Sensitivity analysis and determination of streambed leakance and aquifer hydraulic properties

A nonlinear regression method is used to calculate the hydraulic parameters of a stream-aquifer system using pumping test data. Five parameters, including the horizontal hydraulic conductivity (K-x), aquifer anisotropy (K-a), streambed leakance l, aquifer specific storage S-s, and specific yield S-y, can be calculated. MODFLOW, coupled to the regression method, simulates the groundwater flow that is affected by streams. Sensitivity analyses indicate that for a given stream-aquifer system, the quality of the stream-aquifer test data can be improved through a careful selection of observation and pumping wells, as well as an appropriate test duration. An optimal location of an observation well is where the magnitude of the sensitivities is enhanced and the correlation of the transient sensitivities of two parameters is reduced. Generally, a longer pumping period will increase the sensitivity for l and K-x and reduce the correlation between S-y and K-x and between S-y and l. Results from hypothetical examples and a field test suggest that a two-well analysis of pumping test data can significantly reduce the correlation of sensitivity coefficients; as a result, convergence occurs faster and the estimated standard errors are reduced. (C) 2003 Elsevier B.V. All rights reserved.