Characterisation of a DNAPL source zone in a porous aquifer using the Partitioning Interwell Tracer Test and an...

In this paper, we discuss the results of a Partitioning Interwell Tracer Test (PITT) performed in a large scale experiment with a well-defined TCE spill, and present a novel combined analytical-numerical inverse modelling approach using measured concentration profiles within a TCE plume to predict the distribution of the DNAPL in a virtual vertical plane of the source. The proposed inverse modelling approach assumes local thermodynamic equilibrium of the distribution of TCE between the NAPL phase and the aqueous phase and no decay or sorption of the dissolved TCE concentrations downstream of the spill area. The analytical part of the inverse modelling approach contains two steps. As a first step, the location of the contaminant in a virtual vertical plane of a porous medium is fixed by using measured concentration profiles and considering the dissolution of the organic phase under equilibrium conditions. In the second step, the volume of contaminant entrapped in the source cells is estimated. A multiphase advective-dispersive transport model is used in the final step to adjust the volumes quantified in the second step. The predictions are highly dependent on the quantity and quality of the data in space and time. From the PITT-breakthrough curves measured at the pumping well, a mean TCE saturation in the sweep zone of 0.0004 was derived, which is very low compared to that determined at the local scale. In a second analysis, tracer breakthrough curves available at measuring points placed closely downstream and upstream of the presumed source zone, were used to explain why the globally obtained DNAPL saturation was very low compared to the "real", locally evaluated TCE saturations in the source zone. This was principally caused by the overall travel time compared to the short travel time of the tracers in the source zone. Another reason is that due to bypassing, only part of the volume of tracer injected had been in contact and had eventually interacted with the DNAPL. Furthermore, the quantified TCE volume was nearly 30% higher than the spilled volume; this agrees with the conclusions from other studies emphasizing that calculated volumes can overestimate the measured volumes, particularly in the case of an inhomogeneous distribution of the DNAPL within the soil. A good agreement of the measured and inversed concentration profiles was obtained, highlighting that it is possible to determine the length-averaged distribution of a residual pollution source from dissolved concentration profiles measured downstream of the source zone. The numerically obtained non-uniform distribution of DNAPL entrapped in the vertical plane of the source zone was experimentally confirmed by the TCE saturation values derived from PITT-breakthrough curves at measuring points located 0.75 m downstream of the source zone. However, the sensitivity study showed that the inverse modelling approach provided a rather non-unique solution. More data available may reduce the number of possible representations of the estimated source zone.