Groundwater Model Parameter Estimation Using Response Surface Methodology

摘要

This thesis examined the use of response surface methodology (RSM) to estimatethe parameters of a finite-element groundwater model. An existing two-dimensional, steady-state flow model of a fractured carbonate groundwater system in southwestern Ohio served as the calibration target data set. A Plackett-Burman screening design showed that only four of the ten hydraulic conductivity zones significantly contributed to the output of the finite-element model. Also, the effective porosity parameter did not significantly affect the model's output. Using only the four significant hydraulic conductivity parameters; four two-level, four-factor designed experiments were conducted to exploit the first-order response surface defined by a residual sum of squares function. Additionally, a central composite design and ridge analysis were used to adjust the four parameters and finally arrive at a calibrated model in a grand total of 146 runs. The final calibrated model, which had an average head elevation of 292 meters, matched the calibration target data set with a mean absolute error of only 7 mm over all 524 nodes of the model. RSM provided an effective calibration technique to estimate groundwater flow parameters. (AN).