Identification of Aquifer Parameters from Pumping Test Data with Regard for Uncertainty

Abstract

When fitting hydraulic models of groundwater flow to pumping test data, Bayesian inference provides a framework for quantifying the posterior uncertainty of aquifer parameters estimated from data and the most likely range of parameters that are consistent with the data. In this study, noise-perturbed drawdown data is measured. For clarity, groundwater models with few parameters are considered and Markov chain Monte Carlo is used to quantify uncertainty of transmissivity, storativity, and leakage parameters. These models exhibit many of the features typically encountered in much higher dimensional computational groundwater models like multimodality, failure of least squares algorithms, and poorly determined parameters. For comparison, Bayesian inference is contrasted with least squares model fitting.