ModelCARE 90: Calibration and Reliability in Groundwater Modelling (Proceedings of the conference held in the Hague, September 1990). IAHS Publ. no. 195, 1990. 417 R EEVALUATION OF L ARGE -S CALE D ISPERSIVITIES FOR A W ASTE C HLORIDE P LUME : EFFECTS OF T RANSIENT F LOW DANIEL J. GOODE & LEONARD F. KONIKOW U.S. Geological Survey, Reston, Virginia 22092, USA ABSTRACT This paper investigates the effects of transient ground- water flow on dispersion of a waste chloride plume in the basaltic aquifer beneath the Idaho (USA) National Engineering Laboratory. In an early application of numerical modeling techniques to the two-dimensional simulation of field-scale plumes, previous investigators identified longitudinal and transverse dispersivities using an independently calibrated steady-state flow model and matching contours of observed and simulated concentrations. The unusual result of calibrated transverse dis- persivity (140 m) being significantly larger than longitudinal dispersivity (90 m) has been attributed to spatial heterogeneity, the fractured nature of the aquifer, and to the use of a two-dimensional model. New calibrations of the solute-transport model are performed on point concentration observations using a flow model incorporating transient recharge conditions that cause significant fluctuations in velocity. Under transient flow conditions, lowest calibration errors are achieved with significantly larger dispersivities than previously hypothesized, and with the longitudinal component larger than the transverse component. Unfortunately, the sensitivity of the model calibration error to dispersivity is low. Incorporating transient flow in this two-dimensional porous- media model does not significantly improve our understanding of the processes controlling chloride transport at this site. INTRODUCTION A large-scale chloride plume from injection well disposal has developed in the basalt aquifer beneath the Idaho National Engineering Laboratory (INEL). Robertson et al. (1974) provide background information on the site and its hydrologic conditions, including waste disposal practices and observed groundwater contamination. This information has been continually updated by U.S. Geological Survey reports, including those by Lewis & Goldstein (1982) and Pittman et al. (1988). In one of the first comprehensive numerical transport-modeling studies, Robertson (1974) calibrated a two-dimensional flow and transport model using data from the early 1950's through 1972 and used the calibrated model to predict solute spreading to 2000. The calibrated longitudinal (α L ) and transverse dispersivities (α T ), about 90 and 140 m, respectively, were based on simulations assuming steady flow and no recharge in Big Lost River, a losing river adjacent to the contamination plumes. The characteristic of α T > α L is theoretically unexpected and unique among field-scale case studies. Lewis & Goldstein (1982) compared Robertson's predictions with observed concentrations in 1980, noting differences in the spreading and rate of front movement and discussing possible reasons for these differences, including recharge fluctuations. Duffy & Harrison (1987) examined the relation between temporal fluctuations in tritium concentration at the injection well and at near-field (<500 m downstream) observation wells, and applied a spectral method to estimate α L . The method assumes concentration fluctuations are damped and filtered reflections of mass flux fluctuations at