The motivations of doing large-scale models are very diverse and our inclination to conduct investigations is very often driven by our needs and curiosity. Most of us here design and conduct laboratory and field experiments. And, very likely, most of us use simple but representative computer models to aid these designs. For example, we may run a one-dimensional solute transport code to determine the time and flow rate that may be needed for a reactive column injection experiment. We may run a geochemistry code to determine possible speciations in a field scale multispecies tracer injection and ensure that we are not going to introduce some unwanted reactions that may cause sampling or analytical problems. So, let me start with three of my personal encounters with such needs and curiosity. But, before that, let me recall a little bit the history of the buried wastes at the Oak Ridge National Laboratory (ORNL). The laboratory itself was created by the Manhattan project during the World War II. Over these years, the focus of the laboratory is still largely energy related. But, over the course of this research focus, large amount of low-level nuclear wastes were produced, and up until the 1970s, these wastes were still dumped in open trenches in the shallow B and C horizons without any concrete or leaching prevention measures. The result of this is, in addition to primary source of contaminants in the trench, the leached low level wastes have moved into the surrounding soils and created a so-called secondary contaminant sources in the soil matrix. A lot of tritium and sometimes ⁹⁰Sr can be found in discharge water during high-intensity storm events.

A few years ago, I was approached by two senior investigators in the Energy and Environmental Sciences Divisions at the Oak Ridge National Laboratory (ORNL). Engineers in the Engineering Division were in the process of designing and implementing a zeolite filter that may prevent ⁹⁰Sr from moving into the surface water surrounding the Waste Area Grouping 4 (WAG 4) at ORNL. However, they would need a projection of the subsurface water flux during wet and dry seasons so that filter capacity and replacement schedule can be factored into the design process. A watershed scale model would need to be executed. The difficulties of this task include scarcely available model parameters and the tremendous amount of computational needs to characterize uncertainties of such projection.

A few years ago before that, a debate between engineers and the waste area management team took place on the issue of high density polyethylene surface caps that have been placed on top of some low level waste trenches in the WAG 6 area. The black caps cover a large portion of the waste storage area, but we can still see some lateral flow into the areas below the caps. The caps are supposed to prevent or reduce direct infiltration from precipitation. The main issue here is whether it is necessary to spend millions of dollars to increase the capped area, or to come up with a more economic, cost-effective solution. We therefore conducted a modeling study of the effectiveness of four engineering scenarios: (1) monitoring only (one of my colleagues liked to call it do-nothing), (2) huge polyethylene caps only, (3) French drains only, and (4) both of the polyethylene cap and French drains. The plan to increase cap area was eventually rejected by the management, and in the mean time, our conclusion of the modeling study suggested that capping alone is not going to be effective in reduce the lateral flow.

More recently, my coauthors and I were involved in a research project that is designed to determine the possibility that mass transfer can be a limiting factor of bioavailability during in-situ bioremediaion. If the secondary source is the major concern of remediation efforts, then, we will have to be able to access the pore space in the soil matrix. The idea is that microorganisms may more likely reside in the larger pores of a fractured geological formation or a "macroporous" soil than the rock and soil matrix areas that may contain most of the pore volumes, nutrients and pollutants. In addition to the absorption processes, the availability of substrates (contaminants) and nutrients may be controlled by the diffusive mass transfer process between the fractures and the matrix. This process occurs in a multiscale sense, with the mass transfer and absorption reactions going locally in a scale equal to or smaller than a matrix block but being repeated everywhere in a much larger scale. How to upscale our understanding, in both process and parameter spaces, for watershed or regional applications becomes an issue that needs to be dealt with experimentally and mathematically.