HR: 08:45h
AN: H11E-04 INVITED [Abstracts]
TI: Construction of Objective-oriented Groundwater Models: Worst-Case Parameter Identification and Robust
Experimental Design
AU: * Sun, N
EM: nezheng@ucla.edu
AF: University of California at Los Angeles, 5732L Boelter Hall, UCLA, Los Angeles, CA 90095-1593
United States
AB:
A challenging problem in groundwater modeling is how to construct a reliable model for a real aquifer when its structure is
complex and unknown. The traditional method of model construction uses available data to calibrate a predetermined (or
adjustable) conceptual model. This way can only produce data-driven models: both the complexity of the conceptual model and
the values of its representative parameters are determined only by data. Besides the under- and over-parameterization
problems, the traditional method also has difficulties in reliability analysis and data collection strategy. Without knowing
the complexity and variability in model structure, the Monte-Carlo method for reliability analysis becomes very ineffective,
and the problem of optimal experimental design for model calibration becomes meaningless because the identification of a more
complex structure needs more data.
To avoid the above-mentioned difficulties, a new methodology is presented in this paper that aims at constructing
objective-oriented models instead of data-driven models. The complexity of model structure and the sufficiency of data are
determined by the objectives and accuracy requirements of model application. The mathematical basis of the new methodology
has been established. Sufficient conditions of model structure reducibility and parameter identifiability are derived. Since
the nature of model structure error is systematic rather than random, the presented methodology is a deterministic one that
gives the upper bound of model application error rather than its probability distribution. The worst-case parameter (WCP) of
a structure is a parameter that produces the maximum model application error when the structure is simplified. An effective
algorithm is presented for finding the WCP. The structure error of using one structure to replace another structure now can
be calculated without solving a difficult max-min problem, and the search series for structure optimization now can be
non-nested. Numerical examples show that the WCP of a hydraulic conductivity structure always has a cluster shape and the
homogenization error is insensitive to the increase of structure complexity. This fact may explain why a complex aquifer
structure can often be replaced by a very simple structure for groundwater flow modeling. In certain sense, the paper
presents an approach of global upscaling. The paper also formulates an objective-oriented experimental design problem that is
robust, cost-effective and can provide sufficient data for model calibration. Finally, a step-by-step procedure of
constructing an objective-oriented model is described.
DE: 1829 Groundwater hydrology
DE: 1846 Model calibration (3333)
DE: 1848 Monitoring networks
DE: 1873 Uncertainty assessment (3275)
DE: 3260 Inverse theory
SC: Hydrology [H]
MN: Fall Meeting 2005