Early warning method for important village of mountain flood prevention based on small reservoir water and rainfall condition monitoring

Abstract

This invention discloses a method for early warning of flash floods in important villages based on small reservoir hydrological and rainfall monitoring. The method first collects data from the study area, constructing a multi-source heterogeneous data fusion system and a hydrological-topographic-socioeconomic correlation network for the study area. Then, it constructs spatial location and water flow path correlations between small reservoirs and important villages, generating inundation range maps and delineating risk zones based on the topography, building layout, and set water level thresholds of the important villages. A reservoir-river-village hydrodynamic coupling model is constructed to calculate the correlation between small reservoir monitoring indicators and village disaster events, and this is input into a disaster-causing water level inversion model to obtain the probability distribution of disaster-causing water levels. Finally, a flash flood disaster early warning model is constructed, extracting the probability distribution of disaster-causing water levels as model input to perform multi-timescale risk forecasting and classify early warning levels, developing differentiated early warning schemes for different levels. This improves the scientific rigor and effectiveness of flash flood disaster prevention.

Description

Technical Field

[0001] This invention belongs to the field of flash flood disaster prevention and early warning technology, specifically relating to a flash flood prevention and early warning method for important villages based on water and rainfall monitoring of small reservoirs. Background Technology

[0002] Flash floods are characterized by their suddenness, destructiveness, and high difficulty in prediction, seriously threatening the lives and property of residents in important mountain villages. Existing flash flood early warning technologies have several shortcomings: First, in terms of data utilization, multi-source heterogeneous data are scattered and independent, lacking effective fusion and correlation analysis, making it difficult to uncover potential connections between data and failing to provide comprehensive support for early warning. Second, in terms of model construction, traditional early warning models are mostly based on single physical mechanisms or simple data statistics, failing to fully consider the complex spatial, water flow, and socio-economic relationships between small reservoirs and downstream villages, resulting in insufficient accuracy and targeting of early warnings. Third, in the early warning classification and plan formulation stages, the classification of early warning levels is coarse, lacking differentiated responses to different risk areas and stages of disaster development, making it difficult to meet actual disaster prevention needs. Therefore, an innovative early warning method is urgently needed to integrate multi-source data, construct precise correlation and early warning models, and achieve accurate early warning and effective defense against flash floods. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the purpose of this invention is to provide an early warning method for flood prevention in important villages based on water and rainfall monitoring of small reservoirs, so as to solve the aforementioned problems existing in the prior art.

[0004] This invention is achieved through the following technical solution: a method for early warning of flash floods in important villages based on water and rainfall monitoring of small reservoirs, comprising the following steps:

[0005] Step S1: Collect data from the study area, construct a multi-source heterogeneous data fusion system, and build a hydrological-topographic-socioeconomic correlation network for the study area based on graph neural networks;

[0006] Step S2: Construct spatial location relationships and water flow path relationships for small reservoirs and important villages respectively. Based on the topography and building layout of important villages, set several different water level thresholds, generate inundation range maps, assess the risk level of different areas, and delineate risk zones.

[0007] Step S3: Based on the hydrological-topographic-socioeconomic correlation network of the study area, construct a reservoir-river-village hydrodynamic coupling model, use time series correlation analysis algorithm to calculate the correlation between small reservoir monitoring indicators and village disaster events, and input the pre-constructed disaster-causing water level inversion model to obtain the probability distribution of disaster-causing water level.

[0008] Step S4: Construct a flash flood disaster early warning model, extract the probability distribution of disaster-causing water levels as model input, conduct multi-timescale risk forecasting and classify early warning levels, and formulate differentiated early warning plans for different early warning levels.

[0009] Specifically, step S1 further comprises:

[0010] Step S11: Collect data for the study area, including meteorological data, hydrological data, topographic data, basic information on small reservoirs, historical monitoring data, early warning indicators for important villages, disaster-prone water levels, and risk zone delineation maps;

[0011] Step S12: Use time series interpolation algorithm to synchronize data with different sampling frequencies, and build a unified geographic coordinate system based on GIS spatial indexing technology to obtain a multi-source heterogeneous data fusion system;

[0012] Step S13: Use the spatial influence propagation algorithm to simulate the influence path and intensity of reservoir water level changes on downstream villages through the water system network. Construct a reservoir-river-village relationship graph based on graph neural network, where nodes represent geographical entities and edges represent water flow paths and influence weight relationships.

[0013] Step S14: Construct a population distribution heat map and a building vulnerability matrix; calculate the comprehensive risk index for different regions based on the analytic hierarchy process; identify the spatial coupling relationship between facilities and flash flood risk; and construct a socio-economic impact model.

[0014] Step S15: The hydrological-topographic-socioeconomic relationship network of the study area is constructed by the reservoir-river-village relationship diagram and the socioeconomic impact model.

[0015] Specifically, step S2 further comprises:

[0016] Step S21: Vectorize the small reservoirs, important villages, rivers, and roads to construct a geographic element database, calculate the straight-line distance and shortest path distance between the small reservoirs and each important village, and construct a spatial location relationship map.

[0017] Step S22: Construct a distributed hydrological model. Based on the topographic data and historical hydrological data of the study area, simulate the runoff generation and confluence process of small reservoirs under different rainfall conditions and the evolution path of water flow to important villages. Identify key nodes of water flow and potential inundation areas, and construct the water flow path correlation between small reservoirs and important villages.

[0018] Step S23: Based on the topography and building layout of important villages, set several different water level thresholds, then use a distributed hydrological model to simulate and generate an inundation range map, assess the risk level of different areas, and delineate risk zones.

[0019] Specifically, step S23 further comprises:

[0020] Step S23a: Based on the simulation results of the distributed hydrological model, combined with the topography and building layout of the village, set several different water level thresholds, and then use the distributed hydrological model to simulate and calculate the inundation range of the corresponding important village. Organize and draw the inundation range data corresponding to all different water level thresholds to obtain the inundation range map.

[0021] Step S23b: Using inundation depth, inundation duration, population density, and building seismic resistance as indicators, the analytic hierarchy process (AHP) is used to assess the risk level of different regions under flash flood disasters, and to classify them into high-risk, medium-risk, and low-risk areas.

[0022] Specifically, step S3 further comprises:

[0023] Step S31: Construct a reservoir-river-village hydrodynamic coupling model based on the hydrological-topographic-socioeconomic relationship network of the study area;

[0024] Step S32: Use time-series correlation analysis algorithm to calculate the correlation between monitoring indicators of small reservoirs and disaster events in villages;

[0025] Step S33: Construct a disaster-causing water level inversion model based on physical constraints using a long short-term memory network;

[0026] Step S34: Input the correlation between small reservoir monitoring indicators and village disaster events, along with historical flash flood data, into the disaster-causing water level inversion model to simulate the corresponding disaster-causing water level and calculate the probability distribution of disaster-causing water levels in important villages.

[0027] Specifically, step S31 further comprises:

[0028] Step S31a: Based on the hydrological-topographic-socioeconomic relationship network of the study area, construct a digital terrain model of the river channels and important villages;

[0029] Step S31b: Collect river data to determine the spatial distribution of river roughness coefficient, and use an adaptive grid partitioning algorithm to grid the important village areas;

[0030] Step S31c: Construct a one-dimensional river channel model and a two-dimensional floodplain model respectively. Using the downstream section of the one-dimensional river channel as the interface, establish a real-time data link. The one-dimensional model automatically transmits the section water level and flow rate to the two-dimensional model as the upstream boundary. The two-dimensional model calculates the floodplain and village inundation situation to obtain the reservoir-river-village hydrodynamic coupling model.

[0031] Specifically, step S32 further comprises:

[0032] Step S32a: Use the phase space reconstruction method to map the time series of rainfall, reservoir water level, and river flow to a high-dimensional space;

[0033] Step S32b: Under time window constraints, mine the association rules between monitoring data and disaster-causing events;

[0034] Step S32c: Construct a structural equation model, use Granger causality test to determine the causal relationship between reservoir monitoring indicators and village disaster water level, quantify the influence path of each factor on disaster water level, and obtain the correlation between small reservoir monitoring indicators and village disaster events.

[0035] Specifically, step S34 further comprises:

[0036] Step S34a: Input the correlation between small reservoir monitoring indicators and village disaster events, along with historical flash flood data, into the disaster-causing water level inversion model;

[0037] Step S34b: Construct a 16-grid risk assessment matrix with inundation depth and inundation duration as the horizontal and vertical axes, and determine the risk level corresponding to each cell by combining expert knowledge and historical disaster data.

[0038] Step S34c: Using the Kriging interpolation algorithm, the risk assessment results of discrete points are extended into a continuous risk distribution map, and the probability distribution of disaster-prone water levels in important villages is calculated.

[0039] Specifically, step S4 further comprises:

[0040] Step S41: Couple the hydrodynamic coupling model, the disaster-causing water level inversion model, and the risk zoning model to obtain the flash flood disaster early warning model;

[0041] Step S42: Extract the probability distribution of disaster-causing water levels and input it into the flash flood disaster early warning model to simulate the development trend of flash flood disasters in different future periods, calculate the risk level respectively, make risk forecasts and classify the early warning levels;

[0042] Step S43: Develop differentiated early warning plans for different early warning levels and risk areas.

[0043] The present invention also provides an electronic device, comprising:

[0044] At least one processor; and

[0045] A memory communicatively connected to at least one of the processors; wherein,

[0046] The memory stores instructions that can be executed by the processor to implement the aforementioned method for early warning of flash floods in important villages based on a small reservoir water and rainfall monitoring system.

[0047] The beneficial effects of this invention are as follows: By adopting a flash flood early warning method for important villages based on a small reservoir water and rainfall monitoring system, multi-timescale risk forecasting and precise early warning classification are achieved, improving the accuracy and timeliness of early warning, thereby realizing precise defense, effectively reducing casualties and property losses, and improving the scientificity and effectiveness of flash flood disaster prevention. Attached Figure Description

[0048] Figure 1 This is a flowchart of the present invention;

[0049] Figure 2 This is a flowchart of step S1 of the present invention;

[0050] Figure 3 This is a flowchart of step S2 of the present invention;

[0051] Figure 4 This is a flowchart of step S3 of the present invention;

[0052] Figure 5 This is a flowchart of step S4 of the present invention. Detailed Implementation

[0053] like Figure 1 As shown, the following technical solution is proposed. According to one aspect of this application, a method for early warning of flash floods in important villages based on small reservoir water and rainfall monitoring is provided, comprising the following steps:

[0054] Step S1: Collect data from the study area, construct a multi-source heterogeneous data fusion system, and build a hydrological-topographic-socioeconomic correlation network for the study area based on graph neural networks;

[0055] Step S2: Construct spatial location relationships and water flow path relationships for small reservoirs and important villages respectively. Based on the topography and building layout of important villages, set several different water level thresholds, generate inundation range maps, assess the risk level of different areas, and delineate risk zones. The important villages are determined based on historical disaster situations.

[0056] Step S3: Based on the hydrological-topographic-socioeconomic correlation network of the study area, construct a reservoir-river-village hydrodynamic coupling model, use time series correlation analysis algorithm to calculate the correlation between small reservoir monitoring indicators and village disaster events, and input the pre-constructed disaster-causing water level inversion model to obtain the probability distribution of disaster-causing water level;

[0057] Step S4: Construct a flash flood disaster early warning model, extract the probability distribution of disaster-causing water levels as model input, conduct multi-timescale risk forecasting and classify early warning levels, and formulate differentiated early warning plans for different early warning levels.

[0058] According to one aspect of this application, such as Figure 2 As shown, step S1 further comprises:

[0059] Step S11: Collect data for the study area, including meteorological data, hydrological data, topographic data, basic information on small reservoirs, historical monitoring data, early warning indicators for important villages, disaster-prone water levels, and risk zone delineation maps;

[0060] Comprehensive data collection is fundamental for subsequent analysis and model building. Only by acquiring rich and accurate data can we gain a deep understanding of the natural and social conditions of the study area and provide sufficient information for flash flood warnings. When collecting topographic data, we use UAV oblique photography and ground LiDAR point cloud data fusion technology to obtain high-precision topographic data. Compared with traditional single data sources, this can more accurately reflect topographic details and provide rich data materials for the construction of a multi-source heterogeneous data fusion system and the establishment of a correlation network. This ensures that subsequent analysis and model building can be based on real and comprehensive data, thereby improving the accuracy and reliability of warnings.

[0061] The core of flash flood disaster risk assessment and related network construction is data-driven. The lack of any type of basic data will lead to the distortion of subsequent models. For example, without meteorological data, it is impossible to analyze the triggering effect of rainfall on reservoir water levels. Without disaster-causing water level data, it is impossible to judge the flood risk of villages. Basic data is the premise and underlying support for subsequent data fusion, model construction and relationship inference.

[0062] Step S12: Use time series interpolation algorithm to synchronize data with different sampling frequencies, and build a unified geographic coordinate system based on GIS spatial indexing technology to obtain a multi-source heterogeneous data fusion system;

[0063] Because different data have different sampling frequencies and geographic coordinate benchmarks, direct use will lead to data chaos and make it impossible to perform effective analysis and processing. Therefore, it is necessary to synchronize and spatially integrate the data.

[0064] In this embodiment, for data with different sampling frequencies, a time series interpolation algorithm is used to synchronize the data with different sampling frequencies and unify them to a minute-level time granularity; a unified geographic coordinate system is established based on GIS spatial indexing technology to convert scattered geographic data into a unified geographic coordinate system, and spatial association of data is established through GIS spatial indexing to realize the fusion of multi-source heterogeneous data in time and space, resulting in a multi-source heterogeneous data fusion system;

[0065] Time series interpolation algorithms can effectively solve the problem of inconsistent data sampling frequencies, ensuring data consistency over time; GIS spatial indexing technology can efficiently organize and manage spatial data, enabling spatial integration of multi-source data and providing a foundation for subsequent correlation analysis.

[0066] Hydrological processes are continuous and dynamic processes over time. If reservoir water level data is on a 10-minute scale and rainfall data is on a daily scale, the two cannot be matched to analyze the "rainfall-water level" response relationship. Interpolation algorithms are the core means to achieve alignment in the time dimension.

[0067] GIS spatial indexing technology not only achieves coordinate unification but also enables rapid retrieval of related data surrounding a specific geographic entity, providing an efficient data retrieval method. This embodiment achieves temporal and spatial unification of multi-source heterogeneous data, allowing different types of data to be correlated and analyzed collaboratively, improving data availability and analytical efficiency. It provides a high-quality data foundation for building interconnected networks and early warning models. When establishing a unified geographic coordinate system, a dynamic coordinate transformation algorithm is introduced, which automatically selects the optimal transformation parameters based on the coordinate transformation requirements of different data sources, improving the accuracy and efficiency of coordinate transformation. Compared to traditional fixed-parameter transformation methods, this significantly enhances the quality of data integration.

[0068] The specific implementation process shown in the example below includes the following:

[0069] Three types of data were collected in a certain research area:

[0070] Reservoir water level monitoring: data collected every 10 minutes, over a period of one month;

[0071] Rainfall data: collected once per hour, over a period of one month;

[0072] Village location data: Local independent coordinate system is used.

[0073] Linear interpolation was used to downsample the 10-minute reservoir water level data to the 1-hour level (taking the average of the 10-minute data within each hour), which is consistent with the time scale of the rainfall data;

[0074] The village location coordinates are converted to the same coordinate system as the terrain DEM. Through GIS grid indexing, the villages along the river downstream of the reservoir can be quickly located, and the spatial hierarchy of geographic entities can be determined.

[0075] Step S13: Use the spatial influence propagation algorithm to simulate the influence path and intensity of reservoir water level changes on downstream villages through the water system network. Construct a reservoir-river-village relationship graph based on graph neural network, where nodes represent geographical entities and edges represent water flow paths and influence weight relationships.

[0076] The impact of flash floods is a process of spatial propagation. Traditional statistical models cannot intuitively express the impact path-intensity relationship between geographical entities, while graph neural networks are good at processing the topological structure of nodes and edges, and can accurately depict the nonlinear relationship between reservoirs, rivers and villages.

[0077] Spatial influence propagation algorithms can simulate the propagation process and impact range of water flow based on actual terrain and water system features, accurately reflecting the interaction between geographic entities; graph neural networks are good at processing graph structure data with complex relationships, and can automatically learn the relationship characteristics between geographic entities to build more accurate relationship models.

[0078] In this embodiment, the spatial relationships and influence between reservoirs, rivers, and villages are clearly presented, providing intuitive model support for in-depth analysis of the occurrence and development patterns of flash floods. This helps to predict the impact of flash floods on villages in advance and provides a basis for early warning and defense decisions. At the same time, when constructing the reservoir-river-village relationship map, the graph neural network is trained with historical flood inundation data, enabling the model to learn the relationship characteristics in the actual propagation process of floods. Compared with traditional relationship models based on experience, this model can more accurately reflect the real relationships between geographical entities and improve the predictive ability of the model.

[0079] The specific implementation process shown in the example below includes the following:

[0080] After a small reservoir releases floodwater, the water flows downstream along the main channel. The speed of the flow is affected by the slope and width of the channel. The main channel has a steep slope and the water flows quickly, reaching village A downstream within 2 hours. However, the tributary has a gentle slope and the water flows slowly, so the impact on village B only becomes apparent after 5 hours.

[0081] Collect node data, including: reservoir (node ​​1), main river (node ​​2), tributary (node ​​3), village A (node ​​4), and village B (node ​​5).

[0082] Node 1 → Node 2 (weight 0.9, the main river channel directly receives the flood discharge), Node 1 → Node 3 (weight 0.2, the tributary only receives a small amount of overflow water), Node 2 → Node 4 (weight 0.8, Village A is adjacent to the main river channel), Node 3 → Node 5 (weight 0.4, Village B is some distance from the tributary).

[0083] The resulting relationship diagram shows that the reservoir has a greater impact on village A than on village B.

[0084] Step S14: Construct a population distribution heat map and a building vulnerability matrix; calculate the comprehensive risk index for different regions based on the analytic hierarchy process; identify the spatial coupling relationship between facilities and flash flood risk; and construct a socio-economic impact model.

[0085] In the risk assessment of flash flood disasters, traditional methods often neglect the role of socio-economic factors in the impact of disasters, resulting in incomplete risk assessment results that cannot provide a comprehensive reference for defense decisions. Therefore, it is necessary to construct a model that takes into account the impact of socio-economic factors.

[0086] Hydrological-topographic correlations can only reflect risk paths at the natural level, while the core impacts of flash floods are on population, buildings, and facilities. Without analysis of the socio-economic dimension, the correlation network cannot serve disaster prevention and mitigation decisions.

[0087] This study constructs a population distribution heat map and a building vulnerability matrix, and calculates a comprehensive risk index for different regions based on the analytic hierarchy process (AHP). It identifies the spatial coupling relationship between key infrastructure such as roads, bridges, and communication base stations and flash flood risk, and constructs a socio-economic impact model. The population distribution heat map and building vulnerability matrix can intuitively reflect the distribution and vulnerability of population and buildings. The AHP can comprehensively consider multiple factors, scientifically determine the weight of each factor, and calculate the comprehensive risk index. By identifying the coupling relationship between key infrastructure and flash flood risk, it can comprehensively assess the impact of socio-economic factors on flash flood disasters, providing support for building a more comprehensive risk assessment system.

[0088] This method comprehensively considers the impact of socio-economic factors on flash flood risk, making the risk assessment results more scientific and accurate, providing a more comprehensive basis for formulating reasonable early warning plans and defense measures, and helping to improve the pertinence and effectiveness of flash flood disaster prevention.

[0089] In this embodiment, building information modeling (BIM) technology is introduced when constructing the building vulnerability matrix. By combining information such as building structure and materials, the vulnerability of each building is accurately assessed. Compared with the traditional coarse assessment method based on building type, it can more accurately reflect the actual vulnerability of buildings in flash floods and improve the accuracy of risk assessment.

[0090] The specific implementation process shown in the example below includes the following:

[0091] A certain study area has two villages, village X and village Y;

[0092] Village X: Population density 800 people / km² 2 The houses are mainly made of adobe (vulnerability score 9), and are located close to the main river (50m away). The terrain has a slope of 5°.

[0093] Village Y: Population density 300 people / km² 2 The houses are mainly brick-concrete structures (vulnerability score 4), located 500m from the main river channel, with a terrain slope of 15°.

[0094] The weights of the indicators were determined by AHP, namely: population density (0.4), building vulnerability (0.3), river distance (0.2), and terrain slope (0.1).

[0095] Calculate the risk index for each of the two villages:

[0096] Village X: 0.4×0.8+0.3×0.9+0.2×0.9 (closer distance, higher score)+0.1×0.2 (gentle slope, lower score)=0.81;

[0097] Village Y: 0.4×0.3+0.3×0.4+0.2×0.3 (far away, low score)+0.1×0.8 (sloping steep, high score)=0.38.

[0098] Spatial coupling analysis revealed that village X contains a primary school with a coupling degree of 0.9 with the high-risk area, requiring priority deployment of early warning equipment and evacuation routes; village Y has a low coupling degree and can be used as a temporary resettlement site.

[0099] Step S15: The hydrological-topographic-socioeconomic relationship network of the study area is constructed by the reservoir-river-village relationship diagram and the socioeconomic impact model.

[0100] Decision-making for flood prevention and mitigation requires systematic correlation analysis, rather than isolated natural or social analysis. For example, knowing only the impact path of a reservoir on a river channel does not determine the scope of personnel evacuation; knowing only the comprehensive risk index of a village does not clarify the source of risk. Only by integrating these into a correlation network can we achieve the integration of risk tracing, path tracking, and loss assessment.

[0101] The multi-dimensional, full-chain disaster correlation system constructed during this implementation process can support scenario simulation and decision-making, providing scientific support for the entire process of monitoring, early warning, response, and recovery of flash flood disasters, as shown in the following example:

[0102] Based on the correlation network of a certain study area, the disaster evolution under a once-in-a-century rainstorm scenario is simulated:

[0103] Heavy rain caused the reservoir water level to exceed the warning level, triggering flood discharge;

[0104] The floodwaters spread along the main river channel, reaching village X after 2 hours, and affecting village Y along the tributary after 5 hours.

[0105] Village X has a comprehensive risk index of 0.81, indicating high population density and fragile buildings, with severe expected losses; Village Y has a comprehensive risk index of 0.38, indicating relatively small losses.

[0106] By adjusting the flood discharge plan through the interconnected network and discharging floodwaters in stages (instead of all at once), the flooding time of village X was extended from 2 hours to 8 hours, buying time for the evacuation of people and ultimately reducing the losses by 60%.

[0107] According to one aspect of this application, such as Figure 3 As shown, step S2 further comprises:

[0108] Step S21: Vectorize the small reservoirs, important villages, rivers, and roads to construct a geographic element database, calculate the straight-line distance and shortest path distance between the small reservoirs and each important village, and construct a spatial location relationship map.

[0109] The raw data of geographical elements are mostly unstructured images or paper maps, which cannot be directly used for spatial relationship analysis; while spatial distance is a basic indicator for judging the impact range of a reservoir on a village.

[0110] Vectorization of geographic elements can transform complex geographic information into a computer-processable format, facilitating storage and analysis; GIS spatial analysis functions can accurately calculate the spatial distances and positional relationships between geographic entities, constructing intuitive spatial location relationship maps.

[0111] In this embodiment, when constructing the spatial location correlation map, three-dimensional visualization technology is introduced to display geographical elements in the form of three-dimensional models. Compared with traditional two-dimensional maps, it can more intuitively present the spatial relationship between small reservoirs and villages. Especially in areas with complex terrain, it can more clearly show the impact of terrain on spatial relationships, providing more intuitive visual support for early warning analysis. The spatial location correlation map between small reservoirs and important villages is accurately constructed, intuitively showing the spatial distance and relative position between them. For example, in a mountainous study area, the straight-line distance between reservoir A and village B is 3km, but there are mountains in between, and the shortest path distance along the river is 8km; the straight-line distance between reservoir A and village C is 5km, and the shortest path distance along the river is 6km. Through the spatial correlation map, it can be seen that reservoir A has a more direct impact on the actual water flow of village C. Subsequent hydrological simulations should prioritize the path from reservoir A to village C.

[0112] Step S22: Construct a distributed hydrological model. Based on the topographic data and historical hydrological data of the study area, simulate the runoff generation and confluence process of small reservoirs under different rainfall conditions and the evolution path of water flow to important villages. Identify key nodes of water flow and potential inundation areas, and construct the water flow path correlation between small reservoirs and important villages.

[0113] Distributed hydrological models can consider the impact of various factors such as topography, soil, and vegetation on runoff generation and confluence, and simulate water flow processes more realistically; particle tracking technology can intuitively track water flow trajectories, accurately identify key nodes and potential inundation areas, and provide support for constructing accurate water flow path correlations.

[0114] When constructing a distributed hydrological model, the model parameters are dynamically adjusted by combining real-time monitoring data. For example, by using real-time rainfall and water level data, the soil moisture content and river roughness parameters in the model are adjusted through adaptive algorithms, so that the model can more accurately reflect the current hydrological conditions. Compared with traditional fixed parameter models, this improves the accuracy and real-time performance of water flow path simulation.

[0115] Step S23: Based on the topography and building layout of important villages, set several different water level thresholds, then use a distributed hydrological model to simulate the inundation range, organize and draw the inundation range data corresponding to all different water level thresholds to generate an inundation range map, assess the risk level of different areas, and delineate risk zones.

[0116] According to one aspect of this application, step S23 further comprises:

[0117] Step S23a: Based on the simulation results of the distributed hydrological model, combined with the topography and building layout of the village, set several different water level thresholds, and then use the distributed hydrological model to simulate and calculate the inundation range of the corresponding important village. Organize and draw the inundation range data corresponding to all different water level thresholds to obtain the inundation range map.

[0118] Different water level thresholds correspond to different degrees of disaster severity. A single water level simulation cannot reflect the gradual process of disaster. By setting thresholds in combination with topography and building layout, it can be ensured that the simulation results are consistent with the actual disaster-bearing situation of the village.

[0119] Based on the village's topography (such as village altitude and ground elevation) and building layout (such as house foundation height and location of core facilities), several gradient water level thresholds are set (such as 1m, 2m, and 3m inundation levels, corresponding to light, moderate, and severe inundation).

[0120] By inputting different water level thresholds into the distributed hydrological model, the inundation range of the village under each threshold is simulated (e.g., the farmland on the edge of the village is inundated when the water level is 1m, and the core residential area of ​​the village is inundated when the water level is 3m).

[0121] By integrating the vector data of the inundation range corresponding to each water level threshold, and overlaying elements such as village buildings and roads on the GIS platform, a thematic map of the inundation range with multiple thresholds is drawn.

[0122] Step S23b: Using inundation depth, inundation duration, population density, and building seismic resistance as indicators, the analytic hierarchy process (AHP) is used to assess the risk level of different regions under flash flood disasters, and to classify them into high-risk, medium-risk, and low-risk areas.

[0123] To more scientifically assess the risk level of flash floods to different areas of important villages, it is necessary to set reasonable water level thresholds based on the actual conditions of the villages and divide them into different risk zones so as to take targeted defense measures. In this embodiment, based on the simulation results of the distributed hydrological model and combined with the topography and building layout of important villages, several different water level thresholds are set to calculate the inundation range of the corresponding important villages. Using inundation depth, inundation duration, population density, and building seismic resistance as indicators, the analytic hierarchy process is used to assess the risk level of different areas under flash floods and divide them into high-risk, medium-risk, and low-risk areas.

[0124] Setting water level thresholds based on the actual conditions of villages can more accurately reflect the impact of flash floods on villages. The analytic hierarchy process (AHP) comprehensively considers multiple factors and can scientifically assess the risk levels of different areas. By scientifically dividing risk zones, it is possible to clearly determine the risk levels of different areas of important villages under flash floods, providing a clear basis for formulating differentiated early warning plans and defense measures. At the same time, when assessing risk levels, the fuzzy comprehensive evaluation method is introduced. Considering the uncertainty and fuzziness of each indicator, the indicators such as inundation depth and inundation duration are fuzzily quantified, making the risk level assessment results more consistent with the actual situation. Compared with the traditional precise value assessment method, this improves the accuracy and reliability of risk assessment.

[0125] According to one aspect of this application, such as Figure 4 As shown, step S3 further comprises:

[0126] Step S31: Construct a reservoir-river-village hydrodynamic coupling model based on the hydrological-topographic-socioeconomic relationship network of the study area;

[0127] Step S32: Use time-series correlation analysis algorithm to calculate the correlation between monitoring indicators of small reservoirs and disaster events in villages;

[0128] Step S33: Construct a physical constraint-based LSTM disaster-causing water level inversion model;

[0129] Step S34: Input the correlation between small reservoir monitoring indicators and village disaster events, along with historical flash flood data, into the disaster-causing water level inversion model to simulate the corresponding disaster-causing water level and calculate the probability distribution of disaster-causing water levels in important villages.

[0130] Step S31 further comprises:

[0131] Step S31a: Based on the hydrological-topographic-socioeconomic relationship network of the study area, construct a digital terrain model of the river channels and important villages;

[0132] Traditional terrain data has low resolution and cannot meet the needs of refined hydrodynamic simulation. Digital terrain models constructed based on the fusion of multi-source terrain data from interconnected networks can improve simulation accuracy. In a certain study area, a digital terrain model constructed by fusing DEM data and village elevation data taken by UAVs can accurately identify "low-lying waterlogged areas" in villages. The simulation results show that waterlogging occurs in this area when the water level is 1.5m, which is consistent with the results of the field survey.

[0133] Step S31b: Collect river data to determine the spatial distribution of river roughness coefficient, and use an adaptive grid partitioning algorithm to grid the important village areas;

[0134] Step S31c: Construct a one-dimensional river channel model and a two-dimensional floodplain model respectively. Using the downstream section of the one-dimensional river channel as the interface, establish a real-time data link. The one-dimensional model automatically transmits the section water level and flow rate to the two-dimensional model as the upstream boundary. The two-dimensional model calculates the floodplain and village inundation situation to obtain the reservoir-river-village hydrodynamic coupling model.

[0135] Digital terrain models can accurately reflect the topographic features of rivers and villages, providing a foundation for hydrodynamic simulation; adaptive mesh generation algorithms can use finer meshes in important areas, improving simulation accuracy; coupling one-dimensional river models and two-dimensional floodplain models can comprehensively consider the characteristics of river flow and floodplain flow, more realistically simulating the flood evolution process.

[0136] The construction of the reservoir-river-village hydrodynamic coupling model requires first building a one-dimensional river channel model and a two-dimensional floodplain model, and then achieving coupling through dynamic data interaction: The one-dimensional river channel model starts from the reservoir outlet and ends in the upstream floodplain area of ​​the village. It divides the data into one-dimensional computational units and inputs parameters such as river cross-section and roughness. Based on the Saint-Venant equations, it simulates the longitudinal water level and flow rate changes, using the upstream reservoir discharge flow rate as the boundary. After calibration with historical data, the downstream coupling cross-section is determined. The two-dimensional floodplain model covers the floodplain and village areas, using an unstructured grid (densified in the village area and simplified in open areas). It inputs topographic elevation and surface data. The roughness factor is simulated by a two-dimensional shallow water equation set to simulate the planar diffusion and inundation process of water flow. After calibration with historical inundation traces, an upstream coupling interface is reserved. In the coupling stage, a real-time data link is established with the downstream section of the one-dimensional river channel as the interface. After each step of calculation by the one-dimensional model, the water level and flow rate of the section are automatically transferred to the two-dimensional model as the upstream boundary. The two-dimensional model continuously calculates the floodplain and village inundation based on this. Finally, the typical discharge flow of the reservoir is input to run the global model. The sub-model parameters are adjusted to ensure that the simulation of water flow from the reservoir through the river channel, floodplain and to the village is continuous and without discontinuity, forming a complete "reservoir-river-village" hydrodynamic coupling model.

[0137] The adaptive mesh generation algorithm incorporates artificial intelligence algorithms to automatically learn the flood characteristics of different regions based on historical flood data and real-time monitoring data. It dynamically adjusts the density and accuracy of mesh generation and automatically densifies the mesh in areas prone to flooding. Compared with the traditional fixed mesh generation method, this greatly improves the simulation accuracy and computational efficiency of the model.

[0138] Step S32 further comprises:

[0139] Step S32a: Use the phase space reconstruction method to map the time series of rainfall, reservoir water level, and river flow to a high-dimensional space;

[0140] Step S32b: Under time window constraints, mine the association rules between monitoring data and disaster-causing events;

[0141] By setting a time window (such as 3-6 hours after a sudden increase in reservoir water level), we can explore the correlation rules between monitoring data and disaster-causing events in a high-dimensional phase space.

[0142] Step S32c: Construct a structural equation model, use Granger causality test to determine the causal relationship between reservoir monitoring indicators and village disaster water level, quantify the influence path of each factor on disaster water level, and obtain the correlation between small reservoir monitoring indicators and village disaster events.

[0143] There is a complex nonlinear relationship between monitoring indicators of small reservoirs and disaster events in villages. Traditional methods are difficult to accurately uncover these relationships. Phase space reconstruction can unfold time series data in a high-dimensional space, revealing the inherent structure and patterns of the data. The improved Apriori algorithm can efficiently uncover the association rules between data. Structural equation modeling and Granger causality test can accurately determine the causal relationships and influence paths between variables, providing support for in-depth analysis of the relationship between monitoring indicators and disaster events.

[0144] Accurately identifying the correlation and causal relationship between monitoring indicators of small reservoirs and disaster events in villages provides important input for disaster-causing water level inversion models, which helps improve the accuracy of disaster-causing water level prediction and thus enhances the reliability of flash flood warnings.

[0145] The specific implementation process shown in the example below includes the following:

[0146] The structural equation model of a certain study area shows that the direct influence coefficient of rainfall on the disaster-causing water level is 0.3, the indirect influence coefficient (through reservoir water level) is 0.5, and the total influence coefficient is 0.8. The Granger causality test confirms that "the reservoir water level is a Granger cause of the disaster-causing water level (P<0.01)," indicating that the reservoir water level is a key driving factor for the disaster.

[0147] According to one aspect of this application, step S34 further comprises:

[0148] Step S34a: Input the correlation between small reservoir monitoring indicators and village disaster events, along with historical flash flood data, into the disaster-causing water level inversion model;

[0149] Step S34b: Construct a 16-grid risk assessment matrix with inundation depth and inundation duration as the horizontal and vertical axes, and determine the risk level corresponding to each cell by combining expert knowledge and historical disaster data.

[0150] Step S34c: Using the Kriging interpolation algorithm, the risk assessment results of discrete points are extended into a continuous risk distribution map, and the probability distribution of disaster-prone water levels in important villages is calculated.

[0151] The disaster-causing water level inversion model can simulate disaster-causing water levels based on correlations and historical data; the risk assessment matrix combines multiple factors to comprehensively assess risk levels, making risk classification more scientific; the Kriging interpolation algorithm can reasonably infer the risk situation of continuous areas based on discrete point data, thereby obtaining the probability distribution of disaster-causing water levels and comprehensively reflecting the uncertainty of risk; the core of the Kriging interpolation is to quantify the spatial correlation of discrete points through a semi-variogram function, then construct a weight matrix based on this, and finally generate a continuous risk surface through weighted summation. Specifically:

[0152] The first step is to calculate the semivariogram: ;in Z(x) represents the semivariogram with a lag distance of h, where h is the spatial distance between two points, N(h) is the number of discrete point pairs with a distance of h, and Z(x) is the semivariogram with a lag distance of h. i (x) represents a discrete point. i The risk assessment value at the location, Z(x) i +h) is related to x i Risk assessment value of a point at a distance h;

[0153] The second step is to solve the Kriging equations to obtain the optimal weights;

[0154] The third step involves using the optimal weights to weight and sum the risk values ​​of the discrete points to obtain the risk assessment value of the interpolation points: ;in, The weight of the i-th discrete point; The risk assessment value for interpolation point x0;

[0155] The fourth step, based on the continuous risk surface obtained through interpolation, calculates the probability distribution by combining the correspondence between "risk value - disaster-causing water level". First, the disaster-causing water level threshold for the village is determined, and the grid proportion of the disaster-causing water level corresponding to the interpolated risk value in the area where the village is located is statistically analyzed, which is the preliminary probability. Finally, the preliminary probability is corrected by combining the variance estimated by Kriging to obtain the final probability distribution.

[0156] ; The cumulative distribution function of the standard normal distribution; This represents the risk value corresponding to the disaster-causing water level threshold. Estimate the standard deviation of the Kriging for the i-th grid point;

[0157] By calculating the probability distribution of disaster-causing water levels, we can more comprehensively and accurately assess the risk of flash floods and provide more detailed information for risk forecasting and early warning classification. In this embodiment, a dynamic weight adjustment mechanism is introduced when constructing the risk assessment matrix. Based on different seasons and different prior soil moisture conditions, the weights of factors such as inundation depth and inundation duration in the risk assessment are automatically adjusted. During the rainy season, the weight of inundation duration is increased, making the risk assessment more in line with the actual situation.

[0158] According to one aspect of this application, such as Figure 5 As shown, step S4 further comprises:

[0159] Step S41: Couple the hydrodynamic coupling model, the disaster-causing water level inversion model, and the risk zoning model to obtain the flash flood disaster early warning model;

[0160] A microservice architecture is adopted to reduce model coupling, encapsulating each model as an independent microservice, and enabling data interaction and communication through API interfaces.

[0161] Step S42: Extract the probability distribution of disaster-causing water levels and input it into the flash flood disaster early warning model to simulate the development trend of flash flood disasters in different future periods, calculate the risk level respectively, make risk forecasts and classify the early warning levels;

[0162] The probability distribution of disaster-causing water levels is extracted as input to the flash flood disaster early warning model, and risk forecasts are made for 0-3 hours, 3-12 hours, and 12-24 hours respectively. Based on the disaster-causing water level prediction, inundation range, risk probability and other indicators in the risk forecast results, the early warning level is divided according to the pre-set early warning classification standards. Multi-timescale risk forecasts can meet the early warning needs at different stages. Short-term forecasts can capture sudden changes, while short-term and medium-term forecasts can provide longer-term risk trends. Clear early warning classification standards help to unify the release of early warning information, enabling residents to take corresponding measures according to different levels.

[0163] Step S43: Develop differentiated early warning plans for different early warning levels and risk areas.

[0164] A tiered early warning system is established, with blue indicating low risk, yellow indicating medium risk, orange indicating high risk, and red indicating extremely high risk. The system automatically determines the warning level based on the predicted flood level, inundation area, and risk probability indicators from the risk forecast. Differentiated warning plans are developed for different warning levels and risk areas. For blue warnings, information dissemination is the primary method, reminding residents to pay attention to weather and water conditions via SMS and app push notifications. For yellow warnings, additional defensive measures are suggested, and residents in low-lying areas are organized to prepare for evacuation. For orange warnings, some emergency response mechanisms are activated, and temporary shelters are opened. For red warnings, a full emergency response is implemented, mandating the evacuation of residents along predetermined routes and deploying rescue forces.

[0165] At the same time, a multi-channel dissemination system including SMS, radio, social media, and emergency early warning platform is constructed. Based on the information receiving habits and equipment possession of residents in important villages, the optimal combination of dissemination channels is intelligently selected. For special groups, door-to-door notification, telephone contact and other methods are used to ensure that early warning information can be accurately and efficiently conveyed to every affected resident.

[0166] This application proposes a spatial association construction method of "vector modeling + dual distance constraints", which performs refined vectorization processing on geographical elements such as small reservoirs, villages, rivers, and roads, and simultaneously calculates the dual-dimensional distance index of "straight line distance (geographical proximity) - shortest path distance (actual propagation path)" to construct a spatial location association map containing distance weights;

[0167] Next, a two-level simulation system of "distributed hydrological model + one-dimensional river channel-two-dimensional floodplain coupled hydrodynamic model" is constructed. First, the reservoir runoff generation and confluence process under different rainfall scenarios is simulated through the distributed hydrological model to identify key nodes of water flow and potential inundation areas. Then, based on the high-precision digital terrain model, the "adaptive grid partitioning algorithm" is used to differentiate the village area into grids to achieve a refined simulation of the entire chain of "river flow evolution - village floodplain inundation".

[0168] Then, the problem of "lack of physical rationality" in the data-driven model was solved by "physical constraint embedding", and the problem of "single dimension and strong subjectivity" in traditional risk assessment was solved by "multi-indicator hierarchical analysis". Furthermore, by combining Kriging interpolation and 16-grid risk matrix, the discrete point disaster data was transformed into a continuous risk distribution map, achieving a probabilistic risk assessment that integrates "point-area".

[0169] Next, a progressive correlation mining method of "phase space reconstruction - time window constraint - structural equation model + Granger causality test" is proposed. First, the monitoring time series is mapped to a high-dimensional space, then the correlation rules are mined within the time window, and finally the direct / indirect impact path is analyzed by structural equation model. Combined with Granger causality test, "correlation relationship - causality relationship" is distinguished, and the impact intensity of each monitoring indicator (rainfall, reservoir water level, river flow) on the disaster-causing water level is accurately quantified.

[0170] A fully coupled early warning model consisting of a hydrodynamic coupling model, a disaster-causing water level inversion model, and a risk zoning model is constructed to achieve automated and integrated operation of real-time monitoring data input, disaster-causing water level prediction, risk level assessment, and early warning information output. Based on the probability distribution of disaster-causing water levels and simulation of future multi-period situations, a gradient early warning level is divided, and a two-dimensional differentiated early warning plan of "early warning level - risk area" is formulated in combination with the differences in risk areas (high / medium / low risk areas), clarifying response measures, responsible entities, and time nodes.

[0171] Finally, a geographic feature database construction method based on multi-source data integration, coordinate unification, and attribute expansion was adopted. This method integrates heterogeneous data from multiple sources, including remote sensing imagery, paper maps, DEMs, historical hydrological data, and village architecture data. By unifying the geographic coordinate system through GIS spatial indexing technology, the structured management and efficient retrieval of data are achieved. Through the fusion of heterogeneous multi-source data and the construction of spatial indexes, the "spatiotemporal alignment, attribute completeness, and efficient retrieval" of geographic feature data are realized, providing high-quality data support for full-chain simulation and analysis. This method solves the technical bottleneck of "weak data foundation and ineffective collaboration of multi-source data" in traditional flash flood early warning systems.

[0172] According to another aspect of this application, an electronic device is provided, comprising:

[0173] At least one processor; and

[0174] A memory communicatively connected to at least one of the processors; wherein,

[0175] The memory stores instructions that can be executed by the processor to implement the aforementioned method for early warning of flash floods in important villages based on monitoring water and rainfall conditions in small reservoirs.

[0176] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the specific details of the above embodiments. Within the scope of the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all fall within the protection scope of the present invention.

Claims

1. A mountain flood defense important village early warning method based on small reservoir water regime monitoring, characterized in that, Includes the following steps: Step S1: Collect data from the study area, construct a multi-source heterogeneous data fusion system, and build a hydrological-topographic-socioeconomic correlation network for the study area based on graph neural networks; specifically: Step S11: Collect data for the study area, including meteorological data, hydrological data, topographic data, basic information on small reservoirs, historical monitoring data, early warning indicators for important villages, disaster-prone water levels, and risk zone delineation maps; Step S12: Use time series interpolation algorithm to synchronize data with different sampling frequencies, and build a unified geographic coordinate system based on GIS spatial indexing technology to obtain a multi-source heterogeneous data fusion system; Step S13: Use the spatial influence propagation algorithm to simulate the influence path and intensity of reservoir water level changes on downstream villages through the water system network. Construct a reservoir-river-village relationship graph based on graph neural network, where nodes represent geographical entities and edges represent water flow paths and influence weight relationships. Step S14: Construct a population distribution heat map and a building vulnerability matrix; calculate the comprehensive risk index for different regions based on the analytic hierarchy process; identify the spatial coupling relationship between facilities and flash flood risk; and construct a socio-economic impact model. Step S15: The reservoir-river-village relationship diagram and the socio-economic impact model together constitute the hydrological-topographic-socio-economic relationship network of the study area; Step S2: Construct spatial location relationships and water flow path relationships for small reservoirs and important villages respectively. Based on the topography and building layout of important villages, set several different water level thresholds, generate inundation range maps, assess the risk level of different areas, and delineate risk zones. Step S3: Based on the hydrological-topographic-socioeconomic correlation network of the study area, construct a reservoir-river-village hydrodynamic coupling model, use time series correlation analysis algorithm to calculate the correlation between small reservoir monitoring indicators and village disaster events, and input the pre-constructed disaster-causing water level inversion model to obtain the probability distribution of disaster-causing water level. Step S4: Construct a flash flood disaster early warning model, extract the probability distribution of disaster-causing water levels as model input, conduct multi-timescale risk forecasting and classify early warning levels, and formulate differentiated early warning plans for different early warning levels.

2. The method for early warning of flash floods in important villages based on small reservoir water and rainfall monitoring as described in claim 1, characterized in that, Step S2 further comprises: Step S21: Vectorize the small reservoirs, important villages, rivers, and roads to construct a geographic element database, calculate the straight-line distance and shortest path distance between the small reservoirs and each important village, and construct a spatial location relationship map. Step S22: Construct a distributed hydrological model. Based on the topographic data and historical hydrological data of the study area, simulate the runoff generation and confluence process of small reservoirs under different rainfall conditions and the evolution path of water flow to important villages. Identify key nodes of water flow and potential inundation areas, and construct the water flow path correlation between small reservoirs and important villages. Step S23: Based on the topography and building layout of important villages, set several different water level thresholds, then use a distributed hydrological model to simulate and generate an inundation range map, assess the risk level of different areas, and delineate risk zones.

3. The method for early warning of flash floods in important villages based on small reservoir water and rainfall monitoring as described in claim 2, characterized in that, Step S23 further comprises: Step S23a: Based on the simulation results of the distributed hydrological model, combined with the topography and building layout of important villages, set several different water level thresholds, and then use the distributed hydrological model to simulate and calculate the inundation range of the corresponding important villages. Organize and draw the inundation range data corresponding to all different water level thresholds to obtain an inundation range map. Step S23b: Using inundation depth, inundation duration, population density, and building seismic resistance as indicators, the analytic hierarchy process (AHP) is used to assess the risk level of different regions under flash flood disasters, and to classify them into high-risk, medium-risk, and low-risk areas.

4. The method for early warning of flash floods in important villages based on small reservoir water and rainfall monitoring as described in claim 1, characterized in that, Step S3 further comprises: Step S31: Construct a reservoir-river-village hydrodynamic coupling model based on the hydrological-topographic-socioeconomic relationship network of the study area; Step S32: Use time-series correlation analysis algorithm to calculate the correlation between monitoring indicators of small reservoirs and disaster events in villages; Step S33: Construct a disaster-causing water level inversion model based on physical constraints using a long short-term memory network; Step S34: Input the correlation between small reservoir monitoring indicators and village disaster events, along with historical flash flood data, into the disaster-causing water level inversion model to simulate the corresponding disaster-causing water level and calculate the probability distribution of disaster-causing water levels in important villages.

5. The method for early warning of flash floods in important villages based on small reservoir water and rainfall monitoring as described in claim 4, characterized in that, Step S31 further comprises: Step S31a: Based on the hydrological-topographic-socioeconomic relationship network of the study area, construct a digital terrain model of the river channels and important villages; Step S31b: Collect river data to determine the spatial distribution of river roughness coefficient, and use an adaptive grid partitioning algorithm to grid the important village areas; Step S31c: Construct a one-dimensional river channel model and a two-dimensional floodplain model respectively. Using the downstream section of the one-dimensional river channel as the interface, establish a real-time data link. The one-dimensional model automatically transmits the section water level and flow rate to the two-dimensional model as the upstream boundary. The two-dimensional model calculates the floodplain and village inundation situation to obtain the reservoir-river-village hydrodynamic coupling model.

6. The method for early warning of flash floods in important villages based on small reservoir water and rainfall monitoring as described in claim 4, characterized in that, Step S32 further comprises: Step S32a: Use the phase space reconstruction method to map the time series of rainfall, reservoir water level, and river flow to a high-dimensional space; Step S32b: Under time window constraints, mine the association rules between monitoring data and disaster-causing events; Step S32c: Construct a structural equation model, use Granger causality test to determine the causal relationship between reservoir monitoring indicators and village disaster water level, quantify the influence path of each factor on disaster water level, and obtain the correlation between small reservoir monitoring indicators and village disaster events.

7. The method for early warning of flash floods in important villages based on small reservoir water and rainfall monitoring as described in claim 4, characterized in that, Step S34 further comprises: Step S34a: Input the correlation between small reservoir monitoring indicators and village disaster events, along with historical flash flood data, into the disaster-causing water level inversion model; Step S34b: Construct a 16-grid risk assessment matrix with inundation depth and inundation duration as the horizontal and vertical axes, and determine the risk level corresponding to each cell by combining expert knowledge and historical disaster data. Step S34c: Using the Kriging interpolation algorithm, the risk assessment results of discrete points are extended into a continuous risk distribution map, and the probability distribution of disaster-prone water levels in important villages is calculated.

8. The method for early warning of flash floods in important villages based on small reservoir water and rainfall monitoring as described in claim 1, characterized in that, Step S4 further comprises: Step S41: Couple the hydrodynamic coupling model, the disaster-causing water level inversion model, and the risk zoning model to obtain the flash flood disaster early warning model; Step S42: Extract the probability distribution of disaster-causing water levels and input it into the flash flood disaster early warning model to simulate the development trend of flash flood disasters in different future periods, calculate the risk level respectively, make risk forecasts and classify the early warning levels; Step S43: Develop differentiated early warning plans for different early warning levels and risk areas.

9. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to at least one of the processors; wherein, The memory stores instructions that can be executed by the processor to implement the method for early warning of flash floods in important villages based on monitoring water and rainfall conditions in small reservoirs, as described in any one of claims 1-8.

Images