Prairie Research Institute
Event Detail Information
ISGS Seminar Series: A Bayesian Framework for Uncertainty Quantification and its Implementation using Sparse-Grid Collocation Schemes: with Application to Groundwater Reactive Transport Modeling
In simulating the complex subsurface environment, an obstacle to efficient and effective quantification and reduction of predictive uncertainty is that existing methods of uncertainty quantification are too fragmented and incomplete for understanding and predicting the complex subsurface environment as a whole. To tackle this problem, we developed a comprehensive Bayesian framework from a system perspective. It uses a hierarchical structure to characterize uncertainty in model scenarios, structures, parameters, and data used for the modeling. Variance decomposition is used to quantify relative contribution from the various sources to predictive uncertainty. Based on the variance decomposition, the Sobol’ global sensitivity index is extended from parametric uncertainty to consider model and scenario uncertainty, and individual parameter sensitivity index is estimated with consideration of multiple models and scenarios. The framework is implemented using the Bayesian network, in which different uncertainty sources are described as uncertain nodes. All the nodes are characterized by multiple states, representing their uncertainty in the form of continuous and discrete probability distributions. After building the Bayesian network, we used the sparse-grid collocation schemes to enhance computational efficiency. We demonstrate the use of the developed method for groundwater reactive transport. The example considers three scenarios of precipitation due to climate change, two models that convert precipitation to groundwater recharge, and multiple random parameters of hydraulic conductivity and kinetic reaction rates. While the example is for groundwater reactive transport modeling, our methods are applicable to a wide range of environmental models. The results of uncertainty quantification and sensitivity analysis are useful for environmental management and decision-makers to formulate science-informed policies and strategies.
Bio: Dr. Ming Ye is an Associate Professor in the Department of Scientific Computing at the Florida State University (FSU), Tallahassee, FL. His research is mainly focused on groundwater numerical modeling and uncertainty analysis. He holds a B.S. degree in Geology from the Nanjing University, China. In 2002, he earned his Ph.D. degree in Hydrology from the University of Arizona, Tucson, AZ. Before joining FSU, he worked as a post-doc in the Pacific Northwest National Laboratory and the Assistant Research Professor at the Desert Research Institute. He received the 2012 DOE Early Career Award, and was elected as a Fellow of the Geological Society of America in 2012. He is serving as an Associate Editor of Water Resources Research.