ABSTRACT

The subsurface is inherently heterogeneous, three-dimensional, and commonly composed not only of aquifer materials but also large fractions of intermingled aquitard materials (e.g., silt and clay beds). Molecular diffusion rather than advection can be the dominant physical transport process in the aquitards. Investigators commonly interpret field data on groundwater contaminant concentrations with the help of conceptual models that are less than three dimensional, less heterogeneous than the actual system, and lacking representation of diffusion into aquitards. Our recent research shows how reliance on such models can lead to mischaracterization of not only the plume but also remediation success, natural attenuation, and scale-dependent dispersion.

Development of improved geostatistical methods (TProGS) for modeling hydrostratigraphy and of a fast random-walk particle method (RWHET) for accurate simulation of advection, dispersion and diffusion-dominated processes provide the necessary tools for modeling transport processes in typically heterogeneous media, exemplified by alluvial fan systems of Livermore Valley, eastern San Joaquin Valley, and the South Tahoe basin, California. All of these geostatistical simulations, which are based on field borehole data and geologic concepts, generate heterogeneous representations of aquifers that are extensively connected networks of channel deposits, except where interrupted by paleosol sequence boundaries (Kings River fan, San Joaquin Valley) or other unconformities. In two-dimensional sections, the aquifers typically appear to be disconnected lenses. Three-dimensional transport simulations consistently produce simulated plumes having important characteristics that are lacking in conventional models but consistent with field observations: (1) rapid plume migration along preferred pathways, in both lateral and vertical directions, and (2) substantial retention of solute mass in aquitard materials near the source for decades or centuries, even when the source is an instantaneous pulse. Apparent longitudinal dispersivity (aL) of simulated plumes grows spatially and temporally in accordance with generic data on aL, suggesting that scale-dependent dispersion can be more an artifact of diffusion processes than an indicator of multi-scale heterogeneity. The preferential flow down sinuous pathways produces plumes that would easily be missed by conventional monitoring well networks. Animation of plume evolution with and without the heterogeneity and with wells pumping demonstrates the dominant role of the heterogeneity and groundwater production on plume evolution. Mass accounting among facies demonstrates that most of the mass being sequestered is in the aquitards. Sensitivity analysis shows that relatively small changes in the diffusion coefficient can change the distribution of mass significantly, although the overall plume behavior remains essentially unchanged. Remediation experiments show that the slow release of contaminants by diffusion and advection from low-permeability materials may lead to exceedingly long times (decades to centuries) for pump-and-treat clean up. Results also show that even if advanced remediation technologies could succeed in removing contamination from the aquifers (alluvial channels), back-diffusion and advection out of the aquitards can cause concentrations in the aquifers to increase again to levels of concern for decades into the future.   

FUTURE SEMINARS PAST SEMINARS

[ GG home | SOEST home | GG People ]