Flood forecasting method, device and equipment based on field monitoring and real-time modeling

Abstract

The invention relates to the technical field of flood forecasting, in particular to a flood forecasting method, device and equipment based on field monitoring and real-time modelling, and the method comprises the steps: collecting a topographic data file and an earth surface data text of a target area to construct a sub-meter DEM digital elevation model; radar real-time base data, radar and ground rainfall historical observation data, brightness temperature data and geographical coordinate information of a target area are obtained to solve actual radar real-time base data, rain short-term forecast data and soil moisture data, and then the data are utilized to calculate a preset hydrological model to obtain a mountain torrent gully basin distributed hydrological model result; and establishing a hydrological-hydrodynamic coupling model according to the sub-meter DEM digital elevation model and the torrential flood gully basin distributed hydrological model so as to forecast the current flood state and the future flood state. Therefore, the problems that an existing mountainous area hydrological model cannot completely reflect the current state of a drainage basin, and the simulation error of a hydrodynamic model is large are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of flood forecasting, and in particular to a flood forecasting method, device and equipment based on field monitoring and real-time modeling. BACKGROUND

[0002] With the deepening understanding of the formation mechanism of mountain flood and the maturity of new technologies such as laser radar and X-band radar, the existing hydrological model needs to be improved in terms of prediction ability at the mountain gully scale and coverage range of high-resolution hydrodynamic model, which is reflected in the following three aspects: Firstly, the existing hydrological model is based on the runoff generation mechanism of surface runoff or excess infiltration, while the latest research shows that the dominant runoff generation mechanism of mountain rainstorm flood is shallow subsurface runoff or strong weathering zone, so the current rainfall runoff model cannot fully reflect the formation mechanism of mountain flood, and the existing mountain hydrological model needs to be updated according to the new mechanism.

[0003] Secondly, the existing hydrodynamic model of mountain gully is mostly constructed based on 5m or even 30m grid resolution terrain data, which has large simulation error and cannot accurately describe the dramatic changes of gully morphology, affecting the simulation accuracy of gully water depth, water level and flow velocity, and cannot meet the demand of accurate simulation of mountain flood.

[0004] Thirdly, due to the existing mountain hydrological model cannot fully reflect the formation mechanism of mountain flood and the simulation error of hydrodynamic model is large, the accuracy of the existing mountain flood forecasting model will be low, which cannot guarantee the safety of people's life and the high-quality development of economy and society. SUMMARY

[0005] The present application provides a flood forecasting method, device and equipment based on field monitoring and real-time modeling, which solves the problems of the existing mountain hydrological model that cannot fully reflect the formation mechanism of mountain flood, the simulation error of hydrodynamic model is large, the accuracy of the existing mountain flood forecasting model is low, and cannot guarantee the safety of people's life and the high-quality development of economy and society.

[0006] The first aspect of the present application provides a flood forecasting method based on field monitoring and real-time modeling, comprising the following steps: collecting terrain data files and surface data texts of a target area, and constructing a sub-meter DEM digital elevation model according to the terrain data files and the surface data texts; obtaining radar real-time base data, radar and ground rainfall historical observation data of the target area, and solving actual radar real-time base data and short-term forecast data of rainstorm according to the radar real-time base data, the radar and ground rainfall historical observation data; collecting brightness temperature data and geographic coordinate information of the target area, to solve soil moisture data according to the brightness temperature data and the geographic coordinate information; extracting natural sub-basin, river network and geometric parameters from the terrain text data, embedding a preset runoff generation module into the natural sub-basin to obtain a new natural sub-basin, embedding a preset confluence module into the river network to obtain a new river network, and constructing an initial hydrological model according to the new natural sub-basin, the new river network and the geometric parameters; using the actual radar real-time base data, the short-term forecast data of rainstorm and the soil moisture data to operate the initial hydrological model to obtain a mountain torrent gully distributed hydrological model result; establishing a hydrology-hydrodynamics coupled model according to the sub-meter DEM digital elevation model and the mountain torrent gully distributed hydrological model result; and forecasting current flood state and future flood state of the target area according to the hydrology-hydrodynamics coupled model.

[0007] Optionally, the collecting terrain data files and surface data texts of a target area comprises: controlling a UAV carrying a laser radar to perform a preset flight route and a preset flight number of times over the target area to obtain a trajectory data set; performing trajectory calculation and point cloud calculation on the trajectory data set to obtain a strip map; and performing data filtering on the strip map to generate the terrain data files and the surface data texts.

[0008] Optionally, the obtaining radar real-time base data, radar and ground rainfall historical observation data of the target area, and solving actual radar real-time base data and short-term forecast data of rainstorm according to the radar real-time base data, the radar and ground rainfall historical observation data comprises: The radar real-time base data of the target area, radar and ground rainfall historical observation data, and historical multi-source weather data are acquired; mixed elevation angle data are determined according to the radar real-time base data and the terrain data file; radar quality control is performed on the mixed elevation angle data to obtain a resolution networking mosaic; the radar and ground rainfall historical observation data are used to train a pre-constructed multi-layer perceptron-based radar rainfall inversion model to obtain a trained radar rainfall inversion model; the resolution networking mosaic is input into the trained radar rainfall inversion model to invert the rainfall intensity at the current time, and the radar real-time base data is corrected according to the rainfall intensity at the current time to obtain actual radar real-time base data; the resolution networking mosaic is converted into a low-resolution networking mosaic by using a pre-constructed radar rainfall field downscaling model based on an autoencoder; the historical multi-source weather data and the radar and ground rainfall historical observation data are used to train a pre-constructed deep spatiotemporal representation enhanced mountainous area rainfall nowcasting model to obtain a trained mountainous area rainfall nowcasting model; the low-resolution networking mosaic is input into the trained mountainous area rainfall nowcasting model to predict the future rainfall intensity, and the future rainfall intensity is taken as the short-term rainstorm prediction data.

[0009] Optionally, the collection of the brightness temperature data and the geographic coordinate information of the target area to solve the soil moisture data according to the brightness temperature data and the geographic coordinate information comprises: The brightness temperature data and the geographic coordinate information of the target area are collected by using a drone carrying a microwave radiometer; auxiliary parameter data are extracted from a pre-constructed auxiliary database according to the geographic coordinate information; and the auxiliary parameter data and the brightness temperature data are input into a pre-constructed soil moisture inversion model to generate the soil moisture data.

[0010] Optionally, the deep spatiotemporal representation enhanced mountainous area rainfall nowcasting model comprises a short-term nowcasting module based on a deep convolutional network and a spatiotemporal representation enhancement method, an upsampling module based on spatiotemporal attention, and a downsampling module based on spatiotemporal attention, wherein the short-term nowcasting module based on the deep convolutional network and the spatiotemporal representation enhancement method adopts a U-Net structure, receives radar observation sequences in the historical multi-source weather data and the radar and ground rainfall historical observation data as input to predict future radar sequences; and the upsampling and downsampling modules based on spatiotemporal attention adopt a full convolutional structure and capture spatiotemporal features of radar rainfall images in the radar observation sequences through a spatiotemporal attention mechanism.

[0011] Optionally, the establishment of a hydrology-hydrodynamics coupled model according to the sub-meter DEM digital elevation model and the mountain torrent gully basin distributed hydrological model result comprises: construct a two-dimensional refined topographic model of the mountain torrent gully of the target region according to the sub-meter DEM digital elevation model; extract terrain data, hydrological data and hydrodynamic data in the three-dimensional refined topographic model of the mountain torrent gully; construct a hydrology and hydrodynamics model according to the terrain data, the hydrological data and the hydrodynamic data based on a CPU and GPU heterogeneous parallel acceleration algorithm; and couple the hydrology and hydrodynamics model and the distributed hydrology model result of the mountain torrent gully basin to obtain the hydrology-hydrodynamics coupling model.

[0012] The second aspect of the present application provides a flood forecasting device based on field monitoring and real-time modeling, comprising: a first construction module configured to collect terrain data files and surface data texts of a target region, and construct a sub-meter DEM digital elevation model according to the terrain data files and the surface data texts; a first solving module configured to obtain radar real-time base data, radar and ground rainfall historical observation data of the target region, and solve actual radar real-time base data and short-term forecast data of rain shower according to the radar real-time base data, the radar and ground rainfall historical observation data; a second solving module configured to collect brightness temperature data and geographic coordinate information of the target region, and solve soil moisture data according to the brightness temperature data and the geographic coordinate information; a second construction module configured to extract natural sub-basins, river networks and geometric parameters from the terrain text data, embed a preset runoff generation module into the natural sub-basins to obtain new natural sub-basins, embed a preset confluence module into the river networks to obtain new river networks, and construct an initial hydrology model according to the new natural sub-basins, the new river networks and the geometric parameters; a third construction module configured to operate the initial hydrology model by using the actual radar real-time base data, the short-term forecast data of rain shower and the soil moisture data to obtain a distributed hydrology model result of a mountain torrent gully basin; a fourth construction module configured to establish a hydrology-hydrodynamics coupling model according to the sub-meter DEM digital elevation model and the distributed hydrology model result of the mountain torrent gully basin; and a forecasting module configured to forecast current and future flood states of the target region according to the hydrology-hydrodynamics coupling model.

[0013] The third aspect of the present application provides a mobile flood forecasting device, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor executes the program to implement the flood forecasting method based on field monitoring and real-time modeling as described in the above embodiments.

[0014] The fourth aspect of the present application provides a computer program product, wherein the computer program / instruction is executed by a processor to implement the flood forecasting method based on field monitoring and real-time modeling as described above.

[0015] The fifth aspect embodiment of the present application provides a computer readable storage medium, which stores a computer program, and the program is executed by a processor to realize the flood forecasting method based on field monitoring and real-time modeling.

[0016] The flood forecasting method, device and equipment based on field monitoring and real-time modeling provided by the embodiments of the present application apply the work of measuring terrain by the unmanned aerial vehicle carrying the laser radar to the data post-processing of flood forecasting and early warning exploration for accurate and continuous data correction, splicing and solving processes, and construct a hydrological model for the phenomenon that the dominant runoff mechanism is shallow subsurface runoff in the weathered zone, realize automatic operation in aspects of surface point cloud identification, bridge removal, channel extraction, and improve the efficiency and accuracy of obtaining key hydrological information of mountain torrent basin; complete radar and ground rainfall observation data arrangement, perform mountainous multi-band radar data quality control and networking splicing, mountain storm radar parameter inversion, and realize real-time data fusion of mountain fine rainfall; perform mountain multi-source meteorological data fusion, construct mountain meteorological field downscaling technology and rainfall short-term forecasting technology, and the effect of the rainfall inversion and short-term forecasting is significantly improved, and then the existing mountain torrent gully distributed hydrological model can be better updated to comprehensively reflect the mountain flood formation mechanism; through mountain torrent gully terrain and underlying surface data arrangement, a high-efficiency and high-resolution hydrodynamic model of mountain torrent gully is constructed to realize efficient and large-scale water dynamic simulation calculation and mountain torrent dynamics scene simulation under different working conditions, and mountain flood evolution process simulation; and a high-efficiency and high-resolution hydrodynamic modeling tool of mountain torrent gully is constructed to improve the modeling efficiency and optimize the calculation efficiency; the hydrological-hydrodynamic coupling model established based on the high-efficiency and high-resolution hydrodynamic model of mountain torrent gully and the mountain torrent gully distributed hydrological model improves the prediction accuracy, and can guarantee the life safety of the people and the high-quality development of the economic society.

[0017] Additional aspects and advantages of the present application will be made apparent by the following description and the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0018] The above and / or additional aspects and advantages of the present application will become apparent and be readily understood from the following description, taken in conjunction with the accompanying drawings, in which: Figure 1 A flowchart of a flood forecasting method based on field monitoring and real-time modeling according to an embodiment of the present application is shown in the figure; Figure 2 An unmanned aerial vehicle carrying a laser radar work flowchart according to an embodiment of the present application is shown in the figure; Figure 3 A RIEGL airborne laser radar data preprocessing diagram according to an embodiment of the present application is shown in the figure; Figure 4An eight-belt schematic diagram provided according to an embodiment of the present application; Figure 5 A noise and outlier elimination schematic diagram provided according to an embodiment of the present application; Figure 6 A tree and canopy schematic diagram provided according to an embodiment of the present application; Figure 7 A DJI L1 airborne LiDAR data preprocessing flowchart provided according to an embodiment of the present application; Figure 8 A DJI L1 intelligent map interface diagram provided according to an embodiment of the present application; Figure 9 A rainfall real-time inversion and rainfall short-term forecast flowchart provided according to an embodiment of the present application; Figure 10 A radar data quality control flowchart provided according to an embodiment of the present application; Figure 11 A rainfall real-time data fusion flowchart provided according to an embodiment of the present application; Figure 12 A radar rainfall field downscaling model structure schematic diagram provided according to an embodiment of the present application; Figure 13 A deep spatiotemporal representation enhanced mountainous rainfall short-term forecast model structure schematic diagram provided according to an embodiment of the present application; Figure 14 A structure schematic diagram of a downsampling and upsampling module based on spatiotemporal attention provided according to an embodiment of the present application, wherein (a) is a downsampling module, and (b) is an upsampling module; Figure 15 A real-time automatic high-resolution soil moisture inversion flowchart provided according to an embodiment of the present application; Figure 16 A block schematic diagram of a flood forecasting device based on field monitoring and real-time modeling provided according to an embodiment of the present application; Figure 17 A structure schematic diagram of a field flood monitoring equipment provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0019] Embodiments of the present application are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference signs represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the accompanying drawings are exemplary and are intended to explain the present application, and cannot be understood as a limitation of the present application.

[0020] A flood forecasting method, device and equipment based on field monitoring and real-time modeling of an embodiment of the present application are described below with reference to the accompanying drawings.

[0021] Figure 1 A flowchart of a flood forecasting method based on field monitoring and real-time modeling according to an embodiment of the present application.

[0022] As shown in the drawings, Figure 1 The flood forecasting method based on field monitoring and real-time modeling can be applied to an intelligent mobile flood forecasting vehicle, and includes the following steps: In step S101, topographic data files and surface data texts of a target area are collected, and a sub-meter DEM digital elevation model is constructed according to the topographic data files and the surface data texts.

[0023] In some embodiments, the topographic data files and the surface data texts of the target area are collected, including: A UAV carrying a laser radar is controlled to perform a preset flight path and a preset number of flights over the target area to obtain a trajectory data set; The trajectory data set is trajectory-solved and point cloud-solved to obtain a flight strip map; The flight strip map is data-filtered to generate the topographic data files and the surface data texts.

[0024] In actual execution, the UAV is placed in a relatively flat area, a laser radar and a GPS receiver are installed on the UAV, and satellite search is performed to ensure GPS signal strength. After takeoff, the laser radar performs pre-flight to allow its inertial navigation system (INS) to work normally and accurately; then the aircraft plans a flight path and returns to the home point to replace the battery halfway through the flight, with each flight lasting about 20-25 minutes; this step is repeated until the UAV flies over the entire target mountain torrent valley modeling range, and finally the radar point cloud data (i.e., the trajectory data set) is collected.

[0025] For example, as shown in the drawings, Figure 2 If the target area has 61 channels, the UAV's workflow is roughly as follows: the UAV carrying the laser radar has a work capacity of about 3km 2 / day, and 61 channels of sub-meter high-precision channel topographic data (i.e., L3 level topographic data) need to be surveyed for a total area of 10%, totaling 152km 2 , and the field work duration is about 51 days (not considering weather interference).

[0026] While surveying the terrain, the UAV carries a camera, which has a multi-angle and ultra-fast shooting function, to perform aerial photography and obtain the corresponding two-dimensional orthographic image (i.e., the surface data DOM text) of the survey area, so that the village range and river width can be accurately determined for water dynamics simulation and three-dimensional model construction, to determine the flooded area and make correct predictions.

[0027] Further, as shown in the drawings, Figure 3As shown in the figure, if RIEGL lidar is used to collect trajectory datasets, the trajectory datasets can be solved based on Pospac, or the point cloud data can be solved based on Riprocess combined with the three plug-ins Rriworld, Sdcimport, and Riserver and the trajectory solution file to obtain multiple flight strips. Since there are overlapping areas between adjacent flight strips, the data of the overlapping parts need to maintain consistency and ensure that the accuracy of the point cloud data in the overlapping area is consistent with that of other areas, it is necessary to splice multiple flight strips to eliminate the errors between flight strips.

[0028] The stitched flight strip map was then denoised. The first step in denoising was to configure the drone before takeoff. Based on drone flight experience, engineers concluded that noise concentration is typically within a range of 10 meters vertically downward from the drone. Therefore, the drone was set to automatically not collect reflected signals within 10 meters. Noise was then manually removed using Riprocess. Noise points and anomalies can be clearly observed in the front view.

[0029] like Figures 4-6 As shown, using Riprocess to take a vertical section of a certain width in the aerial view of the flight path chart can obtain a front view of the trees and canopy. Then, Riprocess can be used to remove vegetation. This is a comparison image of the removal in blocks. The left side of the white line is the area where vegetation has been removed. After all vegetation is removed, the corresponding results are exported and the coordinates are converted to the commonly used WGS_1984. During this process, Riprocess can reasonably interpolate areas with defects in the DEM and insufficient point cloud density. Then, Global Mapper is used to create an elevation network for the point cloud, and some areas without data after removing vegetation and buildings are interpolated to obtain the corresponding DEM terrain data file.

[0030] Further, if Figure 7 As shown in the figure, if the DJI L1 lidar is used to collect trajectory datasets, the internal mapping software can be used for processing. The main steps include POS solution (trajectory solution), fusion of point cloud and visible light data (forming a colored point cloud), point cloud output (LAS format, etc.), and job report output, directly obtaining LAS point cloud data with a projected coordinate system.

[0031] Then the LAS point cloud data is used to identify and remove surface buildings, overhead power lines, vegetation, etc. (i.e. data filtering). Considering that the point cloud density after the actual flight is higher than the required 100 points / m 2Much larger, and proven, under the premise of ensuring the density of point cloud, when the point cloud density of the canopy vegetation is relatively low, the effect of removing vegetation and retaining terrain is better. Therefore, when importing MicroStation CONNECT Edition, the dilution process is usually adopted, generally taking one point for every 5-8 points. Subsequently, all point clouds are first divided into a class, and then ground point division is performed based on this class of point cloud. The parameters mainly include the maximum building height (generally set to 60 m), the maximum terrain angle (generally set to 88°), the maximum iteration angle of the relative plane (generally set to 6°), the maximum iteration distance relative to the ground (generally set to 1.4 m), and the iteration angle is reduced when the side length is less than 5 m. As shown in Figure 8 , the above parameters are the results recommended by the software, and it has been verified that the effect of identifying ground objects meets the expectations, and the obtained overall point cloud is compared with the terrain point cloud. Finally, as before, GlobalMapper can be used to create an elevation network to obtain the corresponding DEM result (i.e., DEM terrain data file).

[0032] Finally, a sub-meter DEM digital elevation model can be constructed according to the terrain data file and the ground data text for later water dynamics simulation and three-dimensional model construction.

[0033] In step S102, radar real-time base data of a target area, radar and ground rainfall historical observation data are obtained, and actual radar real-time base data and short-term forecast data of a shower are solved according to the radar real-time base data, radar and ground rainfall historical observation data.

[0034] In some embodiments, radar real-time base data of a target area, radar and ground rainfall historical observation data are obtained, and actual radar real-time base data and short-term forecast data of a shower are solved according to the radar real-time base data, radar and ground rainfall historical observation data, including: Radar real-time base data of a target area, radar and ground rainfall historical observation data, historical multi-source weather data are obtained; The mixed elevation angle data is determined according to the radar real-time base data and the terrain data file; The radar quality control is performed on the mixed elevation angle data to obtain a resolution networking mosaic; The radar and ground rainfall historical observation data are used to train a pre-constructed radar rainfall inversion model based on a multi-layer perceptron to obtain a trained radar rainfall inversion model; The resolution networking mosaic is input into the trained radar rainfall inversion model to invert the rainfall intensity at the current time, and the radar real-time base data is corrected according to the rainfall intensity at the current time to obtain actual radar real-time base data; The pre-constructed radar rainfall field downscaling model based on an autoencoder is used to convert the resolution network mosaic into a low-resolution network mosaic. The pre-constructed deep spatiotemporal representation enhanced mountainous area rainfall nowcasting model is trained using historical multi-source meteorological data, radar, and historical ground rainfall observation data, to obtain a trained mountainous area rainfall nowcasting model. The low-resolution network mosaic is input into the trained mountainous area rainfall nowcasting model to predict future rainfall intensity, and the future rainfall intensity is used as the short-term rainstorm forecast data.

[0035] In actual execution, as shown in Figure 9 , S-band and X-band radar historical and real-time base data are obtained from the meteorological monitoring system of the target area. Since the X-band radar and the S-band radar are both dual-polarization radars, in addition to the reflectivity factor , they can also provide differential reflectivity , differential phase shift , specific differential phase shift , correlation coefficient , and other polarization observation quantities. However, due to the interference of factors such as (1) non-meteorological echoes such as terrain, biology, and electromagnetic radiation; (2) signal attenuation caused by cloud and rain particle scattering; and (3) beam blockage caused by terrain, buildings, and the like, various errors are often generated in the process of observing weather systems by radars. Therefore, quality control of the base data of each radar is required. In addition, multi-band radar network mosaics are used to achieve full coverage of mountainous area radar rainfall observation in the target area, and to maximize the mitigation of the impact of complex mountainous terrain blockage.

[0036] The specific steps of quality control can be: obtaining radar real-time base data from the meteorological monitoring system of the target area, analyzing the blockage of multiple radars in the target area according to the radar real-time base data and terrain data files, determining the minimum elevation angle of each radar with a blockage rate less than 50%, and obtaining the initial mixed elevation angle data. Then, combined with radar observations of multiple years, the minimum elevation angle is adjusted to exclude the influence of buildings, trees, and the like, and the final mixed elevation angle data is obtained. As shown in Figure 10 , the characteristics of differential reflectivity , differential phase shift , correlation coefficient , and other polarization observation quantities generated by rainfall and non-meteorological targets are analyzed, and a fuzzy logic classification is established based on this, which is used to identify and exclude non-meteorological echoes. Then, using the linear programming principle, the differential phase shift is quality controlled to ensure its monotonic increasing property, and based on the quality controlled differential phase shift , the specific differential phase shift is calculated using the linear least squares fitting principle, and the quality controlled differential phase shift Reflectivity factor and differential reflectivity Perform attenuation correction to obtain the resolution network puzzle.

[0037] Furthermore, radar and ground historical rainfall observations were obtained from the meteorological monitoring system in the target area, and deep learning methods were used to establish the relationship between radar parameters and terrain factors as well as precipitation characteristics (particle size spectrum, particle type, rainfall intensity). The calculation parameters suitable for various types of precipitation and local heavy rain in complex mountainous areas were inverted, and the real-time rainfall was quantitatively estimated using the radar parameters of local heavy rain in mountainous areas.

[0038] Among them, Figure 11 As shown in the figure, the inversion process of radar parameters for local heavy rain in mountainous areas can be specifically as follows: the traditional ZR relationship has high uncertainty and cannot consider the spatial structure of cloud clusters during inversion. Therefore, the embodiment of the present invention constructs a radar rainfall inversion model based on a multi-layer perceptron (MLP). The model contains 4 hidden fully connected layers, uses the reflectivity of 4 heights near the ground as the model input, considers the spatial structure of the rainfall cloud cluster, and inverts the rainfall intensity at the current moment. The obtained radar and ground historical rainfall observations are used to construct a rainfall inversion model including radar inversion. and rainfall observed by rain gauges Data set for building the initial correction relationship In real-time operation, the measured rainfall and inversion results The trained radar rainfall inversion model is obtained by integrating the resolution network puzzle into the dataset, updating the relationship every 2 hours, and manually updating and adjusting the relationship during the review after each rainfall. The resolution network puzzle can then be input into the trained radar rainfall inversion model to invert the rainfall intensity at the current moment, and the radar real-time base data can be corrected according to the current rainfall intensity to obtain the actual radar real-time base data.

[0039] Furthermore, if Figure 12 As shown, the embodiment of the present invention adopts an encoder-decoder structure to construct a radar rainfall field downscaling model based on an autoencoder, and uses radar and ground rainfall historical observation data to train encoder and decoder parameters, so that the radar rainfall field downscaling model based on the autoencoder learns the relationship between the high-resolution radar rainfall field and the low-resolution latent variables, so that the trained radar rainfall field downscaling model based on the autoencoder can be used to convert the high-resolution network puzzle into a low-resolution network puzzle.

[0040] The embodiment of the application combines the above downscaling method, and constructs a deep spatiotemporal feature enhancement mountainous rainfall short-term forecast model including a short-term forecast module based on a deep convolutional network and a spatiotemporal feature enhancement method, an upsampling module based on spatiotemporal attention, and a downscaling module based on spatiotemporal attention. The model fully utilizes deep convolutional network technology and spatiotemporal feature enhancement technology, can learn and mine the nonlinear variation law of a local rainfall cloud cluster from historical radar observation data, deduce the growth, evolution, development and extinction process of the rainfall cloud cluster, and realize accurate short-term forecast of rainfall intensity and cumulative rainfall in the next 0-2 hours. That is, low-resolution networked jigsaw puzzles are input into the trained mountainous rainfall short-term forecast model, and short-term forecast data of a rainstorm can be predicted.

[0041] It should be noted that, as Figure 13 shown, the short-term forecast module based on the deep convolutional network and the spatiotemporal feature enhancement method adopts a U-Net structure, receives a historical 30-minute (5-step) radar observation sequence as input, predicts a future 30-minute (5-step) radar sequence, and provides a 2-hour prediction result through multiple iterations. The encoding part of the U-Net contains a 1x1 convolution module for time feature reconstruction and four downscaling modules. The decoding part adopts the same structure, contains four upscaling blocks and a 1x1 convolution module. Multi-scale encoding feature mapping is directly connected to the corresponding decoding feature mapping in the upscaling module through a skip connection mechanism.

[0042] As Figure 14 shown, the upsampling and downscaling modules based on spatiotemporal attention are important components of the short-term forecast module, and the spatiotemporal attention mechanism is used to enhance the ability to capture the spatiotemporal features of the radar rainfall image. The upsampling module adopts a full convolution structure, and the downscaling module is composed of a 2x2 maximum pooling layer, two groups of convolution layers (including a 3x3 convolution layer, a BN layer and a ReLU layer), and a CBAM spatiotemporal attention layer. The upsampling module is composed of a 2x2 bilinear interpolation layer, two groups of convolution layers (including a 3x3 convolution layer, a BN layer and a ReLU layer), and a CBAM spatiotemporal attention layer.

[0043] In step S103, the brightness temperature data and geographic coordinate information of the target area are collected to solve the soil moisture data according to the brightness temperature data and geographic coordinate information.

[0044] In some embodiments, the brightness temperature data and geographic coordinate information of the target area are collected to solve the soil moisture data according to the brightness temperature data and geographic coordinate information, including: The brightness temperature data and geographic coordinate information of the target area are collected by a UAV carrying a microwave radiometer; According to the geographic coordinate information, auxiliary parameter data is extracted from a pre-constructed auxiliary database; The auxiliary parameter data and the brightness temperature data are input into a soil moisture inversion model constructed in advance to generate soil moisture data.

[0045] In actual execution, as shown in the figure, Figure 15 To automatically obtain the brightness temperature data of the target area, the embodiment sets flight parameters such as flight height, flight speed, flight range, flight angle in the flight control software, and uploads these settings to the unmanned aerial vehicle carrying the microwave radiometer through Bluetooth, so as to realize automatic cruising flight of the unmanned aerial vehicle carrying the microwave radiometer, and output brightness temperature data and flight log data (i.e. geographic coordinate information) after flight.

[0046] Further, to ensure that the unmanned aerial vehicle flight data and the microwave radiometer data are at the same time, the brightness temperature data and the geographic coordinate information need to be data arranged and cleaned to obtain preprocessed brightness temperature data and geographic coordinate information. In addition, the embodiment also extracts auxiliary reference data such as soil, vegetation, terrain and soil temperature in the target area according to the flight range of the unmanned aerial vehicle.

[0047] Finally, the auxiliary parameter data and the brightness temperature data are input into a soil moisture inversion model constructed in advance, the brightness temperature is converted into emissivity using the substitute value of the emission layer physical temperature, and then the dielectric constant is determined using the Fresnel equation. Finally, the surface soil moisture is obtained by using the dielectric constant mixing model.

[0048] In step S104, natural sub-basins, river networks and geometric parameters are extracted from the terrain text data, a preset runoff generation module is embedded into the natural sub-basins to obtain new natural sub-basins, a preset confluence module is embedded into the river networks to obtain new river networks, and an initial hydrological model is constructed according to the new natural sub-basins, the new river networks and the geometric parameters.

[0049] In step S105, the initial hydrological model is operated by using actual radar real-time base data, short-term forecast data of array rain and soil moisture data to obtain a result of the mountain torrent gully distributed hydrological model.

[0050] In actual implementation, in the embodiment of the application, the initial hydrological model is constructed based on the THREW model through the mountain torrent gully hydrological data arrangement, the model divides sub-basins based on terrain data, takes each sub-basin as a calculation unit, considers evapotranspiration based on the calculation of canopy interception evaporation, vegetation transpiration and bare soil evaporation based on the sub-basin grid vegetation type, solves soil water movement based on the one-dimensional Richards equation, calculates surface runoff and confluence based on the infiltration formula and the backwater coefficient formula, and adopts the Muskingum method for river confluence. However, the runoff generation mechanism based on the existing hydrological model is the surface runoff mechanism such as storage and overland flow, and the latest research shows that the dominant runoff generation mechanism of mountain torrent rainstorm flood is interflow or shallow groundwater runoff in the strong weathering zone, therefore, the initial hydrological model is updated by using the aforementioned obtained actual radar real-time base data, short-term forecast data of rainstorm and soil moisture data, so as to obtain the mountain torrent gully basin distributed hydrological model meeting the latest scientific cognition of the flood formation mechanism.

[0051] Further, in order to improve the model updating efficiency, the mountain torrent gully basin distributed hydrological model rapid construction tool including the key functions of efficient and automatic natural sub-basin division and rapid extraction, rapid extraction of basin geometric parameters and physical parameters, rapid preprocessing of driving data and efficient optimization of model parameters is developed based on the geographic information spatial analysis method, so as to realize the rapid construction of the mountain torrent gully basin hydrological model.

[0052] It should be noted that the mountain torrent gully basin distributed hydrological model rapid construction tool proposed in the embodiment of the application mainly includes the following contents: (1) The proposed efficient and automatic natural sub-basin division and geometric parameter extraction technology. Quantitative terrain analysis is essential in distributed hydrological models. Natural sub-basins, as the calculation units of distributed hydrological models, have the greatest advantage in that the hydrological processes within and between units are clear, and the unit hydrological model can easily introduce traditional hydrological models, thereby simplifying calculations and shortening model development time. The conventional basin extraction process includes steps such as filling, calculating flow direction, calculating flow accumulation matrix, setting threshold, capturing the point of inclination, river network grading, sub-basin division, and vectorization. However, this method of determining the river network only through the flow accumulation threshold often leads to deviations in the extracted river network from the actual situation, uneven sub-basin areas, and river lengths. The Muskingum method and other hydrological methods require that the lengths of each river section be approximately equal. Therefore, based on the Pfafstetter method, a new natural sub-basin division method is developed to automatically complete the steps of filling, calculating flow direction, calculating flow accumulation matrix, setting threshold, capturing the point of inclination, river network grading, sub-basin division, and vectorization. The method is self-adaptive according to the specified outlet, maximum / minimum basin area, and further encrypts the sub-basins with excessive area. It also appropriately merges areas with excessively small areas and excessively dense divisions. At the same time, the method realizes the rapid extraction of basin geometric and physical parameters such as slope, river slope, river length, and representative unit area, providing support for the rapid construction of mountain flood hydrological models.

[0053] (2) The proposed rapid pre-processing technology for driving data. Based on the collected and real-time updated rainfall station data, radar monitoring data, and weather station data, meteorological data for each mountain flood ditch basin is rapidly produced through interpolation, data fusion, and other techniques. Spatial data statistics methods are used to statistically and organize the meteorological data on the calculation units (sub-basins) in the model, providing support for rolling prediction and real-time correction.

[0054] (3) The proposed efficient parameter optimization algorithm. Parameter optimization is one of the most time-consuming tasks in the process of building a distributed hydrological model. The NSGA-II algorithm is improved, and a more efficient parameter optimization algorithm is developed based on an asynchronous parallel architecture. A high-efficiency parameter optimization program with high convergence, strong compatibility, parallelism, and portability is developed to achieve efficient and batch optimization of model parameters for a single or multiple basins.

[0055] In step S106, a hydrology-hydrodynamics coupled model is established based on the sub-meter DEM digital elevation model and the results of the mountain flood ditch basin distributed hydrological model.

[0056] In step S107, the current and future flood states of the target area are predicted based on the hydrology-hydrodynamics coupled model.

[0057] In some embodiments, a hydrology-hydrodynamics coupling model is established according to a sub-meter DEM digital elevation model and a result of a mountain torrent watershed distributed hydrology model, comprising: A two-dimensional refined topographic model of a mountain torrent in the target region is constructed according to a sub-meter DEM digital elevation model; Terrain data, hydrology data and hydrodynamic data in the three-dimensional refined topographic model of the mountain torrent are extracted; A hydrology-hydrodynamics model is constructed according to the terrain data, the hydrology data and the hydrodynamic data based on a CPU and GPU heterogeneous parallel acceleration algorithm; The hydrology-hydrodynamics model and the mountain torrent watershed distributed hydrology model are coupled to obtain a hydrology-hydrodynamics coupling model.

[0058] In actual execution, the three-dimensional representation precision of the topography of the mountain torrent directly affects the precision and scale of the mountain torrent numerical calculation, because the topography of the mountain torrent is relatively complex. Therefore, on the basis of analyzing the data platform of the mountain torrent watershed, establishing a refined numerical model, an efficient calculation model algorithm and appropriate parameters of the mountain torrent is the key to realizing high-precision and efficient hydrodynamics simulation of the mountain torrent.

[0059] In the embodiment of the application, DEM topographic data of a channel in a target region is used as basic data to construct a sub-meter DEM digital elevation model, and then a three-dimensional refined topographic model of a mountain torrent in the target region is constructed according to the sub-meter DEM digital elevation model. In the construction process, a triangular meshing technique is used to realize adaptive local encryption technology based on the three-dimensional geometric model of the channel, so as to overcome the shortcomings of using DEM data format, such as a large number of meshes and loss of local precision. Based on unstructured meshes, the river and the surrounding area that may be flooded by the flood are used as key areas using encrypted meshes, and the ridges and other areas that will not be flooded are used as sparse meshes. The limited computing resources are used for the calculation of the key areas to maximize the calculation efficiency. At the same time, the LOD2 level DOM of the channel near the village residential area is pasted to the three-dimensional mesh of the terrain to realize the three-dimensional of the village form of the flooding model.

[0060] Based on the three-dimensional refined model of the mountain torrent established above, the simulation of the mountain flood evolution process is carried out. Due to the improvement of the calculation precision of the model, the calculation grid (scale) is greatly increased, and efficient and large-scale calculation is the key to solving the refined hydrodynamic simulation of the mountain torrent. Therefore, in the embodiment of the application, a hydrology-hydrodynamics model based on unstructured meshes is constructed based on the CPU and GPU heterogeneous parallel acceleration algorithm based on the two-dimensional shallow water wave equation as the theoretical basis, and the specific construction process is as follows: Hydrology calculation mainly uses the two-dimensional shallow water wave equation as the main theory, and the finite volume method is used to realize the solution of the algorithm. The two-dimensional shallow water wave equation is as follows:

[0061]

[0062]

[0063] wherein, is the water depth of the flood, , are the flow rates of the flood in the , directions, is the effective rainfall, is the inflow, , are the flow velocities of the flood in the , directions, , is the terrain elevation, is the gravitational acceleration, is the Manning roughness coefficient.

[0064] The flood routing calculation is mainly divided into the following three steps: (1) According to the three-dimensional terrain, flood flow and other input files, initialize the model, and transfer the data from the CPU to the GPU.

[0065] (2) Call the GPU kernel function for parallel calculation, including: discretize and solve the momentum conservation equation according to the water depth , to obtain the flow rates , in the , directions.

[0066] discretize and solve the mass conservation equation according to the flow rates , and the terrain elevation , to obtain the water depth .

[0067] According to whether the water depth is 0, check and update the flow rates , .

[0068] (3) Update the time step, if output is needed, transfer the data from the GPU to the CPU and write to the output file, if output is not needed, proceed to the calculation of the next time step.

[0069] Further, as shown in FIG. 1, the part of the CPU-GPU code that needs a lot of calculation is executed on the GPU through parallel processing, and the other part is executed on the CPU through sequential processing. Figure 16

[0070] ​First, the input data is read and stored in the CPU memory, and then the data used in the kernel of the CUDA C++ implementation is copied to the GPU memory for parallel processing. Next, the shallow water equation is iteratively solved, where the calculations distributed in space are performed in parallel on the GPU, but the time evolution is performed sequentially on the CPU. The specific solving steps can be referred to the previous section.

[0071] Parallel computing is performed by CUDA kernels through processing entities organized in grids, blocks, and threads. Each block contains multiple threads. The GPU called in the program is set to have a maximum of 256 threads per block. Therefore, for a hydraulic model with a grid number less than 256, one block is called for calculation. For a grid number of more than 256, multiple blocks can be used for calculation. The final number of CUDA threads is equal to the number of triangular meshes of the model. When the GPU is called for calculation, each thread is performing calculation, which is equivalent to each mesh being calculated in parallel at the same time. Compared with the loop calculation of the CPU, which must wait for the previous mesh to be calculated before calculating the next mesh, the efficiency is greatly improved.

[0072] Further, the hydrological and hydrodynamic model and the distributed hydrological model of the mountain torrent gully basin are coupled to obtain a hydrological-hydrodynamic coupling model, so that the current flood state and the future flood state of the target region can be predicted according to the hydrological-hydrodynamic coupling model, i.e., the current flow rate, the current water depth, the current inundation range, the future flow rate, the future water depth and the future inundation range of the target region.

[0073] In summary, the flood forecasting method based on field monitoring and real-time modeling according to the embodiments of the present application has the following beneficial effects: (1) The work of measuring terrain by unmanned aerial vehicle carrying laser radar is applied to flood forecasting and early warning, solving the problems of battery and efficiency that exist in the market. In the data post-processing, accurate and coherent data correction, splicing and solving process are explored, and automation is realized in terms of surface point cloud identification, bridge removal, channel extraction and other aspects for the phenomenon that the dominant runoff mechanism is shallow subsurface runoff in the strongly weathered zone, improving the efficiency and accuracy of obtaining key hydrological information of mountain torrent basin; (2) By completing the radar and ground rainfall observation data processing, mountainous multi-band radar data quality control and networking splicing, mountainous rainstorm radar parameter inversion, fine rainfall real-time data fusion in mountainous areas is realized; mountainous multi-source meteorological data fusion is carried out, and the effect of mountainous weather field downscaling technology and rainfall short-term forecasting technology is significantly improved in rainfall inversion and short-term forecasting, so that the existing distributed hydrological model of mountain torrent gully basin can be better updated, and the formation mechanism of mountain flood is fully reflected; (3) By collating the terrain and underlying surface data of mountain torrent gullies, an efficient and high-resolution hydraulic model of mountain torrent gullies is constructed to achieve efficient and large-scale hydrodynamic simulation calculations, as well as simulation of mountain torrent dynamic scenarios under different working conditions and simulation of the evolution process of mountain floods; and a fast modeling tool for efficient and high-resolution hydraulic models of mountain torrent gullies is constructed to improve modeling efficiency and optimize calculation efficiency; (5) The hydrological-hydrodynamic coupling model established based on the high-efficiency and high-resolution hydraulic model of the mountain torrent gully and the distributed hydrological model of the mountain torrent gully basin has improved the accuracy of predictions and can ensure the safety of people's lives and high-quality economic and social development.

[0074] Next, a flood forecasting device based on field monitoring and real-time modeling according to an embodiment of the present invention will be described with reference to the accompanying drawings.

[0075] Figure 16 A block diagram of a flood forecasting device based on field monitoring and real-time modeling provided by an embodiment of the present invention.

[0076] like Figure 16 As shown, the flood forecasting device 160 based on field monitoring and real-time modeling includes: a first construction module 1601, a first solution module 1602, a second solution module 1603, a second construction module 1604, a third construction module 1605, a fourth construction module 1606 and a forecasting module 1607.

[0077] The first construction module 1601 is configured to collect terrain data files and surface data texts of a target region, and construct a sub-meter DEM digital elevation model according to the terrain data files and the surface data texts. The first solving module 1602 is configured to obtain radar real-time base data, radar and ground rainfall historical observation data of the target region, and solve actual radar real-time base data and short-term shower forecast data according to the radar real-time base data, the radar and the ground rainfall historical observation data. The second solving module 1603 is configured to collect brightness temperature data and geographic coordinate information of the target region, so as to solve soil moisture data according to the brightness temperature data and the geographic coordinate information. The second construction module 1604 is configured to extract natural sub-basins, river networks and geometric parameters from the terrain text data, embed a preset runoff generation module into the natural sub-basins to obtain new natural sub-basins, embed a preset confluence module into the river networks to obtain new river networks, and construct an initial hydrological model according to the new natural sub-basins, the new river networks and the geometric parameters. The third construction module 1605 is configured to operate the initial hydrological model by using the actual radar real-time base data, the short-term shower forecast data and the soil moisture data, so as to obtain a result of a mountain torrent valley distributed hydrological model. The fourth construction module 1606 is configured to establish a hydrology-hydrodynamics coupled model according to the sub-meter DEM digital elevation model and the result of the mountain torrent valley distributed hydrological model. The prediction module 1607 is configured to predict a current flood state and a future flood state of the target region according to the hydrology-hydrodynamics coupled model.

[0078] In some embodiments, the first construction module 1601 comprises: A trajectory acquisition unit is configured to control a drone carrying a laser radar to perform a preset flight route and a preset flight number over a target region, so as to obtain a trajectory data set; A solution unit is configured to perform trajectory solution and point cloud solution on the trajectory data set, so as to obtain a strip map; A generation unit is configured to perform data filtering on the strip map, so as to generate terrain data files and surface data texts.

[0079] In some embodiments, the first solving module 1602 comprises: An acquisition unit is configured to acquire radar real-time base data, radar and ground rainfall historical observation data, and historical multi-source meteorological data of a target region; A determination unit is configured to determine mixed elevation angle data according to the radar real-time base data and terrain data files; A quality control unit is configured to perform radar quality control on the mixed elevation angle data, so as to obtain a resolution group network mosaic; A training unit is configured to train a pre-constructed multi-layer perceptron-based radar rainfall inversion model by using radar and ground rainfall historical observation data, so as to obtain a trained radar rainfall inversion model; An inversion correction unit is configured to input the resolution networking puzzle into a trained radar rainfall inversion model to invert a current rainfall intensity, and correct radar real-time base data according to the current rainfall intensity to obtain actual radar real-time base data. A conversion unit is configured to convert the resolution networking puzzle into a low-resolution networking puzzle by using a pre-constructed radar rainfall field downscaling model based on an autoencoder. A training unit is configured to train a pre-constructed deep spatiotemporal representation enhanced mountainous area rainfall nowcasting model by using historical multi-source meteorological data, radar and historical observation data of ground rainfall to obtain a trained mountainous area rainfall nowcasting model. A prediction unit is configured to input the low-resolution networking puzzle into the trained mountainous area rainfall nowcasting model to predict future rainfall intensity, and use the future rainfall intensity as short-term rainstorm forecasting data.

[0080] In some embodiments, the second solving module 1603 includes: A collection of brightness temperature unit is configured to collect brightness temperature data and geographic coordinate information of a target area by using a drone carrying a microwave radiometer; A first extraction unit is configured to extract auxiliary parameter data from a pre-constructed auxiliary database according to the geographic coordinate information; A generation unit is configured to input the auxiliary parameter data and the brightness temperature data into a pre-constructed soil moisture inversion model to generate soil moisture data.

[0081] In some embodiments, the deep spatiotemporal representation enhanced mountainous area rainfall nowcasting model includes a nowcasting module based on a deep convolutional network and a spatiotemporal representation enhancement method, an upsampling module based on spatiotemporal attention, and a downsampling module based on spatiotemporal attention, wherein The nowcasting module based on the deep convolutional network and the spatiotemporal representation enhancement method adopts a U-Net structure, receives radar observation sequences in historical multi-source meteorological data, radar and historical observation data of ground rainfall as input to predict future radar sequences; The upsampling and downsampling modules based on spatiotemporal attention adopt a full convolutional structure, and capture spatiotemporal features of radar rainfall images in the radar observation sequences through a spatiotemporal attention mechanism.

[0082] In some embodiments, the fourth construction module 1606 includes: A three-dimensional model construction unit is configured to construct a two-dimensional refined topographic model of a mountain torrent valley of a target area according to a sub-meter DEM digital elevation model; A second extraction unit is configured to extract topographic data, hydrological data and hydrodynamic data in the three-dimensional refined topographic model of the mountain torrent valley; A hydrodynamic unit is constructed for a CPU and GPU heterogeneous parallel acceleration algorithm, and a hydrology and hydrodynamic model is constructed according to terrain data, hydrology data and hydrodynamic data; A coupling unit is configured to couple the hydrology and hydrodynamic model and the result of the mountain torrent gully basin distributed hydrology model to obtain a hydrology-hydrodynamic coupling model.

[0083] It should be noted that the foregoing explanation of the flood forecasting method based on field monitoring and real-time modeling also applies to the flood forecasting device based on field monitoring and real-time modeling, which will not be repeated here.

[0084] The flood forecasting device based on field monitoring and real-time modeling has the following advantages: (1) The work of measuring terrain by unmanned aerial vehicle carrying laser radar is applied to flood forecasting and early warning, solving the problems of battery and efficiency on the market, exploring accurate and continuous data correction, splicing and solving process in data post-processing, and realizing automatic operation in terms of ground point cloud identification, bridge removal, channel extraction and other aspects for the phenomenon that the dominant runoff mechanism is shallow subsurface runoff in the weathered zone, improving the efficiency and accuracy of obtaining key hydrological information of mountain torrent basin; (2) By completing radar and ground rainfall observation data processing, mountainous multi-band radar data quality control and networking splicing, mountain storm radar parameter inversion, fine rainfall real-time data fusion in mountainous areas is realized; mountainous multi-source meteorological data fusion is carried out, and the effect of mountainous meteorological field scale technology and rainfall short-term prediction technology is significantly improved in rainfall inversion and short-term prediction, so that the existing mountain torrent gully basin distributed hydrology model can be better updated to fully reflect the mountain flood formation mechanism; (3) By processing mountain torrent gully terrain and underlying surface data, a high-efficiency and high-resolution hydrology model of mountain torrent gully is constructed to realize efficient and large-scale hydrodynamic simulation calculation, mountain torrent dynamics scenario simulation under different working conditions, and mountain flood evolution process simulation; and a high-efficiency and high-resolution hydrology model of mountain torrent gully is constructed to improve modeling efficiency and optimize calculation efficiency; (5) The hydrology-hydrodynamic coupling model established according to the high-efficiency and high-resolution hydrology model of mountain torrent gully and the mountain torrent gully basin distributed hydrology model improves the prediction accuracy and can guarantee the safety of people's life and the high-quality development of economy and society.

[0085] Figure 17 The mobile flood forecasting device provided in the embodiment of the present application is shown in the structural schematic diagram. The mobile flood forecasting device can include: The memory 1701, the processor 1702, and the computer program stored in the memory 1701 and executable on the processor 1702.

[0086] The processor 1702 implements the flood forecasting method based on field monitoring and real-time modeling provided in the above embodiments when executing a program.

[0087] Further, the mobile flood forecasting device further comprises: The communication interface 1703 is configured to communicate between the memory 1701 and the processor 1702.

[0088] The memory 1701 is configured to store a computer program executable on the processor 1702.

[0089] The memory 1701 can include a high-speed RAM memory, and can further include a non-volatile memory, for example, at least one disk memory.

[0090] If the memory 1701, the processor 1702 and the communication interface 1703 are implemented independently, the communication interface 1703, the memory 1701 and the processor 1702 can be connected to each other through a bus and complete communication between each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 17 In the figure, only one thick line is used to represent, but it does not mean that there is only one bus or one type of bus.

[0091] Optionally, in a specific implementation, if the memory 1701, the processor 1702 and the communication interface 1703 are integrated on a chip, the memory 1701, the processor 1702 and the communication interface 1703 can complete communication between each other through an internal interface.

[0092] The processor 1702 can be a Central Processing Unit (CPU), or an Application Specific Integrated Circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0093] The embodiments of the present application further provide a computer program product, and the computer program / instruction is executed by the processor to implement the flood forecasting method based on field monitoring and real-time modeling as above.

[0094] The application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the flood forecasting method based on field monitoring and real-time modeling.

[0095] In the description of the specification, the description of the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the application. In the specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or N embodiments or examples. In addition, the person skilled in the art can combine and combine the different embodiments or examples described in the specification and the features of the different embodiments or examples without contradiction.

[0096] In addition, the terms "first", "second" are only for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include at least one of the features. In the description of the application, the meaning of "N" is at least two, for example, two, three, etc., unless otherwise specifically limited.

[0097] Any process or method descriptions in flow charts or otherwise described herein can be understood as representing code modules, segments, or portions of code that include one or more executable instructions for performing a step of a process, and that the scope of the preferred embodiments of the application encompasses alterations, modifications, and variations of these codes, including but not limited to those that do not perform the steps of the process in the order as shown or discussed, including substantially simultaneous or in reverse order. This should be understood by those skilled in the art.

[0098] The logic and / or steps represented in flow diagrams or otherwise described herein, for example, can be considered as a sequence of instructions to implement logic functions, and can be embodied in any computer-readable medium for use by an instruction execution system, apparatus, or device, such as a computer-based system, processor- containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. In the context of this specification, a "computer-readable medium" can be any means that can contain, store, communicate, propagate or transport the program for use by or in connection with the instruction execution system, apparatus, or device. The computer-readable medium can be a machine-readable storage device (e.g., magnetic, optical or other) a machine-readable storage diskette (e.g., floppy, flexible or other), a machine-readable storage card (e.g., ROM, EEPROM, flash memory or other), a machine- readable storage tape (e.g., magnetic, optical or other), a machine-readable storage medium (e.g., a portable electronic device, a computer diskette, a computer memory stick, a computer hard drive, a computer tape, a computer readable storage medium, or other), or a machine-readable wireless transmission (e.g., a radio frequency signal, an infrared signal, a microwave signal, or other). More specific examples (a non-exhaustive list) of the computer-readable medium include the following: an electrical connection (electronic) having one or more wires, a portable computer diskette (magnetic), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can even be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, via, for example, optical scanning of the paper or other medium, then compiled, interpreted, or otherwise processed in a suitable manner, if necessary, and then stored in a computer memory.

[0099] It should be understood that aspects of the application can be implemented in hardware, software, firmware or combinations thereof. In the above embodiments, the N steps or methods can be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. If implemented in hardware and in another embodiment, any of the following technologies known in the art or their combinations can be used: discrete logic circuitry having logic gates for implementing logic functions on data signals, application specific integrated circuits having appropriate combinational logic gates, programmable gate arrays (PGA), field programmable gate arrays (FPGA), etc.

[0100] Those skilled in the art can understand that all or part of the steps carried out by the above-mentioned embodiment methods can be completed by programs instructing related hardware, and the programs can be stored in a computer-readable storage medium. When the programs are executed, they include one of the steps of the method embodiments or a combination thereof.

[0101] In addition, each function unit in each embodiment of the present application can be integrated in one processing module, or each unit can be physically present separately, or two or more units can be integrated in one module. The integrated module can be realized in the form of hardware or in the form of a software function module. When the integrated module is realized in the form of a software function module and sold or used as an independent product, it can also be stored in a computer readable storage medium.

[0102] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it should be understood that the above embodiments are exemplary and should not be construed as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above embodiments within the scope of the present application.

Claims

1. A flood forecasting method based on field monitoring and real-time modeling, characterized in that: The following steps are involved: Collecting terrain data files and surface data text of the target area, and constructing a sub-meter DEM digital elevation model based on the terrain data files and the surface data text; Acquire radar real-time base data and radar and ground rainfall historical observation data of the target area, and solve actual radar real-time base data and shower short-term forecast data based on the radar real-time base data and the radar and ground rainfall historical observation data; collecting brightness temperature data and geographic coordinate information of the target area to solve soil moisture data based on the brightness temperature data and the geographic coordinate information; Extracting natural sub-watersheds, river networks, and geometric parameters from the terrain text data, embedding a preset runoff generation module into the natural sub-watersheds to obtain a new natural sub-watershed, embedding a preset confluence module into the river network to obtain a new river network, and constructing an initial hydrological model based on the new natural sub-watersheds, the new river network, and the geometric parameters; Calculating the initial hydrological model using the actual radar real-time base data, the shower short-term forecast data, and the soil moisture data to obtain a distributed hydrological model result for a mountain torrent gully basin; Establishing a hydrological-hydrodynamic coupling model based on the sub-meter DEM digital elevation model and the distributed hydrological model of the mountain torrent basin; The current flood status and future flood status of the target area are predicted based on the hydrological-hydrodynamic coupling model.

2. The flood forecasting method based on field monitoring and real-time modeling according to claim 1 is characterized in that: The terrain data file and surface data text of the acquisition target area include: Controlling a UAV equipped with a laser radar to execute a preset route and a preset number of flights over the target area to obtain a trajectory data set; Performing trajectory calculation and point cloud calculation on the trajectory data set to obtain a flight path chart; Data filtering is performed on the flight strip chart pair to generate the terrain data file and the surface data text.

3. The flood forecasting method based on field monitoring and real-time modeling according to claim 1 is characterized in that: The step of acquiring radar real-time base data and radar and ground rainfall historical observation data of the target area, and solving actual radar real-time base data and shower short-term forecast data based on the radar real-time base data and the radar and ground rainfall historical observation data, includes: Acquire radar real-time base data, radar and ground rainfall historical observation data, and historical multi-source meteorological data of the target area; Determining mixed elevation angle data according to the radar real-time base data and the terrain data file; performing radar quality control on the hybrid elevation data to obtain a resolution network mosaic; Using the radar and ground rainfall historical observation data, a pre-built radar rainfall inversion model based on a multi-layer perceptron is trained to obtain a trained radar rainfall inversion model; Inputting the resolution network puzzle into the trained radar rainfall inversion model to invert the rainfall intensity at the current moment, and correcting the radar real-time base data according to the rainfall intensity at the current moment to obtain actual radar real-time base data; The resolution network puzzle is converted into a low-resolution network puzzle using a pre-built radar rainfall field downscaling model based on an autoencoder; Using the historical multi-source meteorological data and the radar and ground rainfall historical observation data, a pre-constructed deep spatiotemporal representation enhanced short-term rainfall forecast model for mountainous areas is trained to obtain a trained short-term rainfall forecast model for mountainous areas; The low-resolution network mosaic is input into the trained short-term rainfall forecast model for mountainous areas to predict future rainfall intensity, and the future rainfall intensity is used as the short-term shower forecast data.

4. The flood forecasting method based on field monitoring and real-time modeling according to claim 1 is characterized in that: The collecting of brightness temperature data and geographic coordinate information of the target area to solve soil moisture data according to the brightness temperature data and the geographic coordinate information includes: Using an unmanned aerial vehicle equipped with a microwave radiometer to collect brightness temperature data and geographic coordinate information of the target area; Extracting auxiliary parameter data from a pre-built auxiliary database according to the geographic coordinate information; The auxiliary parameter data and the brightness temperature data are input into a pre-built soil moisture inversion model to generate the soil moisture data.

5. The flood forecasting method based on field monitoring and real-time modeling according to claim 3 is characterized in that: The short-term rainfall forecast model for mountainous areas enhanced by deep spatiotemporal representation includes a short-term forecast module based on a deep convolutional network and a spatiotemporal representation enhancement method, an upsampling module based on spatiotemporal attention, and a downsampling module based on spatiotemporal attention, wherein: The short-term forecast module based on the deep convolutional network and the spatiotemporal representation enhancement method adopts a U-Net structure, receives the radar observation sequence in the historical multi-source meteorological data and the radar and ground rainfall historical observation data as input, and predicts the future radar sequence; The spatiotemporal attention-based upsampling and downsampling module adopts a fully convolutional structure and captures the spatiotemporal features of the radar rainfall image in the radar observation sequence through a spatiotemporal attention mechanism.

6. The flood forecasting method based on field monitoring and real-time modeling according to claim 1 is characterized in that: The hydrological-hydrodynamic coupling model is established based on the sub-meter DEM digital elevation model and the distributed hydrological model results of the mountain torrent gully basin, including: Constructing a two-dimensional refined terrain model of the mountain torrent gully in the target area based on the sub-meter DEM digital elevation model; Extracting terrain data, hydrological data and hydrodynamic data from the three-dimensional refined terrain model of the mountain torrent gully; Based on the CPU and GPU heterogeneous parallel acceleration algorithm, a hydrological and hydrodynamic model is constructed according to the terrain data, the hydrological data and the hydrodynamic data; The hydrological-hydrodynamic model is coupled with the distributed hydrological model results of the mountain torrent gully basin to obtain the hydrological-hydrodynamic coupling model.

7. A flood forecasting device based on field monitoring and real-time modeling, characterized in that: include: The first construction module is used to collect terrain data files and surface data texts of the target area, and construct a sub-meter DEM digital elevation model based on the terrain data files and the surface data texts; a first solving module, configured to obtain radar real-time base data and radar and ground rainfall historical observation data of the target area, and solve actual radar real-time base data and shower short-term forecast data based on the radar real-time base data and the radar and ground rainfall historical observation data; a second solving module, configured to collect brightness temperature data and geographic coordinate information of the target area, and solve soil moisture data based on the brightness temperature data and the geographic coordinate information; A second construction module is configured to extract natural sub-watersheds, river networks, and geometric parameters from the terrain text data, embed a preset runoff generation module into the natural sub-watersheds to obtain new natural sub-watersheds, embed a preset confluence module into the river network to obtain a new river network, and construct an initial hydrological model based on the new natural sub-watersheds, the new river network, and the geometric parameters; A third construction module is configured to calculate the initial hydrological model using the actual radar real-time base data, the shower short-term forecast data, and the soil moisture data to obtain a distributed hydrological model result for a mountain torrent gully basin; A fourth construction module is used to establish a hydrological-hydrodynamic coupling model based on the sub-meter DEM digital elevation model and the distributed hydrological model results of the mountain torrent gully basin; A forecasting module is used to forecast the current flood status and future flood status of the target area based on the hydrological-hydrodynamic coupling model.

8. A mobile flood forecasting device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the flood forecasting method based on field monitoring and real-time modeling as described in any one of claims 1 to 6.

9. A computer program product, characterized in that When the computer program / instruction is executed by a processor, the flood forecasting method based on field monitoring and real-time modeling described in any one of claims 1 to 6 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement the flood forecasting method based on field monitoring and real-time modeling as described in any one of claims 1 to 6.

Images