A nonlinear, least-squares regression technique for the estimation of ground-water flow model parameters was applied to the regional aquifer underlying the eastern Snake River Plain, Idaho. The computer program simulates two-dimensional, steady-state ground-water flow. Hydrologic data for the 1980 water year were used to calculate recharge rates, boundary fluxes, and spring discharges. Groundwater use was estimated from irrigated land maps and crop consumptive-use figures. These estimates of ground-water withdrawal, recharge rates, and boundary flux, along with leakance, were used as known values in the model calibration of transmissivity. Leakance values were adjusted between regression solutions by comparing model-calculated to measured spring discharges. In other simulations, recharge and leakance also were calibrated as prior information regression parameters, which limits the variation of these parameters using a normalized standard error of estimate.
Results from a best-fit model indicate a wide areal range in transmissivity from about 0.05 to 44 feet squared per second and in leakance from about 2.2 x 10-9 to 6.0 x 10-8 feet per second per foot. Along with parameter values, model statistics also were calculated, including the coefficient of correlation between computed and observed head (0.996), standard error of the estimates of head (40 feet), and parameter coefficients of variation (about 10-40 per-cent). Additional boundary flux was added in some areas during calibration to achieve proper fit to ground-water flow directions. Model fit improved significantly when areas that violated model assumptions were removed. Model fit also improved when y-direction (northwest-southeast) transmissivity values were larger than x-direction (northeast-southwest) transmissivity values. The model was most sensitive to changes in recharge, and in some areas, to changes in transmissivity, particularly near the spring discharge area from Milner to King Hill.
