Date of Award
Doctor of Philosophy (PhD)
Dr. Qiushi Chen, Committee Chair
Dr. C. Hsein Juang, Co-chair
Dr. Ronald D. Andrus
Dr. Jie Zhang
The time-averaged shear-wave velocity in the top 30 meters of subsurface material (VS30) is a widely used parameter when estimating the potential for amplification of seismic waves. Situations often arise where a design VS30 value needs to be chosen from multiple proxy-based VS30 models. This dissertation seeks to assist with the problem of model selection and to improve the overall prediction of VS30 through implementation of a Bayesian framework for model ranking. Furthermore, this dissertation investigates the effects of uncertainty on the model ranking results. In this work, probabilistic methods are developed and implemented to assess the performance of multiple proxy-based VS30 models. The methodology utilizes Maximum Likelihood Estimation (MLE) to evaluate how well a model (or set of models) can predict the sample data against which it is being evaluated. Bayesian Information Criterion (BIC) is used to quantify the relative performance of multiple candidate VS30 models. The proposed method can provide a performance ranking for situations when one model is superior as well as when multiple models show comparable levels of performance. With ranking results, a new VS30 database comprised of a superior set of VS30 predictions based on known information can be obtained, and this is illustrated through the development of a new synthetic VS30 database for California. The method is also applied to other regions of the country, specifically the Seattle and Puget Sound area and the Salt Lake City, Ogden, and Provo area to further demonstrate the new method and explore its applicability to areas with limited data. Enhanced site condition maps for those regions are also developed.
To strengthen confidence in predictions and designs, civil engineers have started to explicitly consider uncertainty in their calculations. The Bayesian method for model ranking presented herein is also presented in a modified form to allow users to include appropriate, available uncertainty information. The effects of uncertainty on the updated site conditions map for California are investigated, and recommendations for appropriate use of uncertainty information in model ranking applications are made. Finally, the new synthetic database is used to inform the hazard information needed when performing a CPT-based liquefaction hazard quantification calculation. Its application is explained alongside an illustrative example in the San Francisco Bay area.
Brownlow, Andrew West, "Evaluation and Uncertainty Quantification of VS30 Models Using a Bayesian Framework for Better Prediction of Seismic Site Conditions" (2017). All Dissertations. 1991.