A Probabilistic Framework for Vs30

Date: 4/24/2019

Time: 08:30 AM

Room: Elliott Bay

The time-averaged shear-wave velocity (V_S) in the upper 30 m from the surface (V_S30), is used as an index to quantify seismic site effects and model ground motions. V_S30 is typically derived from in-situ recordings of V_S; however, proxies, as a function of topography, geology, and terrain properties, are used where observations are sparse, or not readily available. Recently, Iwahashi et al (2018) developed a map where global terrain was classified using a 280-m digital elevation model (DEM) down-scaled from the Multi-Error-Removed Improved-Terrain (MERIT) DEM (Yamazaki et al., 2017). In this study, we present a framework to predict the most-probable value of V_S30 for each terrain class. Our framework is grounded in the fundamental principles of geostatistics and probability and uses maximum-likelihood-estimates to optimally identify the non-gaussian distribution of V_S30. We show that a non-gaussian distribution of V_S30 is to be expected and is not a sign of measurement error or sampling bias. This implies that a simplistic approach of using the mean to quantify V_S30 can overestimate the most probable value. Our framework also lends itself to provide probabilistic bounds on the variation of V_S30 for a given terrain class. Quantifying uncertainty in V_S30 distribution can reduce uncertainty involving site effects and ground motion models. Our future work entails using machine learning to generate a synthetic V_S30 database which can complement field measurements by increasing resolution and providing estimates where field observations are not available.

Presenting Author: Utkarsh Mital

Authors