 |
2007 Ground Water Summit |
Wednesday, May 2, 2007 : 1:00 p.m.
Assessing Geostatistical Models in Capturing Geologic Realities
Advanced techniques for combining geological and geostatistical approaches to groundwater modeling have evoked new prospects for capturing geological realities of the subsurface. These approaches produce facies distributions that appear more realistic (i.e. geometrically and juxtapositionally) than those produced using traditional statistical methods. In this project, we explored applying these techniques to a geologically uncertain area, with the hope that one of the simulated scenarios would indicate that the subsurface system had been captured. The modeled region is a ~10,000m2, fluvially-dominated section of the Upper Livermore aquifer of Lawrence Livermore National Laboratory. Successful reproduction of the subsurface was assessed based on simulating a pumping test and matching resultant drawdown curves in observation wells with field data. Multiple realizations were produced using a hierarchical approach of systematically varying two scales of heterogeneity -- large scale at the 100s-meter scale (channel orientations) and small scale at the 10s-meter scale (i.e. channel lengths and widths). Results revealed that ‘successful’ models could be produced from a wide variety of geologic scenarios. These findings suggest that several components must be considered when accepting a model as "reasonable" using these techniques: 1) Models constructed using geostatistics require more robust testing than can be offered by hydraulics matching of observation wells alone. 2) Given enough stochastic simulations, geostatistics are often capable of reproducing hydraulic drawdown curves observed. 3) The geology of a system cannot be determined through model deduction; it must be assessed before modeling.