ABSTRACT

Three-dimensional steady-state and transient models of groundwater flow and advective transport in the eastern Snake River Plain aquifer were developed by the U.S. Geological Survey in cooperation with the U.S. Department of Energy. The steady-state and transient flow models cover an area of 1,940 square miles that includes most of the 890 square miles of the Idaho National Laboratory (INL). A 50-year history of waste disposal at the INL has resulted in measurable concentrations of waste contaminants in the eastern Snake River Plain aquifer. Model results can be used in numerical simulations to evaluate the movement of contaminants in the aquifer.

Saturated flow in the eastern Snake River Plain aquifer was simulated using the MODFLOW-2000 groundwater flow model. Steady-state flow was simulated to represent conditions in 1980 with average streamflow infiltration from 1966–80 for the Big Lost River, the major variable inflow to the system. The transient flow model simulates groundwater flow between 1980 and 1995, a period that included a 5-year wet cycle (1982–86) followed by an 8-year dry cycle (1987–94). Specified flows into or out of the active model grid define the conditions on all boundaries except the southwest (outflow) boundary, which is simulated with head-dependent flow. In the transient flow model, streamflow infiltration was the major stress, and was variable in time and location. The models were calibrated by adjusting aquifer hydraulic properties to match simulated and observed heads or head differences using the parameter-estimation program incorporated in MODFLOW-2000. Various summary, regression, and inferential statistics, in addition to comparisons of model properties and simulated head to measured properties and head, were used to evaluate the model calibration.

Model parameters estimated for the steady-state calibration included hydraulic conductivity for seven of nine hydrogeologic zones and a global value of vertical anisotropy. Parameters estimated for the transient calibration included specific yield for five of the seven hydrogeologic zones. The zones represent five rock units and parts of four rock units with abundant interbedded sediment. All estimates of hydraulic conductivity were nearly within 2 orders of magnitude of the maximum expected value in a range that exceeds 6 orders of magnitude. The estimate of vertical anisotropy was larger than the maximum expected value. All estimates of specific yield and their confidence intervals were within the ranges of values expected for aquifers, the range of values for porosity of basalt, and other estimates of specific yield for basalt.

The steady-state model reasonably simulated the observed water-table altitude, orientation, and gradients. Simulation of transient flow conditions accurately reproduced observed changes in the flow system resulting from episodic infiltration from the Big Lost River and facilitated understanding and visualization of the relative importance of historical differences in infiltration in time and space. As described in a conceptual model, the numerical model simulations demonstrate flow that is (1) dominantly horizontal through interflow zones in basalt and vertical anisotropy resulting from contrasts in hydraulic conductivity of various types of basalt and the interbedded sediments, (2) temporally variable due to streamflow infiltration from the Big Lost River, and (3) moving downward downgradient of the INL.

The numerical models were reparameterized, recalibrated, and analyzed to evaluate alternative conceptualizations or implementations of the conceptual model. The analysis of the reparameterized models revealed that little improvement in the model could come from alternative descriptions of sediment content, simulated aquifer thickness, streamflow infiltration, and vertical head distribution on the downgradient boundary. Of the alternative estimates of flow to or from the aquifer, only a 20 percent decrease in the largest inflow, the northeast boundary underflow, resulted in a recalibrated parameter value just outside the confidence interval of the base-case calibrated value.

Particle-tracking calculations using the particle-tracking program MODPATH were used to evaluate (1) how simulated groundwater flow paths and travel times differ between the steady-state and transient flow models, (2) how wet- and dry-climate cycles affect groundwater flow paths and travel times, and (3) how well model-derived groundwater flow directions and velocities compare to independently derived estimates. Particle tracking also was used to simulate the growth of tritium (3H) plumes originating at the Idaho Nuclear Technology and Engineering Center and the Reactor Technology Complex over a 16-year period under steady‑state and transient flow conditions (1953–68). The shape, dimensions, and areal extent of the 3H plumes were compared to a map of the plumes for 1968 from 3H releases at the Idaho Nuclear Technology and Engineering Center and the Reactor Technology Complex beginning in 1952.

Collectively, the particle-tracking simulations indicate that average linear groundwater velocities, based on estimates of porosity, and flow paths are influenced by two primary factors: (1) the dynamic character of the water table and (2) the large contrasts in the hydraulic properties of the media, primarily hydraulic conductivity. The simulated growth and decay of groundwater mounds as much as 34 ft above the steady-state water table beneath the Big Lost River spreading areas, sinks, and playas, and to a lesser extent beneath the Big Lost River channel lead to non-uniform changes in the altitude of the water table throughout the model area. These changes affect the orientation and magnitude of water-table gradients and affect groundwater flow directions and velocities to a greater or lesser degree depending on the magnitude, duration, and proximity of the transient stress. Simulation results also indicate that temporal changes in the local hydraulic gradient can account for some of the observed dispersion of contaminants in the aquifer near the major sources of contamination at the INTEC and the RTC and perhaps most observed dispersion several miles downgradient of these facilities. The distance downgradient of the INTEC that simulated particle plumes were able to reasonably reproduce the shape and dimensions of the 1968 3H plume extended only to the boundary of zones of abundant sediment, about 4 miles downgradient of the INTEC. This boundary encompasses the entire area represented by the 1968 25,000 picocuries/liter 3H isopleths. Particle plumes simulated beyond this boundary were narrow and long, and did not reasonably reproduce the shape, dimensions, or position of the leading edge of the 3H plume as shown in earlier reports; however, as noted in an assessment of the interpreted plume, few data were available in 1968 to characterize its true areal extent and shape.

First posted August 19, 2010