Abstract Conceptual models of the fracture networks in shale were evaluated at a site contaminated with chlorinated solvents. Prior borehole testing in eight holes under open hole ambient and pumping conditions identified 14 flow zones (140 m bedrock interval) with zero to five zones per hole. Cross-hole testing showed only a few cross-connections between transmissive fractures. The initial conceptual model thus featured a sparse fracture network with few dominant fractures. Detailed profiles (hydraulic head, rock core volatile organic compounds, groundwater volatile organic compounds from packer and multi-level sampling, cross-hole multi-level monitoring of permanganate injections) were collected from several holes and indicated a well-connected fracture network with many hydraulically active fractures not influenced by open hole cross-connection. This contrasting conceptual model contained numerous well-connected horizontal and vertical fractures that allowed chlorinated solvents to penetrate the upper 50–60 m of bedrock as dense non-aqueous phase liquids, followed by diffusion-driven mass transfer from fractures into the porous rock matrix, such that nearly all the contaminant mass resided as dissolved and sorbed phases, measurable in rock core without cross-contamination during drilling. The difference in the two conceptual models has important implications for source zone and plume attenuation.

Abstract This paper reports a long-term field investigation of a fractured dolostone aquifer that was penetrated by a dense non-aqueous phase liquid. High-resolution source zone characterization shows the evolution of deep penetration to the back-diffusion conditions seen at the present day. Metolachlor, a common herbicide, was released into the overburden overlying a fractured dolostone aquifer within a short time window (1978–81). In 2000, the plume front arrived at a municipal supply well located 930 m down-gradient, increasing to a maximum concentration of 2 μg l −1 . Groundwater monitoring with high-resolution, depth-discrete multi-level sampling systems since 1992 shows a clearly delineated bedrock plume. Numerous rock core samples show metolachlor in the low-permeability rock matrix at the bottom of the aquifer. The mass distribution and bedrock hydraulic head pattern strongly suggest that metolachlor entered the bedrock as a free-phase dense non-aqueous phase liquid penetrating to the aquifer bottom, preferentially accumulating in some horizontal fractures, dissolving quickly as a result of the rapid groundwater flow and then diffusing into the rock matrix, where back-diffusion sustains a dilute, persistent and stable plume. Strong plume retardation by matrix diffusion and sorption has greatly mitigated the impact on water quality in the down-gradient supply well, allowing for its continued use, while back-diffusion and degradation maintain a persistent, dilute plume managed by appropriate monitoring.