Constructing a Loess Landslide Run-Out Prediction Input Parameter Database through Multi-Objective Optimization Back Analysis

Loess landslides pose a great physical and environmental threat to communities around the Heifangtai Terrace in Gansu Province, China. Evaluating the potentially affected area provides a basis for loess landslide risk assessment and management. However, rheological input parameters for loess landslide run-out analyses involve large uncertainties that often result in biased predictions. To that end, this study collected 20 loess flowslides occurred from 2015 to 2019 in the Heifangtai Terrace. A friction model embedded in Massflow was used to simulate the run-out behavior of loess landslides. The accumulation depths of several control points located on the accumulation area were employed as observation information. A multi-objective optimization back analysis method using a non-dominated solution genetic algorithm (NSGA-II) was proposed to calibrate the optimal rheological parameters for each landslide, thus obtaining a parameter database including 20 datasets for statistical analyses. The proposed back analysis method and obtained rheological parameter database can be used for conducting future landslide run-out prediction in the Heifangtai area.

Keywords: Loess flowslide, Multi-objective optimization, Back analysis, Rheological parameter database, Statistical analysis