Estimation of aquifer parameters under transient and steady-state conditions

Abstract

The inverse problem of groundwater hydrology is difficult to solve and yet critical for the reliability of model predictions. In this dissertation, the problem is posed within the framework of maximum likelihood estimation by means of an estimation criterion which includes head measurements as well as prior information of the model parameters. Parameters that can be estimated include the principal values and directions of transmissivity (or hydraulic conductivity), storativity (or specific storage), recharge rates, boundary heads or fluxes, leakances, and parameters controlling the error structure of the data. In transient problems, the initial steady-state of the system can also be derived and/or taken advantage of. To minimize the estimation criterion, its gradient is evaluated by a finite element adjoint state method which leads to considerable savings in computer time and storage. The minimization -, performed -by a -combination- of gradient methads, is sbawn, bath theoretically and with examples, to result in stable solutions for a wide range of realistic situations. Synthetic and field examples are included to demonstrate various features of the approach. Applications of the methodology to problems of model selection, network design, and quantification of the worth of data are highlighted.