A method for dynamic evaluation of debris flow disaster susceptibility based on InSAR

Abstract

The application discloses a kind of mud-rock flow disaster proneness dynamic evaluation method based on InSAR, belong to satellite image processing field.For how from large-scale area screening out with mud-rock flow disaster risk gully area, the application proposes a kind of method for screening mud-rock flow disaster risk area based on interferometry synthetic aperture radar and digital elevation model technology.The processing method is that the sentinel-1 remote sensing satellite image data (SAR image) of monitoring area is obtained periodically, the satellite image data of monitoring area is calculated by InSAR technical method, the ground surface deformation rate of monitoring area is obtained, the gully topography of monitoring area is identified by DEM technical method, the area that meets ground surface deformation rate faster and gully topography simultaneously is screened out, as mud-rock flow disaster risk area is monitored.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of debris flow disaster susceptibility evaluation and satellite remote sensing monitoring, in particular to a debris flow disaster susceptibility evaluation method based on InSAR technology. BACKGROUND

[0002] Debris flow disaster is a kind of solid-liquid mixed fluid containing a large amount of water and soil and stone, which often occurs in the river valleys in mountainous areas, and is the result of the comprehensive action of natural factors such as geology, geomorphology, hydrology, meteorology, soil, vegetation and human activity factors. Debris flow disaster is rapid and fierce, with multiple hazards such as collapse, landslide and flood damage. In the process of debris flow disaster outbreak, a large amount of mud and debris will be carried and tilted in a short period of time, which seriously threatens the safety of life and property of people in the downstream, farmland, road traffic and other social infrastructure, destroys the natural ecological environment and seriously restricts social development.

[0003] At present, for the delineation of debris flow disaster hidden danger area in a large range of area, the commonly used method is based on DEM data and deep learning, and after training the delineation model using positive and negative samples, the DEM data of the target area is analyzed to determine whether the area has debris flow disaster risk, or the artificial field exploration method is used to determine the hidden danger area. The deep learning method based on DEM data is limited to the determination of debris flow disaster hidden danger from a single aspect of DEM, and its accuracy is difficult to guarantee. The artificial field exploration method has high accuracy, but its efficiency is too low, and it consumes time and effort, which cannot meet the demand of large-scale monitoring area.

[0004] At present, the data acquisition of debris flow disaster susceptibility evaluation mainly includes qualitative method and quantitative method. The qualitative method mainly refers to the on-site analysis based on expert experience, and the quantitative method mainly refers to statistical analysis method. The qualitative method is difficult to convince people due to the subjective will and professional level of professional people involved, and the evaluation efficiency is low, time-consuming and labor-intensive. In the statistical analysis method of quantitative method, the relationship between rainfall, topography and other influencing factors and debris flow disaster is analyzed by machine learning (such as logistic regression) method. This method does not consider the influence of ground deformation on debris flow disaster under normal circumstances, and it is only suitable for analyzing a specific area, which cannot meet the demand of large-scale monitoring area. In addition, the existing debris flow disaster susceptibility evaluation method mainly uses static influencing factor data for evaluation processing, but debris flow disaster is also affected by dynamic factors, and the existing method usually does not consider the dynamic change of debris flow disaster susceptibility.

[0005] In view of the limitations of the existing debris flow disaster susceptibility evaluation method, the present application proposes a dynamic debris flow disaster susceptibility evaluation method suitable for complex mountainous areas along the railway line in terms of debris flow disaster hidden danger valley delineation, influencing factors and complex mountainous area along the railway line debris flow disaster susceptibility evaluation.

[0006] The debris flow disaster susceptibility evaluation along the region of complex mountain railway is different from that of other regions. Due to the complex geological conditions and topography of the region where the mountain railway passes, it is usually difficult for monitoring personnel to reach the region, and the area is huge. The traditional debris flow disaster susceptibility evaluation model is not suitable for debris flow disaster susceptibility evaluation of a large area along the complex mountain railway.

[0007] At present, the existing debris flow disaster susceptibility evaluation model is usually based on rainfall, topography, soil environment and other influencing factor data for processing. In the absence of online monitoring data or field investigation, the accuracy of such evaluation data is usually low. There may be different topographic and geological conditions in a large area. Debris flow disaster susceptibility evaluation of the entire region cannot guarantee the accuracy and rationality of complex mountain debris flow disaster susceptibility evaluation. The common evaluation method based on logistic regression is sensitive to outliers and may not be highly accurate.

[0008] The existing debris flow disaster susceptibility evaluation method mainly uses static susceptibility evaluation results. Such method usually does not consider the dynamic change of debris flow disaster susceptibility. However, the dynamic influencing factors of debris flow disaster (such as rainfall, soil moisture content, etc.) usually change with time and space. SUMMARY

[0009] The present application solves the problems of how to screen out the gully region with debris flow disaster risk from a large area, the low accuracy of complex mountain debris flow disaster susceptibility evaluation, and how to periodically obtain data and periodically evaluate debris flow susceptibility based on current data.

[0010] For how to screen out the gully region with debris flow disaster risk from a large area, the present application proposes a method for screening debris flow disaster risk region based on interferometric synthetic aperture radar (InSAR) and digital elevation model (DEM) technology. The processing method is to periodically obtain sentinel-1 remote sensing satellite image data (SAR image) of the monitoring region, calculate the satellite image data of the monitoring region by InSAR technology, obtain the ground surface deformation rate of the monitoring region, and identify the gully topography of the monitoring region by DEM technology. The region that meets the conditions of fast ground surface deformation rate and gully topography is screened out as the debris flow disaster risk region for key monitoring. Thus, the technical solution of the present application is: a debris flow disaster susceptibility dynamic evaluation method based on InSAR, which comprises: screening debris flow disaster risk region and debris flow disaster susceptibility dynamic evaluation.

[0011] The method for screening debris flow disaster risk region is:

[0012] Step A1: Collect remote sensing satellite images of the detection area;

[0013] Step A2: Calculate the ground deformation rate;

[0014] (1) Crop the image according to the set area;

[0015] (2) Obtain the orbit data of the satellite that took the image and perform orbit correction, select the main image from all images, and register the main image and auxiliary images to a unified coordinate system;

[0016] (3) Remove the pulse band of all images and perform terrain correction;

[0017] (4) Calculate the main image intensity map and inverted terrain phase, and perform differential interference on the main image and auxiliary images to generate a single main image interferogram set;

[0018] (5) Generate small baseline pair combinations for each image, and perform differential processing on the corresponding interferograms to generate a small baseline interferogram set;

[0019] (6) Process the interferograms in the small baseline interferogram set to obtain an interferogram network;

[0020] (7) Phase unwrapping is performed on the interferogram network, and phase error correction is performed using phase closure;

[0021] (8) Inversion is performed on the corrected interferogram network using SBAS;

[0022] (9) Use the integrated PyAPS open source software package to perform atmospheric layering and ionospheric delay correction on the inverted interferogram network based on the global atmospheric model ERA-5;

[0023] (10) Remove the terrain error from the corrected interferogram network;

[0024] (11) Perform ground deformation time series inversion on the interferogram network to generate the final ground deformation rate;

[0025] Step A3: Obtain the digital elevation model (DEM) file of the monitoring area;

[0026] Step A4: Use ArcGIS tools to calculate and analyze the DEM file of the monitoring area to obtain a valley area identification map of the monitoring area;

[0027] (1) Integrate the target area DEM source files;

[0028] (2) Import the DEM file into ArcGIS, and set the pixel size to be consistent with the DEM layer in the raster analysis;

[0029] (3) Using the fill sink tool in the hydrological analysis of the Spatial Analyst tool in ArcGIS, fill the sinks in the DEM surface grid, remove small defects in the data and create a non-depressed point DEM;

[0030] (4) For the result of filling the sink, use the flow direction tool in the hydrological analysis to create a grid of flow direction from each pixel to its steepest downhill neighbor;

[0031] (5) For the result of flow direction, use the flow accumulation tool in the hydrological analysis to create a grid of the cumulative flow of each pixel;

[0032] (6) Using the raster calculator in the map algebra, use the composite function con to output values greater than 100 in the flow as 1;

[0033] (7) Using the raster to vector tool in the hydrological analysis, convert the grid representing the linear network to a feature representing the linear network; input the flow direction file and the flow data of the raster calculation result respectively, and generate a gully area map;

[0034] Step A5: superimpose the gully area identification map of the monitoring area on the ground deformation rate map, mark the area that simultaneously satisfies the gully area and the ground deformation rate greater than the set threshold value as the debris flow disaster risk area, as the subsequent debris flow disaster monitoring area;

[0035] The method for dynamic evaluation of debris flow disaster susceptibility is:

[0036] Step B1: selecting main influencing factors of debris flow disaster;

[0037] The order of the weights of the main influencing factors of debris flow disaster from high to low is: lithology > rainfall > geological structure > slope > vegetation coverage index > elevation > soil moisture content > surface coverage type > distance from water system > distance from building > distance from road, select the first M influencing factors with the heaviest weight, plus the ground deformation rate, as the main influencing factors of debris flow disaster participating in the evaluation of debris flow disaster susceptibility;

[0038] Step B2: obtaining debris flow disaster influencing factor data;

[0039] According to the main influencing factors determined in step B1, obtain the data of these main influencing factors, and establish a sample database;

[0040] Step B3: dividing the debris flow disaster monitoring area into grid cells;

[0041] Step B4: training a support vector machine using the sample database obtained in step B2,

[0042] Step B5: when the debris flow disaster susceptibility of a certain area is evaluated, the support vector machine trained in step 4 is used to determine the debris flow susceptibility of each grid, and all the grids are spliced to obtain the susceptibility result of the entire monitoring area.

[0043] Further, the input training set {(X i ,y i ), i = 1 ~ N of the support vector machine of the application, wherein y i = +1 or -1 is a positive and negative sample label, and X i is a main influencing factor; all a are solved when the following objective function is maximized:

[0044]

[0045] The constraint condition is:

[0046] Wherein, N represents the size of the training set, Φ(X i ) T represents a kernel function, and C represents a penalty factor.

[0047] In view of the problem that most of the existing debris flow disaster susceptibility evaluation methods are one-time static evaluation, the application proposes a method combining periodic evaluation and automatic data acquisition, uses the latest data to continuously update the evaluation result, overcomes the problem that the accuracy of the traditional susceptibility evaluation decreases over time, and improves the accuracy of the susceptibility evaluation. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 The overall scheme flowchart of the application;

[0049] Figure 2 The satellite image acquisition flowchart;

[0050] Figure 3 The flowchart for calculating the target area ground mark deformation rate map;

[0051] Figure 4 The ground deformation rate map;

[0052] Figure 5 The valley area identification flowchart;

[0053] Figure 6 The valley area identification result;

[0054] Figure 7 The debris flow disaster risk area result;

[0055] Figure 8 The grid division schematic diagram of the debris flow disaster susceptibility monitoring area;

[0056] Figure 9 Training effect diagram for different kernel functions;

[0057] Figure 10 Training time diagram for different kernel functions;

[0058] Figure 11 A flow chart for constructing a debris flow disaster susceptibility evaluation model. DETAILED DESCRIPTION

[0059] Debris flow disaster risk area screening

[0060] The present application proposes to screen debris flow disaster risk areas based on InSAR and DEM technology combination, and the specific implementation is as follows:

[0061] 1. Monitoring area remote sensing satellite image acquisition;

[0062] From the website EarthData (https: / / search.asf.alaska.edu) under the United States Aerospace Administration (NASA), the large area to be studied is framed. The Sentinel-1 image completely containing the region is selected. If the target area is too large, a satellite image cannot completely cover it, and the large area needs to be divided into several small areas that can be completely covered by a single image for processing), and the image is used as a search benchmark to search for the image data of the region. Select the Single Look Complex (SLC) image in the historical SAR image of the region and download it to the local. The image acquisition process of the monitoring area is as shown in Figure 2 .

[0063] 2. Calculate the ground deformation rate;

[0064] Install the interferometric synthetic aperture radar scientific computing environment and the Miami InSAR time series Python software (MintPy); process the SLC image, after image cropping, orbit correction, image registration, de-pulse band, and terrain correction, generate an interferogram. Then, after data conversion, loop closure, baseline inversion, mask time series, and time series filtering, the ground surface deformation rate map of the target area is output. The specific process is as shown in Figure 3 .

[0065] Among them, the image preprocessing steps are as follows:

[0066] (1) Import the SAR image in SLC file format into the tool, and crop the image according to the set region;

[0067] (2) The acquired track data is subjected to track correction, and suitable master images are selected from all images, and the master images and auxiliary images (all SAR images except the master images) are registered to a unified coordinate system;

[0068] (3) The pulse bands of all images are removed, and terrain correction is performed;

[0069] (4) The master image intensity map and the inverted terrain phase are simulated in combination with the DEM file, and the master images and the auxiliary images are subjected to differential interference to generate a single-master-image interferogram set;

[0070] (5) Small baseline pair combinations are generated, and each small baseline corresponding interferogram is subjected to differential processing to generate a small baseline interferogram set;

[0071] The image preprocessing steps are as follows:

[0072] (1) The interferograms in the small baseline interferogram set are processed to obtain an interferogram network;

[0073] (2) The interferogram network is subjected to phase unwrapping, and phase error correction is performed using phase closure;

[0074] (3) The corrected interferogram network is inverted using SBAS;

[0075] (4) The inverted interferogram network is subjected to atmospheric layering and ionospheric delay correction based on the global atmospheric model (ERA-5) using the integrated PyAPS open source software package;

[0076] (5) The corrected interferogram network is subjected to terrain error removal;

[0077] (6) The interferogram network is subjected to surface deformation time series inversion.

[0078] The final generated surface deformation rate map is shown in FIG. 1, wherein the abscissa is longitude, the ordinate is latitude, and the rate unit is cm / year: Figure 4

[0079] 3. Download the digital elevation model of the monitoring area;

[0080] Using this python script, the DEM original file of the target area can be automatically downloaded from the United States Space Agency according to the input parameters and automatically spliced and integrated. The dem.py is used to download the DEM file of the target area.

[0081] 4. Analyze the valley area;

[0082] The downloaded original DEM file is dragged into ArcGIS, and after grid analysis, filling, grid calculator, and grid river network vectorization processing, the target area's valley identification result is obtained.​

[0083] The specific flow chart is as shown in Figure 5

[0084] The detailed steps include:

[0085] (1) Use dem.py to download the target area DEM source file and automatically splice and integrate;

[0086] (2) Import the DEM file into ArcGIS, and in the raster analysis, set the pixel size to be consistent with the DEM layer;

[0087] (3) Use the fill sink tool in the hydrological analysis of the Spatial Analyst tool in ArcGIS to fill the sink in the DEM surface grid, remove small defects in the data, and create a non-pit DEM;

[0088] (4) For the result of filling the sink, use the flow direction tool in the hydrological analysis to create a grid of flow direction from each pixel to its steepest downhill adjacent point;

[0089] (5) For the result of flow direction, use the flow accumulation tool in the hydrological analysis to create a grid of cumulative flow of each pixel;

[0090] (6) Use the raster calculator in map algebra to output values greater than 100 in flow as 1 using the composite function con;

[0091] (7) Use the raster river network vectorization in the hydrological analysis to convert the grid representing the linear network into features representing the linear network. Input the flow direction file and the flow data of the raster calculation result respectively to generate the gully area map.

[0092] The identification result is as shown in Figure 6 , wherein the depth of white represents the height, and the whiter the area, the higher the elevation. The green area represents the gully area.

[0093] 5. Debris flow disaster prone area delineation;

[0094] Use image processing tools to overlay the surface deformation rate map with the gully area identification result generated by ArcGIS. Determine the area with large surface deformation rate and gully area as the debris flow disaster prone area.

[0095] II. Dynamic evaluation of debris flow disaster proneness

[0096] After delineating the debris flow risk area, perform dynamic evaluation of the debris flow disaster proneness of the area, and the processing method is implemented as follows:

[0097] 1. Selection of main influencing factors of debris flow disaster;

[0098] ​There are many influencing factors of debris flow disasters, such as slope, elevation, stratum lithology, geological structure, surface covering type, vegetation coverage index, rainfall, soil moisture content, distance from water, distance from building, distance from road and the like. As for the excavation of the influencing factors of debris flow disasters, other scholars have already studied maturely, and the order of the weight from high to low is respectively: stratum lithology > rainfall > geological structure > slope > vegetation coverage index > elevation > soil moisture content > surface covering type > distance from water system > distance from building > distance from road. The invention selects stratum lithology, rainfall, slope, vegetation coverage number, elevation and soil moisture content as the main influencing factors of debris flow disasters.

[0099] On this basis, the invention proposes to take the surface deformation rate as the main influencing factor to participate in the debris flow disaster susceptibility evaluation.

[0100] 2. Debris flow disaster influencing factor data acquisition;

[0101] The debris flow disaster points in previous years are obtained from the geological disaster disaster report published by the website of the State Natural Resources Department every day, and the disaster points are converted into specific longitude and latitude coordinates through the map website. According to the disaster point coordinates, the meteorological hydrological, soil moisture and the like data of the point are obtained from various data websites, and are summarized into positive samples of the data set. A part of the area not having ever occurred debris flow disaster is selected at random, and the same data is collected, as negative samples of the data set.

[0102] The website of the State Natural Resources Department (https: / / www.mnr.gov.cn / ) will publish the geological disaster disaster report every day, a python script is written to collect the geological disaster disaster report on the website in the past ten years, and the disaster occurrence time, place, disaster type and the like data are extracted from the report. The debris flow disaster data in previous years are screened out, and the debris flow disaster disaster position is obtained. Through the longitude and latitude query positioning website, the text geological information of the disaster point is converted into specific longitude and latitude.

[0103] According to the disaster point longitude and latitude information, other related data are obtained:

[0104] The surface deformation rate information of the disaster point is calculated and obtained.

[0105] The meteorological historical data are obtained from the website system of the U.S. National Environmental Information Center, and the website address is (https: / / www.ncei.noaa.gov / data / global-forecast-system / access / historical / analysis / ). The collected meteorological data include rainfall, soil moisture and the like data.

[0106] The regional topographic data can be obtained from the website of the United States Space Agency (https: / / urs.earthdata.nasa.gov / ), ASTER GDEM 30M resolution original elevation data. A collection program is written to download the regional elevation data to the local, and then a Python program is used to parse the data, and the regional topography is preliminarily analyzed and processed by using the slope and slope direction algorithm calculation formula, and the regional elevation and slope data are extracted and calculated.

[0107] The regional stratum lithology data can be obtained from the website of the National Geological Data Library (http: / / www.ngac.org.cn / DataSpecial / geomap.html). In the national geological map data special service application, find the point corresponding to the latitude and longitude, based on the legend data, manually collect the regional geological lithology data.

[0108] The regional vegetation coverage data is obtained from the website of the European Medium-term Weather Forecasting Center (https: / / www.ecmwf.int / en / forecasts / datasets / reanalysis-datasets / era5) to monitor the leaf area index of the region to represent the vegetation coverage, and a python program is used to call the website data download api to collect the regional leaf area index data.

[0109] 3. Grid cell division of debris flow disaster monitoring area

[0110] The present application aims at the problem that the monitoring area is too large to locate the specific debris flow disaster risk site, and adopts the method of grid cell division to reduce the monitoring area to solve the problem, and the specific operation is as shown in Figure 8

[0111] As shown in the figure, the original monitoring area is the debris flow disaster prone area found after the combination of the ground surface deformation rate map and the valley area identification map, and the area shape is an irregular closed graph. The original monitoring area is divided into unit grids, and the area of each grid is 70m*70m. In this way, the problem of being unable to accurately locate the debris flow disaster risk site can be solved.

[0112] After the grid division of the debris flow disaster monitoring area, the prone evaluation of each unit cell area needs to be carried out, and after the evaluation is completed, the evaluation results of each unit cell are summarized, and the overall prone evaluation of the whole monitoring area is carried out.

[0113]

[0114]

[0115] ​In formula (1) and formula (2), A is a regional overall evaluation result, N is a total number of cells, A i is an evaluation result of the i-th cell.

[0116] 4. Debris flow disaster susceptibility evaluation model construction

[0117] For the divided cells, support vector machines need to be used to evaluate the susceptibility of each cell.

[0118] The training process of the support vector machine inputs the training set {(X i ,y i ), i = 1 ~ N, wherein y i = +1 or -1, which is a positive or negative sample label. The following objective function is solved (all a are solved) to maximize:

[0119]

[0120] Limitations:

[0121] For the support vector machine, the selection of the kernel function is a key part of determining the evaluation accuracy;

[0122] Research has shown that the selection of the support vector machine kernel function directly affects the prediction accuracy and convergence speed of the support vector machine, so in order to obtain a higher prediction accuracy, a suitable kernel function must be selected. Common kernel functions of support vector machines include linear kernel functions, polynomial kernel functions, RBF (Gaussian) kernel functions, and Sigmoid kernel functions. Linear kernel functions, polynomial kernel functions, RBF kernel functions, and Sigmoid kernel functions are respectively used in the sample set for ten-fold cross-validation, and the average prediction accuracy of different kernel functions under the corresponding training samples is as shown in Figure 9 .

[0123] From Figure 8 the test results, it can be seen that the average accuracy of the polynomial kernel function, the RBF kernel function, and the Sigmoid kernel function is relatively high, among which the average prediction effect of the RBF kernel function is the best, and the prediction effect of the linear kernel function is very poor.

[0124] For the analysis of the performance of different kernel function models, because the prediction accuracy of the linear kernel function is too low, it is not meaningful to analyze the performance of the linear kernel function model, so the SVM model is trained using the polynomial kernel function, the RBF kernel function, and the Sigmoid kernel function respectively, and the training time of different kernel functions is as shown in Figure 10 .

[0125] From Figure 10It is shown that the training time will increase with the increase of the number of samples when three kinds of kernel functions are selected. When the training sample is large, the training time required by selecting RBF kernel function changes the least on the whole, which shows that it is reasonable to select RBF kernel function, and it is also in line with the characteristics of nonlinearity of debris flow disaster evaluation. Therefore, in the debris flow disaster susceptibility evaluation model based on SVM, RBF kernel function is selected as the kernel function of SVM.

[0126] The flow chart of the debris flow disaster susceptibility evaluation model is shown in Figure 11

[0127] 5. Debris flow disaster dynamic evaluation

[0128] In scheme one, the InSAR and DEM combination method is used to delineate the debris flow disaster hidden danger valley area. The process of using InSAR to calculate the ground deformation rate can be automatically realized through the script; the process of using DEM to identify the valley area is manually operated by ArcGIS, but because the formation of the valley area usually takes hundreds of millions of years, it can be considered that the valley area will not change in a short time, so the valley identification result does not need to be updated. In scheme (two, 2), the static data of the monitoring area is obtained manually, and the dynamic data is obtained using the python script.

[0129] Based on the advantages of scheme one and scheme (two, 2), only a simple periodic task needs to be set to meet the dynamic susceptibility evaluation of debris flow disaster.

[0130] In order to realize the dynamic evaluation of debris flow disaster susceptibility, a periodic task is added to periodically execute the debris flow disaster dynamic susceptibility evaluation program, and the program process is mainly as follows:

[0131] (1) Run the script to automatically calculate the ground deformation rate by InSAR technology, and automatically update the debris flow disaster risk area delineation result combined with the valley area identification result.

[0132] (2) Run the python script to obtain the latest data of the risk area.

[0133] (3) Add the obtained latest data to the data set, and optimize the model parameters by training the susceptibility evaluation model again.

[0134] (4) Automatically input the latest data into the updated debris flow disaster susceptibility evaluation model to calculate the debris flow disaster susceptibility evaluation result of the current monitoring area.

[0135] (5) End this evaluation process, and wait for the next timing task to return to step (1)

[0136] ​By regularly updating the risk area, updating the evaluation model and the evaluation result, the evaluation accuracy is continuously updated, and high accuracy is maintained.

[0137] In view of the problem that the traditional method for delineating the hidden danger area of debris flow disaster is not good in complex mountainous areas, the application proposes a method for delineating the hidden danger area of debris flow disaster based on InSAR and DEM. Experimental results show that the accuracy of the method for delineating the hidden danger area of debris flow disaster proposed by the application is 91.5%, while the accuracy of the method for delineating the hidden danger area of debris flow disaster based on DEM alone is 88%. The results show that compared with the traditional method, the application can reduce the monitoring cost while ensuring high accuracy.

[0138] In view of the problem that the debris flow disaster susceptibility evaluation is not accurate enough, the application proposes to add the ground surface deformation rate of the monitoring area as one of the important influencing factors into the debris flow disaster susceptibility evaluation model. Experimental results show that after adding the ground surface deformation rate data, the accuracy of the debris flow disaster susceptibility evaluation result is 90.2%, which is higher than 88.5% without adding the ground surface deformation rate.

Claims

1. A method for dynamic evaluation of debris flow disaster susceptibility based on InSAR, the method comprising: Screening debris flow disaster risk area and dynamic evaluation of debris flow disaster susceptibility; The method for screening the debris flow disaster risk area is: Step A1: Collect remote sensing satellite images of the detection area; Step A2: Calculate the ground deformation rate; (1) Crop the image according to the set area; (2) Obtain the orbit data of the satellite of the shooting image for orbit correction, select the main image from all images, and the SAR images other than the main image are auxiliary images; register the main image and the auxiliary image to a unified coordinate system; (3) Remove the pulse band of all images and perform terrain correction; (4) Calculate the main image intensity map and the inverted terrain phase, and perform differential interference on the main image and the auxiliary image to generate a single main image interference graph set; (5) Generate a small baseline pair combination for each image, and perform differential processing on the interference graphs corresponding to each small baseline pair to generate a small baseline interference graph set; (6) Process the interference graphs in the small baseline interference graph set to obtain an interference graph network; (7) Phase unwrapping is performed on the interference graph network, and phase error correction is performed using phase closure; (8) Invert the corrected interference graph network using SBAS; (9) Use the integrated PyAPS open source software package to correct the inverted interference graph network based on the global atmospheric model ERA-5 atmospheric layering and ionospheric delay; (10) Remove the terrain error of the corrected interference graph network; (11) Perform ground deformation time series inversion on the interference graph network to finally generate the ground deformation rate; Step A3: Obtain the digital elevation model DEM file of the monitoring area; Step A4: Calculate and analyze the DEM file of the monitoring area using Arc GIS tools to obtain a gully area identification map of the monitoring area; (1) Splice and integrate the DEM source file of the target area; (2) Import the DEM file into ArcGIS, and set the pixel size in the raster analysis to be consistent with the DEM layer; (3) Use the fill depression tool in the hydrological analysis of the Spatial Analyst tool in ArcGIS to fill the sinks in the DEM surface grid, remove small defects in the data, and create a non-pit DEM; (4) For the fill depression result, use the flow direction tool in the hydrological analysis to create a flow direction grid from each pixel to its steepest downhill adjacent point; (5) For the flow direction result, use the flow accumulation tool in the hydrological analysis to create a grid of the cumulative flow of each pixel; (6) Use the grid calculator in the map algebra to output 1 for values greater than 100 in the flow using the composite function con; (7) Use the raster river network vectorization in the hydrological analysis to convert the raster representing the linear network into features representing the linear network; input the flow direction file and the flow data of the grid calculation result respectively to generate a gully area map; Step A5: Superimpose the gully area identification map of the monitoring area on the ground deformation rate map, and mark the area that simultaneously satisfies the gully area and the ground deformation rate greater than the set threshold as a debris flow disaster risk area, as a subsequent debris flow disaster monitoring area; The method for dynamic evaluation of debris flow disaster susceptibility is: Step B1: select the main influencing factors of debris flow disaster; The order of the main influencing factors of debris flow disaster from high to low is: stratum lithology > rainfall > geological structure > slope > vegetation coverage index > elevation > soil moisture content > surface coverage type > distance from water system > distance from building > distance from road. The first M factors with the highest weight are selected, plus the surface deformation rate, as the main influencing factors of debris flow disaster for the evaluation of debris flow disaster susceptibility; Step B2: obtain the data of debris flow disaster influencing factors; According to the main influencing factors determined in step B1, obtain the data of these main influencing factors, and establish a sample database; Step B3: divide the grid units of the debris flow disaster monitoring area; Step B4: train a support vector machine using the sample database obtained in step B2, Step B5: when evaluating the susceptibility of debris flow disaster in a certain area, use the support vector machine trained in step 4 to determine the susceptibility of each grid, and splice all the grids to obtain the susceptibility result of the whole monitoring area.

2. The InSAR-based dynamic evaluation method of debris flow disaster susceptibility according to claim 1, characterized in that, Input training set of the support vector machine wherein = +1 or -1, positive or negative sample label, is the main influencing factor; all solving the following objective function maximization ； Limitations: ； wherein, denotes the training set size, denotes the kernel function, denotes the penalty factor.

Images