Comparison of Laboratory and Field Measurements of Backfill Hydraulic Conductivity for a Large-Scale Soil-Bentonite Cutoff Wall

Abstract

Soil-bentonite (SB) cutoff walls are commonly employed to control groundwater flow and contaminant migration, and the hydraulic conductivity (k) of the SB backfill is a critical parameter in these applications. In this study, an evaluation of backfill k was conducted for a large-scale SB cutoff wall (approximately 200 m long and 7 m deep) constructed and instrumented for field research. Measurements of k (more than 150 in total) were obtained via (1) laboratory (flexible-wall and rigid-wall) tests on specimens molded from surface grab samples of the field-mixed backfill collected during construction, (2) laboratory (flexible-wall) tests on undisturbed backfill specimens prepared from Shelby tube samples taken from the wall, and (3) in situ (slug) tests conducted at 12 different locations within the wall. The laboratory tests were conducted in stages over a range of applied effective stresses (σ′=6.9, 13.8, 20.7, and 34.5 kPa) that encompassed the range of mean in situ effective stresses measured in the backfill (6–12 kPa) via embedded earth pressure and pore pressure sensors. At a given σ′, flexible-wall k values for the undisturbed specimens were similar to those for the laboratory molded specimens. However, k decreased with increasing σ′ in all of the laboratory tests, as expected, and the differences between k values measured at the lower stresses (6.9–13.8 kPa) and those measured at the higher stresses (20.7–34.8 kPa) were statistically significant. The flexible-wall tests underestimated k relative to the slug tests, but the differences generally were small (on average, within a factor of 2) when considering only the flexible-wall specimens tested at the lower applied stresses (6.9–13.8 kPa) that more closely matched the in situ stresses. The results demonstrate the importance of conducting laboratory k tests on SB backfill at stresses representative of the in situ stresses, which are limited by frictional load transfer along the sidewalls and tend to be lower than expected based on an assumed geostatic stress distribution.