Collapsible loess soils are widely distributed in semiarid and arid regions. The eroded landforms formed by the unique collapsibility of loess may cause environmental damage that puts lives and property at risk. In this work, the collapse potential, quantitative analyses, and qualitative characterization of the microstructures of loess samples were investigated using double oedometer tests, scanning electron microscopy (SEM), and the digital image analysis method to probe the fundamental mechanisms of loess collapsibility. A large number of field investigations have been carried out to study the role of loess collapsibility during the formation of eroded landforms. The results show that the collapse potential of loess first increases and then decreases in response to increasing vertical pressure. Loess from the different study regions exhibits different microstructure types. Loess collapse is mainly caused by the destruction of large and medium pores, although the contribution of the middle pores on the collapse is more significant. The microtopography formed by loess collapse is conducive to surface runoff concentration and infiltration. The eroded landforms resulting from loess collapse are categorized into four different models based on the catchment area and the discharge area. These results and analyses are important to land use planners and engineers and will aid in controlling and mitigating disaster-prone landforms.

Many modified Kozeny–Carman (KC) equations have been used to predict the saturated permeability coefficient (K_s) of porous media in various fields. It is widely accepted that the KC equation applies to sand but does not apply to clay. Little information is available to clarify this point. The effectiveness of the KC equation will be evaluated via laboratory penetration tests and previously published data, which include void ratio, specific surface area (SSA), liquid limit (LL), and permeability coefficient values. This paper demonstrates how to estimate the SSA of cohesive soil from its LL. Several estimation algorithms for determining the effective void ratio (e_e) of cohesive soil are reviewed. The obtained results show that, compared to the KC equation based on porosity and geometric mean particle size (D_g), the KC equation based on the SSA and the e_e estimation algorithm can best predict the K_s of remolded loess. Finally, issues associated with the predictive power of the KC equation are discussed. Differences between measured and the predicted K_s values may be caused by the uniformity of the reconstructed specimen or insufficient control of the test process and errors in the SSA and e_e.