Future Directions

These are what we have so far, and I hope you don’t mind that I give a little bit of preview about what and where we are heading for next.

First of all, the closure of the multiregion theory. So far, we are still struggling with the parameterization of multiregion hydraulic conductivity and water content functions. We have had some progress with the mass transfer function, but the closure of the multiregion theory still needs a lot of work to link small scale structures and processes to large-scale geological structures and processes. To this end, more upscaling works have been planned, in which uncertainty caused by small scale structures and processes will have to be built into the upscaling mechanism.

Secondly, the implementation of macroscale features, such as soils of different types, through geostatistics and scaling factors. Two catchment wide survey of infiltration rates and soil hydraulic properties, including hydraulic conductivities and water retention, were conducted earlier on the Melton site; one by Bob Luxmoore and his coworker, the other by my coauthor Glenn and Phil. Measurements of surface infiltration rates such as this one, reported by Watson and Luxmoore in their SSSAJ paper in late 1980s can be converted to surface, especially, B horizon hydraulic conductivity if one assumes that the vertical hydraulic gradient is uniform over the upper 1 m of soil. One can then use one of the conditional simulation algorithms for geostatistical applications to generate multiple realizations of hydraulic conductivity distributions over the site. For example, the next slide is a realization of hydraulic conductivity distribution based on the sequential Gaussian simulation algorithm included in GSLIB. But, before the simulation, we converted the raw hydraulic parameter to scaling factor. Simulation here was in fact done on scaling factors, which are to be converted back to real hydraulic parameters. Scaling factor is largely based on similar media theory proposed in the 1950s by Miller and Miller. For partially saturated field scale applications where not only hydraulic conductivity functions but also water retention curves are needed to characterize a soil, scaling factors have the advantage of using less amount of data than normal score transformation. Scaling factors of hydraulic parameters at a reference location can be used to predict the hydraulic parameters at non-reference locations. This may reduce data collection costs, but, on the other hand, may introduce unwanted uncertainties at non-reference locations. The pros and cons of this approach still need to be examined. A second factor that we will look at is the assumption that the distributions of the hydraulic parameters are gaussian. Preliminary analysis by my colleagues at ORNL seemed to suggest that the assumption of gaussian may not be valid. If further analysis strongly suggests that this is true, we will use one of the non-gaussian simulation tools such as multifractral simulators to generate the conditional distributions of hydraulic parameters.

With the focus of this research project sifted towards biochemical, we are also looking at the possibility of including microbial transport, in addition to the microbial growth and deposition module, in the code. But, for all our concerns about the physicochemical basis of our biohydrogeochemical applications, we cannot forget the possibility that a monster codes that are trying to do everything right may be doing something wrong only because we did not include the integration and user interface step. So, one of our emphases in the research project is the vertical integration of experiments, modeling, and uncertainty analysis. This integration will be done through a user interface that allows the visualization of experimental and computational data, the comparison of these two, and include an open-ended framework that should later allow the integration of statistical methods for uncertainty analysis. Moreover, because we expect that we would be able to run field scale applications only on large mainframe computers in the near future, a remote steering component is also being developed to allow users remotely tap into the resources at the major computational facilities of the country.