A mountain torrent disaster forecasting and early warning method based on multi-source data fusion

Abstract

This application discloses a method for forecasting and early warning of flash floods based on multi-source data fusion, applicable to the field of geological disaster meteorological risk early warning technology. The method includes: identifying radar monitoring blind spots and mapping high-risk target areas for flash floods; deploying BeiDou monitoring base stations based on the high-risk target area map, constructing a collaborative sensing network complementary to rainfall radar, outputting real-time watershed hydrological, meteorological, and topographic data, and simultaneously generating short-term quantitative precipitation forecast data; inputting the watershed hydrological, meteorological, and topographic data and short-term quantitative precipitation forecast data into a watershed flood simulation model for dynamic flood simulation, matching it with a historical scenario database, and outputting a dynamic flood risk field and a set of similar scenario features; matching dynamic early warning thresholds and automatically determining the warning level and scope. This application not only improves the forecast accuracy of flash floods but also effectively reduces the impact of blind spots, enhancing the timeliness of early warnings and the reliability of decision-making.

Description

Technical Field

[0001] This application relates to the field of geological disaster meteorological risk early warning technology, and more specifically, to a method for forecasting and early warning of flash floods based on multi-source data fusion. Background Technology

[0002] The geological and geomorphological types in mountainous areas are very complex. Influenced by precipitation from the East Asian monsoon region, the amount and spatial and temporal distribution of precipitation exhibit significant variability. In such areas, drought and rainstorms alternate frequently, leading to the frequent occurrence of various types of flash floods and geological disasters. Among them, flash floods caused by precipitation are often accompanied by landslides. Especially in small watersheds in mountainous areas, due to drastic topographic changes, steep riverbed slopes, and rapid water flow, coupled with the weak natural water storage capacity of the region, water flow can easily converge rapidly and cause flood disasters.

[0003] In addition, the frequent artificial slope cutting activities by mountain residents to build houses, combined with torrential rain, greatly increase the probability of disasters. The instability caused by torrential rain and artificial slope cutting makes it easy for flash floods and landslides to interact, forming a chain reaction of disasters, often leading to the outbreak of a series of complex disasters.

[0004] In traditional flash flood disaster early warning, due to the influence of terrain obstruction and meteorological attenuation, there may be blind spots in the radar monitoring area, resulting in missing or inaccurate disaster monitoring information in some areas, which affects the accuracy of risk assessment. Moreover, existing technologies often cannot adjust and optimize flood warnings in real time, which can easily lead to delays or errors, especially under conditions of dynamic weather and terrain changes, resulting in insufficient timeliness of warnings and inability to respond to sudden flash flood disasters in a timely manner.

[0005] No effective solutions have yet been proposed to address the problems in the relevant technologies. Summary of the Invention

[0006] To overcome the above problems, this application aims to propose a flash flood disaster forecasting and early warning method based on multi-source data fusion. The purpose is to solve the problem that existing technologies often cannot adjust and optimize flood warnings in real time, and are prone to delays or errors, especially under conditions of large dynamic weather and terrain changes, resulting in insufficient warning timeliness and inability to respond to sudden flash flood disasters in a timely manner.

[0007] Therefore, the specific technical solution adopted in this application is as follows: A method for forecasting and early warning of flash floods based on multi-source data fusion, the method comprising: S1. Integrate topography, geology, disaster-bearing bodies, historical disaster data and real-time meteorological field of the target mountain small watershed, and use atmospheric refraction correction model to simulate, identify radar monitoring blind spots and draw high-risk target area map of flash flood disaster; S2. Based on the high-risk target area map of flash flood disaster, deploy Beidou monitoring base stations, construct a collaborative sensing network that complements the rainfall radar, output watershed hydrological, meteorological and topographic data in real time and generate short-term quantitative precipitation forecast data simultaneously; S3. Input the basin's hydrological, meteorological, and topographic data and short-term quantitative precipitation forecast data into the basin flood simulation model to perform dynamic flood simulation, and match it with the historical scenario database to output the dynamic flood risk field and the matching similar scenario feature set; S4. Based on the dynamic flood risk field and the matching similar scenario feature set, the dynamic early warning threshold is matched to automatically determine the early warning level and scope, and issue early warning instructions.

[0008] Optionally, the method for identifying radar monitoring blind spots and mapping high-risk target areas for flash floods is as follows: Collect high-precision digital elevation models, geological structure maps, disaster-bearing body distribution maps, and historical flash flood disaster data for the target area, and fuse them with real-time meteorological background fields to form a multi-source basic dataset; By inputting the multi-source basic dataset into the atmospheric refraction correction model, the propagation path and signal attenuation of the radar beam are calculated, and a radar blind zone distribution map is generated. Based on multi-source basic datasets, the slope, confluence path length, soil permeability and historical disaster density within the watershed are calculated to generate a distribution map of the intensity of flash flood disaster risk factors. Spatial overlay analysis of radar blind zone distribution map and flash flood disaster risk factor intensity distribution map is performed to generate a high-risk target area map for flash flood disaster.

[0009] Optionally, a method for generating a radar blind zone distribution map: Based on a multi-source basic dataset, real-time meteorological field data is extracted to calculate the atmospheric refractive index profile; the initial conditions of the propagation model are set according to the inherent parameters of the radar, and an atmospheric refractive correction model is constructed. Run the atmospheric refraction correction model, perform ray tracing from the radar station location in all directions, calculate the propagation path of each ray under a specific refractive index profile, and calculate the signal attenuation by integrating along the path according to the radar equation, and output the attenuation calculation results including the signal strength of each spatial grid point. Analyze the attenuation calculation results, determine and generate the grid point detection status identifier for each spatial grid point based on the preset detection signal power threshold; Based on the grid point detection status indicators, the system comprehensively identifies permanent geometric blind spots caused by terrain and features and time-varying meteorological blind spots caused by meteorological attenuation, and generates a radar blind spot distribution map.

[0010] Optionally, the output includes the attenuation calculation results of the signal strength at each spatial grid point, including: Based on the inherent parameters of the radar and the initial azimuth setting, the radar beam is discretized into multiple independent rays with initial direction, power and polarization, generating an initial ray set to be tracked; For each ray in the initial ray set, perform iterative ray tracing calculations in a pre-configured three-dimensional space, and output the sequence of three-dimensional propagation path points of each ray in space; The total signal attenuation caused by atmospheric gases, clouds, and precipitation particles in the three-dimensional propagation path point sequence is calculated by piecewise integration along the radar equation to obtain the remaining power of the current ray. The remaining power of the current ray is distributed to the three-dimensional spatial grid points it covers using a spatial interpolation method, generating and outputting the attenuation calculation results containing the signal intensity of each spatial grid point.

[0011] Alternatively, the method for creating a high-risk target area map for flash flood disasters is as follows: The radar blind zone distribution map and the risk factor intensity distribution map are standardized, and after unifying the layer resolution, a standardized blind zone layer and risk layer are generated. Weight coefficients are defined for the blind zone layer and the risk layer, and a spatial decision weight matrix is ​​constructed based on the contribution of each element to the risk of flash flood disaster. A spatial decision weight matrix is ​​applied to the standardized layers and then overlaid with a grid to generate a comprehensive risk index representing the intensity of flash flood disaster risk in different regions. Based on the preset risk threshold, the comprehensive risk index is divided and regionally aggregated to generate a high-risk target area map for flash flood disasters.

[0012] Optionally, the expression for generating a comprehensive risk index to represent the intensity of flash flood disaster risk in different regions is as follows: ; In the formula, Represents a grid ( a , b The comprehensive risk index of flash flood disasters at location ) The global weighting coefficient represents the radar blind zone; This represents the interaction strength coefficient between the blind spot and the risk; Represents a grid ( a , b Standardized radar blind zone value at ( ); Indicates the spatial autocorrelation baseline coefficient; This represents the spatial autocorrelation gain coefficient; This indicates the total number of risk factor items within the neighborhood; Represents the neighborhood grid First i Standardized values ​​of risk factors; Represents a grid ( a ,b The set of all risk factor items within the neighborhood; This represents the total number of categories of risk factors for flash floods; Indicates the categories of risk factors for flash floods; Indicates the first i Global weight coefficients for risk factors related to flash floods; Represents a grid ( a , b ) i Standardized values ​​of risk factors; This represents the risk nonlinearity correction coefficient.

[0013] Optionally, the dynamic flood risk field and the matching similar scenario feature set are output, including: Input the basin's hydrological, meteorological, and topographic data, along with short-term quantitative precipitation forecast data, into the basin flood simulation model; The basin flood simulation model is used to calculate the runoff generation and flood evolution process of the basin within a preset time period, and outputs a dynamic flood element simulation sequence. Extract key features from the simulation sequence of dynamic flood elements and match them with the historical scenario database; query similar historical flood scenarios and extract features from the historical scenarios. Based on flood element sequences and historical scenario characteristics, a dynamic flood risk field is generated, and a set of similar scenario features matching the dynamic flood risk field is obtained.

[0014] Optionally, the expression for the flood element sequence is: ; In the formula, This represents a simulation sequence of dynamic flood elements; This represents the core computational operator for the generation, confluence, and evolution of floods in a watershed; Represents the coordinates of time step t. Short-term quantitative precipitation forecast values ​​for the location; Representing coordinates Static comprehensive geographical parameters of the watershed at the location; Indicates the first t -1 time step, coordinates Place, No. k The previous simulation value of flood-like elements; Indicates the first k Historical scenario-corrected weighting coefficients for flood-like elements; Represents the coordinates of time step t. Place, No. k Historical similarity scenario values ​​for flood-like elements; Represents the discrete time step; Indicates the type of flood element; Represents the spatial coordinates of the watershed; Indicates the entire spatial range of the watershed; This indicates the total number of time steps for the preset simulation period; This represents the total number of categories of flood elements; Indicates the first t Time step, coordinates Place, No. k The final simulation values ​​of flood-like elements.

[0015] Optionally, the method for querying similar historical flood scenarios and extracting historical scenario features is as follows: Extract flood peak, peak time, flood rate and spatial distribution pattern indicators from the dynamic flood element simulation sequence to construct a multi-dimensional feature vector describing the current event; The multi-dimensional feature vectors are matched with the feature sets in the historical scenario database to retrieve and generate a list of matching historical scenarios sorted by similarity. Based on the list of matching historical scenarios, the complete original feature data of the corresponding scenarios are retrieved and extracted from the historical scenario database to generate historical scenario features that match the historical scenarios.

[0016] Compared with the prior art, this application has the following beneficial effects: 1. This application, through multi-source data fusion and high-precision model simulation, can identify high-risk areas of flash floods in real time and accurately predict the evolution of floods. By combining topographic, meteorological, and historical disaster data with short-term quantitative precipitation forecasts to generate a dynamic flood risk field, and optimizing the early warning threshold based on historical scenario matching, it can achieve accurate disaster risk assessment and timely early warning. This not only improves the forecast accuracy of flash floods but also effectively reduces the impact of blind spots, and improves the timeliness of early warnings and the reliability of decision-making.

[0017] 2. This application comprehensively considers the effects of terrain obstruction and meteorological attenuation, and uses spatial weighted overlay and risk factor intensity assessment to generate a dynamic risk index, thereby providing comprehensive early warning support for flash flood disasters; through precise radar and meteorological data processing, it reduces the impact of blind spots, improves the accuracy of risk assessment, and provides a reliable basis for emergency response.

[0018] 3. This application combines short-term quantitative precipitation forecasting with historical scenario characteristics to predict flood risks in real time. It can also correct and optimize flood characteristics based on historical data, thereby improving the accuracy and timeliness of flood warnings. Through efficient historical scenario matching and dynamic simulation, it reduces the blind spots in warnings and provides scientific support for the prevention and emergency response to flash floods. Attached Figure Description

[0019] The above-mentioned features, characteristics, and advantages of this application, as well as their implementation methods, will become clearer and more understandable in conjunction with the following description of the embodiments, which are illustrated in detail with reference to the accompanying drawings. Schematic diagrams are shown here: Figure 1 This is a flowchart of a flash flood disaster forecasting and early warning method according to an embodiment of this application. Detailed Implementation

[0020] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0021] According to embodiments of this application, a method for forecasting and early warning of flash floods based on multi-source data fusion is provided. Through multi-source data fusion and high-precision model simulation, it can identify high-risk areas for flash floods in real time and accurately predict the flood evolution process. By combining topographic, meteorological, and historical disaster data with short-term quantitative precipitation forecasts to generate a dynamic flood risk field, and optimizing the early warning threshold based on historical scenario matching, accurate disaster risk assessment and timely early warning are achieved. Figure 1 As shown, the method includes: S1. Integrate topography, geology, disaster-bearing bodies, historical disaster data and real-time meteorological field of the target mountainous small watershed, and use atmospheric refraction correction model to simulate, identify radar monitoring blind spots and draw high-risk target area maps of flash flood disasters.

[0022] Preferably, the method for identifying radar monitoring blind spots and mapping high-risk target areas for flash floods is as follows: Collect high-precision digital elevation models, geological structure maps, disaster-bearing body distribution maps, and historical flash flood disaster data for the target area, and fuse them with real-time meteorological background fields to form a multi-source basic dataset; By inputting the multi-source basic dataset into the atmospheric refraction correction model, the propagation path and signal attenuation of the radar beam are calculated, and a radar blind zone distribution map is generated. Based on multi-source basic datasets, the slope, confluence path length, soil permeability and historical disaster density within the watershed are calculated to generate a distribution map of the intensity of flash flood disaster risk factors. Spatial overlay analysis of radar blind zone distribution map and flash flood disaster risk factor intensity distribution map is performed to generate a high-risk target area map for flash flood disaster.

[0023] Preferably, the method for generating a radar blind zone distribution map is as follows: Based on a multi-source basic dataset, real-time meteorological field data is extracted to calculate the atmospheric refractive index profile; the initial conditions of the propagation model are set according to the inherent parameters of the radar, and an atmospheric refractive correction model is constructed. Run the atmospheric refraction correction model, perform ray tracing from the radar station location in all directions, calculate the propagation path of each ray under a specific refractive index profile, and calculate the signal attenuation by integrating along the path according to the radar equation, and output the attenuation calculation results including the signal strength of each spatial grid point. Analyze the attenuation calculation results, determine and generate the grid point detection status identifier for each spatial grid point based on the preset detection signal power threshold; Based on the grid point detection status indicators, the system comprehensively identifies permanent geometric blind spots caused by terrain and features and time-varying meteorological blind spots caused by meteorological attenuation, and generates a radar blind spot distribution map.

[0024] Preferably, the output includes the attenuation calculation results of the signal strength at each spatial grid point, including: Based on the inherent parameters of the radar and the initial azimuth setting, the radar beam is discretized into multiple independent rays with initial direction, power and polarization, generating an initial ray set to be tracked; For each ray in the initial ray set, perform iterative ray tracing calculations in a pre-configured three-dimensional space, and output the sequence of three-dimensional propagation path points of each ray in space; The total signal attenuation caused by atmospheric gases, clouds, and precipitation particles in the three-dimensional propagation path point sequence is calculated by piecewise integration along the radar equation to obtain the remaining power of the current ray. The remaining power of the current ray is distributed to the three-dimensional spatial grid points it covers using a spatial interpolation method, generating and outputting the attenuation calculation results containing the signal intensity of each spatial grid point.

[0025] Preferably, the method for creating a high-risk target area map for flash flood disasters is as follows: The radar blind zone distribution map and the risk factor intensity distribution map are standardized, and after unifying the layer resolution, a standardized blind zone layer and risk layer are generated. Weight coefficients are defined for the blind zone layer and the risk layer, and a spatial decision weight matrix is ​​constructed based on the contribution of each element to the risk of flash flood disaster. A spatial decision weight matrix is ​​applied to the standardized layers and then overlaid with a grid to generate a comprehensive risk index representing the intensity of flash flood disaster risk in different regions. Based on the preset risk threshold, the comprehensive risk index is divided and regionally aggregated to generate a high-risk target area map for flash flood disasters.

[0026] Preferably, the expression for generating the comprehensive risk index to represent the intensity of flash flood disaster risk in different regions is as follows: ; In the formula, Represents a grid ( a , b The comprehensive risk index of flash flood disasters at location ) The global weighting coefficient represents the radar blind zone; This represents the interaction strength coefficient between the blind spot and the risk; Represents a grid ( a , b Standardized radar blind zone value at ( ); Indicates the spatial autocorrelation baseline coefficient; This represents the spatial autocorrelation gain coefficient; This indicates the total number of risk factor items within the neighborhood; Represents the neighborhood grid First i Standardized values ​​of risk factors; Represents a grid ( a , b The set of all risk factor items within the neighborhood; This represents the total number of categories of risk factors for flash floods; Indicates the categories of risk factors for flash floods; Indicates the first i Global weight coefficients for risk factors related to flash floods; Represents a grid ( a , b ) i Standardized values ​​of risk factors; This represents the risk nonlinearity correction coefficient.

[0027] It should be explained that the basic settings for this instance are as follows: Target area: XX Creek watershed (XX District, XX City, area 200 km²) 2 (A typical small watershed in a mountainous area, prone to flash floods). Grid resolution: uniformly 100m×100m (total number of grids ≈ 20,000); Radar station: Lin'an Meteorological Radar Station (coordinates 119.6°E, 30.2°N, altitude 500m, C-band radar); Simulation time: XX / XX / 20XX (short-term heavy rainfall during the plum rain season, hourly rainfall intensity 80mm, typical flash flood condition). Risk factors: Four core factors were selected (m=4): slope, runoff path length, soil permeability, and historical disaster density; Core parameter presets: The weights and correction coefficients are determined based on the characteristics of flash floods in mountainous areas and expert experience, and comply with the formula constraints; The basic dataset is set as follows: 1. Topographic data: 10m resolution DEM, extracting slope (35° steep slope upstream, 5° gentle slope downstream) and confluence path length (2km upstream, 10km downstream). 2. Geological data: Granite area (soil permeability 0.05 cm / h), sandstone area (0.1 cm / h), geological structure is mainly fault zone; 3. Disaster-bearing areas: 12 villages (mountainous population density 50 people / km²) 2 Towns 200 people / km 2 ), 2 county roads, and 5 township roads; 4. Historical disasters: 18 flash flood disaster sites in the past 10 years, with a disaster density of 0.15 sites / km² in high-incidence areas. 2 Low-incidence areas: 0.02 cases / km 2 ); 5. Real-time weather field: hourly rainfall intensity 80mm (upstream), 20mm (downstream), atmospheric refractive index profile (320N units near the ground, 300N units at 1km altitude). The radar parameters are set as follows: transmit power 500kW, antenna gain 45dB, beamwidth 1°, and ray discretization (14400 initial rays for every 1° azimuth angle + 0.5° elevation angle). In ray tracing, the propagation path is calculated iteratively in three dimensions, starting from the radar station. Example ray (azimuth 30°, elevation 5°): 0-5km (altitude 500-1200m) → 5-10km (altitude 1200-1800m, crossing the Qingliangfeng mountain area, with terrain obstruction). Atmospheric gas attenuation is 0.2 dB / km, precipitation of water particles at 80 mm / h is 1.5 dB / km, and terrain obstruction results in a residual power of 8 dBm (below the detection threshold of 10 dBm). Blind spot markings (2 typical grids): Grid A ( a =500, b =300, upstream Qingliang Peak): Residual power 8dBm < 10dBm → Complete blind zone (standardized c ( a , b =1); Grid B ( a =1000, b =800, downstream towns): Residual power 25dBm>10dBm→weak blind zone (standardized c( a , b =0.1); Permanent geometric blind spots (terrain obstruction, accounting for 15%) + time-varying meteorological blind spots (heavy precipitation areas, accounting for 8%), totaling 23% of the blind spots; The four risk factors are standardized (value range [0, 1]) to generate an intensity distribution map. The factor values ​​for two typical grids are as follows: Table 1 Distribution of Flash Flood Risk Factors

[0028] As shown in Table 1, when setting the radar blind zone weight... w d A value of 0.3 indicates that the blind zone caused monitoring failure, contributing 30%. Risk slope weight w A value of 1 equals 0.25, indicating that terrain slope is the core driving factor for flash floods; Bus length weight w A value of 2 equals 0.2, indicating that the confluence velocity affects flood evolution; Soil permeability weight w 3 equals 0.2, indicating that the infiltration capacity determines the flow rate; Historical disaster density weight w A value of 4 equals 0.05, indicating that historical disasters reflect risk inertia. Blind spots and risk interaction coefficient A value of 0.5 indicates that the risk amplification effect of the blind spot in the mountainous area is moderate; Spatial autocorrelation baseline coefficient A value of 0.8 indicates a high risk concentration in mountainous areas and a high baseline value. Spatial autocorrelation gain coefficient A value of 0.4 indicates that the neighborhood risk has a moderate impact on the target grid gain; Nonlinear correction coefficient A value of 1.5 indicates that the risk of flash floods in mountainous areas increases non-linearly. Total number of items in the neighborhood M It is 36; Neighborhood set A 3×3 grid indicates that the target grid is covered by its 8 neighboring grids plus itself.

[0029] Substitute the above parameters into the comprehensive risk index of flash floods. In the calculation, the comprehensive risk index of flash flood disaster for grid A is obtained. =0.92; Comprehensive risk index of flash flood disaster in grid B =0.07; Aggregate adjacent extremely high-risk grids (grid A and its 12 surrounding grids). >0.8), forming a high-risk target area in the upper reaches of the XX River, Qingliang Peak; The target area is approximately 5 km². 2 (Accounting for 2.5% of the basin), with an average comprehensive risk index of 0.90, and considering the radar blind spots and the distribution of high-risk elements, it is clearly identified as a core area for flash flood warning; The final high-risk target area map for flash flood disasters marks the location, area, and risk level of the target area, and includes plans for manual monitoring and personnel evacuation.

[0030] S2. Based on the high-risk target area map of flash flood disaster, deploy Beidou monitoring base stations, construct a collaborative sensing network that complements the rainfall radar, output watershed hydrological, meteorological and topographic data in real time and generate short-term quantitative precipitation forecast data simultaneously.

[0031] It should be explained that, according to the high-risk target area map of flash flood disasters, BeiDou base stations are prioritized for deployment in high-risk areas (such as mountainous areas, catchment areas, and areas with a high frequency of historical disasters). BeiDou satellite positioning (GNSS) is used for high-precision positioning, which can provide accurate coordinate information for each monitoring point and help track the position of each sensor in real time. The deployed base stations are connected to other monitoring equipment (such as weather stations, flow meters, rain gauges, etc.) through BeiDou satellite communication to ensure smooth data transmission. Real-time hydrological and meteorological data (such as precipitation, water level, flow rate, etc.) and geographical data (such as watershed location, coordinates, etc.) will be uploaded in real time, providing important input information for flood warning. Rainfall radar is used for precipitation monitoring, obtaining the spatial distribution and intensity of precipitation through radar wave reflection data. This is crucial for short-term quantitative precipitation forecasting. Radar can cover a large area, capturing precipitation changes at various points within a watershed and providing real-time data. Rainfall radar and BeiDou base stations each have different monitoring tasks: the former is used for precipitation monitoring, and the latter for hydrological and geographic data collection. By fusing data from rainfall radar and BeiDou base stations and using data fusion algorithms (such as Kalman filtering), more accurate hydrological and meteorological warnings can be provided. For example, by combining real-time radar precipitation data with watershed flow and water level data provided by BeiDou sensors, flood evolution can be dynamically predicted. Based on the deployed BeiDou base stations, meteorological sensors, and other hydrological monitoring equipment, real-time data on water level, flow, precipitation, and meteorological data (such as temperature, humidity, and wind speed) within the basin are collected. Based on real-time precipitation data (collected through rain-measuring radar), short-term quantitative precipitation forecasts are generated using numerical weather prediction models (such as WRF and GFS). These forecasts are typically 1-hour, 3-hour, and 6-hour precipitation predictions and can help assess whether precipitation is sufficient to trigger flash floods.

[0032] S3. Input the basin's hydrological, meteorological, and topographical data and short-term quantitative precipitation forecast data into the basin flood simulation model to perform dynamic flood simulation. Then, match the simulation with the historical scenario database to output the dynamic flood risk field and the matching similar scenario feature set.

[0033] Preferably, the output dynamic flood risk field and the matching similar scenario feature set include: Input the basin's hydrological, meteorological, and topographic data, along with short-term quantitative precipitation forecast data, into the basin flood simulation model; The basin flood simulation model is used to calculate the runoff generation and flood evolution process of the basin within a preset time period, and outputs a dynamic flood element simulation sequence. Extract key features from the simulation sequence of dynamic flood elements and match them with the historical scenario database; query similar historical flood scenarios and extract features from the historical scenarios. Based on flood element sequences and historical scenario characteristics, a dynamic flood risk field is generated, and a set of similar scenario features matching the dynamic flood risk field is obtained.

[0034] Preferably, the expression for the flood element sequence is: ; In the formula, This represents a simulation sequence of dynamic flood elements; This represents the core computational operator for the generation, confluence, and evolution of floods in a watershed; Indicates the first t Time step, coordinates Short-term quantitative precipitation forecast values ​​for the location; Representing coordinates Static comprehensive geographical parameters of the watershed at the location; Indicates the first t -1 time step, coordinates Place, No. k The previous simulation value of flood-like elements; Indicates the first k Historical scenario-corrected weighting coefficients for flood-like elements; Represents the coordinates of time step t. Place, No. k Historical similarity scenario values ​​for flood-like elements; Represents the discrete time step; Indicates the type of flood element; Represents the spatial coordinates of the watershed; Indicates the entire spatial range of the watershed; This indicates the total number of time steps for the preset simulation period; This represents the total number of categories of flood elements; Indicates the first t Time step, coordinates Place, No. k The final simulation values ​​of flood-like elements.

[0035] Preferably, the method for querying similar historical flood scenarios and extracting historical scenario features is as follows: Extract flood peak, peak time, flood rate and spatial distribution pattern indicators from the dynamic flood element simulation sequence to construct a multi-dimensional feature vector describing the current event; The multi-dimensional feature vectors are matched with the feature sets in the historical scenario database to retrieve and generate a list of matching historical scenarios sorted by similarity. Based on the list of matching historical scenarios, the complete original feature data of the corresponding scenarios are retrieved and extracted from the historical scenario database to generate historical scenario features that match the historical scenarios.

[0036] S4. Based on the dynamic flood risk field and the matching similar scenario feature set, a dynamic early warning threshold is matched, and based on the triggering state of the dynamic threshold and the development trend of the risk, the early warning level and scope are automatically determined; and early warning instructions are generated and issued.

[0037] It should be explained that the preferred deployment location for BeiDou base stations is the core of the high-risk target area (Qingliangfeng, around grid A, coordinates 119.58°E, 30.22°N, altitude 1100m), where BeiDou monitoring base station BD-01 will be deployed (prioritizing coverage of mountainous areas / catchment areas / areas with a high incidence of historical disasters). Base station functions: GNSS high-precision positioning (coordinate accuracy ±0.1m), connecting rain gauges, water level gauges, flow meters, and meteorological sensors, and transmitting data via BeiDou satellite communication; Real-time data collection: 1. Hydrology: Rainfall at grid A: 80 mm / h, water level: 0.5 m, flow rate: 5 m³ / h 3 / s; 2. Geography: Base station coordinates (119.58°E, 30.22°N), watershed location (upstream of Changhua River); 3. Weather: Temperature 22℃, humidity 95%, wind speed 3m / s; 4. Rainfall radar (Lin'an station): Inverted grid A rainfall rate of 80 mm / h, downstream grid B rainfall rate of 20 mm / h; 5. Data fusion: After Kalman filtering correction, the precipitation in grid A is 78 mm / h (more accurate); 6. Short-term forecast: generated by WRF model, with a 1-hour forecast of 75 mm / h and a 3-hour forecast of 60 mm / h at grid A.

[0038] Based on the data collected above, the following specific parameters are set: The short-term precipitation forecast value at t=1 is 75 mm / h. Static geographic parameters Medium slope of 35°, soil permeability of 0.05 cm / h, runoff length of 2 km; middle t =0 initial simulation value, =0.5m (water level); =5m 3 / s; =0m (flooding depth); Historical Adjustment Weights middle =0.2、 =0.15、 =0.25; Historical similarity scenario value middle =0.8m =12m 3 / s、 =0.3m; Core computational operators for watershed flood generation, runoff, and evolution middle =0.7m =10m 3 / s、 =0.2m; The total time step T is 24; the number of flood element categories M is 3; the watershed area is... For the entire XX Creek small watershed; Based on the above, a simulation sequence of dynamic flood elements is obtained. ; Based on the above, the peak flow (estimated t=8, water level 2.2m), peak time (2 hours), rise rate (0.36m / 15min), and spatial distribution (high upstream, low downstream) were obtained; the similarity with the XX Creek flash flood was 92% (historical characteristics: peak flow 2.3m, peak time 2 hours, rise rate 0.35m / 15min). Based on the evolution of flash floods, the inundation range, and early warning response records, a set of similar scenario features is formed for simulation. t =8 (2 hours) at grid A ; The judgment result triggers the yellow threshold (which is considered high risk), and combined with the radar blind zone and high risk, it is upgraded to an orange alert; Command details: [Orange Alert for Flash Floods in XX Creek] High-risk target area at XX Peak (within 5km of BD-01 base station) 2 The water level is expected to reach 2.2 meters and the flow rate to exceed 80 cubic meters per second in the next two hours. 3 Immediately relocate 120 people from 3 upstream villages to resettlement sites and close 2 upstream rural roads; send real-time information through township broadcasts and WeChat public accounts, and simultaneously push it to the Lin'an Flood Control Headquarters.

[0039] It should be noted that the calculation formulas and all parameters involved in the calculations in this application have been dimensionless beforehand. The process of dimensionless processing is well known in the industry and will not be described here.

[0040] Although the present application has disclosed the preferred embodiments above, the embodiments are merely examples for the purpose of illustration and are not intended to limit the present application. Those skilled in the art can make some modifications and refinements without departing from the spirit and scope of the present application. The scope of protection claimed by the present application should be determined by the claims.

Claims

1. A flash flood disaster forecasting and warning method based on multi-source data fusion, characterized in that, The method comprises: S1, fuse the terrain, geology, disaster-bearing body, historical disaster data and real-time meteorological field of the target mountainous small watershed, simulate through the atmospheric refraction correction model, identify the radar monitoring blind area and draw a mountain flood disaster high-risk target area map; S2, based on the mountain flood disaster high-risk target area map, arrange the Beidou monitoring base station, construct a collaborative perception network complementary to the rain measuring radar, output the watershed hydro-meteorological terrain data in real time and synchronously generate short-term quantitative precipitation forecast data; S3, input the watershed hydro-meteorological terrain data and short-term quantitative precipitation forecast data into the watershed flood simulation model, perform dynamic simulation of the flood, and match with the historical scenario library to output the dynamic flood risk field and the matched similar scenario characteristic set; S4, based on the dynamic flood risk field and the matched similar scenario characteristic set, match the dynamic early warning threshold to automatically determine the early warning level and range, and issue the early warning instruction.

2. The flash flood forecasting and warning method according to claim 1, characterized in that, The method for identifying the radar monitoring blind area and drawing the mountain flood disaster high-risk target area map comprises: Collect high-precision digital elevation models, geological structure maps, disaster-bearing body distribution maps and historical mountain flood disaster data of the target area, and fuse them with real-time meteorological background fields to form a multi-source basic data set; Input the multi-source basic data set into the atmospheric refraction correction model, calculate the propagation path and signal attenuation of the radar beam, and generate a radar blind area distribution map; Based on the multi-source basic data set, calculate the slope, confluence path length, soil permeability and historical disaster density in the watershed to generate a mountain flood disaster risk factor intensity distribution map; Perform spatial overlay analysis on the radar blind area distribution map and the mountain flood disaster risk factor intensity distribution map to generate a mountain flood disaster high-risk target area map.

3. The flash flood forecasting and warning method according to claim 2, characterized in that, The method for generating the radar blind area distribution map comprises: Based on the multi-source basic data set, extract real-time meteorological field data to calculate the atmospheric refraction rate profile; set the initial conditions of the propagation model according to the inherent parameters of the radar, and construct an atmospheric refraction correction model; Run the atmospheric refraction correction model to perform ray tracing from the radar site position to all directions, calculate the propagation path of each ray under a specific refraction rate profile, and calculate the signal attenuation along the path according to the radar equation to output the attenuation calculation results containing the signal intensity of each spatial grid point; Analyze the attenuation calculation results, determine and generate the grid point detection state identifier of each spatial grid point according to a preset detection signal power threshold; Based on the grid point detection state identifier, comprehensively identify the permanent geometric blind area caused by terrain and object shielding and the time-varying meteorological blind area caused by meteorological attenuation to generate a radar blind area distribution map.

4. The flash flood forecasting and warning method according to claim 3, characterized in that, The output of the attenuation calculation results containing the signal intensity of each spatial grid point comprises: Based on the inherent parameters of the radar and the initial orientation, discretize the radar beam into multiple independent rays with initial direction, power and polarization to generate an initial ray set to be traced; For each ray in the initial ray set, perform step-by-step iterative ray tracing calculation in the pre-configured three-dimensional space to output a three-dimensional propagation path point sequence of each ray in space; According to the radar equation, calculate the total signal attenuation caused by atmospheric gases, clouds and precipitation particles along the path piecewise integral of the three-dimensional propagation path point sequence to obtain the remaining power of the current ray; The remaining power of the current ray is distributed to the three-dimensional spatial grid points it covers using a spatial interpolation method, generating and outputting the attenuation calculation results containing the signal intensity of each spatial grid point.

5. The flash flood forecasting and warning method according to claim 2, characterized in that, The method for creating the high-risk target area map for flash floods is as follows: The radar blind zone distribution map and the risk factor intensity distribution map are standardized, and after unifying the layer resolution, a standardized blind zone layer and risk layer are generated. Weight coefficients are defined for the blind zone layer and the risk layer, and a spatial decision weight matrix is ​​constructed based on the contribution of each element to the risk of flash flood disaster. A spatial decision weight matrix is ​​applied to the standardized layers for grid-weighted overlay to generate a comprehensive risk index representing the intensity of flash flood disaster risk in different regions; Based on the preset risk threshold, the comprehensive risk index is divided and regionally aggregated to generate a high-risk target area map for flash flood disasters.

6. The flash flood forecasting and warning method according to claim 5, characterized in that, The expression for generating the comprehensive risk index to represent the intensity of flash flood disaster risk in different regions is as follows: ； In the formula, represents the flash flood disaster comprehensive risk index at the grid ( a , b ); represents the global weight coefficient of the radar blind area; represents the blind area and risk interaction intensity coefficient; represents the normalized radar blind area value at the grid ( a , b ); represents the spatial autocorrelation reference coefficient; represents the spatial autocorrelation gain coefficient; represents the total number of risk factor items in the neighborhood; represents the normalized value of the risk factor of the class at the grid i neighborhood; represents the set of all risk factor items in the grid ( a , b ) neighborhood; represents the total number of categories of flash flood disaster risk factors; represents the category of flash flood disaster risk factors; represents the global weight coefficient of the i category of flash flood disaster risk factors; represents the normalized value of the a category of risk factors at the grid ( b , i ); represents the risk non-linear correction coefficient.

7. The flash flood forecasting and warning method of claim 1, wherein, The output dynamic flood risk field and the matching similar scenario feature set include: Input the basin's hydrological, meteorological, and topographic data, along with short-term quantitative precipitation forecast data, into the basin flood simulation model; The basin flood simulation model is used to calculate the runoff generation and flood evolution process of the basin within a preset time period, and outputs a dynamic flood element simulation sequence. Extract key features from the simulation sequence of dynamic flood elements and match them with the historical scenario database; query similar historical flood scenarios and extract features from the historical scenarios. Based on flood element sequences and historical scenario characteristics, a dynamic flood risk field is generated, and a set of similar scenario features matching the dynamic flood risk field is obtained.

8. The flash flood forecasting and warning method according to claim 7, characterized in that, The expression for the flood element sequence is: ； In the formula, represents a dynamic flood element simulation sequence; represents a core calculation operator of flood runoff and evolution in a basin; represents a short-term quantitative precipitation forecast value at the t-th time step, coordinates ; represents a static comprehensive geographic parameter of the basin at coordinates ; represents a historical scenario correction weight coefficient of the t-th flood element at the t-th time step, coordinates t ; , and the previous time simulation value of the t-th flood element; k ; represents a historical similar scenario value of the t-th flood element at the t-th time step, coordinates k ; ; ; k ; represents a discrete time step; represents a flood element type; represents a basin spatial coordinate; represents a global spatial range of the basin; represents a total number of time steps of a preset simulation time period; represents a total number of flood element categories; represents a final simulation value of the t-th flood element at the t-th time step, coordinates t ; ; k ; 9. The flash flood forecasting and warning method of claim 8, wherein, The method for querying similar historical flood scenarios and extracting historical scenario features is as follows: Extract flood peak, peak time, flood rate and spatial distribution pattern indicators from the dynamic flood element simulation sequence to construct a multi-dimensional feature vector describing the current event; The multi-dimensional feature vectors are matched with the feature sets in the historical scenario database to retrieve and generate a list of matching historical scenarios sorted by similarity. Based on the list of matching historical scenarios, the complete original feature data of the corresponding scenarios are retrieved and extracted from the historical scenario database to generate historical scenario features that match the historical scenarios.

Images