Optimization of a complez and highly parameterized groundwater system
Groundwater flow predictions in complex geologic environments require the use of models that represent an appropriate number of potentially important physical features. Oversimplification of the flow system may neglect features that are important to understanding the range of possible flow predictions. The Singular Value Decomposition and Tikhonov Regularization techniques implemented in the SVD-Assist toll within PEST enabled optimization of complex and highly parameterized models. These techniques are applied to a finite-element flow model that has 46 layers and 1245 adjustable parameters supported by 302 point observations of heads, head differences, pumping rates, and seepage flow. Optimization improved the model fit significantly as the objective function was reduced by 80％ from its initial value. The optimized model retains sufficient parameter detail to predict flow directions. Moreover, the analysis can be extended to facilitate a probability-based assessment of flow prediction uncertainty.
B.Zhang D.G.Abbey P.J.Martin S.C.James W.W.Wossener R.G.Andrachek C.Gabriel B.W.Arnold
AquaResource Inc.,55 NorthField Dr.East.-Suite 203, Waterloo, Ontario, Canada, N2K 3T6 Sandia National Laboratories, United States of America Department of Geosciences,University of Montana, United States of America MWH Global, United States of America AquaResource Inc.,55 NorthField Dr.East.-Suite 203, Waterloo, Ontario,Canada, N2K 3T6 Department of Geosciences, University of Montana, United States of America
国际会议
武汉
英文
107-110
2009-09-20（万方平台首次上网日期，不代表论文的发表时间）