| dc.description.abstract | Burland (1987) proposed that geotechnical analysis includes the assumptions of three models: Geologic Model, Ground Model and Geotechnical Model. The geologic model expresses the geological conditions of the site, including geological history, geological materials, groundwater and formation conditions, etc. The ground model represents the physical and mechanical properties of the geo-materials. The geotechnical model consists of theoretical or simplified assumptions of a predictive model for the analyzed case. In order to achieve a more representative geotechnical analysis result, the uncertainty of the three models must be considered. In the current study, we have focused on the importance of the geological model in a typical geotechnical analysis.
In this study, the impact of the geological model uncertainty on the National Highway No. 3 (NH-3) landslide event was discussed from the perspective of slope stability and debris run-out parameters. Yeh et al. (2021) obtained the sliding planes dip angles of NH-3 from three sources, which are: (1) the regional geological map of the Central Geological Survey, MOEA; (2) site investigation on the sliding plane after the sliding event; (3) the LiDAR data. Yeh et al. (2021) used the Monte Carlo simulation and the limit equilibrium method to calculate safety factors, and finally obtained the probability of failure of slope sliding, however, without considering the volume change of the sliding block due to the change of the dip angle. In this study, we have incorporated the change of debris volume with dip angle in the Monte Carlo Simulation. The results show that with the change of the dip angle (rotation at slope crest), the failure probability is 26.5% (as compared to Yeh et al. (2021) at 20.6%) with regional geological map information; 4.1% (as compared to Yeh et al. (2021) at 2%) with site investigation information after the sliding event; close to 0% with LiDAR-derived data (similar to Yeh et al. (2021)). The above results indicate that different surveying methods may result in different data accuracy. LiDAR-derived dip angle yielded the smallest standard deviation among the three investigation methods, therefore the probability of failure is relatively small.
To explore the uncertainty of joint distribution on the debris run-out parameters, this study established a set of numerical models that are similar to the landslide case through on-site investigation reports. A number of corresponding numerical models were established by changing the number of the joints along the strike and dip directions of the slope. The debris run-out parameters, including the slope crest displacement, toe run-out distance, debris accumulation width and area of each model are analyzed using point estimate method. Probability Density Function of each parameter is analyzed and the risk assessment (probability of exceedance) of each debris run-out parameter was then calculated. The results show that the uncertainty of dip joints has a greater impact on the slope crest displacement than that of the strike joints; the run-out distance of this case is not affected by the uncertainty of the joints, and there is a very high probability that the highway is directly covered by the landslide debris; the width of the accumulated debris is not affected much by the distribution of joints. Finally, the variability of strike joints has a large effect on the coverage area of the accumulated debris. | en_US |