Applicability of tank models for forecasting landslide risk on individual slopes during heavy rainfall

Abstract

  To utilize tank models in forecasting the risk of landslides on individual slopes during heavy rainfall, it is essential to set model constants that accurately reflect the unsaturated permeability of the ground. This study aimed to verify whether tank models could effectively replicate the unsaturated permeability of various soil samples. To achieve this, a random search method based on genetic algorithms (GA method) was employed to optimize tank model constants. The study focused on two models : the Sugawara model, wherein the seepage rate from the upper tank to the lower tank remains constant, and the Aoki-Hioki model, wherein the seepage rate varies depending on changes in storage height. The reproducibility of vertical one-dimensional unsaturated seepage analysis was then assessed for both tank models. The findings revealed that in the Sugawara model, which is commonly used to calculate soil rainfall indexes, 1) reproducing vertical one-dimensional unsaturated seepage analysis becomes challenging when the unsaturated zone is thick. In contrast ; 2) the Aoki-Hioki model exhibited a relatively high capability to reproduce vertical one-dimensional unsaturated seepage analysis, regardless of the thickness of the unsaturated zone.