Previous Return to Curriculum Vitae Next RECENT RESEARCH ACCOMPLISHMENTS 3. Do available numerical methods and programs solve the forward problem well enough to support the sensitivity analysis, nonlinear regression, and uncertainty analysis methods of my own work and others? How can these deficiencies be remedied? The sensitivity analysis and regression methods described in Recent Research Accomplishments item 1 requires constitutive equations and numerical methods that represent subsurface flow and transport accurately; any model error diminishes the certainty of the results. Laboratory experiments provide a unique opportunity to test numerical models of subsurface transport, as reported by Mehl and Hill (1999, 2000, 2001). The figure below uses results from a laboratory experiment to evaluate the existence and importance of numerical inaccuracies in the solution methods available in the ground-water transport code MT3DMS. Figure: Forward simulations using permeameter values of hydraulic conductivity. The 95% confidence intervals shown for the measured values reflect expected measurement error; coefficients of variation ranged between 4% and 20%. Sequential grid size reductions had indicated that the results of each method were not improved significantly by using a finer grid. For an equivalent problem with constant dispersivity, MOC3D produced results similar to MOC and TVD of MT3DMS; MOC3D had an execution time of 70 minutes. [Computation times using a Linux workstation, Pentium II - 333, 64Mb Ram]. Previous Return to Curriculum Vitae Next mchill@usgs.gov Last Modified: April 14, 2008