Empirical Bayes regionalization techniques for estimation in ground water models
Groundwater contaminant transport risk analysis requires the estimation of parameters exhibiting strong spatial correlation, such as the hydraulic conductivity tensor. Analysis must often be undertaken under conditions of limited data. The estimation process may be strengthened, and the estimation risk reduced by utilizing regional data. The regionalization process is formalized through Parametric empirical Bayes analysis. New Empirical Bayes solutions are proposed for several cases involving spatial stochastic processes. These determine iterative solutions through an Expectation-Maximization algorithm. Tests on synthetically generated data indicate that the Empirical Bayes estimators show performance superior to simple regression and kriging estimators on an average squared prediction error criterion. The methods are applied to the problem of predicting transmissivities in East-Central Illinois glacial drift. Prospects for other useful applications appear good. However, synthetic data test indicate that the Empirical Bayes estimators will have a tendency to underestimate the total process variance because the uncertainty in determining the prior parameter values is ignored.
- Research Organization:
- Duke Univ., Durham, NC (USA)
- OSTI ID:
- 6797965
- Resource Relation:
- Other Information: Thesis (Ph. D.)
- Country of Publication:
- United States
- Language:
- English
Similar Records
NONPARAMETRIC EMPIRICAL BAYES ESTIMATION THROUGH DECONVOLUTION IN PROBABILISTIC RISK ANALYSIS
Predicting Biological Cleanliness: An Empirical Bayes Approach for Spacecraft Bioburden Accounting
Related Subjects
ENVIRONMENTAL TRANSPORT
MATHEMATICAL MODELS
GROUND WATER
CONTAMINATION
ALGORITHMS
FORECASTING
HYDRAULIC CONDUCTIVITY
RISK ASSESSMENT
STOCHASTIC PROCESSES
WATER POLLUTION
HYDROGEN COMPOUNDS
MASS TRANSFER
MATHEMATICAL LOGIC
OXYGEN COMPOUNDS
POLLUTION
WATER
540320* - Environment
Aquatic- Chemicals Monitoring & Transport- (1990-)