Earthquake-induced slope displacement analysis using spatially correlated vector-intensity measure

Realistic prediction of earthquake-induced sliding displacement over a large area is critical for a variety of applications, such as regional-scale risk and loss assessment, landslide-related damage to lifelines, road systems and portfolios of infrastructures. To date, most landslide hazard analyses are only applicable to individual site. The existing methods cannot be applied directly to spatially distributed systems since the key information about the spatial correlation of ground motion intensity measures (such as the peak ground acceleration) cannot be properly considered in the analysis. Compared to a single site analysis, a regional-scale landslide hazard analysis is highly intricate in nature since the spatial distribution of ground motion intensities over the region and interdependency of spatially-distributed components must be considered. The proposed research develops a new analytical and computational framework to rigorously evaluate the earthquake exposure of manmade and natural slopes in a regional scale. The thesis solved the following key technical challenges:;(a) Quantifying the joint occurrence and spatial distribution of various ground motion intensity measures at multiple sites, which is essential to generate regional correlated intensity measures. Observed time history records obtained from a large number of recent earthquake data from California, Japan, Mexico and Taiwan are used to study the relationships between intra-event residuals of intensity measures. Influences of the regional geological condition on the spatial correlation are also investigated for a regional specific application. The importance of spatial correlation is also highlighted using some demonstrated examples.;(b) Developing a computationally efficient method to stochastically simulate all sources of uncertainties in a fully probabilistic analysis. For a fully probabilistic landslide hazard analysis, three levels of variabilities must be rigorously accounted for: uncertainties existed in earthquake scenario level, ground motion intensity level and the predictive displacement level. Conventional Monte Carlo simulation will result in prohibitive computational cost. Using some state-of-the-art data reduction techniques, the proposed method can accurately simulate the spatially-distributed intensity measures over the whole region at a computational cost reduced by three orders of magnitude compared to the conventional Monte Carlo simulation.;(c) Proposing a new one-step Newmark displacement model using seismological variables (like moment magnitude, rupture distance) rather than intensity measures. The new developed model can yield hazard-consistent predicted displacement compared to other displacement models for various earthquake scenarios. In addition, the computational time for this model is dramatically reduced due to its one-stage process.;(d) Developing some improved models to predict the earthquake-induced permanent displacements of various slopes using multiple intensity measures. The Newmark displacement method is not applicable to deeper landslide. For these flexible slopes, a nonlinear coupled stick-slip deformable sliding model is appropriate to estimate the permanent sliding displacement. Some selected intensity measures are identified to represent different characteristics of earthquakes and can significantly improve the seismic slope displacement predictions under a wide range of slope conditions.