Mountain flood identification method and device based on earth big data, storage medium and equipment

Abstract

The application relates to the technical field of mountain torrent identification, and specifically discloses a mountain torrent identification method and device based on earth big data, a storage medium and equipment, the method comprising the following steps: acquiring geographical environment data and meteorological characteristic data of a target river basin; performing grid division on the target river basin based on the geographical environment data to obtain a plurality of grids and construct a grid adjacency matrix; determining actual rainfall values and rainfall prediction values of each grid at each time according to the meteorological characteristic data; for each grid, determining target adjacent grids based on the grid adjacency matrix, and determining water quantity parameters and water flow direction parameters of the grid through a preset rainfall analysis model according to the geographical environment data of the grid, the actual rainfall values and the rainfall prediction values at each time, and the geographical environment data of the target adjacent grids, the actual rainfall values and the rainfall prediction values at each time; determining the mountain torrent grades and the mountain torrent damage ranges of the grids according to the water quantity parameters, and generating a mountain torrent movement track according to the water flow direction parameters.

Description

Technical Field

[0001] This application relates to the field of flash flood identification technology, and in particular to a flash flood identification method, device, storage medium, and equipment based on Earth big data. Background Technology

[0002] Flash floods, characterized by their suddenness and destructive power, pose a serious threat to the lives and property of residents in mountainous areas, as well as to infrastructure. With the acceleration of global climate change and urbanization, the frequency and intensity of extreme weather events have increased significantly, making rapid identification and accurate early warning of flash floods a key need in the field of disaster prevention and mitigation.

[0003] Existing flash flood identification technologies mainly include the single-station critical rainfall method, the regional critical rainfall method, and the watershed model method. The single-station critical rainfall method calculates the critical rainfall value for a single station by comprehensively comparing and filtering long-term monitoring data from different rain gauge stations. This critical rainfall value is then compared with the actual measured rainfall to determine whether a flash flood has occurred. The regional critical rainfall method calculates the regional critical rainfall value by statistically analyzing the critical rainfall data from various rain gauge stations within a specific region for the same time period. This regional critical rainfall value is then combined with the actual measured rainfall to identify flash floods. The watershed model method uses mathematical models and geographic information system (GIS) technology to model and analyze various elements within a watershed, such as topography, hydrology, and meteorology, thereby enabling real-time monitoring, identification, and early warning of flash floods throughout the entire watershed. However, although the single-station critical rainfall method and the regional critical rainfall method judge flash floods by comparing station rainfall with thresholds, their early prediction performance is poor and cannot meet the needs of early flash flood prevention. Moreover, the above methods cannot reflect the spatial differences of flash floods within the watershed, resulting in a lack of accuracy in flash flood impact assessment and failing to meet actual disaster prevention needs. Summary of the Invention

[0004] In view of this, this application provides a method, device, storage medium, and computer equipment for flash flood identification based on Earth big data. It achieves refined grid division of the target watershed using geographic environmental data and captures spatial dependencies using a grid adjacency matrix, overcoming the low spatial resolution of traditional methods. Secondly, by fusing actual rainfall values, predicted rainfall values, and geographic environmental data from multiple moments between the grid and its target adjacent grids, and dynamically calculating the water volume and flow direction parameters of the grid using a deep learning model, it improves data timeliness and accuracy. Finally, based on the water volume and flow direction parameters, it generates flash flood level distribution and movement trajectories, enabling a quantitative assessment of the disaster's scope and severity.

[0005] According to one aspect of this application, a flash flood identification method based on Earth big data is provided, comprising: The target watershed for flash flood identification is determined, and the corresponding geographical environment data and meteorological characteristic data at multiple times are obtained for the target watershed. Based on the geographic environment data, the target watershed is divided into multiple grids, and a grid adjacency matrix is ​​constructed based on the geographic location relationship between the grids. Based on the meteorological characteristic data of the target watershed at multiple times, the actual rainfall value of each grid at each time and the rainfall prediction value for the future preset time period at each time are determined respectively. For each grid, based on the grid adjacency matrix, the target adjacent grid corresponding to the grid is determined. According to the geographic environment data corresponding to the grid, the actual rainfall value and the predicted rainfall value at each time, and the geographic environment data corresponding to the target adjacent grid, the actual rainfall value and the predicted rainfall value at each time, the water volume parameters and water flow direction parameters corresponding to the grid are determined by a preset rainfall analysis model. Based on the water volume parameters corresponding to each grid, the flash flood level of each grid and the flash flood hazard range of the target watershed are determined, and the flash flood movement trajectory of the target watershed is generated based on the water flow direction parameters corresponding to each grid.

[0006] According to another aspect of this application, a flash flood identification device based on Earth big data is provided, comprising: The data acquisition module is used to determine the target watershed for flash flood identification and acquire the corresponding geographical environment data and meteorological characteristic data at multiple times for the target watershed. The grid partitioning module is used to partition the target watershed into multiple grids based on the geographic environment data, and to construct a grid adjacency matrix based on the geographic location relationship between the grids. The rainfall calculation module is used to determine the actual rainfall value of each grid at each time and the rainfall prediction value for a future preset time period at each time, based on the meteorological characteristic data of multiple times corresponding to the target watershed. The model analysis module is used to determine the target adjacent grids corresponding to each grid based on the grid adjacency matrix. Based on the geographic environment data corresponding to the grid, the actual rainfall value and the predicted rainfall value at each time, and the geographic environment data corresponding to the target adjacent grid, the actual rainfall value and the predicted rainfall value at each time, the module determines the water volume parameters and water flow direction parameters corresponding to the grid through a preset rainfall analysis model. The result generation module is used to determine the flash flood level of each grid and the flash flood hazard range of the target watershed based on the water volume parameters corresponding to each grid, and to generate the flash flood movement trajectory of the target watershed based on the water flow direction parameters corresponding to each grid.

[0007] According to another aspect of this application, a storage medium is provided that stores a computer program thereon, which, when executed by a processor, implements the above-described flash flood identification method based on Earth big data.

[0008] According to another aspect of this application, a computer device is provided, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor executes the program to implement the above-described flash flood identification method based on Earth big data.

[0009] By employing the above technical solution, this application provides a flash flood identification method and device, storage medium, and computer equipment based on Earth big data. First, the specific target watershed requiring flash flood identification can be clearly identified. After determining the target watershed, corresponding geographical environment data and meteorological characteristic data at multiple times are acquired. Subsequently, based on the acquired geographical environment data of the target watershed, the target watershed can be divided into grids. After completing the grid division, a grid adjacency matrix can be constructed based on the geographical location relationships between each grid. Further, based on the meteorological characteristic data corresponding to the target watershed at multiple times, the actual rainfall value for each grid at each time can be determined. Simultaneously, rainfall prediction data from meteorological monitoring stations at each time can be used to obtain the rainfall prediction value for each grid at each time. Next, for each grid, based on the constructed grid adjacency matrix, the target adjacent grids corresponding to that grid are determined. Then, based on the geographical environment data corresponding to that grid and its target adjacent grids, the actual rainfall values ​​at each time, and the rainfall prediction values, a preset rainfall analysis model is used to determine the water volume parameters and water flow direction parameters corresponding to that grid. After determining the water volume and flow direction parameters for each grid, the flash flood level of each grid can be determined based on the water volume parameters. Simultaneously, the flash flood hazard range of the target watershed can also be determined based on the water volume parameters for each grid. Furthermore, the flash flood trajectory of the target watershed can be generated based on the flow direction parameters for each grid. This embodiment of the application uses geographic environmental data to perform refined grid division of the target watershed, and combines this with a grid adjacency matrix to capture spatial dependencies, overcoming the low spatial resolution of traditional methods. Secondly, by fusing actual rainfall values, predicted rainfall values, and geographic environmental data from multiple times between each grid and its target adjacent grids, and dynamically calculating the water volume and flow direction parameters of the grid using a deep learning model, the timeliness and accuracy of the data can be improved. Finally, by generating the flash flood level distribution and trajectory based on the water volume and flow direction parameters, a quantitative assessment of the disaster range and hazard level can be achieved.

[0010] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below. Attached Figure Description

[0011] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A flowchart illustrating a flash flood identification method based on Earth big data provided in an embodiment of this application is shown. Figure 2 A flowchart illustrating a method for determining actual rainfall values ​​and predicted rainfall values ​​in a grid, as provided in an embodiment of this application, is shown. Figure 3 This illustration shows a structural schematic diagram of a flash flood identification device based on Earth big data, provided in an embodiment of this application. Figure 4 A schematic diagram of the device structure of a computer device provided in an embodiment of this application is shown. Detailed Implementation

[0012] The present application will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the embodiments of the present application can be combined with each other.

[0013] This embodiment provides a flash flood identification method based on Earth big data, such as... Figure 1 As shown, the method includes: Step 101: Determine the target watershed for flash flood identification, and obtain the corresponding geographical environment data and meteorological characteristic data at multiple times for the target watershed.

[0014] Step 102: Based on the geographic environment data, the target watershed is divided into grids to obtain multiple grids. Based on the geographical location relationship between each grid, a grid adjacency matrix is ​​constructed.

[0015] Step 103: Based on the meteorological characteristic data of multiple times corresponding to the target watershed, determine the actual rainfall value of each grid at each time and the predicted rainfall value for the future preset time period at each time.

[0016] Step 104: For each grid, based on the grid adjacency matrix, determine the target adjacent grid corresponding to the grid. According to the geographic environment data corresponding to the grid, the actual rainfall value and the predicted rainfall value at each time, and the geographic environment data corresponding to the target adjacent grid, the actual rainfall value and the predicted rainfall value at each time, determine the water volume parameters and water flow direction parameters corresponding to the grid through a preset rainfall analysis model.

[0017] Step 105: Based on the water volume parameters corresponding to each grid, determine the flash flood level of each grid and the flash flood hazard range of the target watershed, and generate the flash flood movement trajectory of the target watershed based on the water flow direction parameters corresponding to each grid.

[0018] This application provides a flash flood identification method based on Earth big data, which can accurately analyze and identify the movement trajectory of flash floods and the hazard range of target watersheds, providing detailed disaster information for disaster prevention and mitigation decision-making, thereby effectively reducing flash flood disaster losses.

[0019] First, the specific target watershed for flash flood identification must be clearly identified. After determining the target watershed, corresponding geographic environmental data and meteorological characteristic data at multiple time points are acquired. Geographic environmental data can encompass topographic elevation information of the target watershed (obtained through a digital elevation model (DEM), accurately reflecting the topographic undulations within the target watershed), land use type data (used to understand land use in different areas of the target watershed, such as forests, farmland, and urban areas; different land use types have different impacts on rainwater infiltration and runoff), vegetation data (used to understand vegetation cover in different areas of the target watershed), and historical flash flood records (used to understand past flash flood occurrences in different areas of the target watershed). Meteorological characteristic data at each time point can include radar echo data of the target watershed at that time, real-time rainfall monitoring data from various meteorological monitoring stations at that time, and rainfall forecast data for a preset future time period. This data can be obtained from various data sources such as ground meteorological monitoring stations and weather radars. Selecting meteorological characteristic data from multiple time points ensures accurate capture of the dynamic process of meteorological changes.

[0020] Subsequently, based on the acquired geographic environmental data of the target watershed, the watershed can be gridded. The purpose of gridding is to discretize the complex watershed space into a series of regular or irregular small units, facilitating independent analysis and calculation of each unit. The grid size can be determined based on the actual situation of the watershed and computational needs. After gridding is completed, a grid adjacency matrix can be constructed based on the geographical relationships between the grids. The grid adjacency matrix is ​​a mathematical tool used to describe the adjacency relationships between grids; the elements in the matrix represent the connectivity between grids. For example, if two grids are spatially adjacent, the corresponding element in the matrix has a value of 1; otherwise, it has a value of 0. By constructing the grid adjacency matrix, the topological structure between grids within the target watershed can be clearly expressed, providing a foundation for subsequent information analysis based on adjacency grids.

[0021] Furthermore, the actual rainfall value for each grid at each moment can be determined based on meteorological characteristic data from multiple times corresponding to the target watershed. Specifically, this can be achieved by spatially allocating the meteorological characteristic data, such as real-time rainfall monitoring data and radar echo data, according to the grid. Simultaneously, the rainfall prediction value for each grid at each moment can be obtained using the same method, utilizing rainfall prediction data from meteorological monitoring stations at various times. It is important to note that the rainfall prediction data from a meteorological monitoring station at a specific moment refers to the data obtained by the meteorological monitoring station in predicting rainfall over a predetermined future time period at that moment; similarly, the rainfall prediction value for a grid at a specific moment refers to the predicted rainfall value for that grid over a predetermined future time period at that moment.

[0022] Next, for each grid, based on the constructed grid adjacency matrix, the target adjacent grids are determined. The selection of target adjacent grids can be determined according to specific analysis needs; for example, directly adjacent grids can be selected as target adjacent grids, or a wider adjacency range can be considered. Then, based on the geographic environment data corresponding to the grid and its target adjacent grids, the actual rainfall values ​​at each time point, and the predicted rainfall values, the water volume parameters and flow direction parameters corresponding to the grid are determined through a preset rainfall analysis model. The preset rainfall analysis model can be a mathematical model that comprehensively considers multiple factors. It can determine the water volume parameters (such as water depth H and flow rate Q) corresponding to the grid based on the above data, and it can also determine the flow direction parameters (such as water depth H and flow rate Q) corresponding to the grid based on the above data. (Unit: °, 0° represents true north, angles increase clockwise). Water volume parameters reveal the water depth and flow rate within the grid, reflecting the amount of water and its convergence. Flow direction parameters determine which adjacent grid the water flows from and its approximate direction.

[0023] After determining the water volume and flow direction parameters for each grid, the flash flood level for each grid can be determined based on the water volume parameters. Flash flood levels can be classified according to threshold values ​​of the water volume parameters, such as by water depth and flow rate, categorizing them into mild, moderate, and severe levels. This provides a clear understanding of the severity of flash floods in each grid. Simultaneously, the flash flood hazard range of the target watershed can also be determined based on the water volume parameters for each grid. The hazard range can be determined by statistically analyzing the number and spatial distribution of grids reaching a certain flash flood level, defining the area covered by all grids reaching that level as the flash flood hazard range. This helps assess the size and extent of the area potentially affected by flash floods. Furthermore, the flash flood trajectory for the target watershed can be generated based on the flow direction parameters for each grid. The flash flood trajectory is formed by continuously tracking the flow direction of each grid and connecting them according to the flow direction. By generating flash flood trajectories, the flow direction and path of flash floods within the watershed can be visually displayed, providing important decision support information for early warning and emergency response to flash flood disasters.

[0024] By applying the technical solution of this embodiment, firstly, the specific target watershed requiring flash flood identification can be clearly identified. After determining the target watershed, the corresponding geographic environment data and meteorological characteristic data at multiple times are acquired. Subsequently, based on the acquired geographic environment data of the target watershed, the target watershed can be divided into grids. After completing the grid division, a grid adjacency matrix can be constructed based on the geographic location relationships between each grid. Further, based on the meteorological characteristic data at multiple times corresponding to the target watershed, the actual rainfall value of each grid at each time can be determined. Simultaneously, the rainfall prediction value of each grid at each time can be obtained using the rainfall prediction data from meteorological monitoring stations at each time. Next, for each grid, based on the constructed grid adjacency matrix, the target adjacent grids corresponding to that grid are determined. Then, based on the geographic environment data corresponding to that grid and its target adjacent grids, the actual rainfall value at each time, and the rainfall prediction value, the water volume parameters and water flow direction parameters corresponding to that grid are determined through a preset rainfall analysis model. After determining the water volume parameters and water flow direction parameters corresponding to each grid, the flash flood level of each grid can be determined based on the water volume parameters corresponding to each grid. Simultaneously, based on the water volume parameters corresponding to each grid, the extent of flash flood hazards in the target watershed can be determined. Furthermore, based on the water flow direction parameters corresponding to each grid, the flash flood trajectory of the target watershed can be generated. This application's embodiments utilize geographic environmental data to perform refined grid division of the target watershed, and combine this with a grid adjacency matrix to capture spatial dependencies, overcoming the low spatial resolution of traditional methods. Secondly, by fusing actual rainfall values, predicted rainfall values, and geographic environmental data from multiple moments between the grid and its target adjacent grids, and dynamically calculating the water volume and water flow direction parameters of the grid using a deep learning model, the timeliness and accuracy of the data can be improved. Finally, by generating flash flood level distribution and trajectory based on the water volume and water flow direction parameters, a quantitative assessment of the disaster's extent and severity can be achieved.

[0025] In this embodiment of the application, optionally, step 102, "dividing the target watershed into multiple grids based on the geographic environment data," includes: obtaining a preset grid division shape, and using the preset grid division shape to divide the target watershed into multiple grids to be evaluated; for each grid to be evaluated, calculating the spatial distribution coefficient corresponding to the grid to be evaluated based on the geographic environment data corresponding to the grid to be evaluated, and determining whether the geographic environment corresponding to the grid to be evaluated is uniformly distributed in space based on the spatial distribution coefficient; if the geographic environment corresponding to each grid to be evaluated is uniformly distributed in space, then... Each grid to be evaluated is taken as the final multiple grids. If the geographical environment corresponding to a grid to be evaluated is not uniformly distributed in space, the grid is re-divided based on the multiple grids to be evaluated, according to the preset adjustment strategy and the preset grid division shape, to obtain multiple new grids to be evaluated. The spatial distribution coefficient of each new grid to be evaluated is recalculated. Based on the recalculated spatial distribution coefficient, it is determined whether the geographical environment corresponding to the grid to be evaluated is uniformly distributed in space. The process ends when the geographical environment corresponding to each grid to be evaluated is uniformly distributed in space. The newly divided grids to be evaluated are taken as the final multiple grids.

[0026] In this embodiment, firstly, a preset grid division shape is obtained. The preset grid division shape can be determined based on specific research needs, the characteristics of the target watershed, and the convenience of subsequent analysis. Common grid division shapes include quadrilaterals and hexagons. After obtaining the preset grid division shape, it can be used to perform preliminary grid division of the target watershed, resulting in multiple grids to be evaluated. This discretizes the continuous target watershed space into independent grid cells, laying the foundation for subsequent analysis and processing.

[0027] For each initially defined grid to be evaluated, a spatial distribution coefficient is calculated based on its corresponding geographic environmental data. Geographic environmental data can include various types of data such as topographic elevation data and land use type data. The calculation method for the spatial distribution coefficient can be determined according to the specific geographic environmental data and the analysis objectives. For example, statistical indicators such as variance, standard deviation, and coefficient of variation can be used to measure the dispersion of geographic environmental data within the grid. After calculating the spatial distribution coefficient, a preset threshold is used to determine whether the geographic environment corresponding to the grid to be evaluated exhibits a uniform spatial distribution. If the spatial distribution coefficient is less than the preset threshold, the geographic environment is considered to be uniformly distributed in space; otherwise, it is considered to be unevenly distributed in space. In a specific embodiment, the geographic environment data may include high-precision DEM topographic elevation data, vegetation data (including vegetation cover information), land use type data, and historical flash flood record data within the target watershed. Based on the high-precision DEM topographic elevation data, vegetation data, land use type data, and historical flash flood record data, the target watershed can be divided into grids using either quadrilateral or hexagonal grids. Each grid can be set to a non-uniform size. The division must meet the following conditions: the topographic elevation, vegetation cover, land use type, historical flash flood record, and other information within each grid are spatially uniformly distributed, that is, there is no local concentration or dispersion of information within the grid, ensuring the consistency and representativeness of the grid data.

[0028] If, after assessment, the geographical environment corresponding to each grid to be evaluated is found to be evenly distributed in space, this means that the initially divided grids can well reflect the characteristics of the geographical environment within the target watershed and meet the needs of subsequent analysis. At this point, each grid to be evaluated can be directly used as the final grids for the subsequent flash flood identification and analysis process.

[0029] If the geographic environment corresponding to a grid to be evaluated is not uniformly distributed in space, this indicates that the initially divided grid does not meet the requirements and may affect the accuracy of subsequent flash flood identification. In this case, the grid can be re-divided based on multiple grids to be evaluated, using a preset adjustment strategy and a preset grid division shape. The preset adjustment strategy can be formulated according to specific circumstances, such as adjusting the size and position of the grids to better adapt to the distribution of the geographic environment. After re-dividing, multiple new grids to be evaluated are obtained. Then, the spatial distribution coefficient of each new grid to be evaluated is recalculated, and the uniformity of its corresponding geographic environment is determined based on the recalculated spatial distribution coefficient. The above process is an iterative loop, continuously repeating the steps of re-dividing the grid, calculating the spatial distribution coefficient, and determining uniformity until the geographic environment corresponding to each grid to be evaluated is uniformly distributed in space. Finally, the newly divided grids to be evaluated are used as the final grids to ensure the consistency and representativeness of the grid data, providing reliable spatial units for subsequent flash flood identification.

[0030] In this embodiment of the application, optionally, when the geographical environment data includes historical flash flood record data, after "determining whether the geographical environment corresponding to the grid to be evaluated is uniformly distributed in space", the method further includes: for each grid to be evaluated, determining the flood filling degree corresponding to the grid to be evaluated based on the historical flash flood record data corresponding to the grid to be evaluated; when the flood filling degree is greater than a first preset filling degree or less than a second preset filling degree, determining that the grid to be evaluated meets the flood filling requirement; if all grids to be evaluated meet the flood filling requirement, and the geographical environment corresponding to each grid to be evaluated is uniformly distributed in space. If the geographic environment corresponding to a given geographic grid is not uniformly distributed in space, and / or if a given geographic grid does not meet the flood filling requirements, then based on the given geographic grids, the grids are re-divided according to a preset adjustment strategy and the preset grid division shape to obtain multiple new geographic grids to be evaluated. The spatial distribution coefficient and flood filling degree of each geographic grid to be evaluated are recalculated until all geographic grids to be evaluated meet the flood filling requirements and the geographic environment corresponding to each geographic grid to be evaluated is uniformly distributed in space. The newly divided geographic grids to be evaluated are then used as the final multiple geographic grids.

[0031] In this embodiment, assuming the geographic environment data includes historical flash flood records, the flood fill degree can be determined for each grid to be evaluated based on its corresponding historical flash flood records. The historical flash flood records cover information such as the flood inundation range, flood depth, and flood duration during past flash floods in the grid area. The flood fill degree can be quantified by calculating the proportion of the area within the grid that was covered by floodwater or the proportion of areas where the flood depth reached a certain standard within a specific time period. For example, a standard flood depth can be set, and the ratio of the area within the grid whose actual flood depth reaches or exceeds this standard to the total area of ​​the grid can be used as the flood fill degree. When the calculated flood fill degree is greater than a first preset fill degree or less than a second preset fill degree, the grid to be evaluated is considered to meet the flood fill requirements. The first and second preset fill degrees can be set based on research into the flash flood characteristics of the target watershed and actual analytical needs. In a specific embodiment, the grid after division should have a historical flood coverage rate of more than 85% (first preset coverage rate) or less than 15% (second preset coverage rate) in the grid. A coverage rate of more than 85% indicates that the historical flood coverage area in the grid is high, and a coverage rate of less than 15% indicates that the historical flood coverage area in the grid is low, so as to ensure that the grid effectively represents the distribution of historical floods.

[0032] If, after evaluating each grid to be assessed, it is found that each grid meets the flood filling requirements, and the geographical environment corresponding to each grid is evenly distributed in space, this indicates that the currently divided grid can reasonably reflect the spatial characteristics of the geographical environment and meet the filling requirements of historical flash flood occurrences. In this case, each grid to be assessed can be directly used as the final grid. These grids will provide a reliable spatial basis for subsequent flash flood identification, risk assessment, and other analyses, ensuring the accuracy and reliability of the analysis results.

[0033] If the geographic environment corresponding to a grid to be evaluated is not uniformly distributed in space, and / or if a grid to be evaluated does not meet the flood filling requirements, this means that the currently divided grid has problems and cannot well meet the needs of flash flood identification and analysis. In this case, the grid can be re-divided based on multiple grids to be evaluated, using a preset adjustment strategy and a preset grid division shape. After re-dividing, multiple new grids to be evaluated are obtained, and then the spatial distribution coefficient and flood filling degree of each grid to be evaluated are recalculated. Then, it is determined again whether each grid to be evaluated meets the flood filling requirements and whether the geographic environment corresponding to each grid to be evaluated is uniformly distributed in space. This process is an iterative loop, continuously repeating the steps of re-dividing the grid, calculating indicators, and judging conditions until each grid to be evaluated meets the flood filling requirements and the geographic environment corresponding to each grid to be evaluated is uniformly distributed in space. Finally, the newly divided grids to be evaluated are used as the final grids to ensure that the divided grids can provide accurate and reliable spatial units for flash flood identification and analysis.

[0034] Optionally, in this embodiment of the application, the meteorological characteristic data includes radar echo data corresponding to the target watershed, as well as real-time rainfall monitoring data from each meteorological monitoring station and rainfall forecast data for a preset future time period; such as Figure 2 As shown, step 103 includes: Step 103-1: Determine the first target location of each meteorological monitoring station and the second target location of each grid.

[0035] Step 103-2: For each time moment, using a preset quantitative relationship model between radar echo reflectivity factor and precipitation intensity, the radar echo data at that time is converted into regional precipitation intensity distribution data for the target watershed. The regional precipitation intensity distribution data is then calibrated using real-time rainfall monitoring data from each meteorological monitoring station at that time and the first target location, resulting in calibrated precipitation intensity distribution data for that time moment. Based on the calibrated precipitation intensity distribution data and the second target location of each grid, the first rainfall value for each grid at that time is calculated. Based on the real-time rainfall monitoring data from each meteorological monitoring station at that time, the first target location, and the second target location of each grid, the second rainfall value for each grid at that time is calculated using an interpolation algorithm. Finally, based on the first and second rainfall values, the actual rainfall value for each grid at that time is calculated.

[0036] Step 103-3: For each time moment, based on the rainfall prediction data and first target position of each meteorological monitoring station at that time moment, and the second target position of each grid, the rainfall prediction value for the future preset time period of each grid moment is calculated by interpolation algorithm.

[0037] In this embodiment, for each moment, the meteorological characteristic data can include radar echo data corresponding to the target watershed, as well as real-time rainfall monitoring data from each meteorological monitoring station and rainfall forecast data for a preset future time period. Specifically, the radar echo data can be obtained through radar detection and can cover the entire target watershed.

[0038] First, the primary target location for each meteorological monitoring station and the secondary target location for each grid are determined. The secondary target location for a grid can be determined based on the latitude and longitude coordinates of the grid's center point or representative point.

[0039] For each time point, the radar echo data at that moment is first converted into regional precipitation intensity distribution data for the target watershed using a pre-defined quantitative relationship model between radar echo reflectivity factor and precipitation intensity. Radar echo data reflects the distribution of water droplets or ice crystals in clouds, while the pre-defined quantitative relationship model is established based on extensive radar observation data and actual precipitation measurement data, accurately converting the radar echo reflectivity factor into precipitation intensity. After obtaining the regional precipitation intensity distribution data, the data is calibrated using real-time rainfall monitoring data from each meteorological monitoring station at that moment and the first target location. The real-time rainfall monitoring data from the meteorological monitoring stations has high accuracy and reliability. By comparing and adjusting the precipitation intensity at the meteorological monitoring station location with the precipitation intensity converted from radar echo data, potential errors in the radar echo data can be eliminated, resulting in calibrated precipitation intensity distribution data for that moment. Finally, based on the calibrated precipitation intensity distribution data and the second target location for each grid, the first rainfall value for each grid at that moment is calculated. Specifically, based on the second target location of each grid, the precipitation intensity value corresponding to each grid can be determined in the calibrated precipitation intensity distribution data according to the mapping relationship, and then the precipitation intensity value can be converted into the first rainfall value corresponding to each grid.

[0040] Similarly, for each time moment, based on the real-time rainfall monitoring data and the first target location of each meteorological monitoring station at that moment, as well as the second target location of each grid, the second rainfall value for each grid at that moment is calculated using an interpolation algorithm. An interpolation algorithm is a method for estimating data at unknown points based on data from known points; common interpolation algorithms include inverse distance weighted interpolation and Kriging interpolation. In this embodiment, the real-time rainfall monitoring data from the meteorological monitoring stations is used as known points, and the data from the monitoring stations is extended to each grid using an interpolation algorithm to obtain the second rainfall value for each grid. The second rainfall value fully considers the spatial distribution of the meteorological monitoring station data, and the interpolation algorithm can more reasonably estimate the rainfall value within the grid, complementing the first rainfall value and improving the accuracy of rainfall calculation.

[0041] After obtaining the first and second rainfall values ​​for each grid at each time point, the actual rainfall value for each grid at that time is calculated based on these two values. Specifically, methods such as weighted averaging or arithmetic averaging can be used to fuse the first and second rainfall values ​​to obtain a more accurate and reliable actual rainfall value for each grid. This integrates information from radar echo data and meteorological monitoring station data, fully utilizing the advantages of both data sources and reducing errors that may arise from a single data source. This allows the calculated actual rainfall value for each grid to more accurately reflect the actual rainfall situation within that grid at that time. In a specific embodiment, if a weighted averaging method is used, the weights of the first and second rainfall values ​​can be determined based on the distance between the grid and the meteorological monitoring station: the closer the grid is to the meteorological monitoring station, the greater the weight of the second rainfall value, and vice versa. That is, different grids have different weights.

[0042] Furthermore, for each moment, based on the rainfall forecast data and first target location of each meteorological monitoring station at that moment, as well as the second target location of each grid, an interpolation algorithm can be used to calculate the rainfall forecast value for each grid over a predetermined future time period at that moment. Rainfall forecast data, which is a prediction of rainfall over a future period by the meteorological department based on meteorological models and forecasting techniques, also exhibits spatial distribution characteristics. By extending the rainfall forecast data from the meteorological monitoring stations to each grid using an interpolation algorithm, the rainfall forecast value for each grid over a predetermined future time period can be obtained. This provides crucial data support for subsequent flash flood forecasting and early warning, enabling a more accurate assessment of future rainfall based on the spatial distribution of the grid.

[0043] In this embodiment of the application, optionally, the geographic environment data includes topographic elevation data, vegetation data, land use type data, and historical flash flood information data; step 104, "determining the water volume parameters and water flow direction parameters corresponding to the grid based on the geographic environment data corresponding to the grid, the actual rainfall value at each time, the predicted rainfall value, and the geographic environment data corresponding to the target adjacent grid, the actual rainfall value at each time, and the predicted rainfall value," includes: constructing a first environmental feature vector of the grid based on the topographic elevation data, vegetation data, land use type data, and historical flash flood information data corresponding to the grid; and constructing a second environmental feature vector of the target adjacent grid based on the topographic elevation data, vegetation data, land use type data, and historical flash flood information data corresponding to the target adjacent grid; and constructing a first temporal rainfall feature vector of the grid based on the actual rainfall value at each time and the predicted rainfall value corresponding to the grid. Furthermore, based on the actual rainfall value and predicted rainfall value at each time corresponding to the target adjacent grid, a second temporal rainfall feature vector of the target adjacent grid is constructed; based on the first environmental feature vector of the grid and the second environmental feature vector of the target adjacent grid, the spatial correlation features between the grid and the target adjacent grid are extracted through the spatial feature extraction layer of the preset rainfall analysis model; based on the first temporal rainfall feature vector of the grid and the second temporal rainfall feature vector of the target adjacent grid, the temporal rainfall correlation features between the grid and the target adjacent grid are extracted through the temporal feature extraction layer of the preset rainfall analysis model; the spatial correlation features and the temporal rainfall correlation features are input into the first fully connected layer of the preset rainfall analysis model to obtain the water volume parameters corresponding to the grid; the spatial correlation features and the temporal rainfall correlation features are input into the second fully connected layer of the preset rainfall analysis model to obtain the water flow direction parameters corresponding to the grid.

[0044] In this embodiment, the geographic environment data may include topographic elevation data, vegetation data, land use type data, and historical flash flood information data. First, for each grid, this data can be encoded according to certain rules, and then concatenated to construct the first environmental feature vector for that grid. This transforms the various geographic environment information of the grid into a unified vector form for subsequent model processing and analysis. Simultaneously, for each grid, its target adjacent grids are determined. Target adjacent grids can be grids directly adjacent to the current grid in space, or surrounding grids selected according to certain rules. Similarly, based on the topographic elevation data, vegetation data, land use type data, and historical flash flood information data of each target adjacent grid, a second environmental feature vector for that target adjacent grid is constructed. This step provides the data foundation for subsequent extraction of spatial correlation features between the current grid and its target adjacent grids.

[0045] For each grid cell, the rainfall data from different times can be arranged and encoded chronologically based on the actual and predicted rainfall values ​​at each corresponding moment, constructing the first temporal rainfall feature vector for that grid cell. The actual rainfall values ​​reflect the actual rainfall at each past moment, while the predicted rainfall values ​​are predictions of the rainfall within a predetermined future time period relative to the past moments. This process integrates the rainfall information of the grid cells in the temporal dimension, enabling subsequent models to analyze the temporal changes in rainfall. Similarly, for each target adjacent grid cell, a second temporal rainfall feature vector is constructed based on its corresponding actual and predicted rainfall values ​​at each moment. This step provides data support for subsequent extraction of temporal rainfall correlation features between the current grid cell and its target adjacent grid cells.

[0046] Next, the first environmental feature vector of the constructed grid and the second environmental feature vector of the target adjacent grid are input into the spatial feature extraction layer of the preset rainfall analysis model. The spatial feature extraction layer is a part of the model specifically designed to process spatial correlation information. It can analyze the spatial relationship between two feature vectors using structures such as convolutional neural networks and graph neural networks. For example, convolutional neural networks can capture local spatial patterns in feature vectors through convolution operations, while graph neural networks can perform information transfer and aggregation based on the adjacency relationships between grids. Through these operations, the spatial correlation features between the grid and the target adjacent grid are extracted. These features reflect the mutual influence and correlation between the two grids in terms of the geographical environment.

[0047] In one specific embodiment, the structure of the spatial feature extraction layer can be as follows: Two layers are used. l =2) Graph Convolutional Network (GCN), which calculates the spatial association features between the grid and the target adjacent grid using the following formula: ; in, For the first l The spatial correlation feature vector of the grid v output by the layer graph convolutional network. Let v be the target set, which includes grid v and its corresponding target adjacent grids. The number of grids in the target set. For the first l The weight matrix of the layer, For the first l -1 layer mesh u's environmental feature vector For the first l Layer bias terms, Using the ReLU activation function, this layer is used to extract the spatial correlation features between the grid and its target adjacent grids, capturing the impact of spatial elements such as terrain and vegetation on water volume.

[0048] Furthermore, the first time-series rainfall feature vector of the constructed grid and the second time-series rainfall feature vector of the target adjacent grid can be input into the time feature extraction layer of a preset rainfall analysis model. The time feature extraction layer is a part of the model specifically designed to process time-series correlation information. It can analyze the temporal relationship between two time-series feature vectors using structures such as recurrent neural networks (RNNs) and long short-term memory (LSM) networks. For example, RNNs can process sequence data through recurrent connections, capturing long-term dependencies in time-series feature vectors; LSM networks can better handle both long-term and short-term dependencies in sequence data. Through these operations, the time-series rainfall correlation features between the grid and the target adjacent grid are extracted. These features reflect the mutual influence and correlation between the two grids in terms of rainfall time series.

[0049] In a specific embodiment, the temporal feature extraction layer may include a 3-layer spatio-temporal continuous network (STCN) to process the temporal rainfall feature vector. The convolution kernel size can be set to 3×3 (time step × spatial grid). Through convolution operations, the influence of the dynamic changes in rainfall intensity over time on the formation and development of floods is captured, thereby realizing feature extraction in the temporal dimension.

[0050] Furthermore, the extracted spatial correlation features and temporal rainfall correlation features can be input into the first fully connected layer of the preset rainfall analysis model. The first fully connected layer is a neural network layer in the model that performs linear transformations and nonlinear activation function conversions on the input features, mapping the high-dimensional feature space to a low-dimensional space and outputting a result related to water quantity parameters. Through comprehensive analysis and calculation of these features, the water quantity parameters corresponding to the grid are obtained.

[0051] Furthermore, the extracted spatial correlation features and temporal rainfall correlation features can be input into the second fully connected layer of the preset rainfall analysis model. The second fully connected layer is also a neural network layer, similar to the first fully connected layer, but its output is related to the flow direction parameters. Through the analysis and calculation of these features, the flow direction parameters corresponding to the grid are obtained.

[0052] It should be noted that the training process of the preset rainfall analysis model is similar to the process described above, but the training samples are labeled with water volume parameters and water flow direction parameters, and the loss during the model training process can be calculated using the preset cross-entropy loss function.

[0053] In the process of calculating the water volume and flow direction parameters of each grid through the model, the embodiments of this application can integrate multi-source environmental information such as topography and vegetation of the grid itself and adjacent grids, as well as the actual and predicted time-series rainfall information at each moment, to comprehensively grasp the hydrological characteristics of the grid. With the help of the graph convolutional network structure, it can capture the complex spatial relationships between grids in local adjacent areas and in the global scope, and can also mine the nonlinear relationships between features through activation functions, thereby more accurately reflecting the changing patterns of water volume and flow direction, and providing a reliable basis for flash flood identification, etc.

[0054] In this embodiment of the application, optionally, step 105, "determining the flash flood level of each grid and the flash flood hazard range of the target watershed based on the water volume parameters corresponding to each grid," includes: for each grid, identifying terrain feature data from the geographic environment data corresponding to the grid; constructing a target feature vector corresponding to the grid based on the water volume parameters and terrain feature data corresponding to the grid; performing clustering processing on each grid based on the target feature vectors corresponding to each grid to obtain multiple clusters; for each cluster, determining the flash flood level corresponding to the central grid based on the water volume parameters corresponding to the central grid of the cluster, and using the flash flood level as the flash flood level corresponding to each grid within the cluster; determining the affected objects corresponding to each grid; for each grid, determining the hazard value of the grid based on the water volume parameters corresponding to the grid and the affected objects; and determining the flash flood hazard range of the target watershed based on the hazard values ​​of each grid.

[0055] In this embodiment, firstly, topographic feature data, such as elevation and slope, is accurately identified from the geographic environment data corresponding to each grid. Then, the identified topographic feature data is combined with the water volume parameters corresponding to that grid, and a target feature vector is constructed according to specific rules and formats. This integrates and encodes key factors affecting flash flood severity, presenting the water volume and topographic information of each grid in a unified form, providing standardized data input for subsequent clustering analysis. Subsequently, based on the target feature vectors corresponding to each grid, a suitable clustering algorithm (such as K-means clustering or hierarchical clustering) is used to cluster all grids. The clustering algorithm can group grids with similar water volume and topographic features into the same cluster based on the similarity between the target feature vectors. Through clustering, grids within the target watershed with similar flash flood occurrence conditions and potential hazards can be grouped together, simplifying the subsequent analysis process and helping to discover the patterns of flash flood characteristics in different regions.

[0056] In one specific embodiment, a target feature vector is constructed using water quantity parameters (water depth H, flow rate Q) and terrain features (including DEM elevation, slope, etc., represented by the NDVI index). Where S is the terrain slope. The DBSCAN (density clustering) algorithm is used to identify similar grids. The neighborhood radius is set to 0.2 (after feature vector normalization), and the minimum number of cluster samples is set to 5. This algorithm divides grids with similar target feature vectors into the same cluster, quickly locating grid clusters with similar flash flood characteristics.

[0057] For each cluster, its central grid is identified. The central grid, to a certain extent, represents the overall characteristics of that cluster. Based on the water volume parameters corresponding to the central grid, and combined with pre-defined flash flood level determination criteria (e.g., different water volume ranges correspond to different flash flood levels), the flash flood level corresponding to the central grid is determined. This flash flood level is then assigned to each grid within the cluster. This improves the efficiency of flash flood level confirmation while ensuring a certain level of accuracy.

[0058] In a specific embodiment, the flash flood classification rules can be as follows: Based on the grid water depth (H) and flow rate (Q), combined with the rainfall duration, the flash flood level is divided into four levels, and the specific classification criteria are shown in the table below:

[0059] Next, the affected objects corresponding to each grid are identified, such as residential areas, farmland, and infrastructure. For each grid, based on its corresponding water volume parameters and the specific circumstances of the affected objects (e.g., population density of residential areas, crop value of farmland, importance of infrastructure, etc.), an appropriate calculation method is used to determine the hazard value of that grid. For example, if the water volume is large and the affected objects are densely populated residential areas, then the hazard value of that grid is high. Finally, based on the hazard values ​​of each grid, the flash flood hazard range of the target watershed is determined by setting hazard thresholds or drawing hazard contour lines, clarifying which areas are threatened by flash floods and the degree of threat.

[0060] In one specific embodiment, the affected objects may include population, farmland, and buildings. Then, the hazard value of the grid... R The calculation process is as follows: ; in, For flash flood depth risk (standardized value, , (This refers to the historical maximum water depth of this grid). For population exposure risk ( =Population density within the grid × 0.3, in people / hectare); For the risk of economic loss ( = Farmland / Building Density × Asset Value Coefficient (unit: 10,000 RMB / hectare); Weight =0.5, =0.3, =0.2 (and each weight can be adjusted according to the characteristics of the target watershed area (such as population density, economic development level, etc.).

[0061] In addition, the overall risk value of the target watershed can be calculated: Overall risk value of the target watershed The hazard value is determined by a weighted average of the grid hazard values ​​combined with the proportion of affected grids. The calculation formula is as follows: ; in Let i be the hazard value of the i-th grid. Let be the area of ​​the i-th grid, and n be the total number of grids in the target watershed.

[0062] The warning level for the target watershed can also be calculated using the table below: Among them, the warning level and the overall risk value of the target watershed The correspondence is shown in the table below:

[0063] Optionally, in this embodiment, step 105, "generating the flash flood trajectory of the target watershed based on the water flow direction parameters corresponding to each grid," includes: determining the flash flood initiation grid from each grid based on the second target position of each grid and the water flow direction indicated by the water flow direction parameters of each grid; using the flash flood initiation grid as the target grid, determining the eight neighboring grids corresponding to the target grid from each grid; determining the downstream grid corresponding to the target grid from the eight neighboring grids based on the water flow direction indicated by the water flow direction parameters of the eight neighboring grids; constructing the flash flood trajectory based on the water flow directions of the target grid and the downstream grid; using the downstream grid as the new target grid; re-determining the downstream grid of the new target grid; updating the flash flood trajectory based on the water flow direction of the new downstream grid; and ending when the downstream grid is the edge grid of the target watershed, thus obtaining the flash flood trajectory of the target watershed.

[0064] In this embodiment, based on the second target location of each grid in the target watershed, and combined with the specific water flow direction indicated by the water flow direction parameter of each grid, the flash flood initiation grid can be selected from all grids. For example, grids at the edge of the target watershed that have a water flow direction pointing towards the interior of the target watershed are selected as flash flood initiation grids.

[0065] Subsequently, the determined flash flood initiation grid is used as the target grid, and the corresponding eight-neighbor grid is identified from all grids. The eight-neighbor grid refers to the eight adjacent grids surrounding the target grid, fully considering all possible directions of water flow. Then, based on the water flow direction parameters of each of the eight-neighbor grids, the grids that match the direction of water flow from the target grid are selected and designated as the downstream grids corresponding to the target grid. This process simulates the natural flow direction selection of water between adjacent grids, ensuring that the generated trajectory conforms to the actual water flow movement laws.

[0066] Furthermore, based on the flow direction of the target grid and its corresponding downstream grid, an initial fragment of the flash flood trajectory is constructed. Then, using the downstream grid as the new target grid, its downstream grid is identified from its eight neighboring grids using the same method described above. Next, based on the flow direction of the new downstream grid, the existing flash flood trajectory is updated and extended. This step iterates continuously, gradually extending the flash flood trajectory towards the actual flow direction of the water to fully represent the flash flood's movement path within the watershed.

[0067] The above trajectory update process continues until the newly determined downstream grid is the edge grid of the target watershed, indicating that the main movement path of the flash flood within the watershed has been tracked. At this point, the final flash flood trajectory is the complete movement path of the flash flood from its starting point to the edge of the watershed within the target watershed, which can intuitively show the flow direction and range of the flash flood within the watershed.

[0068] This application embodiment comprehensively considers the multi-directional flow of water using an eight-neighbor grid, accurately determines the downstream grid, and constructs the complete trajectory of flash floods from their origin to the edge of the basin through iterative updates, closely matching the actual water flow movement. This provides an intuitive reference for flash flood disaster early warning and helps reduce disaster losses.

[0069] Furthermore, as Figure 1 In terms of specific implementation, this application provides a flash flood identification device based on Earth big data, such as... Figure 3 As shown, the device includes: The data acquisition module is used to determine the target watershed for flash flood identification and acquire the corresponding geographical environment data and meteorological characteristic data at multiple times for the target watershed. The grid partitioning module is used to partition the target watershed into multiple grids based on the geographic environment data, and to construct a grid adjacency matrix based on the geographic location relationship between the grids. The rainfall calculation module is used to determine the actual rainfall value of each grid at each time and the rainfall prediction value for a future preset time period at each time, based on the meteorological characteristic data of multiple times corresponding to the target watershed. The model analysis module is used to determine the target adjacent grids corresponding to each grid based on the grid adjacency matrix. Based on the geographic environment data corresponding to the grid, the actual rainfall value and the predicted rainfall value at each time, and the geographic environment data corresponding to the target adjacent grid, the actual rainfall value and the predicted rainfall value at each time, the module determines the water volume parameters and water flow direction parameters corresponding to the grid through a preset rainfall analysis model. The result generation module is used to determine the flash flood level of each grid and the flash flood hazard range of the target watershed based on the water volume parameters corresponding to each grid, and to generate the flash flood movement trajectory of the target watershed based on the water flow direction parameters corresponding to each grid.

[0070] Optionally, the mesh division module is used for: Obtain a preset grid division shape, and use the preset grid division shape to divide the target watershed into grids to obtain multiple grids to be evaluated; For each grid to be evaluated, based on the geographic environment data corresponding to the grid to be evaluated, the spatial distribution coefficient corresponding to the grid to be evaluated is calculated, and based on the spatial distribution coefficient, it is determined whether the geographic environment corresponding to the grid to be evaluated is uniformly distributed in space. If the geographical environment corresponding to each grid to be evaluated is uniformly distributed in space, then each grid to be evaluated will be taken as the final multiple grids. If the geographic environment corresponding to a grid to be evaluated is not uniformly distributed in space, then based on the multiple grids to be evaluated, the grid is re-divided according to a preset adjustment strategy and the preset grid division shape to obtain multiple new grids to be evaluated. The spatial distribution coefficient of each new grid to be evaluated is recalculated. Based on the recalculated spatial distribution coefficient, it is determined whether the geographic environment corresponding to the grid to be evaluated is uniformly distributed in space. This process continues until the geographic environment corresponding to each grid to be evaluated is uniformly distributed in space. The newly divided grids to be evaluated are then taken as the final multiple grids.

[0071] Optionally, when the geographic environment data includes historical flash flood records, the grid division module is further configured to: After determining whether the geographical environment corresponding to the grid to be evaluated is uniformly distributed in space, for each grid to be evaluated, the flood filling degree corresponding to the grid to be evaluated is determined according to the historical flash flood record data corresponding to the grid to be evaluated. When the flood filling degree is greater than the first preset filling degree or less than the second preset filling degree, it is determined that the grid to be evaluated meets the flood filling requirements. If each grid to be evaluated meets the flood filling requirements and the geographical environment corresponding to each grid to be evaluated is uniformly distributed in space, then each grid to be evaluated will be taken as the final multiple grids. If the geographic environment corresponding to a grid to be evaluated is not uniformly distributed in space, and / or a grid to be evaluated does not meet the flood filling requirements, then based on the multiple grids to be evaluated, the grid is re-divided according to a preset adjustment strategy and the preset grid division shape to obtain multiple new grids to be evaluated. The spatial distribution coefficient and flood filling degree of each grid to be evaluated are recalculated until all grids to be evaluated meet the flood filling requirements and the geographic environment corresponding to each grid to be evaluated is uniformly distributed in space. The newly divided grids to be evaluated are then taken as the final multiple grids.

[0072] Optionally, the meteorological characteristic data includes radar echo data corresponding to the target watershed, as well as real-time rainfall monitoring data from each meteorological monitoring station and rainfall forecast data for a preset future time period; the rainfall calculation module is used for: Determine the first target location of each meteorological monitoring station and the second target location of each grid; For each time moment, the radar echo data at that time is transformed into regional precipitation intensity distribution data of the target watershed using a preset quantitative relationship model between radar echo reflectivity factor and precipitation intensity. The regional precipitation intensity distribution data is then calibrated using real-time rainfall monitoring data from each meteorological monitoring station at that time and the first target location, resulting in calibrated precipitation intensity distribution data for that time moment. Based on the calibrated precipitation intensity distribution data and the second target location of each grid, a first rainfall value is calculated for each grid at that time moment. Based on the real-time rainfall monitoring data from each meteorological monitoring station at that time, the first target location, and the second target location of each grid, a second rainfall value is calculated for each grid at that time moment using an interpolation algorithm. Finally, based on the first and second rainfall values, the actual rainfall value for each grid at that time moment is calculated. For each moment, based on the rainfall forecast data and first target location of each meteorological monitoring station at that moment, and the second target location of each grid, the rainfall forecast value for the future preset time period of each grid at that moment is calculated by interpolation algorithm.

[0073] Optionally, the geographic environment data includes topographic elevation data, vegetation data, land use type data, and historical flash flood information data; the model analysis module is used for: Based on the topographic elevation data, vegetation data, land use type data, and historical flash flood information data corresponding to the grid, a first environmental feature vector of the grid is constructed; and based on the topographic elevation data, vegetation data, land use type data, and historical flash flood information data corresponding to the target adjacent grid, a second environmental feature vector of the target adjacent grid is constructed; and based on the actual rainfall value and predicted rainfall value at each time corresponding to the grid, a first temporal rainfall feature vector of the grid is constructed; and based on the actual rainfall value and predicted rainfall value at each time corresponding to the target adjacent grid, a second temporal rainfall feature vector of the target adjacent grid is constructed. Based on the first environmental feature vector of the grid and the second environmental feature vector of the target adjacent grid, the spatial correlation features between the grid and the target adjacent grid are extracted through the spatial feature extraction layer of the preset rainfall analysis model; Based on the first temporal rainfall feature vector of the grid and the second temporal rainfall feature vector of the target adjacent grid, the temporal rainfall correlation features between the grid and the target adjacent grid are extracted through the temporal feature extraction layer of the preset rainfall analysis model. The spatial correlation features and the temporal rainfall correlation features are input into the first fully connected layer of the preset rainfall analysis model to obtain the water parameters corresponding to the grid. The spatial correlation features and the temporal rainfall correlation features are input into the second fully connected layer of the preset rainfall analysis model to obtain the water flow direction parameters corresponding to the grid.

[0074] Optionally, the result generation module is used to: For each grid, terrain feature data is identified from the geographic environment data corresponding to the grid, and a target feature vector corresponding to the grid is constructed based on the water quantity parameters and terrain feature data corresponding to the grid. Based on the target feature vector corresponding to each grid, the grids are clustered to obtain multiple clusters; For each cluster, based on the water volume parameter corresponding to the central grid of the cluster, the flash flood level corresponding to the central grid is determined, and the flash flood level is used as the flash flood level corresponding to each grid within the cluster; The affected objects corresponding to each grid are identified. For each grid, the hazard value of the grid is determined based on the water volume parameters corresponding to the grid and the affected objects. Based on the hazard values ​​of each grid, the flash flood hazard range of the target watershed is determined.

[0075] Optionally, the result generation module is further configured to: Based on the second target location of each grid and the water flow direction indicated by the water flow direction parameter of each grid, the flash flood initiation grid is determined from the grids; Using the initial grid of the flash flood as the target grid, determine the eight neighboring grids corresponding to the target grid from each grid. Based on the water flow direction parameters of the eight neighboring grids, determine the downstream grid corresponding to the target grid from the eight neighboring grids. Construct the flash flood trajectory based on the water flow directions of the target grid and the downstream grid, and use the downstream grid as the new target grid. Redetermine the downstream grid of the new target grid, and update the flash flood trajectory based on the water flow direction of the new downstream grid until the downstream grid becomes the edge grid of the target watershed, thus obtaining the flash flood trajectory of the target watershed.

[0076] It should be noted that other corresponding descriptions of the functional units involved in the flash flood identification device based on Earth big data provided in this application embodiment can be found in the following references. Figures 1 to 2 The corresponding descriptions in the method will not be repeated here.

[0077] This application also provides a computer device, which may specifically be a personal computer, a server, a network device, etc. Figure 4As shown, the computer device includes a bus, a processor, memory, and a communication interface, and may also include an input / output interface and a display device. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores location information. The network interface allows communication with external terminals via a network connection. When the computer program is executed by the processor, it implements the steps in the various method embodiments.

[0078] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0079] In one embodiment, a computer-readable storage medium is provided, which may be non-volatile or volatile, having stored thereon a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0080] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0081] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0082] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0083] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0084] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A flash flood identification method based on Earth big data, characterized in that, include: The target watershed for flash flood identification is determined, and the corresponding geographical environment data and meteorological characteristic data at multiple times are obtained for the target watershed. Based on the geographic environment data, the target watershed is divided into multiple grids, and a grid adjacency matrix is ​​constructed based on the geographic location relationship between the grids. Based on the meteorological characteristic data of the target watershed at multiple times, the actual rainfall value of each grid at each time and the rainfall prediction value for the future preset time period at each time are determined respectively. For each grid, based on the grid adjacency matrix, the target adjacent grid corresponding to the grid is determined. According to the geographic environment data corresponding to the grid, the actual rainfall value and the predicted rainfall value at each time, and the geographic environment data corresponding to the target adjacent grid, the actual rainfall value and the predicted rainfall value at each time, the water volume parameters and water flow direction parameters corresponding to the grid are determined by a preset rainfall analysis model. Based on the water volume parameters corresponding to each grid, the flash flood level of each grid and the flash flood hazard range of the target watershed are determined, and the flash flood movement trajectory of the target watershed is generated based on the water flow direction parameters corresponding to each grid.

2. The method according to claim 1, characterized in that, Based on the geographic environment data, the target watershed is divided into multiple grids, including: Obtain a preset grid division shape, and use the preset grid division shape to divide the target watershed into grids to obtain multiple grids to be evaluated; For each grid to be evaluated, based on the geographic environment data corresponding to the grid to be evaluated, the spatial distribution coefficient corresponding to the grid to be evaluated is calculated, and based on the spatial distribution coefficient, it is determined whether the geographic environment corresponding to the grid to be evaluated is uniformly distributed in space. If the geographical environment corresponding to each grid to be evaluated is uniformly distributed in space, then each grid to be evaluated will be taken as the final multiple grids. If the geographic environment corresponding to a grid to be evaluated is not uniformly distributed in space, then based on the multiple grids to be evaluated, the grid is re-divided according to a preset adjustment strategy and the preset grid division shape to obtain multiple new grids to be evaluated. The spatial distribution coefficient of each new grid to be evaluated is recalculated. Based on the recalculated spatial distribution coefficient, it is determined whether the geographic environment corresponding to the grid to be evaluated is uniformly distributed in space. This process continues until the geographic environment corresponding to each grid to be evaluated is uniformly distributed in space. The newly divided grids to be evaluated are then taken as the final multiple grids.

3. The method according to claim 2, characterized in that, When the geographic environment data includes historical flash flood records, after determining whether the geographic environment corresponding to the grid to be evaluated is uniformly distributed in space, the method further includes: For each grid to be evaluated, the flood filling degree corresponding to the grid to be evaluated is determined based on the historical flash flood record data corresponding to the grid to be evaluated. When the flood filling degree is greater than the first preset filling degree or less than the second preset filling degree, it is determined that the grid to be evaluated meets the flood filling requirements. If each grid to be evaluated meets the flood filling requirements and the geographical environment corresponding to each grid to be evaluated is uniformly distributed in space, then each grid to be evaluated will be taken as the final multiple grids. If the geographic environment corresponding to a grid to be evaluated is not uniformly distributed in space, and / or a grid to be evaluated does not meet the flood filling requirements, then based on the multiple grids to be evaluated, the grid is re-divided according to a preset adjustment strategy and the preset grid division shape to obtain multiple new grids to be evaluated. The spatial distribution coefficient and flood filling degree of each grid to be evaluated are recalculated until all grids to be evaluated meet the flood filling requirements and the geographic environment corresponding to each grid to be evaluated is uniformly distributed in space. The newly divided grids to be evaluated are then taken as the final multiple grids.

4. The method according to claim 1, characterized in that, The meteorological feature data includes radar echo data corresponding to the target watershed, real-time rainfall monitoring data from each meteorological monitoring station, and rainfall forecast data for a preset future time period; the step of determining the actual rainfall value of each grid at each moment and the rainfall forecast value for a preset future time period for each moment based on the meteorological feature data corresponding to the target watershed at multiple moments includes: Determine the first target location of each meteorological monitoring station and the second target location of each grid; For each time moment, the radar echo data at that time is transformed into regional precipitation intensity distribution data of the target watershed using a preset quantitative relationship model between radar echo reflectivity factor and precipitation intensity. The regional precipitation intensity distribution data is then calibrated using real-time rainfall monitoring data from each meteorological monitoring station at that time and the first target location, resulting in calibrated precipitation intensity distribution data for that time moment. Based on the calibrated precipitation intensity distribution data and the second target location of each grid, a first rainfall value is calculated for each grid at that time moment. Based on the real-time rainfall monitoring data from each meteorological monitoring station at that time, the first target location, and the second target location of each grid, a second rainfall value is calculated for each grid at that time moment using an interpolation algorithm. Finally, based on the first and second rainfall values, the actual rainfall value for each grid at that time moment is calculated. For each moment, based on the rainfall forecast data and first target location of each meteorological monitoring station at that moment, and the second target location of each grid, the rainfall forecast value for the future preset time period of each grid at that moment is calculated by interpolation algorithm.

5. The method according to claim 1, characterized in that, The geographic environment data includes topographic elevation data, vegetation data, land use type data, and historical flash flood information data; based on the geographic environment data corresponding to the grid, the actual rainfall value at each time, the predicted rainfall value, and the geographic environment data corresponding to the target adjacent grid, the actual rainfall value at each time, and the predicted rainfall value, a preset rainfall analysis model is used to determine the water volume parameters and water flow direction parameters corresponding to the grid, including: Based on the topographic elevation data, vegetation data, land use type data, and historical flash flood information data corresponding to the grid, a first environmental feature vector of the grid is constructed; and based on the topographic elevation data, vegetation data, land use type data, and historical flash flood information data corresponding to the target adjacent grid, a second environmental feature vector of the target adjacent grid is constructed; and based on the actual rainfall value and predicted rainfall value at each time corresponding to the grid, a first temporal rainfall feature vector of the grid is constructed; and based on the actual rainfall value and predicted rainfall value at each time corresponding to the target adjacent grid, a second temporal rainfall feature vector of the target adjacent grid is constructed. Based on the first environmental feature vector of the grid and the second environmental feature vector of the target adjacent grid, the spatial correlation features between the grid and the target adjacent grid are extracted through the spatial feature extraction layer of the preset rainfall analysis model; Based on the first temporal rainfall feature vector of the grid and the second temporal rainfall feature vector of the target adjacent grid, the temporal rainfall correlation features between the grid and the target adjacent grid are extracted through the temporal feature extraction layer of the preset rainfall analysis model. The spatial correlation features and the temporal rainfall correlation features are input into the first fully connected layer of the preset rainfall analysis model to obtain the water parameters corresponding to the grid. The spatial correlation features and the temporal rainfall correlation features are input into the second fully connected layer of the preset rainfall analysis model to obtain the water flow direction parameters corresponding to the grid.

6. The method according to claim 1, characterized in that, The step of determining the flash flood level of each grid and the flash flood hazard range of the target watershed based on the water volume parameters corresponding to each grid includes: For each grid, terrain feature data is identified from the geographic environment data corresponding to the grid, and a target feature vector corresponding to the grid is constructed based on the water quantity parameters and terrain feature data corresponding to the grid. Based on the target feature vector corresponding to each grid, the grids are clustered to obtain multiple clusters; For each cluster, based on the water volume parameter corresponding to the central grid of the cluster, the flash flood level corresponding to the central grid is determined, and the flash flood level is used as the flash flood level corresponding to each grid within the cluster; The affected objects corresponding to each grid are identified. For each grid, the hazard value of the grid is determined based on the water volume parameters corresponding to the grid and the affected objects. Based on the hazard values ​​of each grid, the flash flood hazard range of the target watershed is determined.

7. The method according to claim 1, characterized in that, The step of generating the flash flood trajectory of the target watershed based on the water flow direction parameters corresponding to each grid includes: Based on the second target location of each grid and the water flow direction indicated by the water flow direction parameter of each grid, the flash flood initiation grid is determined from the grids; Using the initial grid of the flash flood as the target grid, determine the eight neighboring grids corresponding to the target grid from each grid. Based on the water flow direction parameters of the eight neighboring grids, determine the downstream grid corresponding to the target grid from the eight neighboring grids. Construct the flash flood trajectory based on the water flow directions of the target grid and the downstream grid, and use the downstream grid as the new target grid. Redetermine the downstream grid of the new target grid, and update the flash flood trajectory based on the water flow direction of the new downstream grid until the downstream grid becomes the edge grid of the target watershed, thus obtaining the flash flood trajectory of the target watershed.

8. A flash flood identification device based on Earth big data, characterized in that, include: The data acquisition module is used to determine the target watershed for flash flood identification and acquire the corresponding geographical environment data and meteorological characteristic data at multiple times for the target watershed. The grid partitioning module is used to partition the target watershed into multiple grids based on the geographic environment data, and to construct a grid adjacency matrix based on the geographic location relationship between the grids. The rainfall calculation module is used to determine the actual rainfall value of each grid at each time and the rainfall prediction value for a future preset time period at each time, based on the meteorological characteristic data of multiple times corresponding to the target watershed. The model analysis module is used to determine the target adjacent grids corresponding to each grid based on the grid adjacency matrix. Based on the geographic environment data corresponding to the grid, the actual rainfall value and the predicted rainfall value at each time, and the geographic environment data corresponding to the target adjacent grid, the actual rainfall value and the predicted rainfall value at each time, the module determines the water volume parameters and water flow direction parameters corresponding to the grid through a preset rainfall analysis model. The result generation module is used to determine the flash flood level of each grid and the flash flood hazard range of the target watershed based on the water volume parameters corresponding to each grid, and to generate the flash flood movement trajectory of the target watershed based on the water flow direction parameters corresponding to each grid.

9. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

10. A computer device, comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 7.

Images