Rainfall-based landslide geological disaster early warning system and method

Abstract

The invention relates to the technical field of disaster early warning, in particular to a rainfall-based landslide geological disaster early warning system and method, and the method comprises the steps: laying a plurality of rainfall monitoring points in a target region based on historical landslide disaster data and historical rainfall data; determining a real-time rainfall risk distribution diagram of the target area through the real-time rainfall early warning threshold value based on a spatial interpolation method; performing overlay analysis on the real-time landslide high-risk cluster and the real-time rainfall risk distribution map to obtain a real-time landslide susceptibility index of each sub-region in the target region; based on the geological parameters of the target area, correcting the susceptibility labels corresponding to the sub-areas in the target area to obtain a landslide susceptibility map; and determining landslide geological disaster early warning information based on the landslide susceptibility map. According to the method, the landslide susceptibility label is corrected through the geological parameters of the target area, the accuracy and reliability of the prediction result can be further improved, and the influence that the geological environment is neglected due to excessive dependence on rainfall data is avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of disaster early warning, and in particular to a landslide geological disaster early warning system and method based on rainfall. Background Art

[0002] Traditional warning systems often rely on historical data and static warning models, lacking the integration of real-time rainfall and disaster data. This lack of real-time data updates prevents traditional warning systems from promptly reflecting sudden rainfall events or new landslide disaster risks, resulting in delayed warning responses and difficulty in taking effective preventive measures. Traditional landslide warning systems often use fixed rainfall thresholds as warning criteria, but factors such as geological conditions, precipitation intensity, and landslide history can vary significantly across regions. Fixed thresholds can cause warning systems to be overly conservative in some cases, missing critical warning times, or issuing warnings too early in other cases, resulting in wasted resources. Traditional systems may also rely solely on precipitation data to predict landslide risk, ignoring the impact of geological characteristics (such as soil type, slope, and vegetation cover) on landslide susceptibility. Changes in the geological environment have a significant impact on the occurrence of landslide hazards, and relying solely on precipitation data can lead to inaccurate predictions, especially in complex geological environments. Traditional systems often use a global risk assessment and lack a detailed breakdown of different subregions. Due to differences in geographical conditions and precipitation, landslide risks can vary significantly across regions. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to overcome the shortcomings of the above-mentioned prior art and provide a landslide geological disaster early warning system and method based on rainfall.

[0004] The technical solution adopted to solve the above technical problems is: a landslide geological disaster early warning system based on rainfall, including:

[0005] A rainfall monitoring unit, the rainfall monitoring unit being configured to deploy a plurality of rainfall monitoring points within a target area based on historical landslide disaster data and historical precipitation data; selecting a plurality of historical landslide disaster points based on the historical landslide disaster data; each of the historical landslide disaster points corresponding to preset geographic coordinates and geological parameters;

[0006] A rainfall collection unit, the rainfall collection unit being configured to collect real-time rainfall data based on the rainfall monitoring points; and determining a real-time rainfall warning threshold corresponding to each of the historical landslide disaster points based on the real-time rainfall data and the geological parameters of each of the historical landslide disaster points based on a dynamic threshold model;

[0007] a risk warning unit, the risk warning unit being configured to determine a real-time rainfall risk distribution map of the target area using the real-time rainfall warning threshold based on a spatial interpolation method; and to determine a real-time high-risk landslide cluster in the target area using the real-time rainfall risk distribution map and the geographic coordinates of the historical landslide disaster points based on a clustering algorithm;

[0008] a landslide analysis unit, configured to perform a superposition analysis on the real-time high-risk landslide clusters and the real-time rainfall risk distribution map to obtain a real-time landslide susceptibility index for each sub-region in the target region; and to perform landslide susceptibility classification on each sub-region in the target region based on the real-time landslide susceptibility index to obtain a susceptibility label corresponding to each sub-region in the target region;

[0009] A region division unit is used to modify the susceptibility labels corresponding to each sub-region in the target region based on the geological parameters of the target region to obtain a landslide susceptibility map; and determine landslide geological disaster warning information based on the landslide susceptibility map.

[0010] Preferably, the rainfall warning threshold includes an hourly rainfall intensity threshold, a cumulative 24-hour rainfall threshold and a cumulative 72-hour rainfall threshold; the susceptibility labels include extremely low susceptibility areas, low susceptibility areas, moderate susceptibility areas, high susceptibility areas and extremely high susceptibility areas; the geological parameters include terrain slope, soil type and land use type.

[0011] Preferably, several rainfall monitoring points are deployed in the target area based on historical landslide hazard data and historical precipitation data, including:

[0012] Acquire a digital elevation model of the target area; calculate a terrain humidity index based on the digital elevation model and the historical precipitation data to obtain a historical regional humidity distribution map corresponding to the target area;

[0013] Performing kernel density estimation on the historical landslide disaster data to obtain a historical landslide disaster density cloud map corresponding to the target area;

[0014] Performing weighted superposition on the historical regional moisture distribution map and the historical landslide disaster density cloud map to obtain a historical landslide sensitivity comprehensive index map;

[0015] Based on the historical landslide sensitivity comprehensive index map, each sub-region in the target area is divided into landslide sensitivity categories to obtain sensitivity labels for each sub-region; wherein the sensitivity labels include high-sensitivity areas, medium-sensitivity areas, and low-sensitivity areas;

[0016] Based on the sensitivity labels of each sub-area, a monitoring point density gradient planning is performed, wherein rainfall monitoring points are arranged at a first spacing in the highly sensitive area, at a second spacing in the medium-sensitive area, and at a third spacing in the low-sensitivity area, so as to obtain a rainfall monitoring point distribution map; wherein the first spacing < the second spacing < the third spacing.

[0017] Preferably, the process of constructing the dynamic threshold model includes:

[0018] Collect rainfall data before historical landslide events and extract the hourly rainfall, cumulative 24-hour rainfall, and cumulative 72-hour rainfall corresponding to each landslide event;

[0019] Based on the geological data of each landslide event, the rainfall data is divided into a plurality of geological condition groups; wherein the geological data includes rock and soil permeability coefficient, slope and vegetation coverage;

[0020] Performing statistical analysis on the rainfall data within each geological condition group to determine the hourly rainfall threshold, the cumulative 24-hour rainfall threshold, and the cumulative 72-hour rainfall threshold corresponding to each geological condition group;

[0021] A nonlinear mapping relationship between the geological data and the rainfall threshold is established based on a support vector machine algorithm to obtain the dynamic threshold model.

[0022] Preferably, determining the real-time rainfall risk distribution map of the target area by using the real-time rainfall warning threshold based on a spatial interpolation method includes:

[0023] Preprocessing the real-time rainfall data collected by the rainfall monitoring point, wherein the preprocessing includes outlier removal and data smoothing;

[0024] Based on the Kriging interpolation algorithm, combined with terrain elevation data and land use type data, the preprocessed rainfall data is spatially interpolated to obtain a real-time rainfall risk distribution map of the target area.

[0025] Preferably, determining the real-time high-risk landslide clusters in the target area based on the real-time rainfall risk distribution map and the geographical coordinates of the historical landslide disaster points based on a clustering algorithm includes:

[0026] Extracting the geographical coordinates of the historical landslide disaster point and the rainfall risk value of the corresponding position in the real-time rainfall risk distribution map;

[0027] Based on the density peak clustering algorithm, the landslide hazard points are clustered and analyzed with the rainfall risk value as a weight to obtain a number of initial clusters;

[0028] Boundary optimization is performed on each of the initial clusters to obtain real-time high-risk landslide clusters in the target area; wherein the boundary optimization includes cluster segmentation and merging based on terrain fault data.

[0029] Preferably, the calculation formula of the landslide susceptibility index is as follows:

[0030] ;

[0031] in, represents the landslide susceptibility index, Indicates the The weight coefficient corresponding to the rainfall warning threshold is: Indicates the The number of times the rainfall warning threshold exceeds the limit, represents the terrain slope correction coefficient, Indicates the stability coefficient corresponding to the soil type.

[0032] Preferably, the susceptibility labels corresponding to the sub-regions in the target region are modified based on the geological parameters of the target region to obtain a landslide susceptibility map, including:

[0033] Acquiring geographic information of the target area and performing standardization processing on the geographic information; wherein the geographic information includes terrain slope, soil type, and land use type;

[0034] Determine the susceptibility label corresponding to the geographic information based on the hierarchical analysis method;

[0035] Based on the weighted superposition method, the standardized geographic information and the landslide susceptibility index are fused to obtain a landslide susceptibility map.

[0036] Preferably, determining landslide geological disaster warning information based on the landslide susceptibility map includes:

[0037] updating the real-time rainfall risk distribution map based on the real-time rainfall data, and extracting the rainfall risk value of each sub-region in the updated rainfall risk distribution map;

[0038] performing coupling analysis on the rainfall risk value and the landslide susceptibility index of the corresponding sub-region in the landslide susceptibility map to obtain a comprehensive landslide risk index;

[0039] Based on the comprehensive landslide risk index, the target area is divided into five warning levels through the natural break method to obtain landslide geological disaster warning information.

[0040] The technical solution adopted to solve the above technical problems is: a landslide geological disaster early warning method based on rainfall, which is applicable to the above-mentioned landslide geological disaster early warning system based on rainfall, comprising:

[0041] Deploying a number of rainfall monitoring points in the target area based on historical landslide disaster data and historical precipitation data; selecting a number of historical landslide disaster points based on the historical landslide disaster data; each of the historical landslide disaster points corresponds to preset geographic coordinates and geological parameters;

[0042] Collecting real-time rainfall data based on the rainfall monitoring points; determining the real-time rainfall warning threshold corresponding to each of the historical landslide disaster points through the real-time rainfall data and the geological parameters of each of the historical landslide disaster points based on a dynamic threshold model;

[0043] Determine the real-time rainfall risk distribution map of the target area through the real-time rainfall warning threshold based on the spatial interpolation method; determine the real-time landslide high-risk clusters of the target area through the real-time rainfall risk distribution map and the geographical coordinates of the historical landslide disaster points based on the clustering algorithm;

[0044] Overlaying and analyzing the real-time landslide high-risk clusters and the real-time rainfall risk distribution map to obtain a real-time landslide susceptibility index for each sub-region in the target area; dividing each sub-region in the target area by landslide susceptibility based on the real-time landslide susceptibility index to obtain a susceptibility label corresponding to each sub-region in the target area;

[0045] Based on the geological parameters of the target area, the susceptibility labels corresponding to the sub-areas in the target area are modified to obtain a landslide susceptibility map; and landslide geological disaster warning information is determined based on the landslide susceptibility map.

[0046] The beneficial effects of the present invention are as follows:

[0047] (1) The present invention generates a real-time rainfall risk distribution map and a landslide susceptibility map based on real-time rainfall data, historical precipitation data, and landslide disaster data. This real-time data update can provide disaster management personnel with immediate landslide risk warnings, so that preventive measures can be taken before landslide disasters occur, thereby improving the timeliness and accuracy of warnings. Moreover, by deploying rainfall monitoring points and combining dynamic threshold models and clustering algorithms, high-risk landslide clusters in the target area can be accurately determined. This allows key resources to be concentrated in landslide-prone areas, optimizing resource allocation for disaster prevention and control.

[0048] (2) The present invention uses a dynamic threshold model, combined with historical rainfall data and geological parameters of landslide hazard sites, to more accurately set the warning threshold for each monitoring point, avoiding warning errors or lags that may be caused by fixed thresholds, thereby improving the sensitivity and accuracy of the system. In addition, the real-time landslide susceptibility index is used to divide each sub-region into landslide susceptibility categories and generate susceptibility labels, which can help governments or management agencies to scientifically manage the risk level of each sub-region and formulate targeted preventive measures.

[0049] (3) The present invention can further improve the accuracy and reliability of prediction results by correcting the landslide susceptibility label through the geological parameters of the target area, especially in areas with complex geological characteristics, avoiding over-reliance on precipitation data and ignoring the impact of the geological environment; and the landslide susceptibility map and early warning information generated by the system can provide a scientific basis for relevant departments, helping decision makers make more effective disaster prevention, emergency response and resource scheduling decisions, thereby reducing the threat of landslide disasters to people's lives and property. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 A schematic diagram of the system architecture of the overall system in an embodiment of the present invention;

[0051] Figure 2 The figure is a flowchart of the steps of the overall method in one embodiment of the present invention. DETAILED DESCRIPTION

[0052] Example 1, as Figure 1 As shown, the present invention proposes a landslide geological disaster early warning system based on rainfall, comprising:

[0053] A rainfall monitoring unit is used to deploy a number of rainfall monitoring points in the target area based on historical landslide disaster data and historical precipitation data; a number of historical landslide disaster points are selected based on the historical landslide disaster data; each historical landslide disaster point corresponds to preset geographical coordinates and geological parameters;

[0054] A rainfall collection unit is used to collect real-time rainfall data based on rainfall monitoring points; based on a dynamic threshold model, the real-time rainfall warning threshold corresponding to each historical landslide disaster point is determined by using the real-time rainfall data and the geological parameters of each historical landslide disaster point;

[0055] The risk warning unit is used to determine the real-time rainfall risk distribution map of the target area based on the spatial interpolation method and the real-time rainfall warning threshold; and to determine the real-time landslide high-risk clusters in the target area based on the real-time rainfall risk distribution map and the geographical coordinates of historical landslide disaster points based on the clustering algorithm;

[0056] The landslide analysis unit is used to overlay and analyze the real-time high-risk landslide clusters and the real-time rainfall risk distribution map to obtain the real-time landslide susceptibility index of each sub-region in the target area; based on the real-time landslide susceptibility index, the sub-regions in the target area are divided into landslide susceptibility categories to obtain susceptibility labels corresponding to each sub-region in the target area;

[0057] The regional division unit is used to modify the susceptibility labels corresponding to each sub-region in the target area based on the geological parameters of the target area to obtain a landslide susceptibility map; and determine landslide geological disaster warning information based on the landslide susceptibility map.

[0058] In the present invention, rainfall monitoring points refer to equipment points set up in the target area for real-time monitoring of precipitation; these monitoring points can collect real-time precipitation data in the area, thereby providing data support for subsequent landslide warnings; historical landslide disaster data refers to relevant data on landslide disaster events that have occurred in the past, including information such as the location, time, scale, and type of the disaster. Through this data, the patterns of landslide occurrence and dangerous areas can be identified; historical precipitation data refers to the precipitation data of a certain area in the past; precipitation is one of the important factors affecting the occurrence of landslides, so historical precipitation data can be used to analyze the relationship between precipitation and landslides.

[0059] In an optional embodiment, the rainfall warning threshold includes an hourly rainfall intensity threshold, a cumulative 24-hour rainfall threshold, and a cumulative 72-hour rainfall threshold; the susceptibility labels include extremely low susceptibility areas, low susceptibility areas, medium susceptibility areas, high susceptibility areas, and extremely high susceptibility areas; and the geological parameters include terrain slope, soil type, and land use type.

[0060] In an optional embodiment, a number of rainfall monitoring points are deployed in the target area based on historical landslide hazard data and historical precipitation data, including:

[0061] Obtain a digital elevation model of the target area; calculate the terrain humidity index based on the digital elevation model and historical precipitation data to obtain a historical regional humidity distribution map corresponding to the target area;

[0062] Perform kernel density estimation on historical landslide hazard data to obtain the historical landslide hazard density cloud map corresponding to the target area;

[0063] The historical regional moisture distribution map and the historical landslide hazard density cloud map are weighted and superimposed to obtain the historical landslide sensitivity comprehensive index map;

[0064] Based on the historical landslide susceptibility comprehensive index map, the landslide sensitivity of each sub-region in the target area is divided to obtain the sensitivity label of each sub-region; the sensitivity label includes high sensitivity area, medium sensitivity area and low sensitivity area;

[0065] Based on the sensitivity labels of each sub-area, the density gradient planning of monitoring points is carried out. Rainfall monitoring points are arranged at a first spacing in the highly sensitive area, at a second spacing in the medium-sensitive area, and at a third spacing in the low-sensitivity area, so as to obtain a rainfall monitoring point distribution map; wherein, the first spacing < the second spacing < the third spacing.

[0066] It should be noted that the Digital Elevation Model (DEM) is a digital representation of the earth's surface (or the terrain of a certain area) obtained through remote sensing technology, LiDAR or other measurement methods; it is usually stored in a grid format and provides altitude information within the area; by obtaining DEM data of the target area, the terrain undulations of the area can be understood, and then the impact on disasters such as landslides can be analyzed; the Terrain Wetness Index (TWI) is calculated using the Digital Elevation Model and precipitation data, and is intended to assess the accumulation of surface moisture, especially the degree of moisture accumulation in low-lying areas; TWI is usually used to analyze which areas are prone to geological disasters such as soil erosion and landslides; its calculation formula combines the slope , watershed area and other factors, reflecting the impact of different terrains on moisture; the humidity index map calculated based on the digital elevation model and historical precipitation data is used to show the accumulation of moisture in the area; areas with higher humidity are usually more prone to landslide disasters, especially when the precipitation is large; kernel density estimation is a method of estimating the probability density function by smoothing data points; in this scenario, it is used to estimate the density of historical landslide disasters; by performing kernel density estimation on historical landslide disaster data points, the distribution of landslide disasters in the target area can be obtained, that is, high-density and low-density areas where landslide events occur; the landslide disaster density map obtained by kernel density estimation shows the occurrence of landslide disasters at different locations in the target area. Density, high-density areas indicate that the frequency of landslide disasters in that area is higher; in geographic information systems (GIS), weighted overlay is a process of overlaying multiple layers and assigning different weights to each layer; in this scenario, weighted overlay of the historical regional moisture distribution map and the historical landslide hazard density cloud map can comprehensively consider the terrain moisture and the density of historical landslide disasters, thereby obtaining a more comprehensive assessment map; by weighted overlaying the historical moisture distribution map and the landslide hazard density cloud map, a comprehensive index map is generated to represent the landslide sensitivity of each part of the region; this map can help identify which areas have higher landslide risks and which areas have relatively lower risks; based on the historical landslide sensitivity comprehensive index map, the target The regions are divided into different levels or labels according to their landslide sensitivity; the density gradient planning of monitoring points refers to adjusting the density of monitoring point layout according to the sensitivity label of each sub-region, so as to strengthen monitoring in high-sensitivity areas and reduce the number of monitoring points in low-sensitivity areas; this can effectively and reasonably allocate resources and achieve accurate monitoring: in high-sensitivity areas: due to the high landslide risk, more dense rainfall monitoring points should be deployed in the area to obtain rainfall data in time and warn of potential landslide risks in advance; in medium-sensitive areas: the precipitation and landslide risk are moderate, and monitoring points can be deployed at the second spacing with moderate spacing; in low-sensitivity areas: the landslide risk is low, and rainfall monitoring points are sparsely deployed with larger spacing, usually at the third spacing;Based on the above monitoring point density gradient planning, rainfall monitoring points are arranged in different regions to form a rainfall monitoring point distribution map. This map shows the layout of rainfall monitoring points in each region, which can help analyze precipitation data and provide important support for landslide risk assessment and early warning.

[0067] In an optional embodiment, the process of constructing the dynamic threshold model includes:

[0068] Collect rainfall data before historical landslide events and extract the hourly rainfall, cumulative 24-hour rainfall, and cumulative 72-hour rainfall corresponding to each landslide event;

[0069] Based on the geological data of each landslide event, the rainfall data were divided into several geological condition groups; the geological data included rock and soil permeability, slope, and vegetation coverage;

[0070] Statistical analysis is performed on the rainfall data within each geological condition group to determine the hourly rainfall threshold, cumulative 24-hour rainfall threshold, and cumulative 72-hour rainfall threshold corresponding to each geological condition group;

[0071] A nonlinear mapping relationship between geological data and rainfall thresholds is established based on the support vector machine algorithm to obtain a dynamic threshold model.

[0072] It should be noted that, before a landslide disaster occurs, rainfall data prior to historical landslide events is collected and recorded. The permeability coefficient of rock and soil reflects the permeability of the rock and soil. The lower the permeability coefficient, the less precipitation penetrates the soil, the higher the risk of water accumulation, and the greater the likelihood of landslides. Slope is a key factor influencing landslide occurrence; the steeper the slope, the higher the likelihood of landslides. Vegetation effectively prevents soil erosion and reduces the probability of landslides. Areas with high vegetation cover are generally more stable. Based on this geological data, rainfall data prior to all historical landslide events is categorized into several geological condition groups. For example, areas with low permeability, high slope, and low vegetation cover may belong to one geological condition group, while areas with high permeability, low slope, and high vegetation cover may belong to another. Within each geological condition group, statistical analysis of rainfall data is performed to determine rainfall thresholds for different geological conditions based on historical landslide data. The Support Vector Machine (SVM) algorithm is a powerful machine learning algorithm commonly used to solve classification and regression problems. In this problem, support vector machines (SVMs) are used to establish a nonlinear mapping between geological data (such as hydraulic conductivity, slope, and vegetation coverage) and rainfall thresholds. The rainfall threshold may not follow a simple linear relationship but rather be influenced by a combination of geological factors. Using a nonlinear kernel function, support vector machines can model the relationship between geological data and rainfall thresholds as a complex function. Based on training data (geological and rainfall data from historical landslide events), SVMs can build a dynamic threshold model that automatically predicts the rainfall threshold for each region based on actual geological conditions (such as rock and soil hydraulic conductivity, slope, and vegetation coverage). This dynamic threshold can change with changing geological conditions, enabling landslide risk prediction in diverse geological environments. By establishing a dynamic threshold model, the likelihood of landslides can be assessed in real time during future rainfall events based on real-time rainfall data and geological conditions. If precipitation exceeds a certain threshold (such as hourly, 24-hour, or 72-hour rainfall), the landslide risk can be predicted, enabling early warning and emergency response. By comprehensively analyzing geological conditions and rainfall data, the support vector machine model can more accurately predict the timing and location of landslides, thereby improving the ability to prevent landslide disasters.

[0073] In an optional embodiment, determining a real-time rainfall risk distribution map of a target area by using a real-time rainfall warning threshold based on a spatial interpolation method includes:

[0074] Preprocessing the real-time rainfall data collected from rainfall monitoring points, including outlier removal and data smoothing;

[0075] Based on the Kriging interpolation algorithm, combined with terrain elevation data and land use type data, the preprocessed rainfall data are spatially interpolated to obtain a real-time rainfall risk distribution map of the target area.

[0076] It should be noted that outliers refer to values ​​in the data that are significantly different from other data points, which are usually caused by measurement errors, equipment failures, etc. The purpose of outlier removal is to ensure the accuracy of the data and prevent abnormal data from affecting subsequent analysis. Some common outlier detection methods include standard deviation-based detection and box plot analysis, which can identify and remove obvious outliers. Data smoothing refers to the use of mathematical methods to remove random fluctuations in the data, making the data smoother and more continuous. Common smoothing techniques include moving average, weighted average, low-pass filtering, etc. In rainfall data, smoothing helps to eliminate noise caused by short-term rainfall fluctuations, making the rainfall trend clearer, thereby facilitating subsequent analysis. Kriging interpolation (Kriging Interpolation is a geostatistical method based on spatial data, which is widely used in fields such as geographic information systems (GIS) and environmental monitoring. Its core idea is to use data from known points to predict values ​​at other locations. Spatial interpolation refers to the use of spatial autocorrelation analysis results. Kriging interpolation can generate a weighted average to estimate the rainfall at unknown locations. The weight coefficient of each point is determined according to its distance from the known points and the spatial autocorrelation relationship. The Kriging method minimizes the prediction error by optimizing the weight coefficient, thereby obtaining a more accurate spatial distribution. Terrain elevation has an important influence on rainfall distribution. For example, high altitude areas may have more precipitation, while low-lying areas are prone to waterlogging. Therefore, Kriging interpolation not only relies on the data of rainfall monitoring points, but also takes into account terrain factors. In the interpolation process, terrain data can be used as auxiliary information to adjust the interpolation results so that the interpolation results are more accurate. The precipitation distribution pattern is more consistent with the actual terrain; land use type data (such as cities, agricultural land, forests, etc.) will also affect the distribution of precipitation; for example, the precipitation in urbanized areas may be larger because the urban heat island effect may lead to increased local precipitation; combining land use data with rainfall data can further refine the spatial interpolation results, making the predicted rainfall risk distribution more realistic and accurate; the real-time rainfall risk distribution map of the target area will show the rainfall intensity and potential risks of each area, such as: high-risk areas: such as areas with heavy rainfall or long duration, these places may experience natural disasters such as floods or landslides; low-risk areas: areas with less precipitation or relatively stable land use and terrain, these areas have lower rainfall risks; the generated rainfall risk distribution map can help relevant departments (such as the Meteorological Bureau, Emergency Management Bureau, etc.) monitor rainfall conditions in real time and provide support for disaster prevention and emergency response.

[0077] In an optional embodiment, a real-time high-risk landslide cluster in a target area is determined based on a clustering algorithm using a real-time rainfall risk distribution map and the geographical coordinates of historical landslide disaster points, including:

[0078] Extract the geographical coordinates of historical landslide hazard points and the rainfall risk values ​​of corresponding locations in the real-time rainfall risk distribution map;

[0079] Based on the density peak clustering algorithm, the landslide hazard points were clustered with rainfall risk value as weight to obtain several initial clusters.

[0080] The boundary of each initial cluster is optimized to obtain the real-time high-risk landslide clusters in the target area; the boundary optimization includes cluster segmentation and merging based on topographic fault data.

[0081] It should be noted that the geographical coordinates of historical landslide hazard points represent the locations where landslide disasters occurred in the past, and their geographical coordinates provide a reference for spatial location; historical hazard points may be collected through satellite images, post-disaster assessment reports or other means; in the above steps, we have generated a real-time rainfall risk distribution map of the target area, which shows the rainfall risks in different areas; the rainfall risk value of each area is the comprehensive result of factors such as precipitation intensity, topography, and land use; through spatial overlapping operations (such as spatial analysis of point and raster data), the rainfall risk value corresponding to the location of the historical landslide hazard point can be extracted from the real-time rainfall risk distribution map; this In this way, we can assign a specific rainfall risk value to each historical disaster point for subsequent cluster analysis; density peak clustering is a clustering algorithm that does not require a pre-set number of clusters; it identifies clusters by calculating the density of each data point and the distance to the point with higher density; areas with high density will form clusters, and areas with low density will be identified as noise; the core of the density peak clustering algorithm is to use rainfall risk value as weight; this means that in the clustering process, landslide disaster points with larger rainfall risk values ​​will be regarded as more important data points, and they will have a greater influence on the final clustering; based on the geographical coordinates of historical landslide disaster points and the corresponding rainfall risk The density peak clustering algorithm can cluster these disaster points into multiple preliminary clusters; each cluster represents an area with a higher rainfall risk, which may be a high-risk area for landslide disasters; after cluster analysis, preliminary clusters will be formed, but due to the complexity of the terrain and other factors, the boundaries of these clusters may not completely conform to the actual situation; therefore, it is necessary to optimize the boundaries of each initial cluster to more accurately define the high-risk area for landslides; boundary optimization includes the following two key steps: topographic fault data refers to the topographic changes in the region, such as mountains, faults, rivers and other natural geographical features; topographic changes are usually closely related to the occurrence of landslide disasters; for example For example, areas with steep slopes and faults are prone to landslides. Based on the guidance of terrain data, the boundaries of clusters are optimized. Terrain faults may cause the boundaries of the original clusters to be inaccurate, so some areas that are greatly affected by the terrain need to be separated, or clusters with similar terrain features need to be merged. In this way, the boundaries of high-risk areas can be obtained that are more consistent with the actual terrain and the laws of landslide occurrence. For example, if a cluster is located on a steep hillside, terrain fault data may indicate that the landslide risk in this area is high, so the boundaries of this cluster may need to be expanded. If a cluster crosses a fault line, it may need to be split into two more accurate clusters.

[0082] In an optional embodiment, the calculation formula of the landslide susceptibility index is as follows:

[0083] ;

[0084] in, represents the landslide susceptibility index, Indicates the The weight coefficient corresponding to the rainfall warning threshold is: Indicates the The number of times the rainfall warning threshold exceeds the limit, represents the terrain slope correction coefficient, Indicates the stability coefficient corresponding to the soil type.

[0085] In an optional embodiment, the susceptibility labels corresponding to the sub-regions in the target region are modified based on the geological parameters of the target region to obtain a landslide susceptibility map, including:

[0086] Obtaining geographic information of the target area and standardizing the geographic information; wherein the geographic information includes terrain slope, soil type and land use type;

[0087] Based on the analytic hierarchy process, determine the susceptibility label corresponding to the geographic information;

[0088] Based on the weighted overlay method, the standardized geographic information and the landslide susceptibility index are fused to obtain the landslide susceptibility map.

[0089] It should be noted that terrain slope refers to the degree of inclination of the surface, usually expressed as an angle; areas with large slopes are more prone to landslides, especially when encountering heavy rain; the slope can be calculated through DEM; different types of soils vary greatly in terms of water infiltration, retention capacity and bearing capacity, and some soils (such as loose sand) may be more prone to landslides; soil types can be obtained through soil classification maps or related databases; land use types reflect the development and use status of land; for example, urban construction, agricultural activities, forest cover, etc. will have an impact on soil stability; land use types are generally obtained through remote sensing images or land use databases; standardization refers to the numerical conversion of the above geographic information to make it They have a unified measurement standard and comparability; since different types of geographic information (such as slope, soil, land use) may vary greatly in numerical range, standardization converts these data into the same scale or standardized values ​​(such as between 0 and 1); Common standardization methods include: minimum-maximum standardization: subtract the minimum value from each data value and divide it by the range of the data (maximum-minimum); Z-score standardization: convert the data into a standard normal distribution by subtracting the mean and dividing it by the standard deviation; Hierarchy Analysis is a multi-level decision analysis method used to evaluate and rank complex problems by decomposing them into different levels and factors; Therefore, we can create a standard normal distribution for each geographic information element (such as terrain slope, The analytic hierarchy process converts these geographical information into a "susceptibility label", that is, the degree of influence of each geographical element on the occurrence of landslides. The evaluation process is to first determine the hierarchical relationship between the target (landslide susceptibility) and the influencing factors; based on expert knowledge or data analysis results, give a relative importance judgment between the influencing factors; for example, slope may have a greater influence on landslides than soil type; the weight of each factor is calculated through the mathematical model of the analytic hierarchy process; a susceptibility label is assigned to each area based on the calculated weight and the numerical value of each factor (such as slope value, soil type code); the weighted overlay method is a technique commonly used in spatial analysis, which combines multiple levels of The data (in this case, standardized geographic information and the landslide susceptibility index) are fused to produce a composite assessment result. In this process, each factor is weighted according to its weight and then superimposed. In weighted superposition, each standardized geographic information is weighted according to its importance in the susceptibility assessment. For example, if the weight of slope is higher, then the slope value will have a greater impact on the final result. The superposition operation combines the weighted geographic information and the landslide susceptibility index to produce a composite score. This usually means combining multiple factors for each area into a final landslide susceptibility score. The landslide susceptibility index is a composite value calculated from different factors, which indicates the risk of landslides in an area.During this process, the standardized geographic information and landslide susceptibility index are weighted and overlaid to generate a final landslide susceptibility map. The higher the susceptibility score for each area, the more likely it is to experience a landslide. The resulting landslide susceptibility map shows the landslide susceptibility of each point or grid within the target area. Each area on the map is assigned a susceptibility value, indicating the area's landslide risk. This map can help governments and disaster management agencies identify high-risk areas, conduct disaster prevention planning, and implement measures to reduce the likelihood of landslides, such as improving land management, strengthening soil protection, and optimizing infrastructure design.

[0090] In an optional embodiment, determining landslide geological disaster warning information based on the landslide susceptibility map includes:

[0091] Update the real-time rainfall risk distribution map based on real-time rainfall data, and extract the rainfall risk value of each sub-region in the updated rainfall risk distribution map;

[0092] The rainfall risk value is coupled with the landslide susceptibility index of the corresponding sub-region in the landslide susceptibility map to obtain a comprehensive landslide risk index.

[0093] Based on the comprehensive landslide risk index, the target area is divided into five warning levels using the natural break method to obtain landslide geological disaster warning information.

[0094] It should be noted that real-time rainfall data reflects the amount of rainfall at present or in the recent period; rainfall is an important triggering factor for landslides, because a large amount of precipitation will increase the moisture of the soil, reduce the soil's anti-slip ability, and lead to the occurrence of landslides; using real-time rainfall data, a rainfall risk distribution map can be generated or updated, which shows the rainfall risks in different areas; the rainfall risk value of each sub-area reflects the current rainfall level in the area. The greater the rainfall, the higher the rainfall risk value and the greater the possibility of landslides; it can be calculated based on rainfall (such as cumulative rainfall, rainfall intensity, etc.); if the rainfall exceeds a certain threshold, some areas may be marked as high-risk areas; the rainfall risk value of each sub-area can be obtained by comparing the rainfall data of the area with pre-set standards; these values ​​will be dynamically calculated based on factors such as the intensity and duration of precipitation, and updated to the rainfall risk distribution map; if a sub-area has received more than 100mm of rainfall in the past 24 hours, its rainfall risk value will be updated to a higher ; At this time, the risk value of the area is extracted and prepared for combined analysis with the corresponding data in the landslide susceptibility map; the rainfall risk value of each sub-area is coupled with the landslide susceptibility index for analysis, that is, the risk of rainfall is combined with the landslide susceptibility of the area itself; through this coupling, it is possible to identify which areas are most prone to landslides under rainfall conditions; for example, although some areas have low landslide susceptibility themselves, their comprehensive risk values ​​may still be high due to the influence of extreme rainfall; the natural break method is a commonly used classification method, which divides the data set into multiple categories so that the data within each category is relatively uniform, while the data differences between different categories are as large as possible; in this application, it is used to divide the target area into different warning levels according to the comprehensive landslide risk index; the divided warning levels will be provided to relevant departments or the public as early warning information for landslide geological disasters for disaster prevention and emergency response; this information can help relevant departments to reasonably arrange emergency responses, take protective measures in advance, and avoid or reduce losses caused by landslide disasters.

[0095] Example 2, as Figure 2 As shown, the present invention proposes a landslide geological disaster early warning method based on rainfall, which is applicable to the above-mentioned landslide geological disaster early warning system based on rainfall, including:

[0096] S1. Deploy several rainfall monitoring points in the target area based on historical landslide disaster data and historical precipitation data; select several historical landslide disaster points based on the historical landslide disaster data; each historical landslide disaster point corresponds to preset geographical coordinates and geological parameters;

[0097] S2. Collecting real-time rainfall data based on rainfall monitoring points; determining the real-time rainfall warning threshold corresponding to each historical landslide disaster point based on the real-time rainfall data and the geological parameters of each historical landslide disaster point based on a dynamic threshold model;

[0098] S3. Determine the real-time rainfall risk distribution map of the target area using the real-time rainfall warning threshold based on the spatial interpolation method; determine the real-time high-risk landslide clusters in the target area using the real-time rainfall risk distribution map and the geographic coordinates of historical landslide disaster points based on the clustering algorithm;

[0099] S4. Overlay analysis is performed on the real-time landslide high-risk clusters and the real-time rainfall risk distribution map to obtain a real-time landslide susceptibility index for each sub-region in the target area; based on the real-time landslide susceptibility index, each sub-region in the target area is divided into landslide susceptibility categories to obtain a susceptibility label corresponding to each sub-region in the target area;

[0100] S5. Modify the susceptibility labels corresponding to each sub-region in the target region based on the geological parameters of the target region to obtain a landslide susceptibility map; and determine landslide geological disaster warning information based on the landslide susceptibility map.

[0101] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A landslide geological disaster early warning system based on rainfall, characterized in that: include: A rainfall monitoring unit, the rainfall monitoring unit being configured to deploy a plurality of rainfall monitoring points within a target area based on historical landslide disaster data and historical precipitation data; selecting a plurality of historical landslide disaster points based on the historical landslide disaster data; each of the historical landslide disaster points corresponding to preset geographic coordinates and geological parameters; A rainfall collection unit, configured to collect real-time rainfall data based on the rainfall monitoring point; Determine the real-time rainfall warning threshold corresponding to each of the historical landslide disaster points based on the dynamic threshold model using the real-time rainfall data and the geological parameters of each of the historical landslide disaster points; a risk warning unit, configured to determine a real-time rainfall risk distribution map of the target area using the real-time rainfall warning threshold based on a spatial interpolation method; Determining the real-time landslide high-risk clusters in the target area based on the real-time rainfall risk distribution map and the geographical coordinates of the historical landslide disaster points based on a clustering algorithm; a landslide analysis unit, configured to perform superposition analysis on the real-time high-risk landslide clusters and the real-time rainfall risk distribution map to obtain a real-time landslide susceptibility index for each sub-region in the target region; Based on the real-time landslide susceptibility index, each sub-region in the target region is divided into landslide susceptibility categories to obtain a susceptibility label corresponding to each sub-region in the target region; a region division unit, configured to modify a susceptibility label corresponding to each sub-region in the target region based on geological parameters of the target region to obtain a landslide susceptibility map; Landslide geological disaster warning information is determined based on the landslide susceptibility map.

2. The landslide geological disaster early warning system based on rainfall according to claim 1 is characterized in that: The rainfall warning thresholds include hourly rainfall intensity thresholds, cumulative 24-hour rainfall thresholds, and cumulative 72-hour rainfall thresholds; the susceptibility labels include extremely low susceptibility areas, low susceptibility areas, moderate susceptibility areas, high susceptibility areas, and extremely high susceptibility areas; the geological parameters include terrain slope, soil type, and land use type.

3. The landslide geological disaster early warning system based on rainfall according to claim 2 is characterized in that: Based on historical landslide hazard data and historical precipitation data, several rainfall monitoring points are set up in the target area, including: Acquire a digital elevation model of the target area; calculate a terrain humidity index based on the digital elevation model and the historical precipitation data to obtain a historical regional humidity distribution map corresponding to the target area; Performing kernel density estimation on the historical landslide disaster data to obtain a historical landslide disaster density cloud map corresponding to the target area; Performing weighted superposition on the historical regional moisture distribution map and the historical landslide disaster density cloud map to obtain a historical landslide sensitivity comprehensive index map; Based on the historical landslide sensitivity comprehensive index map, each sub-region in the target area is divided into landslide sensitivity categories to obtain sensitivity labels for each sub-region; wherein the sensitivity labels include high-sensitivity areas, medium-sensitivity areas, and low-sensitivity areas; Based on the sensitivity labels of each sub-area, a monitoring point density gradient planning is performed, wherein rainfall monitoring points are arranged at a first spacing in the highly sensitive area, at a second spacing in the medium-sensitive area, and at a third spacing in the low-sensitivity area, so as to obtain a rainfall monitoring point distribution map; wherein the first spacing < the second spacing < the third spacing.

4. The landslide geological disaster early warning system based on rainfall according to claim 3 is characterized in that: The process of constructing the dynamic threshold model includes: Collect rainfall data before historical landslide events and extract the hourly rainfall, cumulative 24-hour rainfall, and cumulative 72-hour rainfall corresponding to each landslide event; Based on the geological data of each landslide event, the rainfall data is divided into a plurality of geological condition groups; wherein the geological data includes rock and soil permeability coefficient, slope and vegetation coverage; Performing statistical analysis on the rainfall data within each geological condition group to determine the hourly rainfall threshold, the cumulative 24-hour rainfall threshold, and the cumulative 72-hour rainfall threshold corresponding to each geological condition group; A nonlinear mapping relationship between the geological data and the rainfall threshold is established based on a support vector machine algorithm to obtain the dynamic threshold model.

5. The landslide geological disaster early warning system based on rainfall according to claim 4 is characterized in that: Determining a real-time rainfall risk distribution map of the target area by using the real-time rainfall warning threshold based on a spatial interpolation method includes: Preprocessing the real-time rainfall data collected by the rainfall monitoring point, wherein the preprocessing includes outlier removal and data smoothing; Based on the Kriging interpolation algorithm, combined with terrain elevation data and land use type data, the preprocessed rainfall data is spatially interpolated to obtain a real-time rainfall risk distribution map of the target area.

6. The landslide geological disaster early warning system based on rainfall according to claim 5, characterized in that: Determining a real-time high-risk landslide cluster in the target area based on the real-time rainfall risk distribution map and the geographical coordinates of the historical landslide disaster points based on a clustering algorithm includes: Extracting the geographical coordinates of the historical landslide disaster point and the rainfall risk value of the corresponding position in the real-time rainfall risk distribution map; Based on the density peak clustering algorithm, the landslide hazard points are clustered and analyzed with the rainfall risk value as a weight to obtain a number of initial clusters; Boundary optimization is performed on each of the initial clusters to obtain real-time high-risk landslide clusters in the target area; wherein the boundary optimization includes cluster segmentation and merging based on terrain fault data.

7. The landslide geological disaster early warning system based on rainfall according to claim 6 is characterized in that: The calculation formula of the landslide susceptibility index is as follows: ； in, represents the landslide susceptibility index, Indicates the The weight coefficient corresponding to the rainfall warning threshold is: Indicates the The number of times the rainfall warning threshold exceeds the limit, represents the terrain slope correction coefficient, Indicates the stability coefficient corresponding to the soil type.

8. The landslide geological disaster early warning system based on rainfall according to claim 7 is characterized in that: Modifying the susceptibility labels corresponding to the sub-regions in the target region based on the geological parameters of the target region to obtain a landslide susceptibility map includes: Acquiring geographic information of the target area and performing standardization processing on the geographic information; wherein the geographic information includes terrain slope, soil type, and land use type; Determine the susceptibility label corresponding to the geographic information based on the hierarchical analysis method; Based on the weighted superposition method, the standardized geographic information and the landslide susceptibility index are fused to obtain a landslide susceptibility map.

9. The landslide geological disaster early warning system based on rainfall according to claim 8, characterized in that: Determining landslide geological disaster warning information based on the landslide susceptibility map includes: updating the real-time rainfall risk distribution map based on the real-time rainfall data, and extracting the rainfall risk value of each sub-region in the updated rainfall risk distribution map; performing coupling analysis on the rainfall risk value and the landslide susceptibility index of the corresponding sub-region in the landslide susceptibility map to obtain a comprehensive landslide risk index; Based on the comprehensive landslide risk index, the target area is divided into five warning levels through the natural break method to obtain landslide geological disaster warning information.

10. A landslide geological disaster early warning method based on rainfall, which is applicable to a landslide geological disaster early warning system based on rainfall according to any one of claims 1 to 9, characterized in that: include: Several rainfall monitoring points are deployed in the target area based on historical landslide hazard data and historical precipitation data; Selecting a number of historical landslide disaster points based on historical landslide disaster data; each of the historical landslide disaster points corresponds to preset geographical coordinates and geological parameters; Collecting real-time rainfall data based on the rainfall monitoring points; Determine the real-time rainfall warning threshold corresponding to each of the historical landslide disaster points based on the dynamic threshold model using the real-time rainfall data and the geological parameters of each of the historical landslide disaster points; Determine the real-time rainfall risk distribution map of the target area through the real-time rainfall warning threshold based on the spatial interpolation method; Determining the real-time landslide high-risk clusters in the target area based on the real-time rainfall risk distribution map and the geographical coordinates of the historical landslide disaster points based on a clustering algorithm; Overlaying and analyzing the real-time landslide high-risk clusters and the real-time rainfall risk distribution map to obtain a real-time landslide susceptibility index for each sub-region in the target area; Based on the real-time landslide susceptibility index, each sub-region in the target region is divided into landslide susceptibility categories to obtain a susceptibility label corresponding to each sub-region in the target region; Modifying the susceptibility labels corresponding to the sub-regions in the target region based on the geological parameters of the target region to obtain a landslide susceptibility map; Landslide geological disaster warning information is determined based on the landslide susceptibility map.

Images