Environ. Sci. Technol. 2006, 40, 6770-6781
Pump-and-Treat Remediation of Chlorinated Solvent Contamination at a Controlled Field-Experiment Site M I C H A E L O . R I V E T T , * ,† STEVEN W. CHAPMAN,‡ RICHELLE M. ALLEN-KING,§ STANLEY FEENSTRA,| AND JOHN A. CHERRY‡ School of Geography, Earth & Environmental Sciences, University of Birmingham, Birmingham, B15 2TT, U.K., Department of Earth Sciences, University of Waterloo, Waterloo, Ontario, N2L 3G1, Canada, Department of Geology, 876 Natural Sciences Complex, University at Buffalo, Buffalo, New York, 14260, and Applied Groundwater Research Ltd., 5285 Drenkelly Court, Mississauga, Ontario, L5M 2H7, Canada
Pump-and-treat (P&T) remediation and associated concentration tailing are investigated at the field scale in a mildly heterogeneous sandy aquifer through the extraction of dissolved chlorinated solvent plumes that had developed over 475 d from a multicomponent dense nonaqueous-phase liquid (DNAPL) source intentionally emplaced in the aquifer at the Borden (ON) research site. Extraction was accomplished via a source-containment well located 25 m from the source and two further downgradient plume-centerline wells to remove the advancing high-concentration dissolved plumes. The 550 days of detailed P&T field data demonstrated the following: remediation, albeit slowly, of the leading 25-60 m plume section to around typical drinking water standard concentrations; concentration tailing (reduction) over 4 orders of magnitude in the plume; a steady-state concentration “plateau” in the source-containment well capturing the steadily dissolving DNAPL source; influences of extraction rate changes (concentration rebounds); and, lengthy tailing from inter-well stagnation-zone areas. Much of the contaminant behavior during the P&T appeared to be “ideal” in the sense that with appropriate specification of the source term and pumping regime, it was reasonably predicted by 3-dimensional numerical model (HydroGeoSphere) simulations that assumed ideal (macrodispersion, linear sorption, etc.) transport. Supporting lab studies confirmed nonideal sorption was, however, important at the point sample scale with enhanced PCE (tetrachloroethene) sorption to low- and high-permeability strata and moderate nonlinear and competitive sorption influences. Although there was limited evidence of nonideal tailing contributions to the field data (underprediction of some tailing curve gradients), such contributions to P&T tailing were not easily discerned and appeared to play a relatively minor role within the mildly heterogeneous aquifer studied. * Corresponding author fax: + 44 (0)121 414 4942; e-mail:
[email protected]. † University of Birmingham. ‡ University of Waterloo. § University at Buffalo. | Applied Groundwater Research Ltd. 6770
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 21, 2006
Introduction Pump-and-treat (P&T) remediation was adopted at 75% of Superfund sites during 1982-1992 (1). Toward the end of that period, however, it became apparent that complete aquifer restoration by P&T to low µg/L drinking water standards was generally not feasible, particularly at sites contaminated by dense nonaqueous-phase liquid (DNAPL) sources (1-3). Significant attention has been paid to the challenges of source zone remediation ever since (4, 5); nevertheless, P&T remains prominent, either to hydraulically contain sources, remove dissolved plumes, or as a component of accelerated source removal technologies. Adoption of so-called “smart P&T” practices where greater consideration is given to contaminant removal or control, thorough site investigation, optimized dynamic management of the well field, and setting of realistic cleanup goals (6, 7) is vital, as is underpinning research on processes that cause exponentially declining concentrations (tailing), quasi-stabilization of concentrations (plateau), or even concentration increases (rebound) (2). Understanding such processes is key to reducing remediation costs and effective P&T optimization (8-12). Although P&T data from contaminated field sites from the past two decades are demonstrative of tailing and supportive of the general consensus that P&T alone fails to achieve permanent aquifer restoration due to the presence of DNAPL (13), such data are frequently too ambiguous to draw specific process-related conclusions on concentration tailing (14). Supporting lab and modeling studies indicate slow NAPL dissolution, diffusion from less permeable strata, or slow desorption from aquifer solids are key processes that cause tailing (15-21). Detailed field research studies relevant to P&T have focused upon reverse diffusion from contaminated aquitards (22-25) and the elution of small sub-portions of plumes (26-28). Detailed P&T field research studies at the whole source-plume scale (29) are surprisingly rare. This field research study evaluates processes limiting concentration reduction during P&T remediation of dissolved chlorinated solvent plumes originating from a multicomponent DNAPL source zone. The source had been purposely installed and 3-dimensional plume development was monitored as part of a controlled field experiment (30). The wellknown contaminant source conditions enabled the following research goals: (i) to evaluate the feasibility (or efficacy) of P&T to remediate high-concentration, dissolved plumes in a mildly heterogeneous aquifer through collection and analysis of a high-resolution 3-dimensional P&T field dataset; (ii) to evaluate the ability of a model based on “ideal” transport assumptions (i.e., macrodispersion, linear and local equilibrium sorption) to reasonably capture field-scale P&T behavior for these conditions; and, (iii) to use the model to explore the combined effects of known source and operational activities on concentration tailing. Exponential concentration tailing trends are the expected behavior for any plume flushing operation, even under ideal transport assumptions; e.g., a trailing plume edge is subject to dispersion. Nonideal processes, however, may cause changes in the exponential slope over time and exacerbate tailing. Such processes include nonequilibrium, nonlinear, and competitive sorption; physical heterogeneity that results in greater dispersion than that observed for a nonreactive tracer; or, slow diffusive contaminant release from lowpermeability strata. These may lead to long remediation timescales even when DNAPL sources have been removed or effectively isolated (25). Although nonideal processes 10.1021/es0602748 CCC: $33.50
2006 American Chemical Society Published on Web 09/22/2006
FIGURE 1. Plan view of the emplaced-source (ES) field experiment site and P&T study area (the shown TCM plume position is based upon ES test sampling (30) and extrapolation of pre-P&T sampling data as discussed herein). inevitably occur in all aquifer settings, their influence on P&T in mildly heterogeneous aquifers may not be that significant, or even easily discernible among other contributing factors or uncertainty in those factors at the plume scale. Such issues are examined using ideal transport model simulations of the detailed field dataset and supporting lab work that quantifies some of the nonideal sorption processes relevant to multi-solute plumes spanning wide concentration ranges.
Materials and Methods Site Description. The P&T study was undertaken at the Borden “emplaced-source” site (ES site), a field-research “sub-site” among others located at the Borden research site, Canadian Forces Base Borden, ON (30, 31). The layout of the ES site is shown in Figure 1 with plates in the Supporting Information (SI) Figure 1. The ES site had hosted the ES natural gradient tracer test, a controlled field experiment to study chlorinated solvent dissolution and transport processes in a natural aquifer setting (30-35). The aquifer comprises glaciolacustrine fine-medium sands with mean properties determined at the ES site of hydraulic conductivity (K) 6.34 × 10-3 cm/s, porosity 33%, bulk density 1.75 g/mL, and fraction organic carbon (foc) 0.021%. Based on the average hydraulic gradient of 0.51%, the calculated natural groundwater velocity was 8.5 cm/d (30). The water table was 2-3 m below ground and the lower 2 m of the aquifer contained an inorganic solute-rich leachate plume derived from an upgradient landfill. The ES tracer test and P&T study were conducted in the 6-m-thick, previously clean groundwater zone overlying the landfill plume, and thus in an environment where the DNAPL and downgradient plume was not in contact with an underlying aquitard. Sand beds at Borden are typically cm to tens of cm thick with mm-scale variability visible (36). The mild to modest
degree of geological heterogeneity in the area where the P&T study was undertaken is shown by four K profiles (Figure 2) and a 16-core transect (SI Figure 2). The ln K variance and K range determined from a 794-sample (5-cm vertical interval) set of permeameter measurements obtained from that transect were 0.49 and 1.62 × 10-5 to 3.12 × 10-2 cm/s, respectively. Low and high K samples were usually present at ∼95-96 m elevations (Figure 2a-c) although some profiles demonstrated remarkable uniformity throughout (Figure 2d, SI Figure 2). The DNAPL source was purposely installed to allow field study of plumes generated from a known source zone (30, 31). A 0.75 m3 rectangular block-shaped source was emplaced 1 m below the water table at 95.1-96.1 m site elevation via a controlled dewatered excavation and subsequent backfilling. The source, mixed at the surface and then deposited into a temporary source form in the dewatered aquifer, contained Borden sand mixed with residual DNAPL comprising a multicomponent mixture of TCM (trichloromethane, a.k.a. chloroform), TCE (trichloroethene), and PCE (tetrachloroethene). Gypsum powder was incorporated to provide a conservative sulfate tracer. Initial source (core) mass estimates were as follows: TCM 1.45 ( 0.33 kg; TCE 8.93 ( 1.42 kg; and PCE 12.60 ( 1.61 kg. DNAPL pore saturation was 5.0 ( 0.7% with initial mole fractions as follows: TCM 0.078 ( 0.011; TCE 0.434 ( 0.012; and PCE 0.488 ( 0.020. Based on these data, literature solubility values (37), and a Raoult’s law analogue, initial effective solubilities were TCM 680 mg/L, TCE 610 mg/L, and PCE 120 mg/L (31). Site monitoring methods, plume development, and source dissolution have been previously described (30-32). The natural gradient test lasted 454 d with dissolved plumes continuously generated from the DNAPL source and monitored via 173 Teflon multilevel samplers (2300 sampling points) (Figure 1). By 418 d, just 15% of the solvent DNAPL had dissolved into the aquifer (31). TCM, the most mobile plume, was 60 m long, 7 m wide, and 3 m vertical thickness with peak concentrations >100 mg/L and plume fringe concentrations detected to
