Edinburgh Napier University

  It is well established that the presence of dense non-aqueous phase liquids (DNAPLs) such as trichloroethylene (TCE) in aquifer systems represents a very long-term source of groundwater contamination. Significant effort in recent years has been focussed on developing effective methods to remediate these DNAPL sources, but less on rigorous methods to assess their performance. Rather than concentration/time analysis, mass flux changes in time or travel distance is rapidly gaining acceptance as a treatment performance metric. Where one or more rows of multilevel samplers are installed perpendicular to the direction of contaminant transport (i.e. MLS transects), flux is obtained by integrating discrete concentration data across the transect. This is often accomplished using Theissen Polygon or standard kriging methods. However, both approaches have inherent uncertainty that impact on the reliability of performance considerations. The Theissen Polygon method requires that groundwater velocity is known at each MLS point, while accurate kriging is sensitive to the assumed variogram model. Error contributions also come from analytical imprecision of measured concentrations (~5-15%) and determination of velocity (~10-50%). Often an assumed velocity is used: these can vary over orders of magnitude through the formation at the scale of a MLS transect, and are non-uniformly distributed. Integrating these errors in the interpolation of concentrations should significantly improve the reliability of performance assessment. We applied a robust geostatistical method to interpolate transect concentration data from the Source Area BioREmediation (SABRE) experiment currently underway in the UK. By accounting for analytical error and measured spatial variability of velocity, we can assign a confidence interval to each interpolated value. The result is a "true" concentration value and a probability. Incorporating the probability into concentration integration increases the confidence in treatment performance assessment in heterogeneous aquifers by rigorously attaching uncertainty to mass flux estimates.