Probability assessment method for river blockage caused by the landslide-debris flow geological disaster chain induced by post-earthquake rainfall

Abstract

The present invention discloses a method for evaluating the probability of river blocking in the earthquake-induced rainfall-triggered landslide-debris flow geological disaster chain, comprising the following steps: 1) determining the research area for constructing the landslide occurrence probability evaluation model, and dividing the research area into a plurality of non-overlapping grid cells by grid; constructing the landslide occurrence probability evaluation model; obtaining the landslide susceptibility distribution map; 2) obtaining the joint probability distribution model of the landslide occurrence times and volume in the area to be measured; 3) obtaining the debris flow migration characteristics in the area to be measured; 4) based on the rainfall intensity and distribution data, through multiple Monte Carlo simulations, obtaining the probability of river blocking of the river reach in the grid cell where the river channel is located in the area to be measured. The present invention proposes a complete evaluation process, through the comprehensive calculation of the landslide occurrence probability and the debris flow migration characteristics, combined with the Monte Carlo simulation technology, scientifically quantifying the probability of river blocking of the river reach, and providing an important support for disaster prevention and control.

Description

Technical Field

[0001] The invention relates to the technical field of geological disaster assessment, and in particular to a method for assessing the probability of a river being blocked by a landslide-mudslide geological disaster chain induced by rainfall after an earthquake. Background Art

[0002] The destruction of mountain landforms by earthquakes often causes large-scale loose deposits, providing a rich source of material for debris flows. Especially under the influence of post-earthquake rainfall, these deposits are very likely to cause debris flows, further causing disasters such as river blockage and floods, posing a serious threat to downstream residents and infrastructure. However, the current assessment methods for disaster chains are mostly focused on the analysis of a single disaster type (such as landslides or debris flows). For example, the Chinese invention application with application number CN202311853733.5 discloses a risk prediction method and related products for landslide blocking rivers, realizing the risk prediction of landslide blocking rivers. The Chinese invention application with application number CN202410329944.7 discloses a method and system for determining the blocking of rivers by debris flows, which are all for the determination of the blocking of rivers by a single disaster type, landslide or debris flow, and it is difficult to achieve a systematic assessment of the multi-linkage of regional landslides-debris flows-blocking rivers, and the assessment of the disaster chain integrating multiple disasters is difficult to quantify. Summary of the invention

[0003] In order to overcome the shortcomings of the above-mentioned technology, the purpose of the present invention is to provide a method for evaluating the probability of river blocking caused by a landslide-mudslide geological disaster chain induced by rainfall after an earthquake, to quantitatively analyze the probability of river blocking, and to fill the gap in the regional assessment of regional landslide-mudslide-river blocking multiple links in the prior art.

[0004] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0005] A method for assessing the probability of river blocking caused by a landslide-mudslide geological disaster chain induced by rainfall after an earthquake is characterized in that it comprises the following steps:

[0006] 1) Determine a study area for constructing a landslide probability assessment model, and grid the study area into a plurality of non-overlapping grid units; obtain and pre-process the historical basic data of the study area, and construct a landslide probability assessment model based on the historical basic data; based on the relevant data of the area to be tested, evaluate the landslide probability values of each grid unit in the area to be tested through the landslide probability assessment model and display them in the corresponding grid units in the area to be tested, thereby obtaining a landslide susceptibility distribution map; the study area includes the area to be tested; the historical basic data of the study area and the relevant data of the area to be tested both include rainfall intensity and distribution data;

[0007] 2) Obtain historical landslide event data within the area to be measured and preprocess it; based on the historical landslide event data, obtain the probability density functions of the number of landslides and the landslide volume distribution, and obtain the joint probability distribution model of the number of landslides and the volume.

[0008] 3) Obtain historical debris flow event data within the area to be measured; based on the historical debris flow event data, predict the debris flow migration characteristics; the debris flow migration characteristics include the debris flow migration path, the debris flow migration distance, and the intersection position of the debris flow and the river channel.

[0009] 4) Based on the landslide susceptibility distribution map, the joint probability distribution model of the number of landslides and the volume, the prediction results of the debris flow migration characteristics, the river channel terrain data of the area to be measured, and based on the rainfall intensity and distribution data, obtain the probability of river blockage in the river reach within the grid cell where the river channel is located in the area to be measured through multiple Monte Carlo simulations.

[0010] As a preferred solution, in step 1), the historical basic data of the research area includes historical data related to disaster-causing factors within the research area; the disaster-causing factors include inducing factors; the inducing factors include rainfall intensity and distribution; the preprocessing includes filling and correcting the outliers or missing values in the grid cells by using the interpolation method or the neighborhood averaging method.

[0011] Furthermore, an earthquake has occurred in the past in the area to be measured.

[0012] Furthermore, the rainfall intensity and distribution include one or more of the cumulative amount and spatio-temporal distribution characteristics of heavy rainfall, monthly rainfall, maximum monthly rainfall, and annual rainfall.

[0013] Even further, in the historical basic data, the cumulative amount of heavy rainfall is the cumulative amount and spatio-temporal distribution characteristics of the most recent or current heavy rainfall before the landslide occurs; the monthly rainfall is the monthly rainfall in the year when the landslide occurs; the maximum monthly rainfall and the annual rainfall are the maximum monthly rainfall and the annual rainfall in the year when the landslide occurs.

[0014] In the relevant data of the area to be measured, the cumulative amount of heavy rainfall uses the cumulative amount and spatio-temporal distribution characteristics of the most recent or current heavy rainfall before the evaluation time; the maximum monthly rainfall and the annual rainfall use the average values of the maximum monthly rainfall and the annual rainfall in the years after the earthquake; the monthly rainfall uses the monthly rainfall in the year of the evaluation time.

[0015] Furthermore, the disaster-causing factors also include one or more of topography and geomorphology, hydrological characteristics, geological structure, stratigraphic lithology, co-seismic landslides, post-seismic landslides, and the locations of landslides that occurred in the year before the landslide year to be evaluated.

[0016] As a preferred solution, the method for constructing the landslide occurrence probability evaluation model includes the following steps: constructing a database of landslide historical events in the study area and conducting model training.

[0017] Furthermore, the method for constructing the landslide occurrence probability evaluation model includes the following steps:

[0018] Constructing a database of landslide historical events in the study area: Based on the historical basic data of the study area, construct a database of landslide historical events, and randomly divide the data in the database of landslide historical events into a training set and a test set; the database of landslide historical events includes the location, time, scale of past landslides in the study area, and historical data related to disaster-causing factors at the location where the landslide occurred; the scale is the landslide volume;

[0019] Model training: Based on the historical basic data and the disaster-causing factors, train a machine learning model to obtain a trained model, that is, obtain a landslide occurrence probability evaluation model; in the training, the training set and test set of the database of landslide historical events are used as positive samples, and the relevant data of the locations where landslides did not occur in the historical basic data are used as negative samples;

[0020] The landslide occurrence probability evaluation model outputs the landslide occurrence probability value of each grid cell by inputting the disaster-causing factor data of each grid cell, and the value range is from 0 to 1.

[0021] Furthermore, the method for constructing the landslide occurrence probability evaluation model also includes screening key disaster-causing factors before model training; the key disaster-causing factors include rainfall intensity and distribution; the key disaster-causing factors also include the disaster-causing factors selected from other disaster-causing factors except rainfall intensity and distribution; the disaster-causing factors used in the model training are the key disaster-causing factors obtained by screening, and the trained model outputs the landslide occurrence probability value of each grid cell by inputting the key disaster-causing factor data of each grid cell, and the value range is from 0 to 1.

[0022] Even further, the method for screening the key disaster-causing factors includes the random forest method or the correlation analysis method, and uses the recursive feature elimination technique to optimize factor selection; the correlation analysis method includes the Pearson correlation coefficient method or the information gain method.

[0023] Furthermore, the machine learning model includes the random forest algorithm.

[0024] Furthermore, the model training process also includes sampling the negative samples in combination with the sample sampling method.

[0025] As a preferred solution, the landslide susceptibility distribution map includes the landslide occurrence probability values of each grid unit within the area to be measured; the landslide occurrence probability values are obtained by analyzing the relevant data of the grid units within the area to be measured by inputting them into the landslide occurrence probability assessment model; the landslide occurrence probability values vary according to the real-time data changes of rainfall intensity and distribution.

[0026] Furthermore, the relevant data of the area to be measured also includes the real-time data of other key disaster-causing factors.

[0027] As a preferred solution, the grid division process is repeated multiple times to divide the research area at different resolutions to obtain multiple grids with different resolutions; the model training also includes performing the best resolution analysis on grids with different resolutions by combining multi-scale analysis; the landslide occurrence probability assessment model inputs the disaster-causing factor data of each grid unit within the area to be measured under the best resolution grid division and outputs the landslide occurrence probability values of the corresponding grid units.

[0028] As a preferred solution, in step 2), the historical landslide event data within the area to be measured includes the landslide occurrence time, location, and volume characteristics within the area to be measured; the preprocessing includes establishing a time series dataset of landslide occurrence times and a landslide volume dataset, and filling in missing data and removing abnormal data; the time series dataset of landslide occurrence times includes the number of landslides occurring within a set unit time arranged in chronological order during the historical process; the landslide volume dataset includes the distribution range of the landslide volume characteristics of each landslide event; the volume characteristic is the natural logarithm of the actual landslide volume.

[0029] As a preferred solution, in step 2), the probability assessment method for the number of landslides includes, based on the time series dataset of landslide occurrence times, using the Poisson distribution model to fit the number of landslides occurring within a unit time and obtaining the probability and time series characteristics of the number of landslides occurring through the fitted function; the probability density function of the number of landslides occurring is as follows:

[0030] ;

[0031] In the formula, P ( N ) is the probability of N landslides occurring within a unit time; λ is the average number of landslides occurring within a unit time, obtained through fitting; N is the specific number of landslides occurring;

[0032] The probability assessment method for the landslide volume distribution includes, based on the landslide volume dataset, using the power-law distribution model to fit the distribution of landslide volume characteristics and obtaining the probability of the landslide volume characteristic distribution through the fitted function; the probability density function of the landslide volume characteristic distribution is as follows:

[0033] P ( V' ) = CV' -m

[0034] In the formula, P ([[]] V' ) is the probability that the landslide volume feature is V' ; m is the power-law exponent, which controls the steepness of the distribution and is obtained by fitting; C is the normalization constant to ensure that the probability density integral is 1; V' is the volume feature, which is the natural logarithm of the actual landslide volume;

[0035] The joint probability distribution model of the number of landslides and volume is as follows:

[0036] P ([[]] N, V' ) = P ([[]] N ) P ([[]] V' )

[0037] In the formula, P ([[]] N, V' ) is the joint probability of the number of landslides and volume; P ([[]] N ) is the probability of N landslides occurring per unit time; P ([[]] V' ) is the probability that the landslide volume feature is V' .

[0038] Furthermore, in the probability density function of the number of landslides, the average number of landslides λ occurring per unit time is obtained by fitting with the maximum likelihood estimation method.

[0039] Furthermore, in the probability density function of the landslide volume distribution, the power-law exponent m is obtained by linear fitting with the least squares method.

[0040] As a preferred solution, in step 3), the historical debris flow event data includes the source volume, transportation distance, movement path of the historical debris flow event, and the elevation difference between the landslide trigger point and the adjacent river; the source volume is the landslide volume in the corresponding historical landslide event.

[0041] As a preferred solution, in step 3), the prediction method of the debris flow transportation characteristics includes:

[0042] Debris flow migration path prediction: Calculate the slope of each grid cell in the area to be measured using DEM; Define landslide trigger points in the area to be measured in the GIS environment and use them as the initial positions of debris flows; Based on the slope of each grid cell in the area to be measured, make the debris flow move along the steepest direction to obtain the migration path of the debris flow;

[0043] Debris flow migration distance calculation: Based on historical debris flow event data, establish an empirical formula between landslide volume, elevation difference, and migration distance:

[0044] ;

[0045] In the formula, L is the maximum migration distance; V is the source volume; H is the elevation difference between the landslide trigger point and the adjacent river; a , b and c are parameters obtained by fitting using historical debris flow event data.

[0046] Furthermore, the parameters a , b and c are obtained through regression analysis by the least squares method.

[0047] As an optimal solution, in step 4), the Monte Carlo simulation process includes the following steps:

[0048] Sample according to the landslide susceptibility distribution map of the area to be measured, and extract the grid cells in the grid as the locations where the landslide trigger points are located;

[0049] According to the joint probability distribution model of the number of landslides and volume, randomly assign the number of landslide events and the corresponding volume sizes;

[0050] Based on the grid cell where the landslide trigger point is located, the number of landslides and the volume, simulate the migration path and migration distance of the debris flow;

[0051] Based on the river channel terrain data, determine whether the migration path of the debris flow under each landslide event crosses the river channel, and determine the grid cell where the crossing is located;

[0052] Sample the landslide susceptibility distribution map multiple times and simulate the migration path and migration distance of the debris flow;

[0053] Statistically count the number of times the debris flow crosses the grid cells where the river channel is located during multiple simulation processes N block , and the total number of simulations N total , and calculate the probability of blocking the river in the river section within a single grid cellP block :

[0054] 。

[0055] Furthermore, step 4) further includes: based on the real-time data of rainfall intensity and distribution, dynamically updating the landslide susceptibility distribution map, so as to dynamically update the probability of river blockage in each grid cell.

[0056] Even further, the number of simulation times in the Monte Carlo simulation process is at least 10,000 times.

[0057] Even further, the method of sampling according to the landslide susceptibility distribution map of the area to be measured is as follows: after normalizing the landslide occurrence probability of the grid in the landslide susceptibility distribution map, using the landslide occurrence probability of the normalized grid cell as the sampling probability to sample the grid cell as the location where the landslide trigger point is located; according to the joint probability distribution model of the number of landslide occurrences and volume, randomly assigning the number of landslide events and the corresponding volume size, including normalizing all the obtained joint probability data of the number of landslide occurrences and volume, and using the joint probability of the number of landslide occurrences and volume after normalization as the sampling probability to select the number of landslide events and the corresponding volume size. Setting the sampling probability for sampling, the results are more reasonable and scientific.

[0058] Even further, during the Monte Carlo simulation process, random perturbations are added to the landslide occurrence location, occurrence times, volume distribution, and debris flow path.

[0059] As a preferred solution, step 4) further includes: counting the probability of river blockage in each grid cell to obtain the river blockage probability interval; counting the river width in the area to be measured to obtain the river width interval; dividing the river blockage probability interval and the river width interval into multiple groups for grading; overlaying the grading results of the river blockage probability and the river width with the river section geographic information to generate the final river section river blockage probability grade distribution map.

[0060] Furthermore, the grading includes five levels, which are extremely high, high, medium, low, and extremely low in order of numerical value.

[0061] Even further, the grading uses the natural break method.

[0062] The method of the present invention can combine real-time rainfall data to dynamically update the landslide susceptibility distribution and debris flow path simulation results, ensuring the timeliness and accuracy of the river-blocking probability. Analyze the accumulation effect at the intersection of the debris flow and the upstream river section, and consider the impact of upstream river-blocking events on the water flow and debris flow path in the downstream river section. Introduce the influence coefficient of the river width on the debris flow accumulation behavior. For example, a wider river section may reduce the river-blocking probability. Normalize the river width of the area to be measured to a value in the same interval as the evaluated river-blocking probability, and divide these two sets of numerical values into 5 groups using the natural break method, and classify them into extremely high, high, medium, low, and extremely low levels according to the numerical value. Using the GIS platform, overlay the river-blocking probability and the river section geographic information by the matrix method to generate the final river-blocking probability level distribution map, providing an intuitive reference for decision-makers.

[0063] The present invention also provides an evaluation system for the river-blocking probability of the earthquake-induced rainfall-induced landslide-debris flow geological disaster chain. The special feature is that it implements the above-mentioned evaluation method for the river-blocking probability of the earthquake-induced rainfall-induced landslide-debris flow geological disaster chain, including:

[0064] Grid cell division module, used to divide the research area into grids;

[0065] Landslide occurrence probability evaluation model construction module, used to construct a landslide occurrence probability evaluation model based on the historical basic data of the research area;

[0066] Landslide susceptibility distribution map acquisition module, used to evaluate the landslide occurrence probability values of each grid cell in the area to be measured through the landslide occurrence probability evaluation model based on the relevant data of the area to be measured and display them in the corresponding grid cells, so as to obtain the landslide susceptibility distribution map;

[0067] Joint probability distribution model acquisition module for the number and volume of landslides, used to obtain the joint probability distribution model of the number and volume of landslides based on the historical landslide event data of the area to be measured;

[0068] Debris flow migration characteristics acquisition module, used to predict the debris flow migration characteristics based on the historical debris flow event data of the area to be measured;

[0069] River-blocking probability acquisition module, used to obtain the river-blocking probability of each grid cell containing a river section in the area to be measured through multiple Monte Carlo simulations based on the landslide susceptibility distribution map, the joint probability distribution model of the number and volume of landslides, the predicted results of the debris flow migration characteristics, the river channel terrain data of the area to be measured, and the rainfall intensity and distribution data.

[0070] Furthermore, the landslide occurrence probability evaluation model construction module is also used to determine the optimal grid resolution; the grid cells in the landslide susceptibility distribution map acquisition module and the river-blocking probability acquisition module are the grid cells of the grids divided by the optimal resolution.

[0071] The present invention also provides a computer program product, comprising computer instructions, characterized in that: the computer instructions are used to cause a computer to execute the above-mentioned method for evaluating the probability of river blockage in the earthquake-induced rainfall-triggered landslide-debris flow geological disaster chain.

[0072] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0073] The present invention proposes a complete evaluation process. Through the comprehensive calculation of the landslide occurrence probability and debris flow migration characteristics, combined with the Monte Carlo simulation technology, the river blockage probability of the river section is scientifically quantified, providing important support for disaster prevention and control.

[0074] (1) Realize the systematic evaluation of the entire process of the disaster chain

[0075] Through the multi-link quantitative analysis of the landslide occurrence probability, landslide volume distribution, debris flow migration distance, and river blockage probability, the present invention realizes the full-process modeling and evaluation of the landslide-debris flow-river blockage disaster chain for the first time. Compared with the traditional single-disaster evaluation method, the present invention more comprehensively reflects the dynamic characteristics of the disaster chain and the causal relationship between its multi-links, providing reliable technical support for disaster risk analysis in complex geological environments.

[0076] (2) Improve the accuracy and applicability of the model

[0077] The present invention introduces the random forest algorithm to construct a landslide susceptibility model, accurately identifies the key control factors of landslides in combination with historical data, and significantly improves the accuracy of landslide occurrence probability evaluation. The power-law distribution model is used to describe the landslide volume probability distribution, which can capture the occurrence law of extreme events (large-volume landslides) and enhance the robustness of the model. Based on the digital elevation model (DEM) and the steepest path algorithm to simulate the debris flow migration path, the spatial diffusion process after debris flow triggering is truly restored, especially suitable for complex terrain environments.

[0078] (3) Achieve an innovative breakthrough in the quantitative analysis of the disaster chain

[0079] The Monte Carlo simulation technology is used to handle the uncertainties of each link of the disaster chain, and the river blockage probability of the river section is calculated through large-scale random sampling, significantly improving the scientificity and credibility of the evaluation results. The river blockage probability distribution map of the present invention clearly and intuitively shows the river blockage risks of different river sections, providing an important reference basis for regional disaster prevention and mitigation planning.

[0080] (4) Dynamically predict the evolution process of the disaster chain

[0081] The present invention can adjust the simulation parameters of the landslide-debris flow migration path according to dynamic data such as real-time rainfall intensity and geological conditions, and update the evaluation results of the river-blocking probability in real time. This dynamic response ability provides effective support for early warning and real-time decision-making of disasters, and can help managers quickly identify potential risks and formulate corresponding emergency plans.

[0082] (5) Improve the management efficiency of the disaster chain

[0083] The method of the present invention takes into account both technical accuracy and computational efficiency. By constructing step-by-step models and iterating the results of multiple links, the demand for computing resources is reduced, which is suitable for disaster assessment in large-scale regions. Utilizing the integration characteristics of the GIS platform, combining the landslide susceptibility distribution map, debris flow migration path, and river-blocking probability distribution map output by the model with the regional infrastructure layout provides a visualization tool for the comprehensive management of the disaster chain, significantly improving the disaster management efficiency.

[0084] (6) Provide technical support for disaster reduction planning and emergency response

[0085] The present invention can not only identify high-risk areas, but also quantify the risk levels of each link in the disaster chain, providing a scientific basis for disaster reduction planning. For example: Planning the location of reservoirs or dams: According to the river-blocking probability distribution, optimize the siting of water conservancy facilities to reduce the risk of debris flow blocking the river course. Optimize disaster prevention measures: Such as through landslide treatment and debris flow interception projects to reduce the possibility of triggering the disaster chain. Assist in emergency response decision-making: During the occurrence of a disaster, a list of threatened areas can be quickly generated to help relevant departments accurately locate high-risk areas and optimize the scheduling and allocation of rescue resources.

[0086] (7) Strong technical applicability and high promotion value

[0087] The technical methods adopted by the present invention (such as random forest, power-law distribution fitting, DEM simulation, and Monte Carlo technique) have high generality and applicability, and can be flexibly adjusted according to the geological characteristics, rainfall conditions, and disaster types of different regions.

[0088] (8) Make up for the deficiencies of existing technologies

[0089] Traditional methods mostly focus on the analysis of single disasters (such as landslides or debris flows), and fail to reflect the coupling characteristics of the disaster chain. The present invention makes up for this defect through comprehensive modeling. Existing river-blocking probability assessments mostly use empirical methods, which are difficult to quantify the uncertainties between multiple links. The present invention solves the problem of strong subjectivity of traditional empirical methods through numerical simulation and probability analysis, and the evaluation results are more scientific. Brief description of the drawings

[0090] Figure 1 It is a grid map generated by dividing the research area at a resolution of 150 meters in the embodiment of the present invention;

[0091] Figure 2 This is the evolution diagram of key post-earthquake disaster-causing factors in the embodiments of the present invention;

[0092] Figure 3 This is the probability distribution map of landslide occurrence in area X in the embodiments of the present invention;

[0093] Figure 4 This is the probability distribution map of landslide occurrence volume intervals in area X in the embodiments of the present invention;

[0094] Figure 5 This is the schematic diagram of the downward migration of the steepest path of debris flow in the embodiments of the present invention;

[0095] Figure 6 This is the probability grade distribution map of river blocking in each river section of area X in the embodiments of the present invention. Detailed implementation manners

[0096] To better explain the present invention, the main content of the present invention is further clarified below in conjunction with the accompanying drawings and specific embodiments, but the content of the present invention is not limited to the following embodiments only.

[0097] A method for evaluating the probability of river blocking in a post-earthquake rainfall-induced landslide-debris flow geological disaster chain of the present invention includes the following steps:

[0098] 1) Evaluation of the probability of landslide occurrence

[0099] 1.1) Determine the research area for constructing the landslide occurrence probability evaluation model, and divide the research area into a plurality of non-overlapping grid cells by grid; obtain the historical basic data of the research area and perform preprocessing.

[0100] The historical basic data includes the disaster-causing factors in the research area and their historical data. The disaster-causing factors include inducing factors, and also include one or more of topographic features, hydrological characteristics, geological structures, stratigraphic lithologies, co-seismic landslides, post-earthquake landslides, the location of landslides occurring in the previous year of the landslide year to be evaluated, etc. The historical data is obtained through remote sensing images, field surveys, and literature records.

[0101] The inducing factors are rainfall intensity and distribution, including the cumulative amount and spatio-temporal distribution characteristics of the most recent or current heavy rainfall before the landslide occurrence, the monthly rainfall, the maximum monthly rainfall, and the annual rainfall in the year of the landslide occurrence. The topographic features include one or more of the elevation, slope, aspect, curvature, topographic position index, topographic wetness index, land use, soil moisture, normalized difference vegetation index, vegetation coverage rate, etc. in the research area. The hydrological characteristics include one or more of the river network, river power index, fault distribution, etc. The stratigraphic lithology includes different lithologies, which are used to measure the physical and mechanical properties of the strata such as strength, porosity, etc.

[0102] Among them, the slope and aspect are extracted from the digital elevation model (DEM). The fault distribution includes the spatial distribution of the fault zone and its relative position to the slope surface. The river network includes the river distribution, and different river distributions have different weakening effects on slope stability. The vegetation coverage rate can characterize the influence of vegetation roots on soil reinforcement. The optional range of the research area is the global range or the area to be measured. If the global range is selected as the research area, the obtained landslide occurrence probability assessment model has universality. However, if the area to be measured is selected, the assessment result of the landslide occurrence probability assessment model for the area to be measured has higher accuracy.

[0103] Gridification: The research area is divided into multiple grid units and standardized in a gridified form. Each grid unit represents a geological unit.

[0104] The number of grids is determined according to the size of the research area. If the research area is the global scope, the research area is divided into at least 1000×1000 grids. If the research area is within a province, the research area is divided into at least 30×30 grids.

[0105] The research area is divided into multiple different grids with different resolutions through multiple gridifications, so as to determine the optimal resolution through multi-scale analysis in the subsequent steps and improve the model prediction accuracy. Using the terrain elevation data, the research area is cut into multiple grids with different sizes, so as to select a suitable scale for analysis according to different requirements (such as accuracy, efficiency) in the subsequent steps. For example: The research area is cut into three networks with different resolutions. The resolutions of the three networks are 10 meters, 30 meters, and 50 meters respectively. Through model training, the optimal resolution is screened as 30 meters, and then the grid with a resolution of 30 meters is selected for the subsequent evaluation process.

[0106] Preprocessing: The outliers or missing values in the grid units are filled and corrected by interpolation or neighborhood averaging method.

[0107] 1.2) Based on the historical basic data of the research area, construct a landslide historical event database; and randomly divide the data in the landslide historical event database into a training set and a test set for model training and verification.

[0108] The landslide historical event database includes the location, time, scale of past landslides in the research area, and the relevant historical data of the disaster-causing factors at the location where the landslides occurred.

[0109] 1.3) Obtain the key disaster-causing factors.

[0110] The key disaster-causing factors include rainfall intensity and distribution. Rainfall intensity and distribution are the most critical disaster-causing factors for landslides and debris flows.

[0111] The key disaster-causing factors also include other disaster-causing factors obtained through screening. The screening process includes using the random forest method to screen among other disaster-causing factors except rainfall intensity and distribution; or, using correlation analysis to evaluate the correlation between other disaster-causing factors except rainfall intensity and distribution and landslide occurrence, screening out the disaster-causing factors with the greatest impact on landslide occurrence, such as slope, lithology, etc., and using the recursive feature elimination (RFE) technique to optimize factor selection and avoid the problem of multicollinearity. The correlation analysis methods include the Pearson correlation coefficient method or the information gain method, and the first few factors with a large gap in contribution are screened, which is specifically determined according to the actual situation.

[0112] 1.4) Based on historical basic data and key disaster-causing factors, construct a machine learning model - a landslide occurrence probability assessment model.

[0113] Adopt the Random Forest (RF) algorithm. Use the key disaster-causing factor data in the above training set and test set as positive samples, and the key disaster-causing factor data at the locations where landslides did not occur in the historical basic data as negative samples. Combine sample sampling and multi-scale analysis of resolution to train the model, construct a classification model, and use the cross-validation method to optimize the model hyperparameters to improve the model accuracy. The trained model is the landslide occurrence probability assessment model. The model hyperparameters include the maximum depth of the decision tree, the number of split nodes, etc. The sampling method includes the Bootstrap sampling method to sample the negative samples. The multi-scale analysis performs iterative analysis on the resolution to obtain the optimal resolution.

[0114] For the landslide occurrence probability assessment model, the input parameter is the key disaster-causing factor of the grid cell at the optimal resolution, and the output parameter is the landslide occurrence probability value of the grid cell, with a value range of 0 to 1.

[0115] Random forest has strong non-linear modeling ability and can handle the relationship between complex landslide disaster-causing factors and landslide occurrence probability. The algorithm has good anti-overfitting ability and is suitable for high-dimensional data analysis.

[0116] Use the ROC curve to evaluate the classification performance of the model, and use the AUC (area under the curve) value to quantify the overall prediction ability of the model. Use the confusion matrix to analyze the accuracy, recall rate, and F1 score of landslide prediction to verify the reliability and applicability of the model results.

[0117] 1.5) Based on the relevant data of the area to be measured, generate a landslide susceptibility distribution map.

[0118] Input the real-time data of the key disaster-causing factors in the grid cells of the area to be measured into the landslide occurrence probability assessment model trained in step 1.4), output the landslide occurrence probability value of each grid cell, and input the landslide occurrence probability value into the grid to draw the landslide susceptibility distribution map.

[0119] The relevant data includes the data of the key disaster-causing factors of the current state of the area to be measured. Among them, for the key disaster-causing factors input into the landslide occurrence probability assessment model, the rainfall intensity and distribution include the cumulative amount and spatio-temporal distribution characteristics of the most recent or current heavy rainfall before the assessment time, the average value of the maximum monthly rainfall and the average annual rainfall in the years after the earthquake, and one or more of the monthly rainfall amounts in the year of the assessment time.

[0120] In step 1), the landslide occurrence probability assessment process has the following advantages:

[0121] (1) Application of the random forest algorithm

[0122] The random forest constructs multiple decision trees and votes to obtain the final classification result, overcoming the limitations of a single decision tree. The importance scoring function inside the algorithm can quantify the contributions of various control factors, providing a scientific basis for landslide mechanism analysis. The Bootstrap sampling method is used to handle the problem of uneven distribution of landslide event data, ensuring the robustness of the model.

[0123] (2) Spatial resolution optimization

[0124] Combined with the digital elevation model (DEM), the study area is divided into detailed grids (such as 10-meter or 30-meter resolution) to ensure the spatial accuracy of the landslide susceptibility assessment results. Multiscale analysis is introduced in model training to evaluate the impact of different resolutions on the prediction accuracy and select the best resolution as the model input.

[0125] (3) Temporal dynamic analysis

[0126] The dynamic data of rainfall events is introduced, and the landslide occurrence probability value is dynamically adjusted according to the cumulative effect of rainfall duration and intensity, improving the model's prediction ability for landslide triggering under extreme weather conditions.

[0127] (4) Model verification and performance evaluation

[0128] The ROC curve is used to evaluate the classification performance of the model, and the AUC (area under the curve) value is used to quantify the overall prediction ability of the model. The confusion matrix is used to analyze the accuracy, recall rate, and F1 score of landslide prediction to verify the reliability and applicability of the model results.

[0129] 2) Assessment of the number and volume of landslides:

[0130] Landslide occurrence frequency and volume assessment is one of the key steps of the present invention. The main purpose is to estimate the number of possible landslides and their volume distribution characteristics in the future in the tested area based on the historical data of the tested area. Through statistical analysis and probability distribution fitting of the historical data of landslides in the tested area, accurate parameters can be provided, laying the foundation for subsequent debris flow migration distance simulation and river blocking probability calculation. The specific steps are as follows:

[0131] 2.1) Data collection and preprocessing

[0132] Obtain records of historical landslide events in the area to be tested, mainly including parameters such as landslide occurrence time, location, volume and scale, and establish landslide frequency time series dataset and landslide volume dataset.

[0133] The landslide count time series dataset includes the number of landslides in a set unit time in the historical process, which is arranged in chronological order. The set unit time is annual or quarterly. The landslide volume dataset includes the distribution range of landslide volume characteristics in all landslide events, as well as the maximum, minimum and mean values of all landslide volume characteristics in the dataset, to construct a landslide volume characteristic dataset. V' is the original landslide volume V The natural logarithm of the data is used to reduce the impact of extreme values and make the data more consistent with statistical laws. The data sources include remote sensing images, field surveys, monitoring system records and existing geological disaster databases.

[0134] Preprocessing: missing data are supplemented by inferring the landslide characteristics of the adjacent areas to ensure data integrity. Statistical methods such as the interquartile range method are used to remove outliers in the data set.

[0135] 2.2) Probability assessment of landslide occurrence

[0136] The Poisson distribution model is used to fit the number of landslides per unit time, and the number of landslides is predicted through the fitted function. Poisson distribution is widely used to describe the number of random events per unit time, and its probability density function is:

[0137] (Formula 1)

[0138] P ( N ) is the time unit N The probability of a landslide; λ is the average number of landslides per unit time (Poisson intensity); N is the specific number of landslide occurrences. The parameter is calculated by the maximum likelihood estimation (MLE) method. λ , which is the average occurrence rate of landslides.

[0139] Fit the frequency of landslide events per unit time to determine the temporal characteristics of landslide occurrences. Use the actual number of landslide occurrences data to verify the fitting effect of the Poisson distribution model, and use the chi-square test to evaluate the significance of the model. If the landslide occurrence time shows obvious seasonal variations, a sine function can be further introduced to correct the Poisson distribution to make it more consistent with the periodic changes in rainfall. The formula is as follows:

[0140] λ = λ 0(1 + lsin ( wt + φ )) (Equation 2)

[0141] In the formula, λ is the average number of landslides per unit time (Poisson intensity); λ 0 is the average occurrence rate of the Poisson distribution without seasonal influence; t is the time variable, which can be expressed in days, months, or years; w is the frequency of the sine function, which controls the speed of periodic changes, such as the relationship with the rainfall cycle; φ is the phase of the sine function, which determines the starting point of seasonal changes; l is the amplitude of the sine function, which represents the intensity of seasonal influence.

[0142] 2.3) Probability assessment of landslide volume distribution

[0143] Based on the above landslide volume dataset, a power-law distribution model is used to fit the landslide volume characteristic distribution. Research shows that landslide events in nature mostly follow a power-law distribution, and its probability density function is:

[0144] P ( V' ) = CV' -m (Equation 3)

[0145] P ( V' ) is the probability that the landslide volume characteristic is V' , V' = lnV , V is the landslide volume; m is the power-law exponent, which controls the steepness of the distribution; C is the normalization constant to ensure that the probability density integral is 1. After taking the logarithm of the landslide volume data, linear fitting is performed using the least squares method to obtain the parameters m .

[0146] 2.4) Joint distribution assessment of landslide frequency and volume

[0147] Under the assumption that the number of landslides and the landslide volume are independent, the joint probability is the product of the probabilities of the two:

[0148] P ( N, V' ) = P ( N ) P ( V' ) (Equation 4)

[0149] According to the joint probability distribution model, the number and volume of future landslides are simulated by random sampling to provide input parameters for the subsequent simulation of debris flow migration distance. The Monte Carlo method is used to generate multiple sets of sample data to form possible combinations of the number and volume of landslides.

[0150] Technological innovation and details of this step:

[0151] (1) Combination of seasonality and randomness

[0152] The traditional Poisson distribution assumes that the intensity of event occurrence is constant, but seasonal factors such as rainfall will have a significant impact on the occurrence frequency of landslides. By introducing a modified Poisson distribution, the present invention incorporates the seasonal fluctuations of landslide events into the model, significantly improving the timeliness and accuracy of prediction.

[0153] (2) Joint distribution modeling

[0154] The joint distribution analysis of the number and volume of landslides provides rich input data for landslide-debris flow risk assessment, and at the same time can capture the statistical relationship between the two, laying a foundation for the physical authenticity of the model.

[0155] 3) Simulation of debris flow migration distance

[0156] The simulation of debris flow migration distance is one of the core technical links of the present invention. Its purpose is to accurately predict the migration path and influence range after debris flow is triggered by analyzing the volume of landslide source materials, terrain elevation difference, and historical debris flow movement characteristics. This link provides key spatial parameter support for the assessment of the probability of river blockage. The specific steps are as follows:

[0157] 3.1) Data preparation and feature extraction

[0158] Through on-site investigation, literature records, remote sensing images, etc., collect the volume of source materials, migration distance, movement path of historical debris flow events, and the elevation difference between the landslide trigger point and the adjacent river in the area to be measured.

[0159] Migration distance: Use GIS technology to obtain high-resolution digital elevation data (such as 10-meter or 30-meter resolution) of the area to be measured to provide a terrain basis for the simulation of the migration path. This resolution is the best resolution in step 1).

[0160] Movement path: Extract the location and elevation information of the river network distribution and the vegetation coverage data.

[0161] Source volume: The collected landslide event data provides the mass of the source material, which can be directly used as an input parameter for debris flow simulation.

[0162] Elevation difference: Use the DEM to calculate the elevation difference between the landslide trigger point and the target area; the target area refers to the river where the debris flow reaches.

[0163] 3.2) Determination of debris flow migration path

[0164] The debris flow migration path is driven by gravity, and its path selection usually moves along the steepest slope direction. The path simulation method based on the steepest path algorithm includes: using the DEM to calculate the slope of each grid to simulate the initial flow direction of the debris flow; determining the landslide trigger point in the GIS environment and taking it as the initial position of the debris flow; generating slope data according to the DEM; iteratively updating the migration path of the debris flow along the steepest slope direction until the path crosses the river channel or reaches a flat area.

[0165] 3.3) Calculation of debris flow migration distance

[0166] Based on the historical debris flow event data, the empirical formula between the landslide volume, elevation difference and migration distance is established as follows:

[0167] (Equation 5)

[0168] L is the maximum migration distance, that is, the farthest distance that the debris flow migrates; V is the source volume, that is, the landslide volume in step 2), H is the elevation difference between the landslide trigger point and the target area; a 、 b and c are parameters, and the least squares method is used to perform regression analysis on the historical data to fit the best parameters of the empirical formula.

[0169] 4) Blockage probability assessment:

[0170] The blockage probability assessment is the final link of the present invention. By quantitatively analyzing the possibility of the movement path of the debris flow crossing the river channel, it provides a scientific basis for disaster risk area division and emergency response. The core of the blockage probability assessment is based on the previous landslide and debris flow simulation results, using the Monte Carlo method to simulate multiple landslide-debris flow events, and counting the occurrence frequency of river blockage in the river section, so as to calculate the probability distribution of river blockage in the river section. The specific steps are as follows:

[0171] 4.1) Input data preparation

[0172] Landslide susceptibility distribution map: Generated by the landslide occurrence probability assessment, which marks the possibility distribution of landslide events in the area and provides the probability basis for the landslide trigger point.

[0173] Joint probability distribution model of landslide occurrence times and volume: Characterize the probability of landslide occurrence times and corresponding volume, and determine the source characteristics of different landslide events.

[0174] Simulation results of debris flow migration path: Include the debris flow movement path, migration distance, and intersection position with the river channel for each simulation.

[0175] River channel terrain data: Generate the river channel network and its profile characteristics based on a high-resolution digital elevation model (DEM), and describe the width and height of the river reach.

[0176] 4.2) Monte Carlo simulation process

[0177] According to the landslide susceptibility distribution map, determine the grid cell division resolution of the area to be measured. Sample based on the landslide occurrence probability of each grid cell, and select the location of the landslide trigger point. Among them, the sampling probability of each grid cell, that is, the probability that each grid cell is selected as the landslide trigger point, is the probability obtained by normalizing the landslide occurrence probability within the grid cell. The specific acquisition method is as follows: Normalize the landslide occurrence probability values within the grid; the value obtained by normalizing each grid cell is the sampling probability of each grid cell.

[0178] According to the joint probability distribution model of landslide occurrence times and volume, randomly assign the number of landslide events and the corresponding volume size according to the joint probability of landslide occurrence times and volume (after normalizing the data of landslide occurrence times and volume obtained from the joint probability distribution model of landslide occurrence times and volume, use the normalized data as the sampling probability for sampling).

[0179] Use the debris flow migration model to calculate the debris flow migration distance of each landslide event. The migration path starts from the grid cell where the landslide trigger point is located and moves along the steepest slope direction around the grid cell, while considering the influence of landslide volume, elevation difference, and friction along the way, and dynamically adjusts the flow range of the debris flow.

[0180] Combine the simulated debris flow path with the river channel terrain data to determine whether the debris flow path intersects with the river channel, and determine the grid cell where the intersection is located.

[0181] Count the number of times the debris flow intersects with the grid cells where the river channel is located, that is, the number of river-blocking events N block , and the total number of simulations N total . Calculate the river-blocking probability P block :

[0182] (Equation 6)

[0183] Pblock is the probability of river blockage in a single grid unit; N block is the number of river blocking events in a single grid unit; N total is the total number of Monte Carlo simulations, preferably not less than 10,000 times.

[0184] 4.3) Model optimization and dynamic adjustment

[0185] Random disturbances are added to the location, number of occurrences, volume distribution and debris flow paths of landslides. Combined with real-time rainfall data, the landslide susceptibility distribution and debris flow path simulation results are dynamically updated to ensure the real-time and accuracy of the probability of river blockage. The accumulation effect of the intersection of debris flow and upstream river sections is analyzed, and the impact of upstream river blockage events on the flow and debris flow path of downstream river sections is considered. The influence coefficient of river section width on the accumulation behavior of debris flow is introduced. For example, a wider river section may reduce the probability of river blockage, while a steeper river section may accelerate the scouring of deposits. Using the GIS platform, the probability of river blockage is superimposed on the geographic information of the river section to generate a risk distribution map, providing an intuitive reference for decision makers.

[0186] The specific operation method is as follows:

[0187] Count the probability of river blockage in each grid unit and obtain the probability interval of river blockage;

[0188] Count the width of rivers in the area to be tested and obtain the river width interval;

[0189] The river blocking probability interval and river width interval are divided into five groups using the natural break method for classification, namely very high, high, medium, low and very low.

[0190] The river blocking probability and river width classification results are superimposed on the river section geographic information to generate the final river section blocking probability distribution map on the GIS platform;

[0191] Based on the real-time data of rainfall intensity and distribution, the landslide susceptibility distribution map is dynamically updated, thereby dynamically updating the probability of river blockage in each grid unit and the final river blockage probability level distribution map.

[0192] Technical innovation and details of this step:

[0193] (1) Multi-source data fusion

[0194] By integrating landslide probability distribution, landslide volume model, debris flow path simulation and river terrain characteristics, a multi-dimensional probability analysis framework is constructed to effectively improve the accuracy of river blockage probability assessment.

[0195] (2) Dynamically adjusted Monte Carlo simulation

[0196] Ensure the statistical stability and credibility of the model results through large-scale sampling (such as more than 10,000 simulations).

[0197] (3)Refined analysis of the river-blocking mechanism

[0198] Considering the multiple effects of the movement characteristics of debris flows and river reach characteristics (such as width, slope, vegetation coverage rate) on river-blocking behavior, the rough assumptions of traditional empirical methods are improved.

[0199] The above method can be implemented through a GIS platform.

[0200] The present invention also provides a system for evaluating the probability of river blocking in a landslide-debris flow geological disaster chain induced by post-earthquake rainfall, which implements the above method for evaluating the probability of river blocking in a landslide-debris flow geological disaster chain induced by post-earthquake rainfall, including:[[]]

[0201] A grid cell division module for dividing the study area into grids;

[0202] A landslide occurrence probability evaluation model construction module for determining the optimal grid resolution based on the historical basic data of the study area and constructing a landslide occurrence probability evaluation model;

[0203] A landslide susceptibility distribution map acquisition module for evaluating the landslide occurrence probability values of each grid cell at the optimal grid resolution of the area to be measured through the landslide occurrence probability evaluation model based on the relevant data of the area to be measured and displaying them in the corresponding grid cells, so as to obtain a landslide susceptibility distribution map;

[0204] A joint probability distribution model acquisition module for the number and volume of landslide occurrences, which is used to obtain a joint probability distribution model for the number and volume of landslide occurrences based on the historical landslide event data of the area to be measured;

[0205] A debris flow migration characteristic acquisition module for predicting the debris flow migration characteristics based on the historical debris flow event data of the area to be measured;

[0206] A river-blocking probability acquisition module for obtaining the river-blocking probability of each grid cell containing a river reach at the optimal grid resolution in the area to be measured through multiple Monte Carlo simulations based on the landslide susceptibility distribution map, the joint probability distribution model for the number and volume of landslide occurrences, the predicted results of debris flow migration characteristics, the river channel terrain data of the area to be measured, and the rainfall intensity and distribution data.

[0207] The following takes the X area after an earthquake as an example to evaluate the probability of river blocking in a landslide-debris flow geological disaster chain induced by post-earthquake rainfall. The day of the 15th year after the earthquake in the X area is selected as the prediction time to verify the method of the present invention, that is, the data of the area to be measured at the prediction time is the relevant data (real-time data) of the area to be measured in the method of the present invention.

[0208] 1) Select the area to be measured, i.e., Area X, as the research area, and divide Area X into multiple grid cells through grid division; obtain the historical basic data of each grid cell in Area X, and perform data filling and correction. The grid division is carried out at different resolutions, including 30 meters, 50 meters, 80 meters, 100 meters, and 150 meters, and Area X is divided into multiple grids with different resolutions, as shown in Figure 1 shown.

[0209] In this embodiment, the random forest method is used to analyze the important characteristics related to landslides in Area X. The disaster-causing factors included in the analysis are elevation, slope, aspect, curvature, terrain position index, stratigraphic lithology, terrain humidity index, river network, river power index, fault distribution, location of co-seismic landslides, location of landslides occurred in the previous year, maximum monthly rainfall, annual rainfall, land use, soil humidity, and normalized vegetation index. The analysis results are shown in Figure 2 shown. In the years after the earthquake, the importance of elevation, aspect, normalized vegetation index, location of co-seismic landslides, and location of landslides in the previous year of the evaluated landslide year plays a dominant role. With the evolution of time, by the prediction time, soil humidity, rainfall intensity and distribution dominate, that is, the key disaster-causing factors only remain soil humidity, rainfall intensity and distribution.

[0210] This embodiment uses all the disaster-causing factors including elevation, slope, aspect, curvature, terrain position index, stratigraphic lithology, terrain humidity index, river network, river power index, fault distribution, location of co-seismic landslides, location of landslides occurred in the previous year, maximum monthly rainfall, annual rainfall, land use, soil humidity, and normalized vegetation index to construct a landslide historical event database and construct a landslide occurrence probability evaluation model. The random forest algorithm is adopted, and the sampling method includes the Bootstrap sampling method. Through the multi-scale analysis of the resolution, it is determined that the resolution of 30 meters is the best resolution, which can better obtain the ground terrain feature information of Area X. The cross-validation method is used to optimize the model hyperparameters. After optimization, the maximum depth and the number of splitting nodes of the decision tree are 24. The AUC of the trained model is 0.92, meeting the training requirements of the classification model. The landslide occurrence probability evaluation model inputs the data of all the disaster-causing factors at the best resolution and outputs the landslide occurrence probability value of each grid cell, and the value range is from 0 to 1.

[0211] Generate a landslide susceptibility distribution map: Obtain the data of all the disaster-causing factors of each grid cell in Area X at the prediction time, input the above landslide occurrence probability evaluation model, the model outputs the landslide occurrence probability value, and present the landslide occurrence probability value of the grid in each grid to obtain the landslide susceptibility distribution map as shown in Figure 3 shown. In the figure, the area within the red dotted line represents Area X.

[0212] 2) Evaluation of the number and volume of landslides

[0213] Obtain the records of historical landslide events in area X, mainly including parameters such as the occurrence time, location, volume, and scale of the landslides, and establish a time series dataset of the number of landslides and a dataset of landslide volumes.

[0214] Use the Poisson distribution model to solve the probability of the number of landslides based on the time series dataset of the number of landslides. The parameter λ is 2. Then the probabilities of the landslide occurring 1 time, 2 times, 3 times, 4 times, and 5 times are respectively: 0.27, 0.27, 0.18, 0.09, 0.03. Adopt the power-law distribution model to fit the landslide volume distribution based on the landslide volume dataset. The probability distributions of each volume interval obtained by fitting are as Figure 4 shown. Then the joint probability distribution model function of the number of landslides and volume is the multiplication of the probabilities of different numbers and the probabilities of different volume intervals.

[0215] 3) Debris flow migration distance simulation

[0216] Debris flow migration path: Calculate the slope of each grid using DEM. The debris flow migration path starts from the selected grid and flows to the grid with the lowest elevation among the eight grids around the grid, as Figure 5 shown.

[0217] Debris flow migration distance: Based on the historical debris flow event data, establish an empirical formula between the source volume, elevation difference, and migration distance. Through fitting with historical event data, the relevant parameters a, b, and c are obtained as 0.28, 0.22, and 0.93 respectively. Then the calculation formula for the debris flow migration distance is:

[0218]

[0219] 4) Blockage probability assessment of the river

[0220] Input the landslide susceptibility distribution map, and randomly select landslide trigger points based on the landslide occurrence probability of each grid cell;

[0221] Input the joint probability distribution model of the number of landslides and volume, and randomly assign a landslide volume based on the grid cell where the landslide starting point is located;

[0222] Input the simulation results of the debris flow migration path, and obtain the debris flow movement path and migration distance based on the landslide volume and the elevation of each grid cell;

[0223] Input the river channel terrain data, describe the width, height, and basin characteristics of the river section, and determine whether the debris flow migration path intersects with the river channel.

[0224] Through 10,000 Monte Carlo simulations of landslide - debris flow events, count the number of times the debris flow intersects with the river section in each grid cell, and calculate the blockage probability of the river section in each grid cell.

[0225] The two sets of numerical values of the river width and the river blocking probability in Region X are divided into five groups using the natural break method, and are classified into extremely high, high, medium, low, and extremely low levels according to the numerical values. Using the GIS platform, the matrix method is used to overlay the river blocking probability and the river section geographical information to generate the final river section river blocking probability grade distribution map, providing an intuitive reference for decision-makers. As Figure 6 shown, the extremely high, high, medium, low, and extremely low levels are displayed as red, orange, yellow, light green, and dark green respectively.

[0226] The prediction results show that the river blocking probability of the river section intersecting with the debris flow gully mouth is the highest. By comparing the river blocking probability and the actual river blocking events in the past 15 years after the earthquake, more than 90% of the events occurred in the high river blocking probability grade interval, proving the accuracy of the prediction results, and there is still a 21% river section in Region X with a relatively high river blocking risk in the future.

[0227] Other parts not described are all prior arts.

Claims

1. A method for evaluating the probability of river blockage in a landslide-debris flow geological disaster chain induced by post-earthquake rainfall, characterized in that, Including the following steps: 1) Determine the research area for constructing the landslide occurrence probability assessment model, and divide the research area into multiple non-overlapping grid cells by grid; obtain the historical basic data of the research area, and construct the landslide occurrence probability assessment model; based on the relevant data of the area to be measured, evaluate the landslide occurrence probability values of each grid cell in the area to be measured through the landslide occurrence probability assessment model and display them in the corresponding grid cells of the area to be measured, so as to obtain the landslide susceptibility distribution map; the research area includes the area to be measured; the historical basic data of the research area and the relevant data of the area to be measured both include rainfall intensity and distribution data; 2) Obtain the historical landslide event data in the area to be measured; based on the historical landslide event data, obtain the probability density functions of the number of landslide occurrences and the landslide volume distribution, and obtain the joint probability distribution model of the number of landslide occurrences and the volume; 3) Obtain the historical debris flow event data in the area to be measured; based on the historical debris flow event data, predict the debris flow migration characteristics; the debris flow migration characteristics include the debris flow migration path, the debris flow migration distance, and the crossing position of the debris flow and the river channel; 4) According to the landslide susceptibility distribution map, the joint probability distribution model of the number of landslide occurrences and the volume, the prediction results of the debris flow migration characteristics, and the river channel topographic data of the area to be measured, based on the rainfall intensity and distribution data, obtain the river blocking probability of the river section in the grid cell where the river is located in the area to be measured through multiple Monte Carlo simulations.

2. The method for evaluating the probability of river blockage in the earthquake-induced rainfall-triggered landslide-debris flow geological disaster chain according to claim 1, wherein In step 1), the historical basic data of the research area includes the historical data of disaster-causing factors in the research area; the disaster-causing factors include inducing factors; the inducing factors include rainfall intensity and distribution.

3. The probability assessment method for river blocking by earthquake-induced rainfall-triggered landslide-debris flow geological disaster chain according to claim 2, wherein The construction method of the landslide occurrence probability assessment model includes the following steps: Construct a landslide historical event database in the research area: based on the historical basic data of the research area, construct a landslide historical event database, and randomly divide the data in the landslide historical event database into a training set and a test set; the landslide historical event database includes the locations, times, scales of past landslides in the research area, and the historical data of disaster-causing factors at the locations where the landslides are located; Model training: Based on the historical basic data and disaster-causing factors, train the machine learning model to obtain the trained model, that is, obtain the landslide occurrence probability assessment model; in the training, the training set and test set of the landslide historical event database are used as positive samples, and the relevant data of the locations where no landslides occurred in the historical basic data are used as negative samples; The landslide occurrence probability assessment model outputs the landslide occurrence probability value of each grid cell by inputting the disaster-causing factor data of each grid cell, and the value range is from 0 to 1.

4. The probability assessment method for river blocking by the earthquake-induced rainfall-triggered landslide-debris flow geological disaster chain according to claim 3, wherein, The method for constructing the landslide probability assessment model also includes screening key disaster factors before model training; the key disaster factors include rainfall intensity and distribution; the key disaster factors also include disaster factors screened out from other disaster factors except rainfall intensity and distribution; the disaster factors used in the model training are the screened key disaster factors, and the trained model outputs the landslide probability value of each grid unit by inputting the key disaster factor data of the grid unit, and the value range is 0 to 1.

5. The method for evaluating the probability of river blocking caused by the earthquake-induced rainfall-induced landslide-debris flow geological disaster chain according to claim 3, characterized in that The gridding process is repeated multiple times to divide the study area into different resolutions to obtain multiple grids with different resolutions; the model training also includes performing optimal resolution analysis on grids with different resolutions in combination with multi-scale analysis; the landslide occurrence probability assessment model inputs the disaster factor data of each grid unit in the test area under the optimal resolution grid division, and outputs the landslide occurrence probability value of the corresponding grid unit.

6. The probability assessment method for river blocking of the earthquake-induced rainfall-triggered landslide-debris flow geological disaster chain according to claim 1, wherein In step 2), the probability assessment method of the number of landslide occurrences includes fitting the number of landslide occurrences per unit time using a Poisson distribution model, and obtaining the probability and time series characteristics of the number of landslide occurrences through the fitted function; The probability density function of the number of landslide occurrences is as follows: ； Wherein, P ( N ) is the probability of N landslides occurring per unit time; λ is the average number of landslides occurring per unit time, obtained by fitting; N is the specific number of landslides occurring; The probability evaluation method of the landslide volume distribution includes fitting the landslide volume characteristic distribution by using a power law distribution model, and obtaining the probability of the landslide volume characteristic distribution by the fitted function; the probability density function of the landslide volume characteristic distribution is as follows: P （ V' ） =CV' -m ； In the formula, P ( V' ) is the probability that the landslide volume characteristic is V' ; m is the power-law exponent, which controls the steepness of the distribution and is obtained by fitting; C is the normalization constant to ensure that the probability density integral is 1; V' is the natural logarithm of the actual landslide volume; The joint probability distribution model of landslide number and volume is as follows: P （ N, V' ）= P （ N ） P （ V' ）； wherein, P ( N, V' ) is the joint probability of landslide frequency and volume; P ( N ) is the probability of N landslides occurring per unit time; P ( V' ) is the probability that the landslide volume feature is V' .

7. The method for evaluating the probability of river blockage in the earthquake-induced rainfall-triggered landslide-debris flow geological disaster chain according to claim 1, wherein In step 3), the method for predicting debris flow migration characteristics includes: Prediction of debris flow migration path: Use DEM to calculate the slope of each grid unit in the test area; define the landslide trigger point in the test area in the GIS environment, and use the landslide trigger point as the initial position of the debris flow; based on the slope of each grid unit in the test area, make the debris flow move in the steepest direction, so as to obtain the movement path of the debris flow; Calculation of debris flow migration distance: Based on historical debris flow event data, an empirical formula between landslide volume, height difference and migration distance is established: ； In the formula, L is the maximum migration distance; V is the provenance volume; H is the elevation difference between the landslide trigger point and the adjacent river; a , b and c are parameters obtained by fitting using historical debris flow event data.

8. The method for evaluating the probability of river blockage in the earthquake-induced rainfall-triggered landslide-debris flow geological disaster chain according to claim 1, wherein, Step 4), the Monte Carlo simulation process includes the following steps: Sampling is performed according to the landslide susceptibility distribution map of the area to be tested, and the grid cells in the grid are extracted as the locations of the landslide triggering points; According to the joint probability distribution model of landslide occurrence frequency and volume, the number of landslide events and the corresponding volume size are randomly assigned; Based on the grid cells where the landslide trigger point is located, the number and volume of landslides, the migration path and distance of debris flow are simulated; Based on the river topography data, determine whether the migration path of the debris flow under each landslide event intersects the river channel, and determine the grid unit where the intersection is located; Sampling the landslide susceptibility distribution map multiple times and simulating the debris flow migration path and migration distance; Count the number of times debris flow intersects with each grid cell where the river channel is located during multiple simulation processes N block , as well as the total number of simulations N total , and calculate the probability of river blockage in the river section within a single grid cell P block : 。 9. The probability assessment method for river blocking by the earthquake-induced rainfall-triggered landslide-debris flow geological disaster chain according to any one of claims 1 to 8, characterized in that, Step 4) further includes: statistically analyzing the river blocking probability of the river sections within each grid cell to obtain a river blocking probability interval; statistically analyzing the river width within the area to be measured to obtain a river width interval; dividing the river blocking probability interval and the river width interval into multiple groups for grading; and overlaying the grading results of the river blocking probability and the river width with the river section geographic information to generate a final river section river blocking probability grade distribution map.

10. A probability assessment system for river blocking caused by the landslide-debris flow geological disaster chain induced by post-earthquake rainfall, characterized in that, Implementing the post-earthquake rainfall-induced landslide-debris flow geological disaster chain river blocking probability assessment method according to any one of claims 1 to 9, including: A grid cell division module, configured to divide the research area into a grid pattern. A landslide occurrence probability assessment model construction module, configured to construct a landslide occurrence probability assessment model based on the historical basic data of the research area. A landslide susceptibility distribution map acquisition module, configured to, based on the relevant data of the area to be measured, evaluate the landslide occurrence probability values of each grid cell in the area to be measured through the landslide occurrence probability assessment model and display them within the corresponding grid cells, thereby obtaining a landslide susceptibility distribution map. A joint probability distribution model acquisition module for the number and volume of landslides, configured to obtain a joint probability distribution model for the number and volume of landslides based on the historical landslide event data of the area to be measured. A debris flow migration characteristic acquisition module, configured to predict the debris flow migration characteristics based on the historical debris flow event data of the area to be measured. A river blocking probability acquisition module, configured to, based on the landslide susceptibility distribution map, the joint probability distribution model for the number and volume of landslides, the prediction result of the debris flow migration characteristics, the river channel terrain data of the area to be measured, and the rainfall intensity and distribution data, obtain the river blocking probability of each grid cell containing a river section in the area to be measured through multiple Monte Carlo simulations.

Images