Nonlinear Dynamic Analyses of Perris Dam Using Transition Probability to Model Interbedded Alluvial Strata

Abstract

This case study presents an application of a conditional transition probability method for interpreting subsurface stratigraphy for the interbedded alluvium underlying Perris Dam, and evaluating the effects of stratigraphic uncertainty on nonlinear dynamic analysis (NDA) results for design earthquake loading. The challenges involved in synthesizing information from different sources (i.e., geologic conditions, site investigation tools, lab data, field classifications) into soil categories for interbedded alluvium were examined. The application of conditional transition probability methods for developing three-dimensional (3D) realizations of the upper Holocene and lower Pleistocene alluvial strata over a 305-m-wide interval along the dam alignment is described, including challenges with insufficient data and limitations involved with utilizing a stationary, geostatistical method for approximating nonstationary geologic conditions. Two-dimensional (2D) NDA models of Perris Dam were created by slicing the 3D transition probability realizations into five 2D cross sections. The constitutive models PM4Sand and PM4Silt were used to model the sand and clay soil categories in the alluvial strata, as well as the different zones in the embankment. The deformations and variability in deformations for each cross section were compared, and sensitivity studies were completed to examine the impact of several factors, including impacts of the small-strain shear modulus for the alluvium, mean lengths and sills for the alluvium categories, strengths for each alluvium soil category, and different ground motions. NDA cross sections of Perris Dam with uniformly (noncategorical) distributed properties were performed with and without additional deterministic embedded soil lenses, and the deformations were compared with transition probability models and deterministic models completed by others. The use of conditional transition probability models for NDAs of Perris Dam, along with implications and lessons for practice, are discussed.