The comparison figure below shows the flood inundation simulation results of the five comparison areas in detail. Each area includes a comparison of the results of FABDEM and optimized DEM.

[0065] Area 1 shows the flooding near the viaduct. The simulation results of the optimized DEM show that the flood distribution is more continuous and smooth, eliminating the water flow obstruction caused by the viaduct or highway. Area 2 also shows the flooding near the viaduct. The simulation results of the optimized DEM show a larger and more uniform flooding range, and the details of the ditch are fully preserved.

[0066] Region 3 shows the impact of not removing buildings on the flooding simulation results. The simulation results of the optimized DEM not only show a more continuous and coherent flooded area, but also show a more natural change in water depth. In the upper right corner of the region, the flooded area increases after the dam elevation is optimized. In contrast, the simulation results of FABDEM are rough and affected by abnormal surface elevation.

[0067] Region 4 shows the flooding situation in the flat area due to the impact of road outliers. The optimized DEM results also show a more continuous and extensive flooding area. Region 5 shows the flooding situation in the areas on both sides of the dam. After correcting the dam elevation, the optimized DEM simulates a wider river flood and is not affected by the surface cover in the upper left corner of the area. Overall, the optimized DEM shows higher accuracy and continuity in flooding simulation. It can better reflect the impact of actual terrain on water flow and present a more realistic flooding range and water depth distribution.

[0068] Since the above simulation results are based on data modeling and some parameters have not been fine-tuned, it is necessary to further introduce official data or data from other sources for verification to ensure the accuracy and reliability of the simulation results, thereby providing a basis for the correction and adjustment of model parameters. The study used relevant disaster reports from various townships and combined them with the residential area data extracted in the previous article to verify the simulation results of the maximum flood inundation depth.

[0069] From the inundation simulation results, it can be seen that the places with relatively deep inundation include Shangkou Town, Luocheng Street, Daotian Town, Yingli Town, Tianliu Town, Yangkou Town, and the area connected to Taitou Town and Shuangwang Town. It should be noted that the actual disaster situation of each township is included in but not equal to the flood inundation range. The severity of the disaster reported in the news is closely related to the residential area. There are almost no residential areas in the area between Taitou Town and Shuangwang Town, so it is not listed in the news. Overall, the inundation simulation results can accurately simulate Luocheng Street, Shangkou Town, Yingli Town, Yangkou Town and Hou Town mentioned in the news, while Jitai Town and Daotian Town did not show serious inundation, but in fact they were severely affected. This difference may be due to a variety of factors, such as surface roughness, drainage system, local rainfall intensity, etc., among which surface roughness is the most critical influencing factor. The surface roughness coefficient is an important parameter reflecting the surface roughness, which directly affects the water flow velocity and inundation range. Different land cover types, such as urban built-up areas, farmland, forest land, etc., have different surface characteristics and should be given different Manning coefficient values ​​in the model. Reasonable setting of the Manning coefficient based on the classification of land cover types can initially improve the accuracy of flood simulation. In order to further improve the accuracy of inundation simulation, it is necessary to perform iterative correction based on the actual inundation range or water depth information. Since the flood inundation range is difficult to accurately extract in remote sensing images, the extracted water depth information can be used to correct the Manning coefficient. This application is based on the model Manning coefficient correction method based on water depth information. By comparing the difference between the simulation results and the actual water depth data, the Manning coefficient setting can be iteratively adjusted to improve the simulation accuracy of the model.

[0070] The water depth description information extracted from the text and the water depth information extracted from the image are relative to the real earth surface, while the flood inundation simulation based on the optimized DEM is corrected and the surface features such as ordinary roads and buildings are removed. There is an elevation difference between the two. It is necessary to establish a conversion relationship between the extracted water depth information and the simulated inundation depth. Text-type water depth information generally uses surface reference objects to describe the water depth information. When extracting information from images with reference to people, the people usually stand on the surface of impermeable bodies such as roads or buildings, rather than the optimized DEM with the surface cover removed. Therefore, the relationship chain between the extracted water depth information and the inundation simulation water depth is: real surface (extracted water depth) - COPDEM - optimized DEM (simulated water depth). COPDEM is a general representation of the real surface, and the optimized DEM is based on COPDEM. The three are in a progressive relationship. In order to reduce the error transmission among the three, the study used the altimetry data of ICESat-2 (Ice, Cloud and land Elevation Satellite-2) land observation satellite combined with COPDEM to estimate the residual of COPDEM and the true elevation at the water depth point. Then, according to the elevation difference between the optimized DEM and COPDEM, the conversion relationship between the extracted water depth and the simulated water depth of the relatively optimized DEM was established. The process is as follows: Figure 4 shown.

[0071] ICESat-2 is a new generation of polar ice satellite launched by NASA in September 2018. Its main mission is to monitor changes in the thickness of the polar ice sheet, while also providing global land and water surface elevation data. ICESat-2 is equipped with the Terrain Laser Altimeter (ATLAS), which uses photon counting technology to provide high-resolution and high-precision surface elevation measurement data. The operational characteristics of ICESat-2 include a 91-day revisit cycle, and the earliest available and large-scale data in City S began from December 2018 to January 2019. The study used the ATL03 data product of ICESat-2, which is a global geolocated photon dataset that contains georeferenced photon event location information, including latitude and longitude, elevation, elevation confidence, etc., with a spatial resolution of 0.7 meters and an elevation accuracy of 0.1 meters.

[0072] The study obtained a total of 18 ATL03 trajectories, about 930,000 photon event points, and 78 valid water depth points. During the calibration process, it is necessary to first establish the conversion relationship between the water depth extraction position and the corresponding COPDEM. The steps are as follows:

[0073] 1. ATL03 data screening. With the water depth point as the center, select the ATL03 photon event data within a radius of 5000 meters. Then use the signal_conf_ph field in the ATL03 data to screen out 4-level high-confidence photon events (confidence interval is 0-4, -1 means no signal) as elevation control points. In order to save calculation time, the photon event points within the buffer radius are thinned according to the terrain. 50-100 high-confidence photon event points are selected for each water depth point. The more complex the terrain, the more points are selected.

[0074] 2. COPDEM data extraction. Use a 3x3 pixel window to extract the COPDEM elevation value at the COPDEM corresponding to the water depth point, and calculate the average elevation value of the window as the COPDEM elevation corresponding to the water depth point, denoted as A.

[0075] 3. Interpolation of elevation residuals. Use the ATL03 point elevation minus its corresponding COPDEM elevation to obtain the residual distribution between the ATL03 control point and COPDEM, recorded as B. Use the Kriging interpolation method to interpolate the elevation difference B, and use a 3x3 window to extract the elevation residual C at the water depth point. Add the residual C to the elevation A to get the corrected elevation at the water depth point, and add the residual C to the water depth to get the water depth relative to COPDEM. The process is as follows: Figure 5 shown.

[0076] Next, subtract COPDEM from the optimized DEM to obtain the elevation difference between the optimized DEM and COPDEM. Add the water depth relative to COPDEM to the elevation difference to obtain the water depth relative to the optimized DEM. At this point, the conversion relationship of submerged water depth is established.

[0077] The Manning coefficient is a key parameter that describes the roughness of the surface and directly affects the calculation of water velocity and water depth. By comparing the difference between the simulated water depth and the measured water depth, the Manning coefficient can be iteratively adjusted to optimize the flooding simulation results. Based on the Manning coefficient range table provided by the land cover database, the initial Manning coefficient is set to the median value of each cover type. The Manning coefficient correction process is as follows Figure 6 shown.

[0078] The calculation of water depth difference uses the root mean square error (RMSE) as the main indicator. For each observation point i, we calculate the difference between the simulated water depth (H_sim,i) and the measured water depth (H_obs,i), and then calculate the RMSE of the entire study area. The formula is as follows.

[0079]

[0080] Where n is the total number of observation points. In order to save correction time, 39 points were extracted from 78 valid water depth points as correction points according to the distribution of water depth points, and the remaining 39 points were used as verification points. The iterative adjustment rule of the Manning coefficient is: when the simulated water depth is generally higher than the measured water depth, the Manning coefficient of the corresponding area is increased to increase the surface roughness and slow down the water flow speed; conversely, when the simulated water depth is generally lower than the measured water depth, the Manning coefficient is reduced. The step size of each iterative adjustment is set to 5% of the current value to ensure the stability and convergence of the adjustment. Considering the complexity of the topography and land use in the study area, the Manning coefficient adopts a localized adjustment strategy. The study area is divided into 15 sub-areas according to townships, and each sub-area adjusts the Manning coefficient independently.

[0081] The convergence criterion of the iterative process is set as follows: when the RMSE improvement of three consecutive iterations is less than 1%, or when the maximum number of iterations is reached (set to 20 in this study), the adjustment is stopped. This setting ensures the adequacy of the optimization while avoiding excessive calculation and waste of resources.

[0082] After calculation, the RMSE value of the verification result is 0.32 meters, which is less volatile than the RMSE of the correction point of 0.28 meters, indicating the effectiveness of the Manning coefficient correction method. Further analysis of the comparison between the simulation results and the actual situation shows that the simulation results in the urban area accurately reflect the depth of water accumulation at important intersections; the simulation of the degree of flooding in rural areas is close to the actual report; and the simulation of the water depth in severely flooded areas is also consistent with the actual situation. In general, through the iterative correction of the Manning coefficient, the study has significantly improved the accuracy of flood and inundation simulation results.

[0083] This chapter discusses how to improve the accuracy of flood simulation by fusing VGI data and optimizing DEM to solve the limitations of traditional DEM in flood simulation. First, the study used the HEC-RAS2D unsteady flow model to compare the performance of different DEM data in flood simulation. The results show that the optimized DEM is superior to conventional DEM in detail performance and water depth simulation, and can more accurately reflect the impact of complex terrain on flood diffusion. In order to further improve the simulation accuracy, the study innovatively used the ATL03 data of the ICESat-2 satellite to construct the conversion relationship between the extracted water depth and the simulated water depth, ensuring the consistency of water depth information under different elevation benchmarks. Based on this, a model Manning coefficient correction method based on water depth information is proposed. This method dynamically adjusts the Manning coefficient by iteratively comparing the difference between the simulated water depth and the measured water depth, and using RMSE as the main indicator. After multiple iterations, the simulation results of the study area are closer to the actual flood situation, and the RMSE value of the verification result is reduced to 0.32 meters, which significantly improves the simulation accuracy. Through the above research, this chapter constructs an effective flood inundation simulation process, successfully combines VGI data with optimized DEM, and achieves accurate correction of the Manning coefficient.

[0084] In one embodiment, the optimization sequence is as follows: by superimposing the residential area and tree-covered area data with the outlier data, the overlapping outliers are identified and removed to ensure that the buildings and tree coverage do not interfere with the flood simulation; by creating a river buffer and calculating the average relative height of the road based on COPDEM, the river embankments with interception and storage functions are identified and retained; by spatial intersection analysis of roads and outliers, the roads are identified and removed. A large number of gaps will be generated in the process of outlier removal. This application uses a variety of interpolation methods to deal with the DEM omissions after the outliers are removed. In addition, this application also restores the actual embankment coverage height according to the vegetation coverage height of the embankment to achieve further optimization of FABDEM.

[0085] 1. Remove building coverage outliers. Use the residential area data as a mask file to overlay the slope grid. Delete the grid values ​​that overlap with the residential area and have a slope value greater than 1.5 in the slope grid, and use it as a new mask file, recorded as slope mask 1. Then overlay FABDEM with mask 1 to delete abnormal elevation data. This step generates an optimized DEM file for the residential area, recorded as DEM 1.

[0086] Removing building outliers will produce rectangular or circular gaps. Inverse distance weighted interpolation is a method of interpolation based on inverse distance weights. The core idea is that the value of the interpolation point is the weighted average of the known points around it, where the weight is inversely proportional to the distance from the known point to the interpolation point. The inverse distance weighted interpolation method assumes that points that are close in space have higher similarity, and points that are closer have a greater impact on the interpolation result. Therefore, this application uses the inverse distance weighted interpolation method to fill the gap area after the building coverage outliers are removed.

[0087] 2. Tree cover outlier removal. Tree cover outlier removal is a serial task of building cover outlier removal. Based on slope mask 1 and DEM 1, outliers are removed by superimposing tree cover areas, generating slope mask 2 and DEM 2 with abnormal slopes and abnormal elevations of buildings and trees removed.

[0088] Irregular gaps will be generated after the tree cover outliers are removed. Natural neighbor interpolation is an interpolation method based on the Thiessen polygon principle. This method calculates the value of the interpolation point by determining the natural neighbors of the interpolation point (i.e., the neighboring points that share edges with the interpolation point in the Thiessen polygon) and using the known values ​​of the natural neighbors for weighted averaging. Natural neighbor interpolation can handle irregularly distributed data points, while taking into account the local characteristics of the original data, ensuring that the interpolation result is naturally smooth at the boundary without obvious boundary effects. Therefore, this application uses natural neighbor interpolation to fill the irregular gaps in DEM 2.

[0089] 3. Elevated and ordinary road removal. This task is also a serial task of tree cover outlier removal. The extracted road is essentially the centerline of the road, which does not contain width and elevation information. The extracted road data can be spatially intersected with the slope mask 2. The outliers that intersect with the road in DEM 2 will be removed to obtain the slope mask 3 that removes the outliers of ordinary roads and elevated roads. Slope mask 3 is superimposed with FABDEM, and the abnormal elevation data is removed to obtain the optimized DEM 3.

[0090] Linear gaps will be generated after the road outliers are removed. Spline interpolation uses piecewise polynomial functions to create smooth and continuous surfaces that can capture subtle changes in terrain. Cubic Spline Interpolation is used in DEM processing. This interpolation method ensures the continuity of the first and second order derivatives at the data nodes by constructing a cubic polynomial with continuous first and second order derivatives between each data point. The core idea of ​​spline interpolation is to fit a low-order polynomial between data points while maintaining global smoothness. Therefore, this application uses the spline interpolation method to process the gaps in missing linear features such as roads in DEM 3.

[0091] 4. Levee extraction. Since the levee details are missing in FABDEM, this step is based on COPDEM. This task is parallel to the above three tasks, not serial. Levee extraction is performed based on river buffer and nearest road analysis. First, a 100-meter-wide buffer is created using the river surface data provided by the National Basic Geographic Information Database as the preliminary range of potential levee distribution. The buffer outside the river is then used as a mask to extract the slope raster and extract the abnormal slope values. The extracted road data is loaded. If there are multiple roads in the buffer outside the river, the road with the highest average relative height is selected as the levee road. The levee road and the outliers are spatially analyzed. The outliers that intersect with the levee road are the real levees, and the other outliers are deleted. This step obtains the levee information within 100 meters of the river, recorded as DEM 4.

[0092] After completing the above four processes, we can get DEM 3 with buildings, trees, ordinary roads and elevated outliers removed, as well as DEM 4 of the extracted embankment. DEM 4 is used as updated data to mosaic DEM 3 to get a new DEM 5. DEM5 still has a few outliers because tree coverage, building coverage and road intersections cannot completely cover all outliers. It is necessary to visually determine the outlier type in combination with historical image maps, and then manually delete or retain them and fill them with corresponding interpolation methods. Finally, the optimized DEM is obtained.

[0093] Although the present invention is described in this application with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the present invention. It should therefore be understood that many modifications may be made to the exemplary embodiments and that other arrangements may be devised without departing from the spirit and scope of the present invention as defined by the appended claims. It should be understood that the features described in the various dependent claims and in this application may be combined in a manner different from that described in the original claims. It should also be understood that the features described in conjunction with a single embodiment may be used in other described embodiments.

Claims

1. A method for improving flood simulation accuracy based on fusing VGI data and optimizing DEM, characterized in that: HEC-RAS was selected as the flood inundation simulation model, and the input data included DEM data, river flow data, precipitation data, and land cover type data. According to the Manning coefficient range table provided by the land cover database, the corresponding Manning coefficient was set for different cover types, and it was initially set to the median of each cover type. The Manning coefficient setting was iteratively adjusted by comparing the difference between the simulation results and the actual water depth data. A relationship chain between the extracted water depth information and the inundation simulation water depth was established. The altimetry data of the land observation satellite was combined with COPDEM to estimate the residual of COPDEM and the true elevation at the water depth point, and then a conversion relationship was established between the extracted water depth and the simulated water depth of the relatively optimized DEM based on the elevation difference between DEM and COPDEM.

2. The method for improving flood simulation accuracy based on fusing VGI data and optimizing DEM according to claim 1 is characterized in that: The relationship chain between the extracted water depth information and the submerged simulated water depth is: extracted water depth - COPDEM - simulated water depth. COPDEM is a rough representation of the extracted water depth, and the simulated water depth is based on COPDEM. There is a progressive relationship between the three.

3. The method for improving flood simulation accuracy based on fusing VGI data and optimizing DEM according to claim 1 is characterized in that: The steps to establish the conversion relationship between the extracted water depth and the simulated water depth of the relatively optimized DEM are as follows: Taking the water depth point as the center, select the ATL03 photon event data within the predetermined range; then use the signal_conf_ph field in the ATL03 data to screen out the 4th level high-confidence photon events as elevation control points; the photon event points within the buffer radius are thinned according to the terrain, and 50-100 high-confidence photon event points are selected for each water depth point. The more complex the terrain, the more points are selected; At the COPDEM corresponding to the water depth point, the pixel window is used to extract the elevation value of COPDEM, and the average elevation value of the window is calculated as the COPDEM elevation corresponding to the water depth point, which is recorded as A; Subtract the corresponding COPDEM elevation from the ATL03 point elevation to obtain the residual distribution between the ATL03 control point and COPDEM, denoted as B; use the Kriging interpolation method to interpolate the elevation difference B, and use window extraction to obtain the elevation residual C at the water depth point; Add the residual C to the elevation A to get the corrected elevation at the water depth point, and add the residual C to the water depth to get the water depth relative to COPDEM; Next, subtract COPDEM from the optimized DEM to obtain the elevation difference between the optimized DEM and COPDEM; add the water depth relative to COPDEM to the elevation difference to obtain the water depth relative to the optimized DEM. At this point, the conversion relationship between the submerged water depth is established.

4. The method for improving flood simulation accuracy based on fusing VGI data and optimizing DEM according to claim 1 is characterized in that: The Manning coefficient correction process includes: setting the initial Manning coefficient, running the flooding simulation model, comparing the simulated water depth with the actual water depth, calculating the RMSE of the water depth difference, adjusting the Manning coefficient according to the water depth difference, and outputting the final Manning coefficient if the Manning coefficient reaches the convergence standard; if the Manning coefficient does not reach the convergence standard, returning to running the flooding simulation model.

Images