A method for dynamic analysis of post-earthquake rainstorm mudslide sensitivity

Abstract

This invention relates to a method for dynamic analysis of post-earthquake rainstorm debris flow sensitivity, comprising the following steps: collecting sample datasets of debris flow basins in the study area over several years, wherein the sample datasets contain several factors; resampling the sample datasets to establish multiple sets of sample datasets; performing collinearity detection on the factors in each set of sample datasets, selecting factors for model establishment, thereby establishing a debris flow sensitivity analysis model; selecting an optimal set of sample datasets based on the prediction results of the debris flow sensitivity analysis model; extracting the weight values ​​of the factors in the optimal sample datasets, fitting the dynamic curves of the weight values ​​of each factor changing over time, and establishing a dynamic factor regression equation; and substituting the dynamic factor regression equation into the debris flow sensitivity analysis model to obtain the dynamic analysis model of debris flow sensitivity.

Description

Technical Field

[0001] This invention relates to the field of debris flow disaster prevention and control technology, and in particular to a method for dynamic analysis of the sensitivity of debris flow to post-earthquake rainstorms. Background Technology

[0002] Statistical analysis of debris flow events in Wenchuan County after the Wenchuan earthquake reveals that the catchment area and elevation difference of post-earthquake debris flows gradually increased, while the area density of co-seismic landslides and the average annual rainfall gradually decreased. The overall number of debris flow outbreaks showed a decreasing trend, and the frequency also decreased, indicating a significant change in the weight of factors influencing post-earthquake debris flows. Traditional debris flow dynamic sensitivity analysis only considers source factors, neglecting other factors related to post-earthquake debris flow sensitivity, thus preventing current debris flow analyses from effectively identifying the sensitivity of multiple post-earthquake debris flows.

[0003] Patent application CN110334482A discloses a method for evaluating the dynamic sensitivity of post-earthquake debris flows based on the intensity of source activity. It describes the following steps: (1) By analyzing the dynamic changes of debris flow source bodies in the study area, eight evaluation factors are selected, including changes in source activity; (2) Probabilistic cross-analysis is performed on the evaluation factors and past debris flow events to obtain normalization standards for the evaluation factors; (3) The grey relational analysis method is used, with past debris flow events as a reference column, to calculate the weight values ​​of the evaluation factors; (4) The obtained evaluation model is used to evaluate the current year's debris flow sensitivity, and the current year's debris flow sensitivity is graded to achieve dynamic and effective prediction of the probability of debris flow outbreaks. Due to objective limitations, this paper did not study other factors in evaluating the dynamic sensitivity of debris flows, and the results obtained are only applicable to single-period debris flow sensitivity evaluation.

[0004] Patent application CN107341586A discloses a method for calculating the frequency of geological disasters based on rainfall. It describes how to extract internal influencing factors of landslides using the spatial analysis function of a geographic information system (GIS), and then perform logistic regression analysis using statistical product and service solution software to determine the logistic regression coefficients of each internal influencing factor, thus calculating the spatial probability of landslides within the study area. However, since the internal influencing factors are fixed, the calculated landslide probability is also fixed and cannot meet the requirements of current sensitivity dynamic analysis. Summary of the Invention

[0005] The purpose of this invention is to establish a dynamic analysis model of post-earthquake rainstorm debris flow sensitivity based on a sample dataset of debris flow outbreaks over many years after an earthquake, and to provide a dynamic analysis method for post-earthquake rainstorm debris flow sensitivity, thereby enabling dynamic and effective prediction of multi-stage post-earthquake rainstorm debris flows.

[0006] To achieve the above-mentioned objectives, the embodiments of the present invention provide the following technical solutions:

[0007] A method for dynamic analysis of post-earthquake rainstorm debris flow sensitivity includes the following steps:

[0008] Step 1: Collect sample datasets from several years of post-earthquake debris flow basins in the study area. The sample datasets contain several factors. Resample the sample datasets to establish multiple sets of sample datasets.

[0009] Step 2: Perform collinearity detection on the factors in each group of sample datasets to select the factors used for model building, thereby establishing a debris flow sensitivity analysis model; and select the best sample dataset based on the performance evaluation results of the debris flow sensitivity analysis model.

[0010] Step 3: Extract the weight values ​​of factors in the best sample dataset, fit the dynamic curve of the weight values ​​of each factor changing over time, and establish a dynamic factor regression equation; substitute the dynamic factor regression equation into the debris flow sensitivity analysis model to obtain the debris flow sensitivity dynamic analysis model.

[0011] Step 4: Use the debris flow sensitivity dynamic analysis model to predict future debris flow sensitivity.

[0012] Furthermore, the factors in the sample dataset include topographic factors, sediment source factors, and precipitation factors;

[0013] The topographic factors include watershed area, watershed elevation difference, channel longitudinal gradient, channel length, and Melton intensity index.

[0014] The source factors include lithology, distance from the fault, area density of coseismic landslides in the watershed, average vegetation normalization index of coseismic landslides in the watershed, average vegetation normalization index of coseismic landslides in the watershed, and vegetation restoration index of coseismic landslides in the watershed.

[0015] The rainfall-related factors include the annual average rainfall.

[0016] Furthermore, the step of resampling the sample dataset to establish multiple sets of sample datasets includes: merging sample datasets from adjacent years, selecting the number of years N to be merged, and performing rolling merging of the sample datasets for each year.

[0017] Furthermore, the step of performing collinearity detection on the factors in each group of sample datasets to screen out the factors used for model building, thereby establishing a debris flow sensitivity analysis model, includes:

[0018] For each period of data in each sample dataset, collinearity detection is performed on k factors. The variance inflation coefficient (VIF) of the factors in the sample dataset is calculated. Factors with a VIF greater than a set threshold are deleted, leaving n factors, where n ≤ k.

[0019] A debris flow sensitivity analysis model was established using the remaining n factors via logistic regression:

[0020] (1)

[0021] In equation (1), P represents the sensitivity to debris flow outbreaks, and its value ranges from [0,1].

[0022] Z is a linear fitting equation containing a set of factors related to debris flow outbreaks, and its form is as follows:

[0023] (2)

[0024] In equation (2), b0 is the intercept; b n X represents the partial regression coefficient; n It is a factor variable.

[0025] Furthermore, the step of selecting an optimal set of sample datasets based on the performance evaluation results of the debris flow sensitivity analysis model includes:

[0026] The debris flow sensitivity analysis model constructed for each set of sample datasets was evaluated using cross-validation. Based on the evaluation results, the best set of sample datasets was selected.

[0027] The performance evaluation prediction results include a confusion matrix table, prediction accuracy, and area under the prediction feature curve; the confusion matrix table includes four indicators: debris flow outbreak prediction accuracy (TPR), debris flow non-outbreak prediction accuracy (TNR), debris flow outbreak prediction error rate (FNR), and debris flow non-outbreak prediction error rate (FPR).

[0028] Furthermore, the step of substituting the dynamic factor regression equation into the debris flow sensitivity analysis model to obtain the debris flow sensitivity dynamic analysis model includes:

[0029] If the optimal sample dataset corresponds to n factors, substituting the dynamic factor regression equation of the factors in the optimal sample dataset into equation (2) yields the dynamic linear fitting equation for debris flow:

[0030] (3)

[0031] In equation (3), f0 is the constant term regression equation, and f1~f n These are dynamic factor regression equations for n factors.

[0032] Furthermore, the step of predicting future debris flow sensitivity using a debris flow sensitivity dynamic analysis model includes:

[0033] Input the predicted year, the vegetation normalization index of the predicted coseismic landslide, and the predicted annual average rainfall into the debris flow sensitivity dynamic analysis model, and the model will output the debris flow sensitivity of the predicted year.

[0034] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0035] The present invention provides a dynamic analysis method for post-earthquake rainstorm debris flow sensitivity. By establishing a dynamic analysis model for debris flow sensitivity based on multiple debris flow outbreaks after an earthquake, it is possible to predict the possibility of multiple debris flow outbreaks after an earthquake. This eliminates the drawback of previous analyses that could only analyze the current year, and thus meets the needs of long-term disaster prevention and mitigation planning. Attached Figure Description

[0036] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This is a flowchart of the method in this embodiment;

[0038] Figure 2 This is a box plot of VIF values ​​from a sample dataset used in an example. Figure 2 In the figure, 'a' represents the box plot of VIF values ​​for all factors. Figure 2 In the figure, b is a box plot of VIF values ​​after deleting factors whose VIF values ​​are greater than the set threshold;

[0039] Figure 3 This is a TPR hotspot distribution diagram after cross-training of each training set and each test set in the example.

[0040] Figure 4 The example shows the TNR hotspot distribution after cross-training of each training set and each test set;

[0041] Figure 5 Here is an example of an ACC hotspot distribution diagram after cross-training of each training set and each test set;

[0042] Figure 6 Here is an AUC hotspot distribution diagram after cross-training of each training set and each test set in the example;

[0043] Figure 7Box plots for TPR, TNR, ACC, and AUC in the example. Figure 7 In this diagram, 'a' represents the TPR box plot. Figure 7 In this context, 'b' represents the TNR box plot. Figure 7 In the figure, 'c' represents the ACC box plot. Figure 7 In this context, d represents the AUC box plot;

[0044] Figure 8 A schematic diagram of the dynamic curves of CS and MRN fitted for the example;

[0045] Figure 9 The dynamic curves of LI, LAD, and LAN fitted for the example;

[0046] Figure 10 The dynamic curves of RY and constant terms fitted for the example;

[0047] Figure 11 This is a schematic diagram illustrating the assessment results of debris flow sensitivity dynamic analysis model for the period from 2008 to 2019.

[0048] Figure 12 This is a schematic diagram illustrating the prediction results of the debris flow sensitivity dynamic analysis model for 2025. Detailed Implementation

[0049] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0050] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0051] Wenchuan County was a severely affected area in the 5.12 Wenchuan earthquake. Following the earthquake, it experienced clusters of debris flows in 2010, 2013, and 2019, causing significant damage to the lives, property, and roads of residents in the mountainous area. This embodiment selects 99 watersheds in Wenchuan County as research objects and uses the post-earthquake rainstorm debris flow sensitivity dynamic analysis method of this invention to conduct dynamic sensitivity analysis on these 99 watersheds. This invention is achieved through the following technical solutions, such as... Figure 1 As shown, a dynamic analysis method for post-earthquake rainstorm debris flow sensitivity includes the following steps:

[0052] Step 1: Collect sample datasets from several years of post-earthquake debris flow basins in the study area, resample the sample datasets, and establish multiple sets of sample datasets.

[0053] We collected and produced a multi-year sample dataset of debris flow basins in the study area (2008-2019). The sample dataset includes debris flow outbreak events and factors, including topographic factors, sediment source factors, and precipitation factors.

[0054] Assuming the sample dataset contains k factors, in one embodiment, the topographic factors include watershed area (AC), watershed elevation difference (RF), channel gradient (CS), channel length (CL), and Melton intensity index (MRN). Here, watershed area (AC) refers to the projected area of ​​the watershed on the horizontal plane; watershed elevation difference (RF) refers to the difference between the maximum and minimum elevation values ​​of the watershed; channel gradient (CS) refers to the longitudinal gradient of the channels; channel length (CL) refers to the length of the channels within the watershed; and the Melton intensity index (MRN) characterizes the overall topographic conditions of the watershed and is a dimensionless index of watershed topography, calculated using the following formula:

[0055]

[0056] Provenance factors include lithology (LI), distance from fault (FD), coseismic landslide area density (LAD) within the watershed, mean normalized vegetation index (NL) for coseismic landslides within the watershed, mean normalized vegetation index (NC) for the watershed, vegetation restoration index (LAN) for coseismic landslides within the watershed, and normalized vegetation index (NDVI) for the study area. The NDVI for the study area was obtained using Landsat 7 as the data source and Google Earth Engine as the tool, acquiring the maximum NDVI values ​​from January to June each year. Since lithology (LI) is categorical data and cannot be directly used for factor contribution calculations, the frequency ratio method was used to convert it into continuous data. The formula for the frequency ratio method is:

[0057]

[0058] Where RF represents the frequency ratio of a certain lithology, i represents the lithology category, and D i A represents the number of watersheds where debris flows occur when the lithology is category i. i This represents the total number of basins for debris flows when the lithology is category i.

[0059] Rainfall-related factors include annual average rainfall (RY), and spatial distribution data of annual average rainfall from 2008 to 2019.

[0060] The sample dataset was resampled by increasing the sample size. Specifically, sample datasets from adjacent years were merged, with a selected number of years (N) for merging. This merging process was repeated for each year. For example, with N=3, the original sample dataset consisted of 12 periods: 2008, 2009, ..., 2019. After resampling, the dataset was reorganized into 10 periods: 2008–2010, 2009–2011, ..., 2017–2019. The results of the sample dataset resampling are shown in Table 1.

[0061] Table 1 Sample Dataset Resampling

[0062]

[0063] Step 2: Perform collinearity detection on the factors in each group of sample datasets to select the factors used for model building, thereby establishing a debris flow sensitivity analysis model; and select the best sample dataset based on the performance evaluation results of the debris flow sensitivity analysis model.

[0064] For each period's data in each sample dataset in Table 1, factor collinearity detection was performed separately. The statistical results of factor VIF values ​​for all years are as follows: Figure 2 As shown, from Figure 2 As can be seen in (a), the VIF values ​​of AC, CL, NC, and NL are all greater than 10, and there is significant collinearity between AC and CL, and between NC and NL. Therefore, it is sufficient to delete AC or CL, and NC or NL. Figure 2 As shown in (b), the VIF values ​​of the factors after collinearity detection and deletion are all less than 10. The factors corresponding to each of the five sample datasets shown in Table 1 can be selected.

[0065] Logistic regression was used to link debris flow outbreak sensitivity with the link function. Assuming the link function contains factors that may influence debris flow outbreaks, the relationship between debris flow outbreaks and their dependence on these factors can be expressed as:

[0066] (1)

[0067] Where P represents the sensitivity to debris flow outbreaks, with a value range of [0,1]; Z is the linear fitting equation, which contains a set of factors related to debris flow outbreaks, as shown below:

[0068] (2)

[0069] Where b0 is the intercept; b n These are the partial regression coefficients, i.e., the coefficients of each factor; X nThese are factor variables, i.e., the values ​​of each factor. For example, after performing collinearity detection on k factors, if n factors remain that meet the criteria, then X... n This represents the value of the nth factor that meets the conditions.

[0070] Since five factor sets were selected from the recombined sample dataset, and each factor set may correspond to a different factor set, a debris flow sensitivity analysis model can be established for each factor set, resulting in five debris flow sensitivity analysis models. Each debris flow sensitivity analysis model is trained using cross-validation to obtain performance evaluation results. Based on the performance evaluation results, the best sample dataset can be selected.

[0071] Performance evaluation results include a confusion matrix, prediction accuracy (ACC), and area under the prediction feature curve (AUC).

[0072] The performance evaluation results can be statistically analyzed to include: the number of correctly predicted debris flow events (TP), the number of correctly predicted debris flow non-events (TN), the number of incorrectly predicted debris flow events (FN), and the number of incorrectly predicted debris flow non-events (FP). Based on these four statistical results, four metrics of the confusion matrix can be calculated: the prediction accuracy rate for debris flow events (TPR), the prediction accuracy rate for debris flow non-events (TNR), the prediction error rate for debris flow events (FNR), and the prediction error rate for debris flow non-events (FPR). The confusion matrix is ​​shown in Table 2.

[0073] Table 2 Confusion Matrix

[0074]

[0075] Prediction accuracy (ACC) is the ratio of the number of correct predictions to the total number of input samples. It can be used to represent the overall prediction accuracy of a model. ACC = (TP + TN) / (TP + TN + FP + FN).

[0076] The area under the predictive characteristic curve (AUC) is the integral area under the curve obtained by calculating the prediction accuracy rate (TPR) of debris flow outbreaks and the prediction error rate (FPR) of debris flow non-outbreaks at different thresholds. It is used to indicate the probability that the model correctly identifies whether a debris flow will break out in a debris flow basin compared to random discrimination.

[0077] Each sample dataset is divided into a training set and a test set, such as Figure 3 The image shows the TPR hotspot distribution after cross-training of each training set and each test set. Squares without numbers indicate that an effective debris flow sensitivity analysis model could not be established due to a small sample size. Figure 4 The image shows the TNR hotspot distribution after cross-training of each training set and each test set. Figure 5The image shows the ACC hotspot distribution after cross-training of each training set and each test set. Figure 6 The image shows the AUC hotspot distribution after cross-training of each training set and each test set.

[0078] The TPR, TNR, ACC, and AUC of all valid assessment and prediction results were statistically analyzed, and the results are as follows: Figure 7 As shown, the horizontal axis represents the Nth training set. Figure 7 (a) The vertical axis represents the TPR value. Figure 7 (b) The vertical axis represents the TNR value. Figure 7 (c) The vertical axis represents the ACC value. Figure 7 (d) The vertical axis represents the AUC value.

[0079] As can be seen, the TPR value (mean value) increases overall with the increase of the training set N, indicating that the accuracy of debris flow outbreak prediction increases with the increase of the number of training set samples N. The TPR value reaches its highest point when N=3; however, the TPR value does not increase further when N continues to increase.

[0080] As the training set N increases, the TNR value continuously decreases, indicating that the accuracy of predicting the absence of debris flows gradually decreases with the increase of the training set samples N. It is important to note that when N=1 and 2, the TNR value is abnormally high, which can create the illusion of high overall prediction accuracy. When N is greater than 3, the TNR value is less than 0.7.

[0081] As the training set N increases, the ACC value first increases and then decreases, reaching its maximum when N=2 and 3. However, when N exceeds 3, the ACC value actually decreases. Similarly, as the training set N increases, the AUC value also first increases and then decreases, exhibiting the same trend as the ACC value.

[0082] By statistically analyzing four aspects—TPR, TNR, ACC, and AUC—and training the resampled dataset, it was found that the debris flow sensitivity analysis model has the best predictive ability when N=3. Therefore, the dataset with N=3 was selected as the optimal dataset. It is assumed that the dataset with N=3 corresponds to 6 factors: CS, MRN, LI, LAD, LAN, and RY.

[0083] Step 3: Extract the weight values ​​of the factors in the best sample dataset, fit the dynamic curve of the change of each factor weight value over time, and establish a dynamic factor regression equation; substitute the dynamic factor regression equation into the debris flow sensitivity analysis model to obtain the debris flow sensitivity dynamic analysis model.

[0084] Based on the factors corresponding to the optimal sample dataset selected when N=3, the weight values ​​of these factors are extracted, and the dynamic changes of the factor weight values ​​over time are analyzed to construct dynamic curves, such as... Figure 8 The figure shows the fitted dynamic curves of CS and MRN, as follows: Figure 9 The figure shows the fitted dynamic curves of LI, LAD, and LAN, as follows: Figure 10 The figure shows the dynamic curves of the fitted RY and constant terms. Figures 8-10 R in the figure represents the variance of the fit, indicating the fit effect. It ranges from 0 to 1, with the closer to 1 indicating a better fit.

[0085] Substituting the dynamic factor regression equations of each factor into equation (2), we obtain the dynamic analysis model for debris flow sensitivity:

[0086] (3)

[0087] (4)

[0088] (5)

[0089] (6)

[0090] (7)

[0091] (8)

[0092] (9)

[0093] Substituting equations (4-9) into equation (3), b C The dynamic regression equation has a constant term, b CS For the dynamic factor regression equation of the channel longitudinal slope (CS), b MRN The dynamic factor regression equation for the Melton Intensity Index (MRN) is given by b. LI For the dynamic factor regression equation of lithology (LI), b LAD The dynamic factor regression equation for the area density (LAD) of coseismic landslides within the watershed, b RY The dynamic factor regression equation for the annual average rainfall (RY) is given by equation (4-8). In equation (4-8), x refers to the difference between the year to be analyzed and 2008. For example, if we are analyzing debris flows from 2017 to 2019, then x = (2017 + 2019) / 2 - 2008 = 10.

[0094] Step 4: Use the debris flow sensitivity dynamic analysis model to predict future debris flow sensitivity.

[0095] The debris flow sensitivity was assessed using a dynamic analysis model from 2008 to 2019. The assessment results are as follows: Figure 11As shown, the accuracy of the evaluation results is greater than 70%, and the ACU value is greater than 0.7, indicating that the debris flow sensitivity dynamic analysis model has good predictive performance for debris flow events from 2008 to 2019.

[0096] A debris flow sensitivity dynamic analysis model was used to predict debris flow sensitivity in 2025, covering the entire debris flow basin. The normalized vegetation index (NDVI) and annual average rainfall (RY) of the predicted coseismic landslides in 2025 were input. The prediction results are as follows: Figure 12 As shown in Table 3, the classification criteria for debris flow sensitivity levels are given using the natural discontinuity method. It can be seen that with the evolution of post-earthquake sediment sources, the overall sensitivity of debris flows shows a decreasing trend, but debris flow basins with larger drainage areas have relatively higher sensitivity.

[0097] Table 3 Classification of Debris Flow Sensitivity Levels

[0098]

[0099] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for dynamic analysis of post-earthquake rainstorm debris flow sensitivity, characterized in that: Includes the following steps: Step 1: Collect sample datasets of several years in the post-earthquake debris flow basin of the study area. The sample datasets contain several factors. Resample the sample dataset to create multiple sets of sample datasets; The step of resampling the sample dataset to establish multiple sets of sample datasets includes: merging sample datasets from adjacent years, selecting the number of years N to be merged, and performing rolling merging of the sample datasets for each year. Step 2: Perform collinearity detection on the factors in each group of sample datasets to select the factors used for model building, thereby establishing a debris flow sensitivity analysis model; and select the best sample dataset based on the performance evaluation results of the debris flow sensitivity analysis model. The steps of performing collinearity detection on factors in each group of sample datasets, selecting factors for model building, and thus establishing a debris flow sensitivity analysis model include: For each period of data in each sample dataset, collinearity detection is performed on k factors. The variance inflation coefficient (VIF) of the factors in the sample dataset is calculated. Factors with a VIF greater than a set threshold are deleted, leaving n factors, where n ≤ k. A debris flow sensitivity analysis model was established using the remaining n factors via logistic regression: （1） In equation (1), P represents the sensitivity to debris flow outbreaks, and its value ranges from [0,1]. Z is a linear fitting equation containing a set of factors related to debris flow outbreaks, and its form is as follows: （2） In equation (2), b0 is the intercept; b n X represents the partial regression coefficient; n As a factor variable; The step of selecting an optimal set of sample datasets based on the performance evaluation results of the debris flow sensitivity analysis model includes: The debris flow sensitivity analysis model constructed for each set of sample datasets was evaluated using cross-validation. Based on the evaluation results, the best set of sample datasets was selected. The performance evaluation prediction results include a confusion matrix table, prediction accuracy, and area under the prediction feature curve; the confusion matrix table includes four indicators: debris flow outbreak prediction accuracy, debris flow non-outbreak prediction accuracy, debris flow outbreak prediction error rate, and debris flow non-outbreak prediction error rate. Step 3: Extract the weight values ​​of factors in the best sample dataset, fit the dynamic curve of the weight values ​​of each factor changing over time, and establish a dynamic factor regression equation; substitute the dynamic factor regression equation into the debris flow sensitivity analysis model to obtain the debris flow sensitivity dynamic analysis model. The steps of substituting the dynamic factor regression equation into the debris flow sensitivity analysis model to obtain the debris flow sensitivity dynamic analysis model include: If the optimal sample dataset corresponds to n factors, substituting the dynamic factor regression equation of the factors in the optimal sample dataset into equation (2) yields the dynamic linear fitting equation for debris flow: （3） In equation (3), f0 is the constant term regression equation, and f1~f n These are the dynamic factor regression equations for n factors; Step 4: Use the debris flow sensitivity dynamic analysis model to predict future debris flow sensitivity.

2. The method for dynamic analysis of post-earthquake rainstorm debris flow sensitivity according to claim 1, characterized in that: The factors in the sample dataset include topographic factors, sediment source factors, and precipitation factors; The topographic factors include watershed area, watershed elevation difference, channel longitudinal gradient, channel length, and Melton intensity index. The source factors include lithology, distance from the fault, area density of coseismic landslides in the watershed, average vegetation normalization index of coseismic landslides in the watershed, average vegetation normalization index of coseismic landslides in the watershed, and vegetation restoration index of coseismic landslides in the watershed. The rainfall-related factors include the annual average rainfall.

3. The method for dynamic analysis of post-earthquake rainstorm debris flow sensitivity according to claim 1, characterized in that: The steps for predicting future debris flow sensitivity using a debris flow sensitivity dynamic analysis model include: Input the predicted year, the vegetation normalization index of the predicted coseismic landslide, and the predicted annual average rainfall into the debris flow sensitivity dynamic analysis model, and the model will output the debris flow sensitivity of the predicted year.

Images