Examination of solute transport in highly heterogeneous media at the Macrodispersion Experiment (MADE) Site, Columbus, MS

Abstract

The Macrodispersion Experiment (MADE) Site in Columbus, MS was developed to examine the reliability and predictive power of macrodispersion transport theories, used to describe the hydrodynamic spreading of solutes at the field-scale. Prior to the initial investigation reported in 1992, it was believed that the classical advection dispersion equation (ADE) was a valid model for solute transport if the subsurface hydraulic conductivity (K) was sufficiently characterized. However, the injected plume at the MADE Site displayed anomalous behavior and the classical ADE was unable to capture the early arrival time and the heavy tailing seen in the heterogeneous aquifer. At highly heterogeneous sites tracer often returns quickly after extraction begins. This quick arrival of tracer creates an early peak, likely a result of channelized preferential flow paths. Heterogeneous sites also display late-time heavy tails, which seemingly represent low K zones or dead-end pores where dissolved solute becomes relatively immobile. The resultant asymmetric breakthrough curves (BTC) led some investigators to show that non-local models may be more appropriate for modelling transport within the MADE aquifer. A temporal non-local model, the time fractional advection dispersion equation (t-FADE) may be more suitable for highly heterogeneous media because it is capable of matching the power-law tails exhibited in the MADE BTCs. This work aims to examine the accuracy of the ADE and t-FADE for the singlewell injection-withdrawal (SWIW) test conducted by Liu et al. (2010). The SWIW test was completed in the Intensively Cored Area (ICA), and used the conservative tracer NaBr. Within this area there were 4,962 recorded K values, with samples retrieved every 1.5 cm from 9 cores. This study is unique in that two components will be investigated: (1) the inclusion of the vadose zone which is often neglected in unconfined aquifers, and (2) examination of various subsurface configurations from a homogeneous domain to a detailed K-field. The inclusion of the vadose zone may account for the lowering of the water table during the extraction phase, seen in other modelling efforts (Liu et al., 2010 and Ronayne et al., 2010), and tracer being trapped in previously unsaturated areas. Also, the effects of refining the K detail will be examined, i.e., what improvements does each simulation gain by increasing the fineness of the K.