T140 Complexity in Modeling: How Much Is Too Much?

Monday, 10 October 2011
1:30 PM - 5:30 PM
Room 101DE

SIMPLIFYING COMPLEXITY--COMPARISON OF PARTICLE AND GRID METHODS WITH COARSE GRIDS

Paper number: T140-11
Program: Technical Sessions
Time: 4:10 PM – 4:25 PM

Authors

Abstract

Upward trends in dissolved solids are occurring in some areas in the basin-fill aquifer in Salt Lake Valley, Utah. Possible sources of dissolved solids include mineral dissolution in native and recently recharged water, surface streams, and application of road de-icing chemicals. Because of complex interactions between sources of water and historical pumpage, a groundwater simulation model is being used to understand the trends. Groundwater flow is simulated using an existing transient finite-difference model, and the advective transport of solute is simulated using particle- and grid-based methods. Model parameters are optimized using nonlinear regression to match tritium concentrations in samples from public-supply wells, and the model is applied to the problem of dissolved solids trends. The accuracy of the model is limited because of a coarse model grid; however, the model may be sufficient to understand the causes of the trends.

Both advective transport methods have inaccuracies. Advective grid methods on coarse grids are inaccurate because of numerical dispersion. Advective particle-tracking methods on coarse grids are inaccurate because concentration will vary on a much finer scale than simulated groundwater velocity. Also, on a coarse grid, there is no mixing of convergent path lines near pumping wells. In this study, a velocity refinement technique is used to simulate radial flow toward a pumping well to more accurately represent path lines near wells.

A component of optimal nonlinear regression parameter estimates is related to method inaccuracy. For example, in the grid method, parameters of mass transfer in dual- porosity-domain medium are estimated, but the estimates may include compensation for numerical dispersion. We compare the effect of numerical method on predicted values and prediction uncertainty. Because of inaccuracy of methods on coarse grids, predictions of trends may not be reliable. We present a simpler method in which each solute source is represented by a hypothetical solute with source strength of unity. Breakthrough curves at observed trend locations show the pattern of source mixing has changed over time. This approach helps explain observed trends in water quality in the basin and is less reliant on the choice of numerical methods.