ESTIMATING SPATIALLY VARIABLE REPRESENTATIVE ELEMENTARY SCALES IN A FRACTURED AQUIFER USING HYDRAULIC HEAD OBSERVATIONS, TURKEY CREEK BASIN, COLORADO

Characterizing fluid movement within fractured aquifers is an unresolved challenge of increasing scientific and societal importance. Hydrologic characterization of large-scale fractured aquifers is typically performed using numerical simulation, which in addition to model parameterization requires an appropriate representation of the hydraulically conductive network. While discrete feature network simulation is conceptually robust and has proven useful for many applications, it can be computationally limiting for large-scale systems. For large-scale aquifers continuum representation offers significant benefits such as simplified geometry, which bypasses the need for the exact geometrical configuration of fractures and rock matrix. A major challenge for the continuum model is determining an equivalent continuum representation of a discretely fractured system. While it is possible to arbitrarily impose any continuum element geometry, variability in fracture connectivity must be accurately represented. It is computationally advantageous to employ continuum elements at the representative elementary volume, where the assumption of global element connectivity is conceivable. Otherwise, all possible variations of connectivity should be considered, amounting to 2^n analyses, where n is the total number of model elements. Further, fracture connectivity can be spatially variable and it is therefore reasonable to expect that representative elementary scales, and thus continuum representation, should vary spatially. We discuss our methodology and rationale for implementing hydraulic head as an indicator of spatially variable representative elementary scales (RES). We compare hydraulic head estimations of RES (HYRES) to structurally based RES predictions, and examine HYRES sensitivity to varying system conditions. Multiple realizations of conditionally simulated hydraulic head are employed to estimate heterogeneity and significantly improve HYRES predictive accuracy. We summarize our HYRES methodology, data optimization, and comparisons to structural measures of RES, now under review for publication, and present current results of implementing these techniques in a fractured watershed of Turkey Creek Basin, Colorado.