Because different database and functional forms are used in developing these empirical prediction models, it is not surprising that these models would yield different predicted results. Uncertainty associated with the Newmark displacement models can be classified as two categories: epistemic uncertainty and aleatory uncertainty. Epistemic uncertainty is due to lack of knowledge. In principal, it can be reduced by using sufficient data or improved regression techniques. The epistemic uncertainty can be approximately evaluated by the variation of different model predictions. Aleatory variability, on the other hand, represents inherent randomness that can not be reduced. Aleatory variability is usually quantified by variation of the observed data against the model prediction. Recently, Douglas (2010, 2012) studied the consistency of


Newmark displacement model is commonly used to estimate the seismic performance of slopes during earthquake (Newmark, 1965). The Newmark displacement model assumes the sliding mass is rigid, and sliding occurs on a predefined interface. The critical acceleration (ac) represents the resistance of the slope against sliding. It can be determined by the strength of material and the slope angle etc. Sliding is initialized when the shaking acceleration exceeds the critical acceleration, and the block displaces plastically along the interface. The permanent displacement D is calculated by double integrating the exceeded accelerations with respect to time (Fig. 1). Although the simple rigid-plastic model does not consider the deformation of the block itself during shaking, this method has been widely used to evaluate earthquake-induced displacement for natural slopes (Jibson 2007).