ANN deformation prediction model for deep foundation pit with considering the influence of rainfall

Deep foundation pits involving complex soil-water-structure interactions are often at a high risk of failure under heavy rainfall. Predicted deformation is an important index for early risk warning. In the study, an ANN model is proposed based on the Wave Transform (WT), Copula method, Convolutional Neural Network (CNN) and Long Short-Term Memory Neural Network (LSTM). The total deformation was firstly decomposed into low and high frequency components with WT. The CNN and LSTM were then used for prediction of the two components with rolling training and prediction. The input variables of the CNN and LSTM were determined and optimized based on the correlations analysis of Copula method of the two components with different random variables, especially with the rainfall. And finally, the predicted total deformation was obtained by adding the two prediction components. A deep foundation pit in Chengdu, China was taken as a case study, of which the horizontal deformation curves at different measuring points shows three types of developed trend, as unstable, less stable, and stable types. The predictions of the deformations of different development types by the proposed ANN model show high accuracies with a few input variables and can accurately prompt risk warning in advance.

Transient motion of the largest landslide on earth, modulated by hydrological forces

Sea-level rise of the Caspian Sea (CS) during the early Khvalynian (approximately 40–25 ka BP) generated hundreds of giant landslides along the sea’s ancient coastlines in western Kazakhstan, which extended hundreds of kilometers. Although similar landslides have been observed along the present-day coastlines of the CS in the area of a prominent high escarpment, it remains unclear whether some of these ancient landslides are still active and whether the movement is slow or catastrophic, as previously suggested. The present study is the first to show evidence proving that the geomorphic responses to sea-level changes of the CS that were triggered in the Pleistocene are currently active. Using interferometric synthetic aperture radar (InSAR) data, we show that one of these giant landslides occurring along the western shore of the Kara-Bogaz-Gol (KBG) lagoon of the CS presents active transient motion, which makes it the world’s largest active landslide reported thus far. Extending more than 25 km along the eastern coast of the inundated KBG depression in a N–S direction with maximum landward expansion of 5 km from the shoreline to the flat Ustyurt Plateau, this landslide conveys ~ 10 × 10⁹ m³ rocks toward the lagoon at a rate of ~ 2.5 cm/year. This event releases a nearly episodic aseismic moment of 6.0 × 10¹⁰ Nm annually, which is equivalent to the response of an Mw 5.1 earthquake. We analyze the present-day evolution of this giant coastal landslide at high temporal and spatial resolutions using Sentinel-1 radar images acquired on descending and ascending modes every 12 days between 2014 and 2020. Modelling with elastic dislocations suggests that the KBG landslide was accommodated mostly by a shallow basal décollement with a nearly horizontal listric slip plane. Moreover, our analysis reveals week-long accelerating slip events at changing amplitudes that occur seasonally with slow, lateral spreading rather than sudden catastrophic motion. A strong correlation between the episodic slip events and seasonal water-level changes in the KBG lagoon suggests a causative mechanism for the transient accelerating slip events. Although water-level changes are widely acknowledged to trigger transient motion on a land mass, such movement, which is similar to a silent earthquake, has not been observed thus far at this mega scale; on an extremely low-angle detachment planes at < 5° with modulation by sea-level changes. This study suggests that present-day sea-level changes can reactivate giant landslides that originated 40–25 ka.

Review of Satellite Interferometry for Landslide Detection in Italy

Landslides recurrently impact the Italian territory, producing huge economic losses and casualties. Because of this, there is a large demand for monitoring tools to support landslide management strategies. Among the variety of remote sensing techniques, Interferometric Synthetic Aperture Radar (InSAR) has become one of the most widely applied for landslide studies. This work reviews a variety of InSAR-related applications for landslide studies in Italy. More than 250 papers were analyzed in this review. The first application dates back to 1999. The average production of InSAR-related papers for landslide studies is around 12 per year, with a peak of 37 papers in 2015. Almost 70% of the papers are written by authors in academia. InSAR is used (i) for landslide back analysis (3% of the papers); (ii) for landslide characterization (40% of the papers); (iii) as input for landslide models (7% of the papers); (iv) to update landslide inventories (15% of the papers); (v) for landslide mapping (32% of the papers), and (vi) for monitoring (3% of the papers). Sixty-eight percent of the authors validated the satellite results with ground information or other remote sensing data. Although well-known limitations exist, this bibliographic overview confirms that InSAR is a consolidated tool for many landslide-related applications.

TSM—Tracing Surface Motion: A Generic Toolbox for Analyzing Ground-Based Image Time Series of Slope Deformation

Passive sensors such as multi-spectral (e.g., Single Lens Reflex, SLR) cameras are increasingly being used for geohazards monitoring (landslides, cliffs affected by rock falls, ice glaciers, and volcano flanks) because of their low cost compared to expensive terrestrial laser scanner (TLS) or radar imaging (GB-InSAR) systems. Indeed, due to the large consumer market, sensor resolution and quality (e.g., gain, dynamic range, and geometry) are increasing rapidly. For gravitational processes, such as landslides, recent research has focused on the development and implementation of image correlation techniques to estimate the spatial shift between at least a pair of images by maximizing a cross-correlation function. A generic and fully automated pipeline is proposed for the processing of long image time series acquired for several site configurations. The system associates modules for 1) the selection of the image sequences, 2) the registration of the image stacks and the correction of the camera movements, and 3) the calculation of the terrain motion using change detection approaches. The system is based on the open-source photogrammetric library MicMac and tailored for the processing of monoscopic images. A sensitivity analysis is conducted to design and test the image processing for two use cases respectively the Chambon landslide (Isère, France) characterized by slow motion (< 10 cm.day−1), and the Pas de l’Ours landslide (Hautes-Alpes, France) characterized by moderate motion (> 50 cm.day−1). Four categories of parameters are tested: the image modality, the image matching parameters, the size of the stable area used in the co-registration stage, and the strategy used to combine the images in the time series. The application of the pipeline on the two use cases provides information about the kinematics and the spatial behavior of the landslides.

A novel method for landslide displacement prediction by integrating advanced computational intelligence algorithms

Landslide displacement prediction is considered as an essential component for developing early warning systems. The modelling of conventional forecast methods requires enormous monitoring data that limit its application. To conduct accurate displacement prediction with limited data, a novel method is proposed and applied by integrating three computational intelligence algorithms namely: the wavelet transform (WT), the artificial bees colony (ABC), and the kernel-based extreme learning machine (KELM). At first, the total displacement was decomposed into several sub-sequences with different frequencies using the WT. Next each sub-sequence was predicted separately by the KELM whose parameters were optimized by the ABC. Finally the predicted total displacement was obtained by adding all the predicted sub-sequences. The Shuping landslide in the Three Gorges Reservoir area in China was taken as a case study. The performance of the new method was compared with the WT-ELM, ABC-KELM, ELM, and the support vector machine (SVM) methods. Results show that the prediction accuracy can be improved by decomposing the total displacement into sub-sequences with various frequencies and by predicting them separately. The ABC-KELM algorithm shows the highest prediction capacity followed by the ELM and SVM. Overall, the proposed method achieved excellent performance both in terms of accuracy and stability.

Quick clay and landslides of clayey soils

We study the rheology of quick clay, an unstable soil responsible for many landslides. We show that above a critical stress the material starts flowing abruptly with a very large viscosity decrease caused by the flow. This leads to avalanche behavior that accounts for the instability of quick clay soils. Reproducing landslides on a small scale in the laboratory shows that an additional factor that determines the violence of the slides is the inhomogeneity of the flow. We propose a simple yield stress model capable of reproducing the laboratory landslide data, allowing us to relate landslides to the measured rheology.