Typhoon and seawater backflow induced landslide damming prediction and early warning method

Abstract

The application discloses a typhoon and seawater backflow comprehensive induced landslide blocking river dam forming prediction and early warning method and belongs to the technical field of disaster prediction. The specific steps are as follows: acquiring in-sea river bank slope data, constructing a landslide background trend degree calculation model based on a FIS-YOLOv8 model; calculating a typhoon rainfall influence coefficient and a seawater backflow enhancement coefficient on a prediction day, and calculating a typhoon and seawater backflow comprehensive induced landslide occurrence probability; determining a landslide body migration loss index of each dangerous slope, calculating a typhoon and seawater backflow comprehensive induced landslide blocking river dam forming probability, and performing early warning. The application comprehensively considers various inducing factors of a disaster environment, simultaneously adopts the FIS-YOLOv8 model to calculate the landslide background trend degree, and improves the accuracy of the calculation result.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of disaster prediction, and particularly relates to a typhoon and seawater backflow comprehensive induced landslide blocking river damming prediction and early warning method. BACKGROUND

[0002] Typhoon is a kind of intense tropical cyclone, usually accompanied by strong wind, heavy rain, storm surge and other phenomena, causing serious damage to the affected areas. The existing landslide prediction method only considers the influence of rainfall on the slope, but does not consider the change of slope structure caused by seawater backflow. For example, the patent with publication number CN116151483B discloses a regional rainfall landslide probabilistic prediction method and prediction terminal, which selects characteristic factors suitable for different underlying surface conditions, different time phases and different spatial resolutions, and uses a descaling algorithm with strong generalization ability and high robustness to overcome the time lag effect of vegetation and low visibility weather on seasonal rainfall and the problem of abnormal rainfall distribution, thereby improving the accuracy of rainfall data descaling. The damage caused by the typhoon disaster chain is much greater than the typhoon disaster itself. Typhoon causes seawater backflow in the river into the sea. Under the comprehensive action of rainfall infiltration, the groundwater level of the slope along the river into the sea rises, the mechanical strength and deformation properties of the rock-soil body decrease, and the probability of occurrence of geological disasters such as landslides increases. The landslide body piles up into a dam and blocks the river, and after the dam breaks, it may cause downstream flood disasters.

[0003] The existing technology also has the following problems:

[0004] (1) The existing method selects disaster-causing factors by manual or feature extraction algorithm when calculating the landslide background trend degree, which cannot comprehensively and objectively reflect all elements of the disaster environment.

[0005] (2) The existing method only considers the influence of typhoon rainfall infiltration on the tensile, shear and tensile-shear coupling strength of the slope rock-soil body when analyzing the induction mechanism of typhoon on landslide, however, the seawater backflow caused by typhoon rainfall can increase the groundwater level of the slope and reduce the stability of the slope, and the existing method ignores the comprehensive action of typhoon rainfall and seawater backflow.

[0006] (3) Typhoon often leads to continuous rainfall, and even continuous rain in the north. The existing method generally uses process rainfall or landing day rainfall to represent the typhoon rainfall grade, which exaggerates the role of typhoon rainfall and ignores the influence of underlying surface factors, thereby reducing the success rate of disaster chain prediction and the working efficiency of comprehensive prevention and control measures.

[0007] (4) The landslide body volume that may flow into the river is highly uncertain due to the influence of factors such as landslide type, topography, sliding force and sliding surface depth, and the existing method regards it as a fixed value, which does not conform to the complex mutual feedback mechanism of landslide blocking river damming and disaster environment. SUMMARY

[0008] The application provides a typhoon and seawater backflow comprehensive induced landslide river blocking dam forming prediction and early warning method, and the specific steps are as follows:

[0009] Step S1: acquiring the data of the slope along the river into the sea, collecting the Bijie landslide data set and the southwest landslide data set, preparing the historical disaster point data set and the dangerous slope data set;

[0010] Step S2: constructing a landslide background trend degree calculation model based on the FIS-YOLOv8 model, and calculating the landslide background trend degree of each dangerous slope through the trained model ;

[0011] Step S3: calculating the potential landslide volume of each dangerous slope;

[0012] Step S4: counting the landslide starting rainfall threshold of each dangerous slope, predicting the daily effective rainfall, determining the typhoon rainfall triggering coefficient and the seawater backflow enhancement coefficient ;

[0013] The seawater backflow enhancement coefficient calculation formula is as follows:

[0014] ;

[0015] Wherein: is the river section type adjustment coefficient of the slope, is the average river bottom longitudinal slope, is the water level rising height caused by seawater backflow at the slope, is the river bottom longitudinal slope at the slope, is the historical extreme value of the water level rise at each slope;

[0016] Step S5: calculating the typhoon and seawater backflow comprehensive induced landslide occurrence probability, and the calculation formula is as follows:

[0017] ;

[0018] Wherein: is the typhoon and seawater backflow comprehensive induced landslide occurrence probability;

[0019] Step S6: calculating the landslide volume required for each dangerous slope to occur landslide river blocking dam forming;

[0020] Step S7: determining the landslide body migration loss index of each dangerous slope ;

[0021] Step S8: calculating the typhoon and seawater backflow comprehensive induced landslide river blocking dam forming probability of each dangerous slope and performing early warning.

[0022] Preferably, the in-sea river bank slope data includes historical disaster point data and dangerous bank slope data;

[0023] The historical disaster point data includes the number of historical disaster points , longitude and latitude, and satellite images;

[0024] The dangerous bank slope data includes the number of dangerous bank slopes , longitude and latitude, satellite images, and river data at each bank slope;

[0025] The river data at the dangerous bank slope includes the river bottom longitudinal slope, river bed width, river bed depth, potential sliding direction angle with the river, and average river bottom longitudinal slope;

[0026] The Bijie landslide data set, the southwest landslide data set, and the historical disaster point data are combined to form a model training data set, including 1270+ satellite images of 2003 non-landslide, augmented by rotation (90°, 180°, 270°), mirror flipping (horizontal, vertical), brightness transformation (lightening, darkening), wherein the input value of each sample is the satellite image, the output value of the landslide disaster point is 1, and the output value of the non-landslide disaster point is 0.

[0027] Preferably, the FIS-YOLOv8 model includes a backbone network, a neck network, and a head network arranged in sequence, the backbone network adopts YOLOv8, the Bottleneck structure of the C2f module in YOLOv8 is replaced by a more lightweight FasterNetBottleneck, and the iRMB attention mechanism and the EMA attention mechanism are combined in YOLOv8 to introduce the iRMB-EMA fusion attention mechanism, and the CIoU loss function in YOLOv8 is replaced by the Shape-IoU loss function, thereby obtaining the FIS-YOLOv8 model.

[0028] The augmented model training data set is input into the FIS-YOLOv8 model, and the satellite image of the dangerous bank slope at the position is input into the trained model, and the output value is the landslide background trend degree .

[0029] Preferably, the potential landslide volume is estimated by depicting the through sliding surface through the cracks on both sides of the dangerous bank slope and the back wall crack, and estimating the volume of the rock-soil mass above the through sliding surface V .

[0030] Preferably, the typhoon rainfall trigger coefficient is determined according to the difference between the predicted daily effective rainfall of each dangerous bank slope and the landslide starting rainfall threshold . ​​ The determination method is shown in Table 1.

[0031] Table 1 Typhoon Rainfall Triggering Coefficient Determining the method

[0032] ;

[0033] Predicted daily effective rainfall The calculation formula is as follows:

[0034] ;

[0035] in: To predict the amount of rainfall that can infiltrate during a typhoon, For the front Rainfall that can be infiltrated by typhoons;

[0036] and The calculation method is shown in the following formula:

[0037] ;

[0038] in: To predict daily typhoon rainfall, For the front Rainfall during the typhoon.

[0039] Preferably, the landslide initiation rainfall threshold The determination method is as follows: Determine The landslide initiation rainfall threshold at historical disaster sites was determined by interpolating the above thresholds using the Glikin interpolation method of ArcGIS 10.2 spatial analysis function, thus obtaining the landslide initiation rainfall threshold for each dangerous slope.

[0040] The methods for determining the rainfall threshold that triggers landslides at historical disaster sites include the following two scenarios:

[0041] (1) If the effective rainfall on the day a certain historical disaster occurs is the highest value in the past 50 years, then the second highest effective rainfall in the past 50 years is determined as the landslide initiation rainfall threshold;

[0042] (2) If the effective rainfall on the day a historical disaster occurs is not the highest value in the past 50 years, then the effective rainfall on the day of occurrence is determined as the landslide initiation rainfall threshold at that disaster point.

[0043] Preferred methods for calculating daily typhoon rainfall over the past 50 years include the following two scenarios:

[0044] (1) If a rainfall monitoring station is located within 3 km of the historical disaster site and dangerous slope, the monitoring value is determined as the typhoon rainfall.

[0045] (2) If there is no rainfall monitoring station within 3 km of the historical disaster point and the dangerous slope, determine the number of all rainfall monitoring stations within 30 km of the historical disaster point and the dangerous slope as , count the occurrence time distribution of the historical disasters in days, count the rainfall of each rainfall monitoring station in the above-mentioned days, and obtain a total of rainfall data;

[0046] In rainfall monitoring stations, randomly select 1 as a reference rainfall station, and the remaining as a correction rainfall station, determine the rainfall of the reference rainfall station on each disaster occurrence day , the average rainfall of the correction rainfall station , and the difference and as a group of data, a total of groups of data;

[0047] A G1-entropy weight coupled TOPSIS model is constructed, the elevation, slope, aspect, NDVI (normalized vegetation coverage index), plane curvature, profile curvature, SPI (topographic humidity index), STI (water flow intensity index), TWI (sediment transport index), atmospheric precipitable water, enhanced vegetation coverage index, total primary productivity, net primary productivity, and surface specific humidity of the reference rainfall station on the disaster occurrence day are used as model inputs, and groups of data are used to train the model to obtain a trained model;

[0048] The elevation, slope, aspect, NDVI, plane curvature, profile curvature, SPI, STI, TWI, atmospheric precipitable water, enhanced vegetation coverage index, total primary productivity, net primary productivity, and surface specific humidity of any dangerous slope on any day are used as model inputs, and the sum of the model output based on the trained G1-entropy weight coupled TOPSIS model and the average rainfall of each rainfall monitoring station on that day is the typhoon rainfall of the dangerous slope on that day.

[0049] Preferably, is a river section type adjustment coefficient, 1.25 for a gorge curved river section, 1.00 for a gorge straight river section and an open curved river section, and 0.75 for an open straight river section. The river section type is divided according to the relevant provisions of “Code for Hydrological Survey and Design of Highway Engineering” (JTGC30-2015) and “Code for Survey and Design of Highway Bridge Location” (JTJ 062-91). ​​​​​

[0050] Preferably, the volume of landslide mass required for landslide blocking river to form dam V g The calculation method is as follows:

[0051] ;

[0052] In the formula: is the volume of landslide mass required for landslide blocking river to form dam, is the depth of river bed at the slope, is the width of river bed at the slope, c is the cohesion of landslide mass, is the internal friction angle of landslide mass, , b is the undetermined parameter;

[0053] The recorded 249 times of landslide blocking river to form dam data in China are used to carry out regression analysis on , b to determine the calculation formula of .

[0054] Preferably, the landslide mass transport loss index is related to the shortest distance from the front edge of landslide mass to the river bank, and the calculation formula is as follows:

[0055] ;

[0056] In the formula: is the shortest distance from the front edge of landslide mass to the river bank.

[0057] Preferably, the volume of landslide mass capable of rushing into the river is set to obey normal distribution with as the mean value and 1 as the variance, and the calculation formula of the occurrence probability of landslide blocking river to form dam induced by typhoon and seawater backflow is as follows:

[0058] ;

[0059] In the formula: is the occurrence probability of landslide blocking river to form dam induced by typhoon and seawater backflow, V is the volume of potential landslide mass, is the angle between the potential sliding direction and the river.

[0060] Preferably, according to the maximum value of the occurrence probability of landslide blocking river to form dam induced by typhoon and seawater backflow p cb,max , the grading early warning is carried out, and the early warning strategy and early warning measures are determined.

[0061] The early warning grades, early warning strategies and early warning measures are shown in Table 2.

[0062] Table 2 Warning level, warning strategy and warning measures

[0063] ;

[0064] Therefore, the typhoon and seawater backflow comprehensive induced landslide damming and damming prediction and early warning method has the beneficial effects as follows:

[0065] (1) The landslide background trend degree is no longer determined by artificial or feature extraction algorithm, but the satellite images of landslides and non-landslides are directly input into the model, which comprehensively and accurately reflects the disaster environment information and reduces the calculation complexity; at the same time, the FIS-YOLOv8 model is used to calculate the landslide background trend degree, which improves the accuracy of the calculation result.

[0066] (2) The landslide background trend degree, typhoon rainfall influence coefficient and seawater backflow enhancement coefficient are introduced when calculating the typhoon and seawater backflow comprehensive induced landslide occurrence probability, which comprehensively considers various factors of the disaster environment, and is more in line with the disaster chain mechanism and evolution mechanism of "typhoon→landslide→damming and damming".

[0067] (3) The present application characterizes the time sequence characteristics of typhoon rainfall by effective rainfall, which can accurately reflect the induction mechanism of typhoon rainfall on landslide.

[0068] (4) In order to accurately depict the uncertainty of the volume of landslide body rushing into the river, the present application adopts normal distribution with mean value of sin θ · k 3· V and variance of 1 when calculating the typhoon and seawater backflow comprehensive induced landslide damming and damming occurrence probability, which overcomes the shortage of not being able to describe the parameter dispersion in revealing the evolution mechanism of disaster chain.

[0069] The technical solutions of the present application will be further described in detail below with the help of the drawings and examples. DESCRIPTION OF DRAWINGS

[0070] Figure 1 is a flow chart of a typhoon and seawater backflow comprehensive induced landslide damming and damming prediction and early warning method of the present application;

[0071] Figure 2 is a network structure diagram of FIS-YOLOv8 model. DETAILED DESCRIPTION

[0072] The embodiments of the present application will be described in detail below with the help of the drawings.

[0073] EMBODIMENT

[0074] For example Figure 1As shown, a typhoon and sea water backflow comprehensive induced landslide damming prediction and early warning method, the specific steps are as follows:

[0075] Step S1: Obtain the data of the coast of the river into the sea, collect Bijie landslide data set and southwest landslide data set, prepare historical disaster point data set and dangerous slope data set.

[0076] This embodiment is aimed at Fufang River, with a total length of 71.8km, a drainage area of 1040km 2 , the average annual precipitation is 923.5mm, the annual runoff depth is 347.8mm, and the equivalent annual runoff is 362 million m 3 , the river network density is 0.37km / km 2 . The estuary is located in Jiacang Village, Qushan Street, Rizhao Economic Development Zone, Shandong Province, with a width of 800m, and the Fufang River basin experiences 3.7 typhoons per year, accompanied by sea water backflow during typhoon. The terrain along the coast of Fufang River is rugged and complex, with the development of landslides and other geological disasters, which has the natural conditions for the formation and evolution of the disaster chain of "typhoon→landslide→damming".

[0077] The data of the coast of the river into the sea is obtained by literature review, field reconnaissance and unmanned aerial vehicle, etc. The literature review mainly includes Rizhao Statistical Yearbook, relevant news reports and published papers, etc. Field reconnaissance uses tape and total station instrument measurement, and the total station instrument model is Topcon GTS2002, and the unmanned aerial vehicle model is DJI Mavic 3.

[0078] The historical disaster point data are as follows:

[0079] Number =182;

[0080] The latitude and longitude is between 35°17'29.51"N~35°38'10.57"N, 119°18'08.24"E~119°26'18.95"E;

[0081] The satellite image is Landsat 8 image of 182 historical disaster points.

[0082] The dangerous slope data are as follows:

[0083] Number =479;

[0084] The latitude and longitude is between 35°17'27.36"N~35°43'31.45"N, 119°16'53.64"E~119°27'04.53"E;

[0085] The satellite image is Landsat 8 image of 479 dangerous slopes.

[0086] The river bottom longitudinal slope at 479 dangerous slopesi riverbed width B r riverbed depth H r angle between potential sliding direction and river θ The extreme values are as follows:

[0087] i min = 0, i max = 1.24%, B r,min = 83.15m, B r,max = 746.98m, H r,min = 5.23m, H r,max = 17.63m, θ min = 0, θ max = 90°;

[0088] average riverbed longitudinal slope i m = 0.23%.

[0089] The Landsat 8 images of the Bijie landslide data set, the southwest landslide data set and the historical disaster point data are merged to form a model training data set, containing 1452 landslide satellite images and 2003 non-landslide satellite images, and are augmented by rotation (90°, 180°, 270°), mirror flipping (horizontal, vertical), brightness transformation (lightening, darkening), a total of 12548 satellite images, wherein the output value of the landslide disaster point is 1 and the output value of the non-landslide disaster point is 0.

[0090] Step S2: constructing a landslide background trend degree calculation model based on the FIS-YOLOv8 model, and calculating the landslide background trend degree of each dangerous slope through the trained model As shown in Figure 2 , the Bottleneck structure of the C2f module in YOLOv8 is replaced by a more lightweight FasterNet Bottleneck, and the new module is named FC Block, a new iRMB-EMA fusion attention mechanism is introduced by combining the iRMB attention mechanism and the EMA attention mechanism, the original CIoU loss function is replaced by the Shape-IoU loss function, and thus the FIS-YOLOv8 model is constructed.

[0091] The FIS-YOLOv8 model is trained using the augmented model training data set, and the Landsat 8 image of the 479 dangerous slopes is input into the trained model, and the output value is the landslide background trend degree k 0, the results show: k 0,min =0.054, k 0,max =0.972.

[0092] Step S3: Calculate the potential landslide volume of each dangerous slope. The cracks on both sides and the back wall of the 479 dangerous slopes are determined through field reconnaissance. The sliding surface of each dangerous slope is depicted according to the cracks, and the volume of the rock-soil mass above the sliding surface is estimated to obtain the potential landslide volume V , the results show: V min =1834m 3 , V max =125427m 3 .

[0093] Step S4: Statistics of the landslide starting rainfall threshold of each dangerous slope, prediction of daily effective rainfall, determination of typhoon rainfall triggering coefficient and seawater backflow enhancement coefficient .

[0094] There are =9 rainfall monitoring stations within 30 km of the 182 historical disaster points and 479 dangerous slopes =126 days. Statistics of the typhoon rainfall data of 9 rainfall monitoring stations on the day of landslide occurrence show that the Jiacanghekou monitoring station = 183.54mm, the Xihexia monitoring station =192.34mm, the Gaojiacun monitoring station =189.27mm, the Xiaoguzhen West monitoring station =185.69mm, the Caijiatanzhu monitoring station =204.52mm, the Songshuyuan East monitoring station =201.89mm, the Pingtailing monitoring station =192.38mm, the Feijiacun monitoring station =190.46mm, the Dongsun village monitoring station =197.18mm.

[0095] According to statistics, there are rainfall monitoring stations within 3 km of 69 historical disaster points and 127 dangerous slopes, and the monitoring values are directly used as the typhoon rainfall of the corresponding historical disaster points and dangerous slopes.

[0096] For the remaining 113 historical disaster points and 352 dangerous slopes, the following method is used to determine the typhoon rainfall:

[0097] Select the Jiacanghekou monitoring station as the reference rainfall station, and the other monitoring stations as the corrected rainfall stations, to determine the rainfall of the reference rainfall station on the 126 disaster occurrence days , the average rainfall of the 8 corrected rainfall stations on the 126 disaster occurrence days and and - , a total of 126 groups of data, the results show that: ( - ) min = -42.73 mm, ( - ) max = 53.46 mm.

[0098] A G1-entropy weight coupled TOPSIS model is constructed, with the elevation, slope, aspect, NDVI, planar curvature, profile curvature, SPI, STI, TWI, atmospheric precipitable water, enhanced vegetation cover index, gross primary productivity, net primary productivity, and surface specific humidity of the Jiacanghekou monitoring station on the disaster occurrence day as the model input, and - as the model output, the model is trained using 126 groups of data, and the trained model is obtained.

[0099] The elevation, slope, aspect, NDVI, planar curvature, profile curvature, SPI, STI, TWI, atmospheric precipitable water, enhanced vegetation cover index, gross primary productivity, net primary productivity, and surface specific humidity of the 113 historical disaster points on the disaster occurrence day and the 352 dangerous slopes on any day are used as the model input, and the G1-entropy weight coupled TOPSIS model is used to obtain the model output, and the sum of the model output and the average daily rainfall of the 9 rainfall monitoring stations is the typhoon rainfall of the historical disaster point and the dangerous slope on that day.

[0100] The daily effective rainfall of the 182 historical disaster points in the past 50 years is calculated, and the results show that the effective rainfall of 143 disaster points on the disaster occurrence day is the highest in the past 50 years, and the landslide initiation rainfall threshold of the 143 disaster points is determined as the second highest value of the effective rainfall in the past 50 years; the effective rainfall of the remaining 39 disaster points on the disaster occurrence day is not the highest in the past 50 years, and the landslide initiation rainfall threshold of the 39 disaster points is determined as the effective rainfall on the disaster occurrence day.

[0101] ​The rainfall threshold of 182 historical landslide points was analyzed by using the ArcGIS10.2 spatial analysis function of the Gringold interpolation method, and the rainfall threshold of 479 dangerous slopes was obtained. The results show that: Q d,min =180.56mm, Q d,max =213.17mm.

[0102] The effective rainfall of 479 dangerous slopes in the typhoon influence period was calculated by taking Typhoon No. 3 in 2024 “Goni” affecting Rizhao City in Shandong Province as an example. The results show that: Q l,min =204.25mm, Q l,max =297.68mm.

[0103] The typhoon rainfall triggering coefficient of 479 dangerous slopes was determined , and the results show that: =0, =0.6.

[0104] The calculation formula of the seawater backflow enhancement coefficient is as follows:

[0105] ;

[0106] Among them: is the river section type adjustment coefficient, is the average river bottom slope, is the water level rise height caused by seawater backflow, is the slope river bottom slope, is the historical extreme value of water level rise at each slope.

[0107] River section type adjustment coefficient Take 1.25 for canyon curved river section, 1.00 for canyon straight river section and open curved river section, and 0.75 for open straight river section. The river section type is divided according to the relevant provisions of “Code for Hydrological Survey and Design of Highway Engineering” (JTG C30-2015) and “Code for Survey and Design of Highway Bridge Site” (JTJ 062-91). The river section type of 479 dangerous slopes was determined, and the results show that: 43 are canyon curved river sections, 82 are canyon straight river sections, 150 are open curved river sections, and 204 are open straight river sections.

[0108] According to engineering experience and typhoon weather forecast, the water level rise value Δ h caused by seawater backflow at 479 dangerous slopes on the forecast day was estimated. Taking Typhoon “Goni” as an example, the minimum value of 479 dangerous slopes is 0, and the maximum value is 0.60m.

[0109] The flood trace elevation and flood surface line of 479 slopes caused by seawater intrusion are determined by historical flood investigation method, and the historical extreme value of water level rise is determined = 1.75 m.

[0110] The seawater intrusion enhancement factor of 479 dangerous slopes in typhoon 'Goni' is calculated The results show that: = 0, = 0.246.

[0111] Step S5: Calculate the probability of landslide induced by typhoon and seawater intrusion. The probability of landslide induced by typhoon and seawater intrusion of 479 dangerous slopes is calculated, and the calculation formula is as follows:

[0112] ;

[0113] The calculation results show that: p l,min = 0, p l,max = 0.543.

[0114] Step S6: Calculate the landslide volume required for each dangerous slope to block the river and form a dam. The landslide volume required for each dangerous slope to block the river and form a dam V g The calculation method is shown in the following formula:

[0115] ;

[0116] In the formula: is the depth of the river bed at the slope, is the width of the river bed at the slope, c is the cohesion of the landslide, is the internal friction angle of the landslide, , b is a to-be-determined parameter.

[0117] The values of 279 dangerous slopes c , are obtained by field sampling and laboratory test, and the results show that 20.14 kPa ≤ c ≤ 25.07 kPa, 27.12° ≤ ≤ 30.65°.

[0118] The regression analysis of a , b is carried out by using the recorded 249 times of blocking river and dam disasters in China, and the results show that: , b = 0.526.

[0119] The volume of landslide mass required for landslide to block river and form dam at 479 dangerous slopes V g The results show that: V g,min = 2987 m 3 , V g,max = 116444 m 3 .

[0120] Step S7: Determine the landslide mass transport loss index of each dangerous slope The shortest distance from the front edge of 479 dangerous slopes to the river bank is counted, and the results show that: L q There are 254 dangerous slopes with a distance of ≤ 10 m, 10 m < d < 30 m for 107 dangerous slopes, L q There are 83 dangerous slopes with a distance of ≤ 50 m, and 30 < d < 50 m for 83 dangerous slopes, L q There are 35 dangerous slopes with a distance of > 50 m, i.e. 254 dangerous slopes with a distance of ≤ 10 m, L q = 1.0, 107 dangerous slopes with a distance of 10 m < d < 30 m, = 0.8, 83 dangerous slopes with a distance of 30 m < d < 50 m, = 0.6, and 35 dangerous slopes with a distance of > 50 m, = 0.4.

[0121] The average volume of landslide mass that can be flushed into the river channel at 479 dangerous slopes is calculated, and the results show that the minimum value is 1320 m 3 and the maximum value is 125427 m 3 .

[0122] Step S8: Calculate the probability of landslide blocking river and dam formation induced by typhoon and seawater backflow at each dangerous slope and give early warning. Taking typhoon "Gomei" as an example, the probability of landslide blocking river and dam formation induced by typhoon and seawater backflow at 479 slopes is calculated p cb The results show that: p cb,min = 0, p cb,max = 0.413. Since p cb,max = 0.413, the second-level early warning of landslide blocking river and dam formation induced by typhoon and seawater backflow should be started, and the early warning strategy and measures are shown in Table 3.

[0123] Table 3 Early warning strategy and measures of this embodiment

[0124] ;

[0125] ​Based on the above method, a typhoon and seawater backflow comprehensive induced landslide blocking river and dam forming prediction and early warning system along the coast of the river into the sea is developed by ArcGIS Engine platform, system data contains all spatial information and attribute information of 182 historical disaster points, 479 dangerous slopes and 9 rainfall stations, wherein, the spatial information includes river distribution direction and longitude and latitude of historical disaster points, dangerous slopes and rainfall stations, the attribute information includes river section type, k 0、 k 1、 、 、 、 、 、 、 、 、 Q l 、 Q d 、 、 、 n 3、 n 4、 、 、 、 、 、 c 、 、 a 、 b 、 、 、 p cb,max The system can realize real-time map query functions of early warning level, early warning strategy and early warning measure.

[0126] Therefore, the typhoon and seawater backflow comprehensive induced landslide blocking river and dam forming prediction and early warning method is adopted, early warning levels are divided into four levels, and the prediction and early warning system is developed based on the ArcGIS Engine platform. The feedback mechanism of landslide occurrence and background trend degree, typhoon rainfall and seawater backflow can be revealed, and the system is connected with disaster risk assessment and monitoring in front and engineering treatment measure decision-making in back, so that personnel casualties and economic losses caused by the disaster chain of typhoon, landslide, blocking river and dam forming are greatly reduced, and a perfect landslide disaster risk prevention and control system is formed.

[0127] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the present application but not to limit it, although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents, and these modifications or equivalent replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present application.

Claims

1. A typhoon and seawater backflow comprehensive induced landslide river-blocking dam- forming prediction and early warning method, characterized in that, The specific steps are as follows: Step S1: Obtain the data of the slope along the coast of the river flowing into the sea, collect the Bijie landslide data set and the southwest landslide data set, prepare the historical disaster point data set and the dangerous slope data set; Step S2: Construct a landslide background trend degree calculation model based on the FIS-YOLOv8 model, and calculate the landslide background trend degree of each dangerous slope through the trained model ; Step S3: Calculate the potential landslide volume of each dangerous slope; Step S4: Statistics of the landslide starting rainfall threshold of each dangerous slope, prediction of daily effective rainfall, determination of typhoon rainfall triggering coefficient and seawater backflow enhancement coefficient ; The calculation formula of the seawater backflow enhancement coefficient is as follows: ； wherein: is the type adjustment coefficient of the river section at the slope, is the average longitudinal slope of the river bed, is the water level rise height caused by seawater backflow at the slope, is the longitudinal slope of the river bed at the slope, is the historical extreme value of the water level rise at each slope. Step S5: Calculate the comprehensive landslide occurrence probability induced by typhoon and seawater backflow, and the calculation formula is as follows: ； wherein: is the probability of landslide occurrence induced by typhoon and seawater intrusion; Step S6: Calculate the landslide volume required for each dangerous slope to occur landslide and block river and form dam; Step S7: determining the landslide body migration loss index of each dangerous slope ; Step S8: Calculate the typhoon and seawater backflow comprehensive landslide blocking river and forming dam occurrence probability of each dangerous slope and carry out early warning.

2. The typhoon and seawater backflow induced landslide damming and damming prediction and early warning method according to claim 1, characterized in that, The slope data along the coast of the river flowing into the sea includes historical disaster point data and dangerous slope data; The historical disaster point data includes historical disaster point quantity , longitude and latitude, and satellite image; The dangerous slope data includes the number of dangerous slopes , latitude and longitude, satellite images, and river data at each slope; The river data at the dangerous slope includes river bottom longitudinal slope, river bed width, river bed depth, potential sliding direction and river angle, and average river bottom longitudinal slope; The Bijie landslide data set, the southwest landslide data set and the historical disaster point data set are combined to form a model training data set and are augmented, wherein the input value of each sample is a satellite image, the output value of a landslide disaster point is 1, and the output value of a non-landslide disaster point is 0.

3. The typhoon and seawater backflow induced landslide damming and damming prediction and early warning method according to claim 1, characterized in that, The FIS-YOLOv8 model includes a backbone network, a neck network and a head network arranged in sequence, the backbone network adopts YOLOv8, the Bottleneck structure of the C2f module in YOLOv8 is replaced by a lighter FasterNetBottleneck, the iRMB attention mechanism is introduced into YOLOv8, the iRMB-EMA fusion attention mechanism combined with the EMA attention mechanism is introduced, the CIoU loss function in YOLOv8 is replaced by the Shape-IoU loss function, and thus the FIS-YOLOv8 model is obtained. The augmented model training data set is input into the FIS-YOLOv8 model to train the FIS-YOLOv8 model, and the output value is the landslide background trend degree The satellite image of the dangerous slope is input into the trained model, and the output value is the landslide background trend degree .

4. The typhoon and seawater backflow induced landslide damming and damming prediction and early warning method according to claim 1, characterized in that, The potential landslide volume is obtained by estimating the volume of the rock-soil mass above the through sliding surface.

5. The typhoon and seawater backflow induced landslide damming and damming prediction and early warning method according to claim 1, characterized in that, The typhoon rainfall triggering factor is determined according to the difference between the predicted daily effective rainfall of each dangerous slope and the landslide starting rainfall threshold When <10mm, When ≥535mm, When , increases with the increase of and ;​​ Predicted daily effective rainfall The calculation formula is as follows: ； wherein: is the predicted daily typhoon infiltrable rainfall, is the previous daily typhoon infiltrable rainfall; and The calculation method is shown in the following formula: ； wherein: to predict typhoon rainfall, to predict typhoon rainfall, typhoon rainfall; Landslide initiation rainfall threshold The determination method is to determine The landslide initiation rainfall threshold at the historical disaster point is obtained by using the Gridding interpolation method of the spatial analysis function of ArcGIS10.2 to interpolate the above threshold value, thereby obtaining the landslide initiation rainfall threshold of each dangerous slope. The landslide starting rainfall threshold value determination method of the historical disaster point includes the following two cases: (1) If the effective rainfall on the day of the historical disaster is the highest value in the past 50 years, the second highest value of the effective rainfall in the past 50 years is determined as the landslide starting rainfall threshold value; (2) If the effective rainfall on the day of the historical disaster is not the highest value in the past 50 years, the effective rainfall on the day of the disaster is determined as the landslide starting rainfall threshold value of the disaster point.

6. The typhoon and seawater backflow induced landslide damming and damming prediction and early warning method according to claim 5, characterized in that, The calculation method of the typhoon rainfall of each day in the past 50 years includes the following two cases: (1) If there is a rainfall monitoring station within 3 km of the historical disaster point and the dangerous slope, the monitoring value is determined as the typhoon rainfall; (2) If there are no rainfall monitoring stations within 3 km of the historical disaster site and the dangerous slope, then the number of all rainfall monitoring stations within 30 km of the historical disaster site and the dangerous slope shall be determined as follows: ,statistics The timing of historical disasters is distributed as follows: On that day, statistics were collected from various rainfall monitoring stations. The total rainfall for the day was [amount missing]. Rainfall data; In 1 of the 1,000 rainfall monitoring stations was randomly selected as a reference rainfall station, and the remaining stations were corrected rainfall stations. The rainfall on each disaster day was determined for the reference rainfall station , the average rainfall of the corrected rainfall stations , and the difference between the two . - This resulted in a total of sets of data. A G1-entropy weight coupled TOPSIS model is constructed, and the elevation, slope, aspect, normalized vegetation cover index, plane curvature, profile curvature, terrain humidity index, water flow intensity index, sediment transport index, atmospheric precipitation, enhanced vegetation cover index, total primary productivity, net primary productivity and surface specific humidity of the benchmark rainfall station on the disaster occurrence day are taken as the model inputs, and the - model outputs are taken as the model outputs, and the model is trained by using a set of data to obtain the trained model; The elevation, slope, aspect, homogenized vegetation cover index, plane curvature, profile curvature, terrain moisture index, water flow intensity index, sediment transport index, atmospheric precipitation, enhanced vegetation cover index, total primary productivity, net primary productivity and surface specific humidity of any day of any dangerous slope are used as the model input, the model output is obtained based on the trained G1-entropy weight coupled TOPSIS model, and the sum of the model output and the average rainfall of each rainfall monitoring station on the day is the typhoon rainfall of the dangerous slope on the day.

7. The typhoon and seawater backflow induced landslide damming and damming prediction and early warning method according to claim 1, characterized in that, Volume of landslide mass required to block a river and create a dam The calculation method is as follows: ； wherein: V is the volume of the landslide mass required to block the river and create a dam, h is the depth of the river bed at the slope, w is the width of the river bed at the slope, c c is the cohesion of the landslide mass, φ is the internal friction angle of the landslide mass, , b is a parameter to be determined; Using existing data on river damming and dam construction , b Perform regression analysis to determine The calculation formula.

8. The typhoon and seawater backflow induced landslide damming and damming prediction and early warning method according to claim 1, characterized in that, Landslide travel loss index The shortest distance from the front of the landslide to the riverbank is related, and the calculation formula is as follows: ； In the formula: is the shortest distance from the front edge of the landslide to the river bank.

9. The typhoon and seawater backflow induced landslide damming and damming prediction and early warning method according to claim 7, characterized in that, The volume of the landslide that can flow into the river channel is set to be subject to the following formula The calculation formula of the probability of landslide dam formation induced by typhoon and seawater intrusion is as follows: ； In the formula: is the probability of landslide dam formation induced by typhoon and seawater intrusion, V is the volume of the potential landslide, is the angle between the potential sliding direction and the river.

10. The typhoon and seawater backflow induced landslide damming and damming prediction and early warning method according to claim 1, characterized in that, According to the maximum value of the probability of landslide dam formation induced by typhoon and seawater intrusion p cb,max Hierarchical early warning is carried out, and early warning strategies and measures are determined.

Images