Proceedings and Articles: ModelCARE99 Mary C. Hill Predictive Modeling of Flow in a Two-Dimensional Intermediate-Scale, Heterogeneous Porous Media GILBERT R. BARTH, MARY C. HILL U.S. Geological Survey, Water Resources Division, 3215 Marine St., Boulder, CO 80303-1066 email: barthg@bechtel.colorado.edu TISSA H. ILLANGASEKARE Division of Env. Sciences and Eng., Colorado School of Mines, Golden, CO 80401-1887 HARIHAR RAJARAM Department of Civil and Env. Eng, University of Colorado, Boulder, CO 80309-0428 1999, IAHS Publication no. 265 ABSTRACT To better understand the role of sedimentary structures in flow through porous media, and to determine how small-scale laboratory-measured values of hydraulic conductivity relate to in situ values this work deterministically examines flow through simple, artificial structures constructed for a series of intermediate-scale (10-m long), two-dimensional, heterogeneous, laboratory experiments. Nonlinear regression was used to determine optimal values of in situ hydraulic conductivity, which were compared to laboratory-measured values. Despite explicit numerical representation of the heterogeneity, the optimized values were generally greater than the laboratory-measured values. Discrepancies between measured and optimal values varied dependent on the sand sieve size, but their contribution to error in the predicted flow was fairly consistent for all sands. Results indicate that, even under these controlled circumstances, laboratory-measured values of hydraulic conductivity need to be applied to models cautiously. Fig. 1 Intermediate-scale tank design showing (a) the dimensions of the entire tank including a simplified representation of the sand-cell packing and the pressure and sampling ports, (b) the frequency distribution of the five sands packed in the tank categorized using the natural log of their permeameter values, and (c) an enhanced gray-scale detail of the tank packing. Each sand lens is 2.54-cm tall and 25.4-cm long, and is tapered at the ends as shown in (c). Discussion/Conclusions Despite careful construction of the porous media, the detailed information available and an exhaustive consideration of possible errors, predictions of tank effluent using mean column-measured hydraulic conductivity were 13% less than the observed effluent, while the highest column-measured values closely reproduced the effluent rate. For the mean constant-head column values the scaled discrepancies were comparable for all sands indicating that the measured values of each sand had similar contributions to the under-prediction of flow. The data presented illustrate limitations on the application of laboratory-measured hydraulic-conductivity values to predictive modeling of heterogeneous systems, even in the absence of scaling and zonation issues. Acknowledgments Financial support was provided by the U. S. Geological Survey, U.S. EPA Great Plains/Rocky Mountain Hazardous Substance Research Center and the U.S. Army Research Office at Research Triangle Park, North Carolina. mchill@usgs.gov Last Modified: August 15, 2000