Data error and highly parameterized groundwater models

摘要

Strengths and weaknesses of highly parameterized models, in which the number of parametersexceeds the number of observations, are demonstrated using a synthetic test case. Results suggest that theapproach can yield close matches to observations but also serious errors in system representation. It isproposed that avoiding the difficulties of highly parameterized models requires close evaluation of:(1) model fit, (2) performance of the regression, and (3) estimated parameter distributions. Comparisons tohydrogeologic information are expected to be critical to obtaining credible models.