Modeling is the art of representing and interpreting observations derived from the conceptualization of the physical world. Because of human’s needs to make decisions on the basis of limited information and the belief that phenomena in the physical world can be described by fundamental physicochemical laws, modeling has become not only a practice of interpretation but also a method of prediction. However, because of the "inaccurate," simplistic representation in models, uncertainties in prediction and interpretation becomes inevitable. In this regard, modeling is therefore also a practice of characterizing uncertainty and risk, which is one of the objectives of the research.

Reducing uncertainties of model predictions or determining the associated health, monetary, and environmental risks has been one of the active research areas in groundwater hydrogeology. In a typical model of subsurface mass transport phenomena, one can always approach the uncertainty issue from the parameter and the process spaces. In addition to the characterization of uncertainties, how we approach these two major spaces of subsurface models is also one of the subjects that we will discuss in just a few minutes.

Conceptualization of processes and characterization of uncertainties aside, computational efficiencies of methodology and numerical algorithms are also getting more attention these days. This may largely result from our needs in regional water resources management and our insatiable curiosity of discovering a pattern that may emerge from a consortium of physicochemical and biological processes intentionally or unintentionally included in a sometimes monstrous computer model. Therefore, we will also discuss a few computational issues in watershed scale modeling, in particular the role of parallel supercomputers. The study we are presenting today was supported by the Subsurface Sciences Program of the Department of Energy under the direction of Dr. Frank Wobber and is currently being supported by the Environmental Technology Partnership Program of DOE under the direction of Dr. Paul Bayer.

Throughout this presentation, we will use data collected and modeling conducted on the Melton Branch Watershed at ORNL in a period of about 10 years. We believe the lessons we learned can be equally applied to watershed scale modeling in other areas and to other types of macroporous soils.