Advective Transport: Importance to Groundwater Remediation and Illustration of Basic Transport Concepts to Introductory Audiences

Abstract

Developing effective transport models with meaningful results requires a solid understanding of transport concepts and the underlying mechanics the model, and how data can be best integrated into the model. This research makes contribution to each of these three requirements: concept education, understanding simulated mechanics, and integrating data into models. (1) How can groundwater flow and transport processes be well communicated to introductory audiences, while providing a foundation for complex model development and interpretation? The first part of this work presents GroundWaterTutor, a freely available computer module for groundwater education. GroundWaterTutor provides a simple, interactive environment for learning how key system characteristics affect hydraulic heads and the flow of tracer particles. The software was developed using MATLAB in conjunction with MODFLOW 2005 and MODPATH 6, and thus provides a solid foundation from which students can expand to simulating more complex situations. GroundWaterTutor is distributed as a set of freely available standalone executables. (2) How do simulated advection interact with dispersion in groundwater remediation simulations? This question is addressed in the context of the following research question: How well do advection-based metrics for assessing the effectiveness of active in situ groundwater remediation strategies work? Results are important to developing an efficient optimization framework for in-situ active remediation systems. (3) Can heteroscedastic data, like concentration data, be integrated into models, such as groundwater models, without log-transformations, which make results hard for many users to interpret? Here the use of error-based weighting methods are investigated, which provide more intuitive regression models than log-transformation in the presence of highly variable (e.g. heteroscedastic) data. For this problem, log-transformation produced good model fit, while the error-based weighting formulations tested worked poorly.