Practical Advances in Groundwater Modeling with Analytic Elements

The analytic element method is a relatively new solution technique for groundwater flow modeling. The AEM differs from traditional numerical techniques in that the model solution is achieved by superimposing solutions for many analytic functions, providing an analytically accurate flow field. AEM models are most useful when modeling large regional flow problems. The modeler controls the behavior at the perimeter of the study area by applying boundary conditions far from the study region. Numerical models require that the flow conditions at the edge of the model mesh be specified by the modeler. Since the conditions at the model perimeter are almost never known a priori, the use of AE models can improve modeling practice.

Though the AEM provides many benefits for the practitioner, it is still limited in its ability to model certain flow features and processes. For example, it is currently impossible to include continuously varying aquifer properties in an AEM model. It is also difficult to simulate three dimensional flow and transient flow.

This thesis describes several new tools and techniques which advance the applicability of the AEM as a practitioner's tool. In particular:

• I have developed a tool that allows the construction of local numerical model domains from regional AEM models. This screening model approach improves the use of AEM models by providing a reliable method for local modeling. It also improves numerical models by providing perimeter boundary conditions that are consistent with the regional flow field.

• I have investigated the potential for parallel execution of AEM codes and developed a modular, massively-parallel research code, ModAEM. ModAEM will be documented and released into the public domain, and may be enhanced by future researchers to become a generic modeling tool.

• I have used ModAEM to evaluate the practicality of modeling three-dimensional flow using simpler two-dimensional analytic functions, in a manner similar to numerical models. It appears that this approach will be useful if efficient computational algorithms can be devised.

These practical developments can improve modeling practice and may help to stimulate interest in and development of the analytic element method.