HR: 1340h
AN: H13D-1354 [Abstracts]
TI: Semivariogram Estimation Using Ant Colony Optimization and Ensemble Kriging Accounting for Parameter
Uncertainty
AU: * Cardiff, M A
EM: mcardiff@stanford.edu
AF: Stanford University
Department of Civil and Environmental Engineering, Terman Engineering Center M42, Stanford, CA 94305
United States
AU: Kitanidis, P K
EM: peterk@stanford.edu
AF: Stanford University
Department of Civil and Environmental Engineering, Terman Engineering Center M42, Stanford, CA 94305
United States
AB:
In this presentation we revisit the problem of semivariogram estimation and present a modular, reusable, and encapsulated set
of MATLAB programs that use a hybrid Ant Colony Optimization (ACO) heuristic to solve the "optimal fit" problem. Though the
ACO heuristic involves a stochastic component, advantages of the heuristic over traditional gradient-search methods, like the
Gauss-Newton method, include the ability to estimate model semivariogram parameters accurately without initial guesses input
by the user. The ACO heuristic is also superiorly suited for strongly nonlinear optimization over spaces that may contain
several local minima. The presentation will focus on the application of ACO to existing weighted least squares and restricted
maximum likelihood estimation methods with a comparison of results.
The presentation will also discuss parameter uncertainty, particularly in the context of restricted maximum likelihood and
Bayesian methods. We compare the local linearized parameter estimates (or Cramer-Rao lower bounds) with modern Monte Carlo
methods, such as acceptance-rejection. Finally, we present ensemble kriging in which conditional realizations are generated
in a way that uncertainty in semi-variogram parameters is fully accounted for. Results for a variety of sample problems will
be presented along with a discussion of solution accuracy and computational efficiency.
DE: 0510 Agent-based models
DE: 1848 Monitoring networks
DE: 1873 Uncertainty assessment (3275)
DE: 3252 Spatial analysis (0500)
DE: 3260 Inverse theory
SC: Hydrology [H]
MN: Fall Meeting 2005