Machine learning-based methods are commonly used for landslide susceptibility mapping. Most of the recent publications focused on quantitative analysis, i.e., improving data processing methods, comparing and perfecting the data-driven model itself, but rarely taking the qualitative aspects of the local landslide occurrences into consideration and the further analysis of the key features was always lacking. This study aims to combine qualitative and quantitative analysis and examine its effect on mapping accuracy; based on the feature importance ranks and the related literature, the key features for identifying landslide/non-landslide points of different sub-zones were further analyzed. Before modeling, the study area Yunyang County, Chongqing City, China, was manually divided into four sub-zones based on the information from geological hazards exploration in Chongqing, including the mechanism of landslide formation and sliding failure and geomorphic unit characteristics. Upon the qualitative analysis basis, five grid searches tuned random forest models (one for the whole region and four for the sub-zones independently) were established by 1654 data points and 20 conditioning features. Compared with the conventional data-driven method, the integrated quantitative evaluation based on the qualitative analysis results showed higher reliability, which not only improved the mapping accuracy but also increased the AUC values of all four sub-models, which were 8.8%, 2.3%, 1.9% and 9.1% higher than that of the parent model. Moreover, the quantitative evaluation based on the qualitative analysis revealed the key factors affecting local landslide formation. Therefore, qualitative analysis is recommended in future landslide susceptibility modeling with the additional combination of data-driven methods.

This paper is motivated by the Time Capsule Project (TCP) of the International Society for Soil Mechanics and Geotechnical Engineering (ISSMGE). The historical developments of geotechnical risk and reliability are reviewed for the past six decades. The key features distinguishing geotechnical and structural engineering are the natural origin of the ground and the lack of sufficient data to characterize the ground using the more familiar frequentist interpretation of probability. For the first feature, random field theory is applied to model spatial variability and the random finite element method or other methods are proposed for solving soil-structure interaction problems in spatially variable soil. For the second feature, compilation of databases is essential to serve as priors for Bayesian updating and more recently for Bayesian machine learning. There is a gradual evolution towards reliability-based design because probabilistic methods offer a pathway to address big data and implement data-centric geotechnics as one step towards digital transformation. Given the complexity of the natural ground (known unknowns can be large and there are unknown unknowns), engineering judgment remains important to bridge the gap between theory and practice. However, the role of engineering judgment needs to be updated as modern machine learning methods become more powerful.

In this study, a probabilistic approach for evaluating the bearing capacity of surface footings is discussed. The evaluation is based on a kinematic approach. The considered substrate consists of two different layers of soil: a top layer formed of medium or dense sand followed by a layer of soft clay. The sand layer is assumed to be homogenous, whereas the undrained shear strength of the soft clay layer is assumed to be spatially variable, described by a lognormal random field. The random field is discretized according to Vanmarcke’s spatial averaging along dissipation regions in the considered failure mechanism. The mechanism utilizes plane strain conditions; however, due to consideration of the soil spatial variability in three dimensions, the impact of the length of the foundation on the random bearing capacity evaluation is considered in this study. As a result of the discretization procedure, a set of correlated random variables is obtained (each associated with an individual dissipation region in the failure mechanism). A series of numerical analyses are performed for two thicknesses of the first layer and a set of anisotropic correlation structures for the spatial variability of the undrained shear strength. The proposed method is computationally efficient and allows consideration of three-dimensional spatial variability in soil strength properties. The results are discussed and compared with those obtained by other methods.