A typhoon rainstorm landslide warning method considering extreme rainfall and environmental variables

Abstract

The present invention relates to a typhoon rainstorm landslide warning method considering extreme rainfall and environmental variables. For the first time, the influence of different rainfall return periods is considered in landslide risk warning, and the coupling analysis of the topographic factors (DEM) of landslides and the time probability (return period) of rainfall is carried out. Based on the physical model FSLAM model, the influence of both antecedent rainfall and extreme rainfall on landslide stability is considered. The response sensitivity of grids to two different conditions of antecedent rainfall and extreme rainfall is analyzed based on the landslide risk judgment matrix, and the influence of unconditionally stable and unconditionally unstable regions is considered, so as to realize landslide warning considering rainfall infiltration on multiple time scales. It can carry out regional-scale landslide risk warning more reasonably and effectively, and play a certain guiding role in the disaster prevention and mitigation work of typhoon rainstorm-induced landslides.

Description

Technical Field

[0001] The present invention belongs to the field of engineering geology research, and particularly relates to a typhoon rainstorm landslide early warning method considering extreme rainfall and environmental variables. Background Art

[0002] Landslides are one of the most common types of natural disasters, causing a large number of casualties and property losses every year, and posing a huge threat to natural landscapes and infrastructure construction. In China, more than 70% of the territory is mountainous and hilly areas, and the population is dense. Therefore, the losses caused by frequent landslide disasters are often incalculable. According to the statistics of the Ministry of Natural Resources of China, a total of 4,772 landslides occurred across the country in 2021, resulting in 80 deaths, 11 missing persons, and direct economic losses of 3.2 billion yuan. Among them, 4,403 landslides were induced by rainfall, accounting for about 92% of the total number of landslides. Thus, rainfall is one of the important factors inducing landslides. For the southeast coastal areas of China, landslides are closely related to extreme rainfall caused by typhoons. On average, 9 typhoons land in the southeast coastal areas of China every year, including Zhejiang, Fujian, Guangdong, Taiwan and other places. Typhoon rainstorms usually induce "cluster-style" occurrence of landslides, and due to small pre-deformation signs and sudden outbreaks, the disaster-causing effect is particularly significant. Therefore, how to effectively carry out the risk assessment and reasonable early warning of typhoon rainstorm-induced landslides plays a very important role.

[0003] In current research, the risk assessment and early warning of rainstorm-induced landslides are mainly carried out by coupling landslide spatial susceptibility and critical rainfall thresholds. Landslide susceptibility (i.e., the spatial probability of landslide occurrence) is mainly reflected in the establishment of the non-linear relationship between landslides and their internal geological influencing factors. The process mainly calculates "where" landslides are likely to occur by statistically analyzing the correlation between the locations of historical landslides and different geological factors. The critical rainfall threshold of landslides is mainly reflected in the establishment of the non-linear relationship between landslides and their external triggering factors (mainly rainfall). By calculating the occurrence situation (or probability) of landslides under different rainfall amounts, it determines "what rainfall amount (or rainfall intensity) can induce landslides". Huang et al. (Huang Faming et al., Regional rainfall-induced landslide hazard warning based on landslide susceptibility mapping and a critical rainfall threshold) respectively used machine learning models and effective rainfall models to evaluate the spatial and temporal probabilities of historical landslide occurrences in Xunwu County, Jiangxi Province, and then carried out landslide risk early warning. However, only the critical rainfall threshold was considered, and the return periods of rainfall on different time scales were not taken into account. Lee et al. (Lee et al., Temporal prediction modeling for rainfall-induced shallow landslide hazards using extreme value distribution) selected the antecedent rainfall in the previous 72 hours as an indicator to evaluate the landslide risk in Jinbu City, South Korea under the rainfall return periods of 20 - 200 years. However, the effective infiltration of cumulative antecedent rainfall and the influence of environmental variables were not considered, and the research object was not typhoon rainstorm-induced landslides.

[0004] Landslides induced by typhoon rainstorms are mainly affected by two different time-scale rainfall conditions: one is the rainfall before the typhoon rainstorm, i.e., the antecedent rainfall condition, which provides initial conditions such as soil moisture and pore water pressure for the instability and failure of landslides. This condition generally has a time scale of one month to several months. The other is the typhoon rainstorm condition (which can also be called the extreme rainfall condition), and the time scale involved is usually several hours to several days, providing critical conditions for the final occurrence of landslides. However, the existing methods have deficiencies in comprehensively considering the rainfall infiltration involving different time scales (cumulative antecedent rainfall and extreme rainfall), and there is no effective method to consider the influence of environmental variables such as temperature and evapotranspiration on rainfall infiltration.

[0005] To solve the above problems, the present invention provides a typhoon rainstorm landslide warning method considering extreme rainfall and environmental variables. By introducing a physical model for evaluating landslide failure probability and a landslide risk judgment matrix, rainfall infiltration at different time scales is incorporated into landslide risk warning. The research results are of great significance for spatio-temporal warning of landslides under future climate change conditions. Summary of the Invention

[0006] To solve the above technical problems, the present invention provides a typhoon rainstorm landslide warning method considering extreme rainfall and environmental variables, which solves the drawback that current warnings of typhoon rainstorm-induced landslides cannot simultaneously consider antecedent rainfall infiltration and extreme rainfall conditions, and adds the influence of environmental variables to the method, making the technical method process more reasonable and closer to the actual situation.

[0007] Specifically, the technical solution adopted by the present invention is as follows:

[0008] A typhoon rainstorm landslide warning method considering extreme rainfall and environmental variables, comprising the following:

[0009] S1. Statistically analyze the antecedent rainfall P in the two months before the occurrence of landslides in the current area, a and statistically analyze the typhoon rainstorm amount P that induces the occurrence of landslides. e The typhoon rainstorm amount is also called the extreme rainfall amount.

[0010] Use the daily actual atmospheric rainfall and various environmental variables during the two months before the occurrence of landslides as input data to calculate the antecedent rainfall effective infiltration amount q in the EASY_BAL software. a The unit is mm / d.

[0011] The environmental variables include the daily maximum temperature, minimum temperature, average temperature, and evapotranspiration.

[0012] Based on the physical model FSLAM model, calculate the landslide risk index of typhoon rainstorm-induced landslides, and calibrate the FSLAM model parameters using the antecedent rainfall and typhoon rainstorm amounts during the two months before the occurrence of landslides to obtain the calibrated FSLAM model.

[0013] S2. Obtain the daily actual atmospheric rainfall and various environmental variables within the study area over the years, and use the EASY_BAL software to calculate the antecedent rainfall P a and the antecedent rainfall effective infiltration amount q a for each month; at the same time, statistically analyze the maximum rainfall for three consecutive days in each month, and regard the maximum rainfall for three consecutive days in each month as the extreme rainfall amount for that month.

[0014] Set the recurrence period duration, and use the effective infiltration of the previous rainfall in each month as the input to obtain the effective infiltration value of the previous rainfall in the study area under the recurrence period by using the Gumbel distribution calculation formula; use the monthly extreme rainfall as the input to obtain the extreme rainfall in the study area under the recurrence period by using the calculation formula of the Gumbel distribution.

[0015] S3. Input the effective infiltration q a of the previous rainfall in the study area and the extreme rainfall P e obtained in S2 into the calibrated FSLAM model in S1, and use the calibrated FSLAM model to calculate the landslide risk indices under three working conditions respectively: Working condition 1: without considering rainfall; Working condition 2: only considering the effective infiltration q a of the previous rainfall; Working condition 3: considering both the effective infiltration q a of the previous rainfall and the extreme rainfall P e ;

[0016] Calculate the difference in risk indices for each grid between Working condition 2 and Working condition 1, and the difference in risk indices for each grid between Working condition 3 and Working condition 2 based on the ArcGIS software. The difference in risk indices in the first case reflects the sensitivity of the grid to the previous rainfall response, and the difference in risk indices in the second case reflects the sensitivity of the grid to the extreme rainfall response; count the distribution of the number of grids with the difference in risk indices in both cases. According to the changing trend of the number of grids, divide each case into four different levels, namely, small response to cumulative previous rainfall, moderate response to cumulative previous rainfall, large response to cumulative previous rainfall, very large response to cumulative previous rainfall, small response to extreme rainfall, moderate response to extreme rainfall, large response to extreme rainfall, and very large response to extreme rainfall. Construct a landslide risk judgment matrix with the different degrees of response to extreme rainfall infiltration and previous rainfall infiltration as rows and columns. The landslide risk indices corresponding to the rows and columns are in the landslide risk judgment matrix. Then, based on the landslide risk judgment matrix, divide the risk warning levels for each grid to obtain the initial landslide risk warning level map for the whole area.

[0017] S4. Considering the terrain factor at the same time, use the calibrated FSLAM model to calculate the probabilities of unconditional instability and unconditional stability of the grids in the study area respectively. Define the area with an unconditional instability probability greater than 0.8 as the unconditional instability area, and the area with an unconditional stability probability greater than 0.8 as the unconditional stability area. Define the grids located in these areas as the extremely high risk area and the extremely low risk area respectively, and mark the landslide risk indices formed by considering the terrain factor in these grids.

[0018] S5. Superimpose the extremely high-risk areas and extremely low-risk areas divided considering topographic factors on the results of the landslide risk warning levels divided using the landslide risk judgment matrix in S3 to obtain the final landslide risk warning level map for the whole region. When superimposing, use the "Erase" tool in ArcGIS. First, erase the grids in the initial landslide risk warning level map for the whole region obtained in S3 that coincide with the unconditional unstable areas and unconditional stable areas in S4. Then, use the "Merge" tool to merge the map obtained after erasing with the unconditional unstable areas and unconditional stable areas, and update the landslide risk index to the corresponding landslide risk index formed considering topographic factors to obtain the final landslide risk warning level map for the whole region, which is used for typhoon and rainstorm landslide warnings.

[0019] In the present invention, there is no requirement for the execution order of S3 and S4. It is also possible to execute S4 first and then S3.

[0020] In a further optimized solution, the FSLAM model is a physical model that can quickly evaluate the risk of typhoon and rainstorm-induced landslides. It includes two basic parts, namely a geotechnical model based on the infinite slope theory and a simplified groundwater model. Its stability coefficient FS is calculated by the following formula:

[0021]

[0022] In the formula, C r is the apparent cohesion generated by soil roots, C s is the effective cohesion of the soil, g is the acceleration due to gravity, ρ s is the saturated soil density, z is the soil depth, a is the upstream catchment area, b is the grid resolution, K is the soil layer permeability coefficient, θ is the slope angle, ρ w is the density of water, is the angle of internal friction.

[0023] The FSLAM model considers the uncertainty of parameters by using a stochastic method. When the input parameters have a certain statistical distribution form, the calculated stability coefficient will also inherit this statistical property and have its own probability distribution. At this time, the landslide risk index can be calculated. During the iterative calculation process, the proportion of the number of times the stability coefficient is less than 1 in the total number of times is defined as the landslide risk index. The parameters related to the distribution of the stability coefficient are determined by the following formulas:

[0024]

[0025]

[0026]

[0027]

[0028] Among them, μ FS and μ C are respectively the average values of the FS, C (either C r or C s of one kind) distribution, σ FS and σ C are respectively the standard deviations of the FS distribution and the C (either C r or C s of one kind) distribution, h is the groundwater level height, z is the soil depth, γ w and γ s are respectively the specific weights of water and soil, A and D have no specific meanings and are respectively expressed by formulas (4) and (5).

[0029] The input data of the FSLAM model includes 5 raster files and 2 text files. The raster files are respectively the digital elevation model DEM, soil, land use, effective infiltration of previous rainfall, typhoon rainstorm (extreme rainfall). The first text file contains the root cohesion values of each land use type and the corresponding runoff curve numbers, and the other text file contains the physical and mechanical parameter values (K, z, C s , ρ s , z, ) of each specific soil.

[0030] For the further optimization scheme, the Gumbel distribution can describe the probability of extreme events occurring in a specific area during a specific period, and is used to evaluate the probability of one or more rainfall events exceeding a certain specific rainfall threshold within a specified time. Its formula is as follows:

[0031] F(x) = e -e(f) (6)

[0032]

[0033] Among them, x is a random variable representing the size of the extreme event, μ is a location parameter representing the distribution average value, σ is a scale parameter representing the distribution variability, f has no specific meaning and is expressed by formula (7), and F represents a function of the variable x.

[0034] The landslide risk judgment matrix is divided into five risk levels, namely extremely low landslide risk, low landslide risk, medium landslide risk, high landslide risk and extremely high landslide risk, as shown in Table 1.

[0035] Table 1 Landslide risk judgment matrix

[0036]

[0037] Compared with the prior art, the beneficial effects of the present invention are:

[0038] (1) Typhoon rainstorm-induced landslides are affected by the combined action of antecedent rainfall and extreme rainfall. Since the infiltration of these two types of rainfall has different time scales, it is difficult to conduct a coupled analysis. Based on the physical model FSLAM model, the present invention simultaneously considers the effects of antecedent rainfall and extreme rainfall on landslide stability. Based on the landslide risk judgment matrix, the sensitivity of the grid to the two different conditions of antecedent rainfall and extreme rainfall is analyzed, and the influence of unconditional stable and unconditional unstable regions is considered, thereby realizing landslide warning considering multi-time scale rainfall infiltration.

[0039] (2) Existing technical methods cannot consider the influence of environmental variables on rainfall infiltration. The present invention uses EASY_BAL software to calculate the effective infiltration amount of antecedent rainfall considering environmental variables including temperature and evapotranspiration, which can more reasonably and effectively carry out regional-scale landslide risk warning and play a guiding role in the disaster prevention and mitigation work of typhoon rainstorm-induced landslides.

[0040] (3) The influence of different rainfall return periods is considered in landslide risk warning for the first time, and the coupling analysis of the topographic factors (DEM) of the landslide and the time probability (return period) of rainfall is carried out. The present invention comprehensively considers the spatial and time probabilities of the occurrence of typhoon rainstorm-induced landslides, quantitatively evaluates and warns the landslide risk under the influence of multi-time scale rainfall infiltration, which is a very beneficial supplement to the research of such technical methods. Description of the Drawings

[0041] Figure 1 Schematic diagram of multi-time scale rainfall infiltration of typhoon rainstorm-induced landslides.

[0042] Figure 2 Effective infiltration amount of antecedent rainfall under different return periods.

[0043] Figure 3 Extreme rainfall amounts under different return periods.

[0044] Figure 4 q under 50-year return period a Relationship diagram between the number of grids and the risk change amount under the condition.

[0045] Figure 5 P under 50-year return period e Relationship diagram between the number of grids and the risk change amount under the condition. Detailed Implementation Modes

[0046] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the drawings and embodiments. Of course, the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0047] A typhoon rainstorm landslide warning method considering extreme rainfall and environmental variables according to the present invention includes the following steps:

[0048] S1. Statistically analyze the antecedent rainfall (P a ) in the two months before the landslide occurred in the current area, and statistically analyze the typhoon rainstorm amount (which can also be called extreme rainfall amount) (P e ) that induced the landslide. The data of both can be obtained through a meteorological station;

[0049] Use the daily actual atmospheric rainfall and various environmental variables during the two months before the landslide occurred as input data to calculate the antecedent rainfall effective infiltration amount q a (unit: mm / d) in the EASY_BAL software;

[0050] The environmental variables include the daily maximum temperature, minimum temperature, average temperature, and evapotranspiration amount;

[0051] Considering two types of rainfall infiltration, namely antecedent rainfall and typhoon rainstorm amount, during the two months before the landslide occurred, calculate the landslide risk index of typhoon rainstorm-induced landslides based on the physical model FSLAM model. Use the landslide risk index to produce a risk distribution map for landslide warning. The higher the landslide risk index, the higher the warning level. Conduct accuracy analysis on the obtained landslide risk index results to calibrate the FSLAM model parameters and obtain the calibrated FSLAM model;

[0052] The FSLAM model parameters include the cohesion C s of the rock and soil mass, the internal friction angle soil depth z, soil permeability coefficient K, soil porosity n, saturated soil density ρ s additional apparent cohesion C r generated by soil roots, and the runoff curve number CN, etc.;

[0053] The accuracy analysis process is as follows: Use the area AUC value under the receiver operating characteristic curve (ROC) to measure the quality of the calculation results. If the accuracy is greater than 70%, it indicates that the parameter selection is reasonable; if the accuracy is less than 70%, the parameters need to be adjusted and recalculated, and the iterative calculation is performed until the accuracy is higher than 70%. The finally obtained parameters are the accurate calculation parameters calibrated for this landslide. In the receiver operating characteristic curve (ROC), the horizontal axis is FPR, the vertical axis is TPR, FPR is the false positive rate, and TPR is the true positive rate.

[0054] S2. Obtain the daily actual atmospheric rainfall and various environmental variables within the study area over the years. The environmental variables are the same as those in S1, including the maximum temperature, minimum temperature, average temperature, evapotranspiration amount, etc. Use the EASY_BAL software to statistically analyze the antecedent rainfall P a of each month and calculate the antecedent rainfall effective infiltration amount qa Meanwhile, count the maximum rainfall over three consecutive days each month, and regard the maximum rainfall over three consecutive days in each month as the extreme rainfall in that month.

[0055] Then, based on the effective infiltration amount q of the previous rainfall each month a and the maximum rainfall over three consecutive days in each month, use the calculation formula of the Gumbel distribution developed by Matlab software to calculate the effective infiltration amount q of the previous rainfall in the study area under different return periods a and the extreme rainfall P e . Specifically: Set the return period duration, obtain the effective infiltration amount of the previous rainfall in the study area under the return period with the effective infiltration amount of the previous rainfall each month as the input, and obtain the extreme rainfall in the study area under the return period with the monthly extreme rainfall as the input; The return period is the impact of the rainfall event probability, and different return periods refer to once in 5 years, once in 50 years, or once in 200 years, etc. The probability of once in 5 years is 1 / 5 = 0.2, and the probability of once in 100 years is 1 / 100 = 0.01.

[0056] Taking the calculation of the effective infiltration amount of the previous rainfall in each month under different return periods by the Gumbel distribution as an example, the user sets the return period duration to be calculated and needs to count the effective infiltration amount of the previous rainfall each month. Arrange these data from largest to smallest and input them into the calculation formula of the Gumbel distribution in turn. For example, if the user wants to calculate the rainfall once in 200 years at most, set the duration to 1 - 200, and then calculate 200 numbers arranged from smallest to largest, which are the effective infiltration amounts of the previous rainfall in each year from 1 to 200 years return period. The calculation of the extreme rainfall under different return periods is the same.

[0057] S3. Input the data of the effective infiltration amount q of the previous rainfall in the study area under the return period obtained in S2 a and the extreme rainfall P e into the calibrated FSLAM model in S1, and use the calibrated FSLAM model to calculate the landslide risks under three working conditions respectively: Working condition 1: Without considering rainfall, Working condition 2: Only considering the effective infiltration amount of the previous rainfall (q a ), Working condition 3: Considering both the effective infiltration amount of the previous rainfall (q a ) and the typhoon rainstorm amount (P e ).

[0058] Calculate the difference in risk index for each grid between Case 2 and Case 1, and the difference in risk index for each grid between Case 3 and Case 2 using ArcGIS software. The difference in risk index in the first case reflects the sensitivity of the grid to the response of previous rainfall, and the difference in risk index in the second case reflects the sensitivity of the grid to the response of extreme rainfall. Statistically analyze the distribution of the number of grids with respect to the difference in risk index in both cases. According to the changing trend of the number of grids, each case is divided into 4 different levels, namely small response to cumulative previous rainfall, moderate response to cumulative previous rainfall, large response to cumulative previous rainfall, very large response to cumulative previous rainfall, small response to extreme rainfall, moderate response to extreme rainfall, large response to extreme rainfall, and very large response to extreme rainfall. Construct a landslide risk judgment matrix with the different degrees of response to extreme rainfall infiltration and previous rainfall infiltration as rows and columns. The landslide risk index in the landslide risk judgment matrix corresponds to the row and column. Then, based on the landslide risk judgment matrix, divide the risk warning level for each grid to obtain the initial landslide risk warning level map for the entire region. In the warning level map, set different colors according to the warning level to make it clearer.

[0059] S4. Considering the terrain factor, use the calibrated FSLAM model to calculate the probabilities of unconditional instability and unconditional stability for the grids in the study area respectively. Define the area with an unconditional instability probability greater than 0.8 as the unconditional instability area, and the area with an unconditional stability probability greater than 0.8 as the unconditional stability area. Define the grids located in these areas as the extremely high risk area and the extremely low risk area respectively, and mark the landslide risk index formed considering the terrain factor within these grids.

[0060] S5. On the basis of S4, overlay the results of the landslide risk levels divided using the landslide risk judgment matrix in step S3 to obtain the final landslide risk warning level map for the entire region. When overlaying, use the "Erase" tool in ArcGIS. First, erase the grids that coincide with the unconditional instability area and the unconditional stability area in the initial landslide risk warning level map for the entire region obtained in S3. Then, use the "Merge" tool to merge the map obtained after erasing with the unconditional instability area and the unconditional stability area, update the landslide risk index to the corresponding landslide risk index formed considering the terrain factor, and keep the landslide risk index in step S3 for the rest that do not coincide. Obtain the final landslide risk warning level map for the entire region to achieve typhoon and rainstorm landslide warning.

[0061] Furthermore, the calculation formula of the Gumbel distribution is the formula under the first type of extreme value distribution (k = 0).

[0062] Embodiment

[0063] (1) Calculate the effective infiltration amount of the antecedent rainfall. Based on the physical model, the FSLAM model, calculate the landslide risk index for typhoon rainstorm-induced landslides, and then use the ROC curve for accuracy analysis to calibrate the FSLAM model parameters. Landslides induced by typhoon rainstorms usually show the result of the coupling of antecedent rainfall and extreme rainfall ( Figure 1 ).

[0064] The example is located in the southwestern part of Wenzhou City, Zhejiang Province, between Wencheng County, Ruian City, Pingyang County, and Taishun County. In 2016, Typhoon "Megi" induced hundreds of shallow landslides in the local area. Long-term monitoring data of the meteorological station in the study area (27°47′N, 120°39′E, H = 39.7 m) from 1981 to 2015 were obtained from the local meteorological department. Input the daily atmospheric rainfall and daily data of environmental variables (the maximum, minimum, average of daily temperature, and daily evapotranspiration) in the two months before the landslide occurred into the EASY_BAL software to calculate the effective infiltration amount of the antecedent rainfall in these two months, and the effective infiltration amount of the cumulative antecedent rainfall is 0.75 mm / d.

[0065] The DEM of the study area is from the Geospatial Data Cloud (https: / / www.gscloud.cn / ); the land use types include 13 categories, namely forest land, sparse forest, other forest land, shrubs, high-coverage grassland, medium-coverage grassland, low-coverage grassland, paddy field, dry land, bare land, town, rural area, water; the soil types include 13 types, namely red soil, yellow soil, yellow-red soil, paddy soil, gley paddy soil, percolated paddy soil, acidic purple soil, flooded paddy soil, grey tide soil, brown soil, dark yellow-brown soil, leached chernozem, cinnamon soil; the effective infiltration amount of the antecedent rainfall adopts the calculated numerical results, and a unified value is adopted for the whole region; the extreme rainfall adopts the rainfall raster file of Typhoon "Megi" that induced the landslide. For the parameter values in the two text files, first determine their initial values according to expert experience and literature, and then input all the data into the FSLAM model for multiple iterative calculations. Conduct ROC curve accuracy analysis on the landslide risk results of the study area obtained after calculation. The historical landslide data used in this study is in the form of landslide surfaces. Therefore, extract the maximum landslide risk probability of each landslide surface. The AUC value of the final calculation result is 70.1%, indicating that the existing parameters can accurately reflect the actual situation of the study area. Therefore, the parameter values used in the FSLAM model are determined, and the calibrated FSLAM model is obtained.

[0066] (2) Use the calculation formula of the Gumbel distribution developed by Matlab software to calculate the effective infiltration amount of the antecedent rainfall and extreme rainfall in the study area under different return periods.

[0067] Using the EASY_BAL software and meteorological station monitoring data, the effective infiltration amount of the previous month's rainfall in the study area from 1981 to 2015 was calculated. At the same time, the maximum value of the continuous 3-day rainfall per month during this period was statistically analyzed. Since typhoon rainstorms generally last for 2-3 days, the maximum value of the continuous 3-day rainfall per month can be regarded as the extreme rainfall of that month. Through the calculation code of the Gumbel distribution developed by the Matlab software, the effective infiltration amount of the previous rainfall (q a ) and extreme rainfall (P e ) in the study area under the return periods of 2-100 years were calculated respectively. The calculation results are as shown in Figure 2 and Figure 3 . When the return period is 50 years, the effective infiltration amount of the previous rainfall is 7.3 mm / d, and the extreme rainfall is 415 mm; when the return period increases to 100 years, the effective infiltration amount of the previous rainfall is 8.6 mm / d, and the extreme rainfall is 466 mm.

[0068] (3) Based on the ArcGIS software, the sensitivity of the study area grid to the response of the previous rainfall and extreme rainfall was analyzed, and the landslide risk warning level of the grid was divided according to the landslide risk judgment matrix to obtain the initial landslide risk warning level map of the whole area.

[0069] Based on the calibrated FSLAM model, the landslide risks of three working conditions under the return periods of 50 years and 100 years were calculated respectively, which are: Working condition 1: not considering rainfall; Working condition 2: only considering the effective infiltration of the previous rainfall (q a ); Working condition 3: considering both the effective infiltration of the previous rainfall (q a ) and the infiltration of typhoon rainstorm (P e ). The difference in the risk index of the first case reflects the sensitivity of the grid to the response of the previous rainfall, and the difference in the risk index of the second case reflects the sensitivity of the grid to the response of the extreme rainfall. Subsequently, the distribution of the grid number with the change in grid risk (difference in risk index) under each response condition was statistically analyzed. According to the change trend of the grid number, the change range of the grid risk was divided into 4 different levels. For example, under the return period of 50 years, the response degree of the grid to the previous rainfall was divided into 4 categories (as shown in Figure 4 ), which are the grid risk change amount of 0-0.01 (the grid risk has a small response to the previous rainfall), 0.01-0.08 (the grid risk has a moderate response to the previous rainfall), 0.08-0.18 (the grid risk has a large response to the previous rainfall), 0.18-0.99 (the grid risk has a very large response to the previous rainfall). The change range of the grid risk for the response to extreme rainfall was also divided into 4 categories ( Figure 5) They are 0 - 0.01 (the grid risk has a small response to extreme rainfall), 0.01 - 0.11 (the grid risk has a moderate response to extreme rainfall), 0.11 - 0.19 (the grid risk has a large response to previous rainfall), and 0.19 - 0.99 (the grid risk has a very large response to previous rainfall). Corresponding this result to the different degrees of the grid's response to the two types of rainfall infiltration in Table 1, and then classifying and predicting the landslide risk warning levels of the grid according to the corresponding relationship in Table 1, the initial landslide risk warning level map of the whole region was obtained, and different colors were set for different risk levels. The very low landslide risk is set to green, the low landslide risk is set to blue, the medium landslide risk is set to yellow, the high landslide risk is set to orange, and the extremely high landslide risk is set to red.

[0070] Table 1 Landslide Risk Judgment Matrix

[0071]

[0072] (4) Overlay the unconditional instability and unconditional stability regions to obtain the final landslide risk warning level map of the whole region.

[0073] The FSLAM model was used to calculate the probabilities of unconditional instability and unconditional stability of the grids in the study area respectively. The area where the unconditional instability probability is greater than 0.8 is the extremely high risk area, and the area where the unconditional stability probability is greater than 0.8 is the very low risk area. In the ArcGIS software, these two regions were overlaid into the result of obtaining the landslide risk warning level in step S3 to obtain the final landslide risk warning level map of the whole region.

[0074] When overlaying, use the "Erase" tool in ArcGIS. First, erase the grids that coincide with the unconditional instability region and the unconditional stability region in the initial landslide risk warning level map of the whole region obtained in S3, and then use the "Merge" tool to merge the map obtained after erasing with the unconditional instability region and the unconditional stability region, and update the landslide risk index to the corresponding landslide risk index formed by considering the terrain factors to obtain the final landslide risk warning level map of the whole region, realizing the typhoon rainstorm landslide warning.

[0075] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

[0076] Matters not described in the present invention apply to the prior art.

Claims

1. A typhoon rainstorm landslide warning method considering extreme rainfall and environmental variables, including the following: S1. Statistically analyze the pre-rainfall P in the two months before the landslide occurred in the current area a , and statistically analyze the typhoon rainstorm amount P that induced the landslide e , and the typhoon rainstorm amount is also called the extreme rainfall Use the actual daily atmospheric rainfall and various environmental variables during the two months before the landslide as input data to calculate the effective infiltration of antecedent rainfall q in the EASY_BAL software a , with the unit of mm / d; The environmental variables include the daily maximum temperature, minimum temperature, average temperature, and evapotranspiration; Based on the physical model FSLAM model, calculate the landslide risk index induced by typhoon rainstorms. Use the two types of rainfall infiltration, namely the antecedent rainfall and typhoon rainstorm amount during the two months before the landslide occurred, to calibrate the FSLAM model parameters and obtain the calibrated FSLAM model; S2. Obtain the daily actual atmospheric rainfall and various environmental variables in the study area over the years, and use the EASY_BAL software to calculate the antecedent rainfall P for each month a and the effective infiltration amount q of the antecedent rainfall a ; at the same time, count the maximum rainfall of consecutive 3 days in each month, and regard the maximum rainfall of consecutive 3 days in each month as the extreme rainfall of that month; Set the recurrence period duration. Use the Gumbel distribution calculation formula with the effective infiltration amount of antecedent rainfall per month as the input to obtain the effective infiltration amount value of antecedent rainfall in the study area under the recurrence period; Use the calculation formula of the Gumbel distribution with the monthly extreme rainfall as the input to obtain the extreme rainfall in the study area under the recurrence period; S3. Input the effective infiltration amount q of the previous rainfall in the study area and the extreme rainfall amount P obtained in S2 into the calibrated FSLAM model in S1, and use the calibrated FSLAM model to calculate the landslide risk indices under three working conditions respectively: Working condition 1: Without considering rainfall; Working condition 2: Only considering the effective infiltration amount q of the previous rainfall a and the extreme rainfall amount P e ; Working condition 3: Considering both the effective infiltration amount q of the previous rainfall a and the extreme rainfall amount P a ; e ​ Based on the ArcGIS software, calculate the difference in risk index for each grid between Case 2 and Case 1, and the difference in risk index for each grid between Case 3 and Case 2. The difference in risk index in the first case reflects the sensitivity of the grid to the antecedent rainfall response, and the difference in risk index in the second case reflects the sensitivity of the grid to the extreme rainfall response; count the distribution of the number of grids with the difference in risk index in both cases. According to the change trend of the number of grids, each case is divided into four different levels, namely small response to cumulative antecedent rainfall, moderate response to cumulative antecedent rainfall, large response to cumulative antecedent rainfall, very large response to cumulative antecedent rainfall, small response to extreme rainfall, moderate response to extreme rainfall, large response to extreme rainfall, and very large response to extreme rainfall. Construct a landslide risk judgment matrix with the different degrees of response to extreme rainfall infiltration and antecedent rainfall infiltration as rows and columns. The landslide risk index corresponding to the row and column is in the landslide risk judgment matrix. Then, based on the landslide risk judgment matrix, divide the risk warning level for each grid to obtain the initial landslide risk warning level map of the whole area; S4. Considering the terrain factor at the same time, use the calibrated FSLAM model to calculate the probabilities of unconditional instability and unconditional stability of the grids in the study area respectively. Define the area with an unconditional instability probability greater than 0.8 as the unconditional instability area, and the area with an unconditional stability probability greater than 0.8 as the unconditional stability area. Define the grids located in these areas as the extremely high risk area and the extremely low risk area respectively, and mark the landslide risk index formed by considering the terrain factor in these grids; S5. Overlay the extremely high risk area and extremely low risk area divided by considering the terrain factor with the result of the landslide risk warning level divided by using the landslide risk judgment matrix in S3 to obtain the final landslide risk warning level map of the whole area; During the overlay, use the "Erase" tool in ArcGIS. First, erase the grids that coincide with the unconditional instability area and unconditional stability area in S4 from the initial landslide risk warning level map of the whole area obtained in S3. Then use the "Merge" tool to merge the map obtained after erasing with the unconditional instability area and unconditional stability area, and update the landslide risk index to the corresponding landslide risk index formed by considering the terrain factor to obtain the final landslide risk warning level map of the whole area for typhoon rainstorm landslide warning.

2. The early warning method according to claim 1, wherein The FSLAM model parameters include the effective cohesion C of the soil s , the internal friction angle soil depth z, soil permeability coefficient K, soil porosity n, saturated soil density ρ s , the additional apparent cohesion C generated by soil roots r and the runoff curve number CN; The accuracy of the calibrated FSLAM model meets the following: the area under the receiver operating characteristic curve (ROC), denoted as AUC value, is used to measure the goodness of the calculation results. If the accuracy is greater than 70%, it indicates that the parameter selection is reasonable; if the accuracy is less than 70%, the parameters need to be adjusted and recalculated, and the iterative calculation is carried out until the accuracy is higher than 70%. The finally obtained parameters are the accurate calculation parameters of the landslide after calibration.

3. The warning method according to claim 1, wherein The FSLAM model includes two basic parts, namely a geotechnical model based on the infinite slope theory and a simplified groundwater model, and its stability coefficient FS is calculated by the following formula: Where C r is the apparent cohesion generated by soil roots, C s is the effective cohesion of the soil, g is the acceleration due to gravity, ρ s is the saturated soil density, z is the soil depth, a is the upstream catchment area, b is the grid resolution, K is the soil layer permeability coefficient, θ is the slope angle, ρ w is the density of water, and is the angle of internal friction; The input data of the FSLAM model includes 5 raster files and 2 text files. The raster files are Digital Elevation Model (DEM), soil, land use, effective infiltration of previous rainfall, typhoon rainstorm respectively. The first text file contains the root cohesion values and corresponding runoff curve numbers for each land use type, and the other text file contains the physical and mechanical parameter values for each specific soil. The physical and mechanical parameters include K, C s , ρ s , z, 4. The early warning method according to claim 1, wherein The landslide risk judgment matrix is divided into 5 risk levels, namely extremely low landslide risk, low landslide risk, medium landslide risk, high landslide risk and extremely high landslide risk.

Images