Ying Cao

Conventional landslide susceptibility mapping (LSM) mainly describes long-term predisposing conditions, but it is less sensitive to recent slope deformation. Interferometric synthetic aperture radar (InSAR) provides dynamic deformation evidence, yet valid observations are often spatially discontinuous in mountainous reservoir terrain. To address this limitation, this study proposes an interpretable framework for completing sparse InSAR-derived dynamic activity and incorporating it into landslide susceptibility enhancement in the Changping area of the Three Gorges Reservoir Area. The framework first uses 12 conditioning factors and ensemble learning to generate first-level susceptibility. Small baseline subset interferometric synthetic aperture radar (SBAS-InSAR) deformation is then projected into the slope-oriented direction, and deformation-rate intensity and deformation trends are combined to derive observed activity labels. For activity completion, the active class was constructed from observed active InSAR cells, whereas the non-active class combined 1882 low-activity InSAR cells with randomly sampled proxy-reference cells from the unlabelled background. Extreme Gradient Boosting (XGBoost) was selected to estimate a continuous relative dynamic-activity score across the study area, achieving a mean receiver operating characteristic area under the curve (ROC-AUC) of 0.9002 across five independent proxy-reference sampling repeats. The completed activity layer was integrated with first-level susceptibility at the slope-unit scale and evaluated against a separately compiled supplementary survey inventory comprising 27 landslide records and 25 affected slope units. The enhanced moderate-to-high susceptibility classes captured more affected slope units than the corresponding first-level static classes, while occupying a similar proportion of all slope units. The enhanced moderate-to-high susceptibility classes captured 17 of the 25 affected slope units, compared with 12 for the static classes, a descriptive improvement while classifying a similar proportion of all slope units as moderate-to-high. A conditional comparison within static susceptibility classes further indicated incremental discriminatory information from the InSAR-conditioned activity layer. The completed activity layer was therefore described as an InSAR-conditioned relative dynamic-activity score constrained by the shared environmental predictors. SHapley Additive exPlanations (SHAP) and directed-dependence analyses further characterized the predictive contributions and structural associations of road proximity, built-up intensity, vegetation cover, lithological background, and slope geometry with the completed activity pattern. This framework provides a practical and interpretable route for using discontinuous InSAR observations in susceptibility enhancement.

Landslide displacement prediction is an essential component in landslide early warning system. The displacement prediction based on in-suit monitoring performs excellently but is expensive, which limited its promotion in less-developing regions. In this study, we propose a cost-effective landslide displacement prediction method with the combination of Multi-Temporal Interferometric Synthetic Aperture Radar (MT-InSAR) technique and Artificial Bee Colony and Kernel-based Extreme Learning Machine (ABC-KELM). Two large slow-moving landslides in the Three Gorges Reservoir area, namely Shuping landslide and Muyubao landslide, are selected as study cases. We first extract the surface displacement sequences of both landslides through the MT-InSAR and the spaceborne Copernicus Sentinel-1 SAR imagery. The original displacement sequences are decomposed into trend terms, periodic terms, and noise using wavelet analysis. The modelling inputs of trend and periodic displacements are determined by analyzing the relationship between their influencing factors and deformation. The trend and periodic displacement are respectively predicted using ABC-KELM, and summing both predicted displacement to get total displacement. By comparing the displacement obtained by the Global Positioning System in both landslides, we find the MT-InSAR can monitor landslide displacement accurately. Prediction results demonstrate that the ABC algorithm can effectively optimize the parameters of the KELM. ABC-KELM outperforms the commonly used algorithms of extreme learning machine and support vector machine. Its root mean square error, relation coefficient, and mean absolute percentage error is 5.460, 0.022, and 0.990, respectively. Our proposed method is cost-effective in landslide displacement prediction, which can be recommended in susceptible regions.

In recent years, several catastrophic landslide events have been observed throughout the globe, significantly affecting the loss of lives, infrastructure, everyday life and livelihood. To minimize the impact of landslides and issue early warnings, landslide susceptibility maps (LSM) are essential. Aim to improve the accuracy of LSM, this study applied a random selection of non-landslide samples and low accuracy of individual classifiers using machine learning (ML) techniques, coupled with ensemble learning and ML, for LSM. China's Zigui-Badong section of the Three Gorges Reservoir area (TGRA) was considered a case study. Twelve influencing factors were selected as inputs for modelling, and the relationship between each causal factor and landslide spatial development was quantitatively analyzed. A total of 179 landslides were identified in the present study. About 70 % of the landslide pixels were randomly considered for training, and the remaining 30 % were used for validation. The Logistic Regression model (LR) was applied to produce an initial susceptibility map, and the non-landslide samples were selected within the classified low-susceptibility area. Subsequently, two ML classifiers – the Classification and Regression Tree (CART), and the Multi-Layer Perceptron (MLP), and four coupling models – the CART-Bagging, CART-Boosting, MLP-Bagging, and MLP-Boosting, were utilized for LSM. Finally, the receiver operating characteristics (ROC) curve and statistical analysis were applied for accuracy assessment. The results show that elevation and distance to rivers were the main causal factors of landslide development in the study area. The modeling accuracy of LR-MLP was calculated approx. 0.901, which is higher than the LR-CART (0.889). The LR-MLP-Boosting performed the best with an accuracy of 0.986 followed by the LR-CART-Bagging (0.973), LR-CART-Boosting (0.981), and LR-MLP-Bagging (0.978). The accuracy has been improved compared with the NO-CART, NO-MLP, NO-CART-Bagging, NO-CART-Boosting, NO-MLP-Bagging, and NO-MLP-Boosting models. Four ensemble models outperformed their corresponding classifiers, while Boosting outperforms Bagging. Overall, the combination of ensemble learning and ML effectively improved the accuracy of LSM. The LR model can effectively constrain the selection range of non-landslide samples and enhance the quality of sample selection. Our results show promise to map susceptible landslides locations which will help to monitor for an early warning of the landside.

Landslide susceptibility mapping (LSM) can provide valuable information for local governments in landslide prevention and mitigation. Despite significant improvements in the predictive performance of LSM, it remains a challenge to be carried out in areas with limited availability of data. For example, in the early stage of road construction, landslide inventory data can be particularly scarce, while there is a high need to have a susceptibility map. This study aims to set up a novel procedure for coupling the knowledge-driven and data-driven models for LSM in an area with limited landslide inventory data. In particular, we propose a two-step approach. The first step consists of applying four data-driven models (logistic regression, decision tree, support vector machines, and random forest (RF)) to derive a regional susceptibility map. In the second step, the application of a heuristic model (analytic hierarchy process, AHP) is proposed to calculate a local susceptibility map for the areas with incomplete landslide inventories. The final landslide susceptibility map is obtained by merging the most accurate regional map (RF) with the local map. We apply this novel procedure to a landslide-prone region with developed road construction (National Highway G69) in Wanzhou district, where landslide inventory is difficult to update due to timely recovery from landslide-induced road damage. Results show that the proposed methodology allows identifying new landslide-prone areas, and improving LSM predictive performance, as demonstrated by the fact that two new landslides developed along G69 were perfectly classified in the highly susceptible areas. The results show that implementing the landslide susceptibility assessment with different geographical settings and combining them into best-sensitivity partitions is more accurate than focusing on creating new models or hybrid models.

Landslides are a common natural hazard that causes casualties and unprecedented economic losses every year, especially in vulnerable developing countries. Considering the high cost of in-situ monitoring equipment and the sparse coverage of monitoring points, the Sentinel-1 images and Interferometric Synthetic Aperture Radar (InSAR) technique were used to conduct landslide monitoring and analysis. The Muyubao landslide in the Three Gorges Reservoir area in China was taken as a case study. A total of 37 images from March 2016 to September 2017 were collected, and the displacement time series were extracted using the Stanford Method for Persistent Scatterer (StaMPS) small baselines subset method. The comparison to global positioning system monitoring results indicated that the InSAR processing of the Muyubao landslide was accurate and reliable. Combined with the field investigation, the deformation evolution and its response to triggering factors were analyzed. During this monitoring period, the creeping process of the Muyubao landslide showed obvious spatiotemporal deformation differences. The changes in the reservoir water level were the trigger of the Muyubao landslide, and its deformation mainly occurred during the fluctuation period and high-water level period of the reservoir.

The monitoring and prediction of the landslide groundwater level is a crucial part of landslide early warning systems. In this study, Tangjiao landslide in the Three Gorges Reservoir area (TGRA) in China was taken as a case study. Three groundwater level monitoring sensors were installed in different locations of the landslide. The monitoring data indicated that the fluctuation of groundwater level is significantly consistent with rainfall and reservoir level in time, but there is a lag. In addition, there is a spatial difference in the impact of reservoir levels on the landslide groundwater level. The data of two monitoring locations were selected for establishing the prediction model of groundwater. Combined with the qualitative and quantitative analysis, the influencing factors were selected, respectively, to establish the hybrid Genetic Algorithm-Support Vector Machine (GA-SVM) prediction model. The single-factor GA-SVM without considering influencing factors and the backpropagation neural network (BPNN) model were adopted to make comparisons. The results showed that the multi-factor GA-SVM performed the best, followed by multi-factor BPNN and single-factor GA-SVM. We found that the prediction accuracy can be improved by considering the influencing factor. The proposed GA-SVM model combines the advantages of each algorithm; it can effectively construct the response relationship between groundwater level fluctuations and influencing factors. Above all, the multi-factor GA-SVM is an effective method for the prediction of landslides groundwater in the TGRA.

In view of the customer churn problem faced by banks, this paper will use the Python language to clean and select the original dataset based on real bank customer data, and gradually condense the 626 customer features in the original dataset to 77 customer features. Then, based on the pre-processed bank data, this paper uses logistic regression, decision tree and neural network to establish three bank customer churn warning models and compares them. The results show that the accuracy of the three models in predicting bank loss customers is above 92%. Finally, based on the logistic regression model with better evaluation results, this paper analyses the characteristics of the lost customers for the bank, and gives the bank management suggestions for the lost customers.

Landslides are destructive geological hazards that occur all over the world. Due to the periodic regulation of reservoir water level, a large number of landslides occur in the Three Gorges Reservoir area (TGRA). The main objective of this study was to explore the preference of machine learning models for landslide susceptibility mapping in the TGRA. The Wushan segment of TGRA was selected as a case study. At first, 165 landslides were identified and a total of 14 landslide causal factors were constructed from different data sources. Multicollinearity analysis and information gain ratio (IGR) model were applied to select landslide causal factors. Subsequently, the landslide susceptibility mapping using the calculated results of four models, namely, support vector machines (SVM), artificial neural networks (ANN), classification and regression tree (CART), and logistic regression (LR). The accuracy of these four maps were evaluated using the receive operating characteristic (ROC) and the accuracy statistic. Results revealed that eliminating the inconsequential factors can perhaps improve the accuracy of landslide susceptibility modelling, and the SVM model had the best performance in this study, providing strong technical support for landslide susceptibility modelling in TGRA.