Search

Datasets containing recorded landslide and non-landslide samples can greatly influence the performance of machine learning (ML) models, which are commonly used in landslide susceptibility mapping (LSM). However, the non-landslide samples cannot be directly obtained. In this study, a pattern-based approach is proposed to improve the LSM process, constructing unsupervised machine learning (UML)–supervised machine learning (SML) collaborative models in which the non-landslide samples can be reasonably selected. Two UML models, the Gaussian mixture model (GMM) and K-means, are introduced to sample the non-landslide datasets with four sampling selections (abbreviated as A, B, C and D, respectively). Then non-landslide patterns recognised by the UML models are learned by the random forest (RF). A new sensitivity index, accuracy improvement ratio (AIR), is defined to evaluate the superiority of these sampling selections. Compared with the GMM-RF model, the K-means-RF model is more capable of recognising non-landslide patterns and providing sufficient and reliable non-landslide samples. The sampling selection A of the K-means-RF with an AIR value of 2.3 is regarded as the best selection. The results indicate that the UML-SML model based on the pattern-based approach offers an effective strategy to find the non-landslide samples and has a better solution to the LSM. © 2022 Informa UK Limited, trading as Taylor & Francis Group.

Loess slopes along the Yellow River in China are prone to failure in response to intensive irrigation. Existing studies found that the wetting front depth subjected to intensive farming water could barely reach 6 m in loess terrace. Long-term field monitoring revealed that irrigation water takes a short time to recharge the groundwater. Yet the dynamic process and mechanism of irrigation water percolation through thick unsaturated loess remains unrevealed. To address this problem, the Heifangtai terrace was selected as the study area and a series of unique full-scale controlled field irrigation experiments were carried out combined with numerical modelling. The variations in water content, pore-water pressure, and electrical resistivity over time were monitored. From the field testing, the unsaturated stratified loess layer could be divided into two vadose zones based on the water percolation characteristics, namely, transient flow zone (0–6.5 m) and steady flow zone (>6.5 m). Variation in water content presented a hysteresis effect in the transient flow zone compared with the steady flow zone. In the whole seepage process, the vertical percolation rate of 0.5 m/day was double that of the horizontal percolation rate. Furthermore, the irrigation water percolating process was modelled using a finite element code, PANDAS, which was validated by the field testing. Numerical simulation results demonstrated that the movement of wetting front was hampered by the air outflow within transient flow zone and that irrigation water reached the deep groundwater table through the thick unsaturated stratified loess within 100 days, which was consistent with the field monitoring data. This research aids in a better understanding of the irrigation water percolation process in unsaturated stratified loess in the loess region. © 2021 Elsevier B.V.

On February 7, 2021, a catastrophic landslide occurred in Chamoli, India on the southern hills of the Himalayas (30.37 degrees N, 79.73 degrees E). About 28 x 10(6) m(3) of landslide mass detached from the mountain face, entrained deposits in the valley and riverbed, and generated a huge debris flood along the Dhauliganga River. In this study, the geomorphological and volumetric characteristics of the disaster chain were interpreted from satellite images. The full process of the disaster chain and the erosion, deposition and flow discharge processes were reproduced using a cell-based analysis program, Erosion-Deposition Debris Flow Analysis (EDDA). The results indicate a peak flow discharge of the debris flood at Tapovan Hydropower Station of about 25,000-28,000 m(3)/s. The main erosion and deposition zones are distributed along the valley floor and the Dhauliganga River, respectively. This study serves as basis for understanding the disaster chain dynamics in high mountain areas.