Phone: 510-495-8232
Fax: 510-486-5686
Email: xiaoyiliu@lbl.gov
My research focuses on heterogeneity characterization of subsurface media and quantification of the level and consequence of the associated uncertainty. A detailed subsurface characterization is crucial for accurate prediction of the dynamics of subsurface fluids such as water, contaminants, CO2, and brine, and a successful characterization relies on the effective use of measurement data from multiple physical processes as well as the use of advanced computing techniques and technologies such as large-scale optimization and High Performance Computing (HPC). When the characterization results are fed to management models such as a remediation optimization model or a CO2 injection optimization model, I am interested in not only the level of uncertainties (i.e., variance or higher moments) associated with the characterization, but the consequence of uncertainties for the management model. For the same characterization, the consequence varies with the management model; hence a context-specific measure of uncertainty is needed. However, traditional measures such as statistical moments are not context-specific and not suitable for this purpose.
Earth Science Division, Lawrence Berkeley National Laboratory - Jul 2011 - present
Postdoctoral Research Fellow
Geologic CO2 storage in the deep subsurface such as depleted oil reservoirs, with focuses on
Environmental Fluid Mechanics & Hydrology, Stanford University - Sep 2007 – Jun 2011
Graduate Research Assistant
Conducted research on subsurface imaging, model calibration, remediation optimization and value of information in groundwater and environmental problems. More specifically, my research included:
Iowa Institute of Hydraulic Research, University of Iowa - Aug 2004 - Aug 2007
Graduate Research Assistant
Conducted research on subsurface imaging with validation, at lab scale (sandbox) and field scale (Mizunami Underground Research Site), with both hydraulic data (hydraulic tomography) and tracer data (partitioning/non-partitioning tracer tomography).