Weighting observations in the context of calibrating ground-water models

M. C. Hill; C. R. Tiedeman

Home/Archive/USGS

USGS · 70024248

Weighting observations in the context of calibrating ground-water models

M. C. Hill·C. R. Tiedeman

View original source ↗Export metadata ↓

Abstract

This paper investigates four issues related to weighting observations in the context of ground-water models calibrated with nonlinear regression: (1) terminology, (2) determining values for the weighting, (3) measurement and model errors, and (4) the effect weighting can have on the accuracy of calibrated models and measures of uncertainty. It is shown that the confusing aspects of weighting can be managed, and are not a practical barrier to using regression methods.

Explore related subjects

Acta Universitatis Carolinae, Geologica Article

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

M. C. Hill, C. R. Tiedeman. 2002. Weighting observations in the context of calibrating ground-water models. https://pubs.usgs.gov/publication/70024248

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related USGS reports

USGS2002

Evaluation of longitudinal dispersivity estimates from forced-gradient tracer tests in heterogeneous aquifers

Converging radial-flow and two-well tracer tests are simulated in two-dimensional aquifers to investigate the effects of heterogeneity and forced-gradient test configuration on longitudinal dispersivity (??L) estimates, and to compare ??L estimates from forced-gradient tests with ??L values that characterize solute spreading under natural-gradient flow. Results indicate that in mildly heterogeneous aquifers, for tests with relatively large tracer transport distances, ??L estimates from the two test types are generally similar, and are also similar to ??L values determined from natural-gradient tracer simulations. In highly heterogeneous aquifers, ??L estimates from two-well tests are generally larger than those from radial-flow tests, and the ??L estimates from both test types are typically smaller than the ??L values determined from natural-gradient simulations.

Acta Universitatis Carolinae, Geologica↗

USGS2002

Parameter estimation using carbon-14 ages: Lessons from the Danube-Tisza interfluvial region of Hungary

Parameter estimation was conducted on a groundwater model of the Danube-Tisza interfluvial region of Hungary. The model was calibrated using 300 water levels and 48 14C ages. The model provided a test of regression methods for a system with a large number of observations. Up to 103 parameters representing horizontal and vertical hydraulic conductivities and boundary conductances were assigned using point values and bilinear interpolation between points. The lowest errors were obtained using an iterative approach with groups of parameters, rather than estimating all of the parameters simultaneously. The model with 48 parameters yielded the lowest standard error of regression.

Acta Universitatis Carolinae, Geologica↗

USGS2002

Conceptual uncertainty in crystalline bedrock: Is simple evaluation the only practical approach?

A simple evaluation can be used to characterise the capacity of crystalline bedrock to act as a barrier to releases of radionuclides from a nuclear waste repository. Physically plausible bounds on groundwater flow and an effective transport-resistance parameter are estimated based on fundamental principles and idealised models of pore geometry. Application to an intensively characterised site in Sweden shows that, due to high spatial variability and uncertainty regarding properties of transport paths, the uncertainty associated with the geological barrier is too high to allow meaningful discrimination between good and poor performance. Application of more complex (stochastic-continuum and discrete-fracture-network) models does not yield a significant improvement in the resolution of geologic-barrier performance. Comparison with seven other less intensively characterised crystalline study sites in Sweden leads to similar results, raising a question as to what extent the geological barrier function can be characterised by state-of-the art site investigation methods prior to repository construction. A simple evaluation provides a simple and robust practical approach for inclusion in performance assessment.

Acta Universitatis Carolinae, Geologica↗