HR: 08:30h
AN: H51J-03 INVITED     [Abstracts]
TI: A New Paradigm for Groundwater Modeling
AU: * Li, S
EM: lishug@egr.msu.edu
AF: Department of Civil and Environmental Engineering, Michigan State University, East Lansing, MI 48824 United States
AU: Liu, Q
EM: liuqu@msu.edu
AF: Department of Civil and Environmental Engineering, Michigan State University, East Lansing, MI 48824 United States
AU: Afshari, S
EM: afshari1@egr.msu.edu
AF: Department of Civil and Environmental Engineering, Michigan State University, East Lansing, MI 48824 United States
AB: In this talk, we present interim results from an on-going integrated NSF project to develop a new paradigm and a computational steering environment for real-time multi-scale groundwater modeling, data assimilation, information integration, and visualization. The new paradigm has the potential to significantly improve our ability to characterize and predict water and solute pathways and fluxes through geological formations and the fluxes across surface water and groundwater interfaces, to quantify the scale interactions that affect the response of aquifer systems to local and external forcing factors, and to develop model-based tools for integrated resources management. Despite the dramatic growth of computational capability over the past two decades - one that has allowed computational science to become a powerful tool for scientific discovery, our ability to model complex, 3D systems is still severely limited because of the following computational and conceptual challenges: The machine bottleneck. The computation in large-scale modeling, analysis, and visualization increases exponentially with the problem size and the level of details simulated and quickly becomes prohibitive for large problems. This is especially the case for multi-scale and coupled processes modeling, inverse modeling, and uncertainty analysis. The algorithmic bottleneck. Three-dimensional modeling based on a single numerical representation of a multi-scaled system faces an algorithmic bottleneck. The high dimensionality, especially when combined with distorted grids representing multiple scales of heterogeneity, anisotropy, complex stratigraphy, and singular stresses, translates into ill conditioned matrix systems, and causes a host of numerical problems. The scale and data assimilation problem. A fundamental problem in the analysis of groundwater systems is the interplay of data and modeling. Improving how data and models are used, especially across a multitude of scales, has proven to be exceedingly difficult. The prevalent ways of modeling do not properly account for scale interactions and the disparity between model, data, and management scales and are unable to make effective use of the available measurements, resulting in a loss of valuable information. In this on going project, we address these fundamental difficulties systematically. In particular, we develop a "hierarchical and patch dynamics paradigm" (HPDP) for modeling flow and transport across multiple scales - one that may significantly alleviate the computational and conceptual difficulties and an object-oriented "parallel computing" approach that makes the implementation of the HPDP practical. Specifically, the HPDP and the object-oriented "parallel computing" approach: - allows modeling large systems in high resolution without having to solve large matrix systems and alleviate significantly the infamous curse of dimensionality in large scale modeling; - allows modeling complex dynamics incrementally (one scale at a time), obviates the need to use highly distorted grids representing multi-scale variability, and significantly alleviates the algorithmic bottleneck; - provides a scaling ladder to link data and models across multiple scales and to assimilate information from disparate sources; - provides dynamic real-time steering and integration of hierarchical computations, analyses, and visualizations, frees modelers from having to interact offline with subscale modeling patches, and eliminates the associated human bottleneck. A sophisticated hierarchical navigation process and on the fly information integration and visualization would be an invaluable tool for understanding fundamental processes and for practical investigations.
DE: 1805 Computational hydrology
DE: 1829 Groundwater hydrology
DE: 1830 Groundwater/surface water interaction
DE: 1832 Groundwater transport
DE: 1839 Hydrologic scaling
SC: Hydrology [H]
MN: Fall Meeting 2005