Skip to main content
U.S. flag

An official website of the United States government

Locally refined block-centered finite-difference groundwater models: Evaluation of parameter sensitivity and the consequences for inverse modelling and predictions

January 1, 2002

Models with local grid refinement, as often required in groundwater models, pose special problems for model calibration. This work investigates the calculation of sensitivities and performance of regression methods using two existing and one new method of grid refinement. The existing local grid refinement methods considered are (1) a variably spaced grid in which the grid spacing becomes smaller near the area of interest and larger where such detail is not needed and (2) telescopic mesh refinement (TMR), which uses the hydraulic heads or fluxes of a regional model to provide the boundary conditions for a locally refined model. The new method has a feedback between the regional and local grids using shared nodes, and thereby, unlike the TMR methods, balances heads and fluxes at the interfacing boundary. Results for sensitivities are compared for the three methods and the effect of the accuracy of sensitivity calculations are evaluated by comparing inverse modelling results. For the cases tested, results indicate that the inaccuracies of the sensitivities calculated using the TMR approach can cause the inverse model to converge to an incorrect solution.

Publication Year 2002
Title Locally refined block-centered finite-difference groundwater models: Evaluation of parameter sensitivity and the consequences for inverse modelling and predictions
Authors S. Mehl, M. C. Hill
Publication Type Article
Publication Subtype Journal Article
Series Title Acta Universitatis Carolinae, Geologica
Index ID 70023815
Record Source USGS Publications Warehouse