USGS Home
Contact USGS
Search USGS

U.S. GEOLOGICAL SURVEY
Scientific Investigations Report 2007–5044

Back to Table of Contents

Summary and Conclusions

The ground-water flow model presented in this report is a component of a comprehensive study initiated by the Idaho Department of Water Resources, the Washington Department of Ecology, and the U.S. Geological Survey to improve the understanding of ground-water flow in the Spokane Valley-Rathdrum Prairie (SVRP) aquifer and of the interaction between ground water and surface water. The model was developed by the Modeling Team formed within the comprehensive study. The Modeling Team consisted of staff and personnel working under contract with the Idaho Department of Water Resources, personnel working under contract with the Washington Department of Ecology, and staff of the U.S. Geological Survey. To arrive at a final model that has the endorsement of all team members, decisions on modeling approach, methodology, assumptions, and interpretations were reached by consensus. The primary purpose of the model is to serve as a tool for analyzing aquifer inflows and outflows, simulating the effects of future changes in ground-water withdrawals from the aquifer, and for evaluating aquifer management scenarios. The scale of the model and the level of detail are intended for analysis of aquifer-wide water-supply issues.

The SVRP aquifer model encompasses an area of approximately 326 square miles. For the most part, the model extent coincides with the 2005 revised extent of the SVRP aquifer as defined in previous reports. However, the model excludes Spirit and Hoodoo Valleys because of uncertainties about the ground-water flow directions in those valleys and the degree of hydraulic connection between the valleys and northern Rathdrum Prairie. In addition, the model excludes three areas, one in northern Rathdrum Prairie and two in the vicinity of Five Mile Prairie, because the aquifer sediments in those areas likely are unsaturated.

The SVRP aquifer is considered to be a single hydrogeologic unit except in Hillyard Trough and the Little Spokane River Arm. In those areas, a continuous clay layer divides the aquifer into an upper, unconfined unit and a lower, confined unit. Both the upper and lower units extend from Hillyard Trough toward the west through the Little Spokane River Arm. The model terminates at the east end of Long Lake, a reservoir on the Spokane River. At this model boundary, the upper unit is in direct hydraulic connection with Long Lake. Available field data indicate that the clay layer and lower unit extend beyond the model boundary for an unknown distance. However, the confining clay layer eventually pinches out, allowing ground water in the lower unit to discharge into Long Lake.

The SVRP aquifer model includes all known components of inflows to and outflows from the aquifer. Inflows to the SVRP aquifer include (1) recharge from precipitation, (2) inflows from tributary basins and adjacent uplands, (3) subsurface seepage and surface overflows from lakes that border the aquifer, (4) flow from losing segments of the Spokane River to the aquifer, (5) return percolation from irrigation, and (6) effluent from septic systems. Outflows from the SVRP aquifer include (1) ground-water withdrawals from wells, (2) flow from the aquifer to gaining segments of the Spokane River, (3) aquifer discharge to the Little Spokane River, and (4) subsurface outflow from the lower unit at the western limit of the model area near Long Lake.

The ground-water flow model MODFLOW-2000 is used to simulate ground-water flow in the SVRP aquifer. The finite-difference model grid consists of 172 rows, 256 columns, and 3 layers. Ground-water flow is simulated for September 1990 through September 2005 using 181 stress periods of 1 month each. Aquifer heads at the start of the simulation period are unknown, and the ground-water flow system was assumed to be at a steady state during the first stress period. Because aquifer heads simulated during this stress period serve as the initial conditions for the rest of the transient simulation, the first 5 years of the simulation period is considered to be a start-up period. No attempt is made to fit simulated quantities to measured quantities during the start-up period. Instead, the model is calibrated using water-level and flow data for October 1995 to September 2005.

Boundary conditions representing inflow and outflow components are implemented using packages in MODFLOW-2000. The Recharge Package is used to simulate recharge from precipitation. The Well Package is used to simulate withdrawals from wells, return percolation from irrigation, and effluent from septic systems. The Flow and Head Boundary Package is used to simulate flows to the aquifer from tributary basins and from all lakes except Lake Pend Oreille and Coeur d’Alene Lake. The River Package is used to simulate the Little Spokane River, Lake Pend Oreille, and Coeur d’Alene Lake. The Streamflow-Routing Package is used to simulate the interaction between the Spokane River and the aquifer. The General-Head Boundary Package is used to simulate ground-water outflow from the lower unit at the west end of the Little Spokane River Arm. The spatial distribution of aquifer properties such as hydraulic conductivity and specific yield are represented by dividing the aquifer into zones. Within each zone, the aquifer property is assumed to be uniform.

The parameter estimation program PEST is used to calibrate the SVRP aquifer model. PEST implements a nonlinear least-squares regression method to estimate model parameters by minimizing the sum of squared weighted residuals. Calibration data include 1,573 measurements of water levels and 313 measurements of streamflow gains and losses along segments of the Spokane and Little Spokane Rivers. Weights for the measurements initially are determined from errors associated with the measurements and subsequently adjusted to balance the influence of water-level measurements and flow measurements on the regression.

A total of 38 model parameters are estimated during calibration. These parameters include 22 values of horizontal hydraulic conductivity, 3 values of specific yield, 10 values of vertical hydraulic conductivity of riverbed sediments of the Spokane River, and 3 values of hydraulic conductances for riverbed sediments in the Little Spokane River and for lakebed sediments in Lake Pend Oreille and Coeur d’Alene Lake. An additional four model parameters are not estimated by calibration because the calibration data are insensitive to these parameters. Instead, the parameters are assigned reasonable values.

Model calibration gives the following results. In the central part of the aquifer in Rathdrum Prairie and in Spokane Valley, estimated horizontal hydraulic conductivity (K_h) values range from about 6,200 to 22,000 feet per day. In Hillyard Trough, the Little Spokane River Arm, and Western Arm, estimated K_h values range from about 2,000 to 3,000 feet per day. In the Coeur d’Alene area, the estimated K_h value is 1,290 feet per day. For side valleys and regions of shallow bedrock along the margins of the aquifer, the estimated K_h values range from 4 to 137 feet per day. Estimated specific yield values range from 0.10 to 0.21. For the Spokane River bed sediments, estimated values of vertical hydraulic conductivity range from 0.01 to 10 feet per day.

In general, the simulated water levels and flows (streamflow gains and losses) are in good agreement with the measured water levels and flows throughout most of the aquifer. The greatest discrepancies between measured and simulated quantities occur in the two wells (wells 99 and 115) completed in the lower unit. These discrepancies indicate that the lower unit might not be represented accurately by the model. For southern Rathdrum Prairie (including the Coeur d’Alene area), simulated water-level fluctuations during 2004–05 do not agree with the observed measured fluctuation. This discrepancy indicates that the temporal distribution of recharge to southern Rathdrum Prairie might not be represented accurately in the model for 2004–05.

The calibrated model gives a 10-year average flow of 67 cubic feet per second (ft³/s) from Lake Pend Oreille to the aquifer, 138 ft³/s from Coeur d’Alene Lake to the aquifer, and 27 ft³/s from the lower unit to Long Lake. To examine the assumptions in the SVRP aquifer model, five alternative models are analyzed. In each alternative model, one aspect of the calibrated model is modified and the alternative model is recalibrated. Results of these alternative model analyses show that changes in certain model parameter values can result in changes to certain simulated flow components even though the overall fit of the alternative model to the measured quantities is nearly as good as the calibrated model. This suggests some degree of nonuniqueness in the ground-water flow simulated by the calibrated model.

The SVRP model presented in this report is calibrated using significantly more data than are used in previous models. The relatively good fit between simulated and measured quantities indicates that the overall simulated ground-water flow is a reasonable representation of ground-water flow in the SVRP aquifer. Nonetheless, the model has limitations. In particular, there is insufficient hydrologic information to determine ground-water inflow from Spirit and Hoodoo Valleys to the SVRP aquifer. In Hillyard Trough and the Little Spokane River Arm, ground-water flow in the lower unit is not well understood, and simulated water levels do not fit measured water levels as well as in other parts of the aquifer. There also is significant uncertainty in the simulated seepages from Lake Pend Oreille and Coeur d’Alene Lake. Further investigations in these parts of the SVRP aquifer could provide valuable knowledge that can be used to improve the model in the future.

Back to Table of Contents