[en] Accidental release of light non aqueous phase liquids (LNAPLs, such as gasoline) into the subsurface can be a significant health and environmental concern. LNAPLs mostly reside in the zone of water table fluctuation and therefore, pose hazards in air (vapour) and water phases. The recovery and removal of LNAPLs through appropriate methods is a basic step in remediation of contaminated sites. Any technical approach such as pilot and field-scale trials to determine appropriate recovery methods for a particular site and condition may be expensive and time consuming. In addition, the long-term consequence and the endpoint of the recovery is not easily predictable. Simulation provide an avenue to assess this for a range of remediation options. Here we simulate field pilot trials to show how a multi-phase and multi-component framework can be used to determine the effective endpoint of LNAPL remedial options. (author)

[en] We aimed to quantify the possible scales of previous LNAPL petroleum fuel releases (mainly gasoline) into the subsurface at a site with a paucity of information on the possible sources and timing of releases. These incidents occurred prior to or around the late 1990s. Previous studies during the period from 1998 to 2007 and the more recent site investigation in 2014 were considered in framing the simulations used to elucidate the location and scale of the releases along with the fate of the source and dissolved petroleum hydrocarbons in the groundwater at the site. (author)

[en] Highlights: • Reflective optical imaging and adapted IPA framework generated 2-D observation data. • Multiphase numerical model used for Monte Carlo analysis on HPC infrastructure. • Calibration led to plausible results for glass bead and natural sand scenarios. • Calibrated retention curve parameters $n_{\mathit{nw}}$ and ${\alpha }_{\mathit{nw}}$ are unsteady in time. We conducted multiple laboratory trials in a robust and repeatable experimental layout to study dense non-aqueous phase liquid (DNAPL) source zone formation. We extended an image processing and analysis framework to derive DNAPL saturation distributions from reflective optical imaging data, with volume balance deviations < 5.07%. We used a multiphase flow model to simulate source zone formation in a Monte Carlo approach, where the parameter space was defined by the variation of retention curve parameters. Integral and geometric measures were used to characterize the source zones and implemented into a multi-criteria objective function. The latter showed good agreement between observation data and simulation results for effective DNAPL saturation values > 0.04, especially for early stages of DNAPL migration. The common hypothesis that parameters defining the DNAPL-water retention curves are constant over time was not confirmed. Once DNAPL pooling started, the optimal fit in the parameter space was significantly different compared to the earlier DNAPL migration stages. We suspect more complex processes (e.g., capillary hysteresis, adsorption) to become relevant during pool formation. Our results reveal deficits in the grayscale-DNAPL saturation relationship definition and laboratory estimation of DNAPL-water retention curve parameters to overcome current limitations to describe DNAPL source zone formation.