A shallow landslide susceptibility prediction method and system based on a single landslide triggering event

Abstract

The present application belongs to the technical field of disaster prevention and reduction, aiming at the problems of the current shallow landslide susceptibility prediction research being not rigorous, not accurate, not objective and the like, a shallow landslide susceptibility prediction method and system based on single landslide triggering event are disclosed, a new logic for predicting susceptibility is proposed, by first constructing a probability risk model of triggering landslide by rainfall / earthquake, and then eliminating the influencing factors of the event, the true susceptibility prediction result of the shallow landslide in the region is obtained. The present application adopts the reverse thinking method of deducing the landslide susceptibility model from the landslide probability risk model, which improves the prediction accuracy of the susceptibility of the regional shallow landslide.

Description

Technical Field

[0001] This invention relates to the field of disaster prevention and mitigation technology, and in particular to a method and system for predicting the susceptibility of shallow landslides based on a single landslide triggering event. Background Technology

[0002] In existing techniques for landslide susceptibility assessment, a common approach is to use historical landslide data as samples and combine various machine learning techniques to train predictive models. For example, Multilayer Perceptron (MLP), Random Forest (RF), Support Vector Machine (SVM), and deep learning models are widely used. These methods construct models to predict potential landslide areas by analyzing the relationship between historical landslide data and environmental factors. However, these approaches have limitations, primarily due to the inconsistent time spans of historical landslide events, leading to significant differences in data detail across different regions. Furthermore, the selection of positive and negative samples significantly impacts model accuracy and can result in overfitting. These issues limit the model's universality and accuracy across different regions and conditions.

[0003] Traditional landslide susceptibility prediction methods struggle to accurately assess the susceptibility of shallow landslides in a region. First, traditional methods rely on historical landslide samples or landslide remains visually interpreted from high-resolution satellite imagery. The timing and type of these landslides are often unclear, making accurate prediction of shallow landslide susceptibility difficult. Second, most existing methods focus on collecting landslide sample data and then training various machine learning models to compare performance, without considering the logical rigor of landslide susceptibility prediction. This leads to numerous studies that confuse susceptibility and hazard prediction. Furthermore, no accurate method for predicting shallow landslide susceptibility has yet been found in the current research field. Summary of the Invention

[0004] This invention addresses the problems of current research on shallow landslide susceptibility prediction being unrigorous, imprecise, and lacking objectivity. It proposes a method and system for predicting shallow landslide susceptibility based on a single landslide triggering event. By utilizing landslide data triggered by a single rainfall / earthquake event and combining it with multiple environmental factors, a landslide probabilistic hazard model is first constructed for the region. Then, the influence of rainfall or earthquake events is eliminated, enabling the model to more accurately predict the shallow landslide susceptibility of the region.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] This invention proposes a method for predicting the susceptibility of shallow landslides based on a single landslide triggering event, comprising:

[0007] Step 1: Based on high-resolution, low-cloud-coverage optical remote sensing images of the area affected by a heavy rainfall / earthquake event, acquire recent rainfall-induced landslide disaster data / earthquake-triggered landslide disaster data for the target area, and complete the construction of the regional landslide database;

[0008] Step 2: Select multiple landslide influencing factors based on the regional landslide geological hazard development characteristics, and then normalize each landslide influencing factor to establish a landslide influencing factor dataset;

[0009] Step 3: Select landslide sample points and non-landslide sample points within the landslide area based on the regional landslide database;

[0010] Step 4: Based on the landslide impact factor dataset and the selected landslide sample points and non-landslide sample points, analyze the multicollinearity and contribution of each landslide impact factor, and establish a landslide training sample set and a test sample set.

[0011] Step 5: Based on the established landslide training sample set, train a shallow landslide probability hazard model built based on a machine learning model;

[0012] Step 6: Evaluate the performance of the constructed landslide probabilistic hazard model using a landslide test sample set, and obtain the optimal landslide probabilistic hazard model;

[0013] Step 7: Input all landslide impact factor data of a single landslide triggering event into the optimal landslide probability hazard model to obtain the landslide probability hazard results for regional heavy rainfall / earthquake events;

[0014] Step 8: Set the factor variables of landslide triggering events in the landslide impact factors to uniform values, and input them again into the optimal landslide probability hazard model for calculation to obtain the susceptibility results of shallow landslides in the target area.

[0015] Furthermore, the landslide influencing factors include: historical earthquake parameters, elevation, slope, aspect, profile curvature, slope position, topographic relief, topographic humidity index, land use type, vegetation cover, stratigraphic lithology, distance from active fault, distance from river, and average annual rainfall over the past five years.

[0016] Further, step 3 includes:

[0017] Points were randomly selected evenly throughout the entire rainfall / earthquake-affected area. Points falling within the landslide area were designated as landslide sample points, while points falling outside the landslide area were designated as non-landslide sample points.

[0018] Further, step 3 includes:

[0019] Landslides are extracted to points and all of them are used as landslide sample points. The selection of non-landslide sample points depends on the ratio of landslide area to non-landslide area. The ratio of the number of non-landslide sample points to the number of landslide sample points is consistent with the ratio of non-landslide area to landslide area.

[0020] Further, step 4 includes:

[0021] The variance inflation factor analysis method was used to analyze the multicollinearity of each landslide influencing factor, and landslide influencing factors with variance inflation factors not greater than 10 were selected to establish landslide training sample set and test sample set.

[0022] Further, step 4 includes:

[0023] The contribution of landslide influencing factors to the probability of landslide occurrence is analyzed by using a multiple linear regression model. The importance of each landslide influencing factor is determined by the standardized regression coefficients. The larger the absolute value of the coefficient, the greater the contribution of the factor to the landslide.

[0024] Furthermore, in step 5, the machine learning model includes: gradient boosting decision tree, extreme gradient boosting tree, random forest, and logistic regression algorithm.

[0025] Further, step 8 includes:

[0026] The susceptibility of shallow landslides was classified into five levels according to the natural discontinuity method, corresponding to extremely low susceptibility areas, low susceptibility areas, medium susceptibility areas, high susceptibility areas, and extremely high susceptibility areas, respectively.

[0027] Another aspect of this invention proposes a shallow landslide susceptibility prediction system based on a single landslide triggering event, comprising:

[0028] The first data construction module is used to acquire recent rainfall-induced landslide disaster data / earthquake-triggered landslide disaster data of the target area based on high-resolution low cloud coverage optical remote sensing images of the area affected by a heavy rainfall / earthquake event, and to complete the construction of the regional landslide database;

[0029] The second data construction module is used to select multiple landslide influencing factors based on the development characteristics of regional landslide geological hazards, and then normalize each landslide influencing factor to establish a landslide influencing factor dataset.

[0030] The sample point selection module is used to select landslide sample points and non-landslide sample points within the landslide area based on the regional landslide database.

[0031] The data analysis module is used to analyze the multicollinearity and contribution of each landslide influencing factor based on the landslide influencing factor dataset and selected landslide and non-landslide sample points, and to establish landslide training and test sample sets.

[0032] The model training module is used to train a shallow landslide probability hazard model based on a machine learning model, using an established landslide training sample set.

[0033] The model evaluation module is used to evaluate the performance of the constructed landslide probability hazard model through a landslide test sample set, and to obtain the optimal landslide probability hazard model.

[0034] The first model prediction module is used to input all landslide impact factor data of a single landslide triggering event into the optimal landslide probability hazard model to obtain the landslide probability hazard results of regional heavy rainfall / earthquake events.

[0035] The second model prediction module is used to set the factor variables of landslide triggering events in the landslide influencing factors to uniform values, and then input them again into the optimal landslide probability hazard model for calculation to obtain the susceptibility results of shallow landslides in the target area.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] This invention employs a reverse thinking approach, deriving a shallow landslide susceptibility model from a shallow landslide probabilistic hazard model, thus solving the problems of inaccurate and subjective regional shallow landslide predictions. Compared to traditional methods, this invention makes landslide susceptibility model research more precise, effectively avoiding the confusion between landslide susceptibility models and landslide probabilistic hazard models. Furthermore, this invention provides a clear technical process for predicting landslide probabilistic hazard under different rainfall or earthquake scenarios, based on the construction of a landslide probabilistic hazard model. Attached Figure Description

[0038] Figure 1 This is a flowchart illustrating a method for predicting the susceptibility of shallow landslides based on a single landslide triggering event, according to an embodiment of the present invention.

[0039] Figure 2 This is a schematic diagram of the architecture of a shallow landslide susceptibility prediction system based on a single landslide triggering event, according to an embodiment of the present invention. Detailed Implementation

[0040] The present invention will be further explained below with reference to the accompanying drawings and specific embodiments:

[0041] like Figure 1 As shown, a method for predicting the susceptibility of shallow landslides based on a single landslide triggering event includes:

[0042] Step S101: Based on high-resolution low-cloud-coverage optical remote sensing images of the area affected by a heavy rainfall / earthquake event, acquire recent rainfall-induced landslide disaster data / earthquake-triggered landslide disaster data for the target area, and construct a regional landslide database;

[0043] Step 102: Select multiple landslide influencing factors based on the development characteristics of landslide geological hazards in the region, and then normalize each landslide influencing factor to establish a landslide influencing factor dataset;

[0044] Step 103: Select points within the landslide area as sample points. It's important to note that the selection of landslide and non-landslide sample points has strict extraction strategies. Two specific methods exist: First, randomly select points evenly across the entire rainfall-affected area. Points falling within the landslide area are considered landslide sample points, while those falling outside are considered non-landslide sample points. Second, extract all landslide points, using them all as landslide sample points. The selection of non-landslide sample points depends on the ratio of landslide area to non-landslide area. The ratio of non-landslide sample points to landslide sample points must be consistent with the ratio of non-landslide area to landslide area.

[0045] Step 104: Based on the landslide impact factor dataset and selected landslide and non-landslide sample points, analyze the multicollinearity and contribution level of each landslide impact factor. As one possible implementation, select landslide impact factors with a variance inflation factor (VIF) no greater than 10 to establish a landslide training and testing sample set;

[0046] Step 105: Based on the established landslide training sample set and test sample set, train a landslide probability hazard model constructed based on a machine learning model;

[0047] Step 106: Evaluate the performance of the constructed model using evaluation metrics such as precision, accuracy, F1 score, and ROC curve, and then construct the optimal model through methods such as model optimization;

[0048] Step 107: At this point, inputting all the influencing factor data into the optimal model will yield the probability hazard results of landslides under regional heavy rainfall / earthquake conditions;

[0049] Step 108: Set the factor variables of landslide triggering events in the influencing factors to uniform values, and input the optimal model calculation results again. The result obtained at this time is the susceptibility result of shallow landslides in this area. As an implementation method, the susceptibility result of shallow landslides can be divided into 5 levels as needed.

[0050] This invention proposes a novel logic for predicting landslide susceptibility. By first constructing a probabilistic hazard model of a landslide triggered by a rainfall / earthquake, and then eliminating the influencing factors of the event, the true susceptibility prediction result of shallow landslides in the region is obtained. This invention employs a reverse thinking method—deriving a landslide susceptibility model from a landslide probabilistic hazard model—which improves the prediction accuracy of shallow landslide susceptibility in the region.

[0051] Furthermore, in step S101, the high-resolution low cloud coverage optical remote sensing images generally select image data with a resolution greater than 6 meters, cloud coverage less than 10%, and no cloud or fog obstruction.

[0052] Furthermore, the landslide database described in step S101 should include only landslide data triggered by rainfall / earthquakes, and the data should include, but is not limited to, spatial and attribute data, such as landslide location, extent boundaries, area, quantity information, etc.

[0053] Furthermore, the landslide database format described in step S101 includes, but is not limited to, “.shp”, “.kml”, and “.kmz”.

[0054] Furthermore, the multiple landslide influencing factors mentioned in step S102 include, but are not limited to: historical earthquake parameters, elevation, slope, aspect, profile curvature, slope position, topographic relief, topographic humidity index, land use type, vegetation cover, stratigraphic lithology, distance from active fault, distance from river, and average annual rainfall over the past five years.

[0055] Furthermore, in step S102, the spatial resolution of all impact factor data in the earthquake landslide impact factor dataset includes, but is not limited to, 12.5m, 30m, and 100m.

[0056] Furthermore, the landslide sample selection method described in step S103 includes, but is not limited to: 1) converting landslide surface features into raster data and extracting each pixel of the converted raster data as a landslide sample point; 2) converting landslide surface features into point features, with one landslide corresponding to one point as a landslide sample.

[0057] Furthermore, the non-landslide sample selection method described in step S103 includes, but is not limited to: 1) extracting the non-landslide area within the study area and randomly selecting points within this range; 2) constructing a fishing net point matrix from the non-landslide area and randomly selecting points from it.

[0058] Furthermore, in step S103, the sample size of the landslide samples triggered by rainfall is selected according to the ratio of landslide to non-landslide area, including but not limited to: 1) taking a large number of random points in the entire study area, and then counting whether the area where the point is located belongs to the landslide area, and labeling each point as a landslide or non-landslide sample; 2) first randomly selecting an appropriate amount of point data in the landslide area, and then selecting corresponding sample points in the non-landslide area according to the ratio of landslide to non-landslide area. As an implementation method, any landslide sampling method that can achieve the area ratio can be used.

[0059] Further, in step S104, the multicollinearity analysis methods for landslide influencing factors include, but are not limited to: 1) Variance Inflation Factor (VIF) analysis: By calculating the variance inflation factor (VIF) value of each influencing factor, the degree of collinearity among variables is determined. When the VIF value is higher than a specific threshold (such as 5, 10, or 70), it indicates that the variables are collinear. 2) Correlation coefficient matrix analysis: The correlation coefficients among landslide influencing factors are calculated to generate a correlation coefficient matrix. High correlation coefficients (close to 1 or -1) indicate that there may be collinearity among variables. 3) Condition number analysis: By calculating the condition number of the landslide influencing factor characteristic matrix, the collinearity problem is determined. When the condition number is greater than 30, it indicates that there is severe collinearity.

[0060] Further, in step S104, the contribution methods for landslide influencing factors include, but are not limited to: 1) Multiple regression analysis: Analyzing the contribution of landslide influencing factors to the probability of landslide occurrence using a multiple linear regression model, and using standardized regression coefficients to determine the importance of each factor; the larger the absolute value of the coefficient, the greater the contribution of the factor to the landslide. 2) Random forest feature importance: Calculating the feature importance of each influencing factor using a random forest algorithm; factors with higher importance contribute more to landslide prediction. 3) Information gain: Evaluating the contribution of each influencing factor to landslide classification information using the information gain method; the larger the information gain, the higher the importance of the factor in landslide occurrence.

[0061] Furthermore, in step S104, the methods for dividing the landslide training set and test set include (random and sequential methods, etc.), and the division ratios include 9:1, 8:2, 7:3, and 6:4.

[0062] Furthermore, in step S105, the labeling methods in the earthquake landslide training test sample set include: landslide labels are labeled as "1", non-landslide labels are labeled as "2", or they may be "1" and "0" or two other different characters.

[0063] Further, the machine learning models described in step S105 include, but are not limited to: 1) Gradient Boosting Decision Tree (GBDT): A classic gradient boosting decision tree algorithm that optimizes the model's residuals by progressively building multiple decision trees, with each tree learning the error predicted in the previous step. GBDT performs well in regression and classification tasks. 2) Extreme Gradient Boosting Tree (XGBoost): An improved gradient boosting tree algorithm with higher efficiency and scalability. XGBoost prevents overfitting through parallel processing and regularization, and allows for custom objective functions. 3) CatBoost: A gradient boosting algorithm optimized for categorical features, which processes categorical features through a special objective encoding method, reducing model bias and variance and preventing overfitting; and combined algorithms of these types. 4) Random Forest: Compared to other algorithms, it has advantages such as high accuracy, high generalization ability, and strong noise resistance. 5) Logistic Regression Algorithm: Compared to other algorithms, it has advantages such as high computational efficiency, applicability to linearly separable data, and ease of expansion.

[0064] Furthermore, the model evaluation methods described in step S106 include, but are not limited to, methods such as accuracy, precision, F1 score, and ROC curve.

[0065] Furthermore, in step S107, the value range of the landslide probability hazard result includes 0-1 or other ranges greater than 0 and less than 1; the specific value depends on the calculation result.

[0066] Furthermore, in step S107, all landslide impact factor data include the actual parameters of a single landslide triggering event, and at this point, no modifications must be made to the input data.

[0067] Furthermore, in step S108, the factor variable of the landslide triggering event in the landslide impact factor is set to a uniform value. This value can be any value, as long as it is the same, and the values ​​of other variables are kept unchanged.

[0068] Furthermore, in step S108, the shallow landslide susceptibility result is no longer a landslide susceptibility in the traditional sense, but only represents the shallow landslide susceptibility within the area affected by the event.

[0069] Furthermore, in step S108, the influencing factor variables of the model can be entered again beyond the study area, but the event-related factors still need to be kept at the same set value.

[0070] Furthermore, in step S108, the factor variable of the landslide triggering event in the landslide impact factor is set to a uniform value in order to eliminate the impact of the triggering event. Other methods that can achieve the same purpose are also acceptable.

[0071] Furthermore, in step S108, the zoning method for shallow landslide susceptibility includes, but is not limited to: the natural discontinuity method, which divides the susceptibility level into 5 levels: extremely low susceptibility zone, low susceptibility zone, medium susceptibility zone, high susceptibility zone and extremely high susceptibility zone; and the equal interval classification method, which divides the 0-1 interval into 5 susceptibility levels at intervals of 0.2.

[0072] Based on the above embodiments, such as Figure 2 As shown, this invention also proposes a shallow landslide susceptibility prediction system based on a single landslide triggering event, comprising:

[0073] The first data construction module is used to acquire recent rainfall-induced landslide disaster data / earthquake-triggered landslide disaster data of the target area based on high-resolution low cloud coverage optical remote sensing images of the area affected by a heavy rainfall / earthquake event, and to complete the construction of the regional landslide database;

[0074] The second data construction module is used to select multiple landslide influencing factors based on the development characteristics of regional landslide geological hazards, and then normalize each landslide influencing factor to establish a landslide influencing factor dataset.

[0075] The sample point selection module is used to select landslide sample points and non-landslide sample points within the landslide area based on the regional landslide database.

[0076] The data analysis module is used to analyze the multicollinearity and contribution of each landslide influencing factor based on the landslide influencing factor dataset and selected landslide and non-landslide sample points, and to establish landslide training and test sample sets.

[0077] The model training module is used to train a shallow landslide probability hazard model based on a machine learning model, using an established landslide training sample set.

[0078] The model evaluation module is used to evaluate the performance of the constructed landslide probability hazard model through a landslide test sample set, and to obtain the optimal landslide probability hazard model.

[0079] The first model prediction module is used to input all landslide impact factor data of a single landslide triggering event into the optimal landslide probability hazard model to obtain the landslide probability hazard results of regional heavy rainfall / earthquake events.

[0080] The second model prediction module is used to set the factor variables of landslide triggering events in the landslide influencing factors to uniform values, and then input them again into the optimal landslide probability hazard model for calculation to obtain the susceptibility results of shallow landslides in the target area.

[0081] The above description is only a preferred embodiment of the present invention. It should be noted that those skilled in the art can make several improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for predicting the susceptibility of shallow landslides based on a single landslide triggering event, characterized in that, include: Step 1: Based on high-resolution, low-cloud-coverage optical remote sensing images of the area affected by a heavy rainfall / earthquake event, acquire recent rainfall-induced landslide disaster data / earthquake-triggered landslide disaster data for the target area, and complete the construction of the regional landslide database; Step 2: Select multiple landslide influencing factors based on the regional landslide geological hazard development characteristics, and then normalize each landslide influencing factor to establish a landslide influencing factor dataset; Step 3: Select landslide sample points and non-landslide sample points within the landslide area based on the regional landslide database; Step 4: Based on the landslide impact factor dataset and the selected landslide sample points and non-landslide sample points, analyze the multicollinearity and contribution of each landslide impact factor, and establish a landslide training sample set and a test sample set. Step 5: Based on the established landslide training sample set, train a shallow landslide probability hazard model built based on a machine learning model; Step 6: Evaluate the performance of the constructed landslide probabilistic hazard model using a landslide test sample set, and obtain the optimal landslide probabilistic hazard model; Step 7: Input all landslide impact factor data of a single landslide triggering event into the optimal landslide probability hazard model to obtain the landslide probability hazard results for regional heavy rainfall / earthquake events; Step 8: Set the factor variables of landslide triggering events in the landslide impact factors to uniform values, and input them again into the optimal landslide probability hazard model for calculation to obtain the susceptibility results of shallow landslides in the target area.

2. The method for predicting shallow landslide susceptibility based on a single landslide triggering event according to claim 1, characterized in that, The landslide influencing factors include: historical earthquake parameters, elevation, slope, aspect, profile curvature, slope position, topographic relief, topographic humidity index, land use type, vegetation cover, stratigraphic lithology, distance from active fault, distance from river, and average annual rainfall over the past five years.

3. The method for predicting shallow landslide susceptibility based on a single landslide triggering event according to claim 1, characterized in that, Step 3 includes: Points were randomly selected evenly throughout the entire rainfall / earthquake-affected area. Points falling within the landslide area were designated as landslide sample points, while points falling outside the landslide area were designated as non-landslide sample points.

4. The method for predicting shallow landslide susceptibility based on a single landslide triggering event according to claim 1, characterized in that, Step 3 includes: Landslides are extracted to points and all of them are used as landslide sample points. The selection of non-landslide sample points depends on the ratio of landslide area to non-landslide area. The ratio of the number of non-landslide sample points to the number of landslide sample points is consistent with the ratio of non-landslide area to landslide area.

5. The method for predicting shallow landslide susceptibility based on a single landslide triggering event according to claim 1, characterized in that, Step 4 includes: The variance inflation factor analysis method was used to analyze the multicollinearity of each landslide influencing factor, and landslide influencing factors with variance inflation factors not greater than 10 were selected to establish landslide training sample set and test sample set.

6. The method for predicting shallow landslide susceptibility based on a single landslide triggering event according to claim 1, characterized in that, Step 4 includes: The contribution of landslide influencing factors to the probability of landslide occurrence is analyzed by using a multiple linear regression model. The importance of each landslide influencing factor is determined by the standardized regression coefficients. The larger the absolute value of the coefficient, the greater the contribution of the factor to the landslide.

7. The method for predicting shallow landslide susceptibility based on a single landslide triggering event according to claim 1, characterized in that, In step 5, the machine learning models include: gradient boosting decision tree, extreme gradient boosting tree, random forest, and logistic regression algorithm.

8. The method for predicting shallow landslide susceptibility based on a single landslide triggering event according to claim 1, characterized in that, Step 8 includes: The susceptibility of shallow landslides was classified into five levels according to the natural discontinuity method, corresponding to extremely low susceptibility areas, low susceptibility areas, medium susceptibility areas, high susceptibility areas, and extremely high susceptibility areas, respectively.

9. A shallow landslide susceptibility prediction system based on a single landslide triggering event, characterized in that, include: The first data construction module is used to acquire recent rainfall-induced landslide disaster data / earthquake-triggered landslide disaster data of the target area based on high-resolution low cloud coverage optical remote sensing images of the area affected by a heavy rainfall / earthquake event, and to complete the construction of the regional landslide database; The second data construction module is used to select multiple landslide influencing factors based on the development characteristics of regional landslide geological hazards, and then normalize each landslide influencing factor to establish a landslide influencing factor dataset. The sample point selection module is used to select landslide sample points and non-landslide sample points within the landslide area based on the regional landslide database. The data analysis module is used to analyze the multicollinearity and contribution of each landslide influencing factor based on the landslide influencing factor dataset and selected landslide and non-landslide sample points, and to establish landslide training and test sample sets. The model training module is used to train a shallow landslide probability hazard model based on a machine learning model, using an established landslide training sample set. The model evaluation module is used to evaluate the performance of the constructed landslide probability hazard model through a landslide test sample set, and to obtain the optimal landslide probability hazard model. The first model prediction module is used to input all landslide impact factor data of a single landslide triggering event into the optimal landslide probability hazard model to obtain the landslide probability hazard results of regional heavy rainfall / earthquake events. The second model prediction module is used to set the factor variables of landslide triggering events in the landslide influencing factors to uniform values, and then input them again into the optimal landslide probability hazard model for calculation to obtain the susceptibility results of shallow landslides in the target area.

Images