HR: 11:50h
AN: H11I-07    [PDF]
TI: Parameter estimation via risk-based optimization
AU: * Tartakovsky, D M
EM: dmt@lanl.gov
AF: Theoretical Division, Los Alamos National Laboratory, Group T-7, MS B284, Los Alamos National Laboratory, Los Alamos, NM 87545
AU: Berndt, M
EM: berndt@lanl.gov
AF: Theoretical Division, Los Alamos National Laboratory, Group T-7, MS B284, Los Alamos National Laboratory, Los Alamos, NM 87545
AB: Most hydrologic systems are inherently heterogeneous and are characterized by parameters that can be sampled at a few selected locations. Yet, numerical simulations of system behavior require that the system parameters be specified at every point of a computational domain. Traditionally, this is done by means of statistical interpolation schemes, such as kriging, that produce the system parameter fields that are much smoother than their true counterparts. This yields predictions of the system states that provide little, if any, insight into the likelihood of a system failure (the so-called rare events). This problem arises in a variety of applications that range from flood prediction, to contaminant transport in groundwater, to oil and gas extraction or water supply. To resolve this issue, we present a new paradigm for parameter estimation. It is based on risk-based optimization, thus providing decision-makers with best and worst case scenarios of the system behavior.
DE: 1829 Groundwater hydrology
DE: 1832 Groundwater transport
DE: 1869 Stochastic processes
DE: 3210 Modeling
DE: 3260 Inverse theory
SC: Hydrology [H]
MN: 2003 Fall Meeting