A longitudinal analysis of collapsibility with predictions over the southeastern Loess Plateau in China

Abstract

AbstractLoess presents very unique collapsible behaviour due to its special under-compactness, weak cementation and porousness. Many environmental issues and geological hazards including subgrade subsidences, slope collapses or failures, building cracking and so on are directly caused by the collapsible deformation of loess. Such collapsible behaviour may also severe accidents due to sinkholes, underground caves or loess gullies. Moreover, with the increasing demand of construction and development in the loess areas, an in-depth research towards effective evaluation of loess collapsibility is urged. Currently no studies have made attempts to explore a rather complete and representative area of Loess Plateau. This paper thus provides a novel approach on spatial modelling over Jin-Shan Loess Plateau as an extension to experimental studies. The in-lab experiment results have shown that shown that the porosity ratio and collapsibility follow a Gaussian distribution and a Gamma distribution respectively for both sampling areas: Yan’an and Lv Liang. This establishes the prior intuition towards spatial modelling which provides insights of potential influential factors on loess collapsibility and further sets a potential direction of the loess studies by considering an extra dimension of spatial correlation. Such modelling allows robust predictions taken into account of longitudinal information as well as structural parameters and basic physical properties. Water contents, dry densities, pressure levels and elevations of samples are determined to be statistically significant factors which affect the loess collapsibility. All regions in Lv Liang area are at risk of high collapsibility with average around 0.03, out of which roughly a third of them are predicted to be at high risk. Clear spatial patterns of higher expected collapsibility in the southwest comparing to the northeast are shown adjusting for influential covariates. On reference guidelines for potential policy makings, county-level regions with the highest expected loess collapsibility are also identified.