HR: 0800h
AN: H11A-1248 [Abstracts]
TI: Simulating Fluid Movement in Saturated Heterogeneous Fractured Watersheds: Exploring the Influence of
Data Distribution, Aquifer Structure, and Element Size on Model Accuracy
AU: * Wellman, T P
EM: tpwellma@mines.edu
AF: Colorado School of Mines, 1516 Illinois St Berthoud Hall, room 121a, Golden, CO 80401
United States
AU: Poeter, E P
EM: epoeter@mines.edu
AF: Colorado School of Mines, 1516 Illinois St Berthoud Hall, room 121a, Golden, CO 80401
United States
AB:
Caused in part by population rise and subsequent development, fractured watersheds are increasingly relied upon as primary
water resources. Yet due to their complexity, accurate model predictions are often beyond reach. Methods must be developed
that aid in the creation of viable models, which is the focus of this study. In light of the computational expense of
discrete fracture models and limited ability for characterizing hydraulically conductive fractures, continuum models have
remained the preferred tool for simulating hydrologic processes in large-scale fractured aquifers. The major challenge for
continuum representation is determining the size of continuum (i.e. element size), if any, that can accurately represent
fracture-controlled fluid movement. A common approach is to employ the representative elementary volume in three dimensional
systems, which we generically refer to as representative elementary scale (RES). We present an energy-based, multi-scale
approach for estimating spatially variable RES, developed in a previous phase of our research. Rather than evaluating
fracture structure directly, we spatially analyze the effective fluid energy at varying scales using hydraulic head
observations. Building upon this initial framework, we present a method for determining prediction uncertainty in RES
selection. Our approach employs Tikhonov regularization, direct inversion, conditioned random walks, and nonparametric
bootstrapping. In comparison to geostatistical simulation, we show our method is faster computationally, does not require
variogram construction, needs fewer input parameters, and produces reasonably accurate predictions. Although resolving
near-field RES resulting from small-scale features may not be possible for many systems, macroscopic continuum structure is
shown to be reasonably approximated and useful in developing large-scale hydrologic models. We apply our method of RES
estimation and RES uncertainty analysis under varying data distributions and fracture architectures to illustrate the
resulting effects on model accuracy. Stochastically-generated, three-dimensional, finite element, discrete fracture models
are compared with equivalent two-dimensional continuum models constructed using our RES analysis, which in sum illustrates
the interplay between data distribution, aquifer structure, and scale(s) of continuum representation and their effects on
model accuracy. Though we focus on fractured aquifers, this study has direct implications to spatially correlated
heterogeneous porous media, and should be of interest to researchers who focus in related areas.
DE: 1829 Groundwater hydrology
DE: 1839 Hydrologic scaling
DE: 1879 Watershed
DE: 1880 Water management (6334)
DE: 1894 Instruments and techniques: modeling
SC: Hydrology [H]
MN: Fall Meeting 2005