ALBERT

Notes: While the discussion of model uncertainty has centered on spatial heterogeneity, it is possible that ground water models have not enjoyed much success as predictive tools often because the sources that were eventually imposed in the field differed from those represented in the simulations. This is because deterministic prediction of future conditions is often inaccurate due to the random nature of contaminant sources, in terms of their timing, location, and magnitude. This paper presents a stochastic framework for accommodating random contaminant sources in conventional, deterministic advection-dispersion transport models. The contaminant sources are first classified into two types: those occurring continuously with a deterministic component and random variations and those occurring randomly at instantaneous discrete-time intervals. For the first type, the governing partial differential equation (PDE) is replaced by a stochastic PDE. The random variations are modeled by Gaussian noise or Brownian motion, and the solution is obtained by using Ito's integration technique. For the second type, Markovian analysis is used for discrete-time contamination events. Both approaches use a deterministic transport model to generate response functions at any observation location and time. The response functions are then integrated to yield probabilistic description of contaminant transport, from which key statistical properties such as mean, standard deviation, and confidence interval can be drawn.

Notes: Several recent studies at the Macrodispersion Experiment (MADE) site in Columbus, Mississippi, have indicated that the relative preferential flowpaths and flow barriers resulting from decimeter-scale aquifer heterogeneities appear to have a dominant effect on plume-scale solute transport. Numerical experiments are thus conducted in this study to explore the key characteristics of solute transport in two-dimensional flow fields influenced by decimeter-scale preferential flowpaths. A hypothetical but geologically plausible network of 10 cm wide channels of high hydraulic conductivity is used to represent the relative preferential flowpaths embedded in an otherwise homogeneous aquifer. When the hydraulic conductivity in the channels is 100 times greater than that in the remaining portion of the aquifer, the calculated concentration distributions under three source configurations all exhibit highly asymmetrical, non-Gaussian patterns. These patterns, with peak concentrations close to the source and extensive spreading downgradi-ent, resemble that observed at the MADE site tracer tests. When the contrast between the channel and nonchannel hydraulic conductivities is reduced to 30:1 from 100:1, the calculated mass distribution curve starts to approach a Gaussian one with the peak concentration near the central portion of the plume. Additional analysis based on a fieldscale model demonstrates that the existence of decimeter-scale preferential flowpaths can have potentially far-reaching implications for ground water remediation. Failure to account for them in numerical simulation could lead to over-estimation of the effectiveness of the remedial measure under consideration.

Notes: Benzene, toluene, ethyl benzene and xylene (BTEX) dissolved into ground water and migrated from a light nonaqueous phase liquid (LNAPL) source in a sandy aquifer near a petroleum, oil, and lubricants (POL) facility at Hill Air Force Base (AFB), Utah. Field observations indicated that microbially mediated BTEX degradation using multiple terminal electron-accepting processes including aerobic respiration, denitrification, Fe(III) reduction, sulfate reduction, and methanogenesis has occurred in the aquifer. To study the transport and transformation of dissolved BTEX compounds under natural conditions, a reactive flow and transport model incorporating biochemical multispecies interactions and BTEX was developed. The BTEX, oxygen, nitrate, Fe(II), sulfate, and methane plumes calculated by the model agree reasonably well with field observations. The first-order biodegrada-tion rate constants, estimated based on model calibration are 0.051, 0.031, 0.005, 0.004, and 0.002 day−1 for aerobic respiration, denitrification, Fe(III), sulfate reduction, and methanogenesis, respectively. The results of a sensitivity analysis show that the saturated aquifer thickness, hydraulic conductivity, and reaction rate constants are the most critical parameters controlling the natural attenuation of BTEX at this site. The hydraulic conductivity and aquifer thickness were found to be the key factors affecting the restoration of oxygen, nitrate, and sulfate after their interaction with the BTEX plume. The multispecies reactive transport modeling effort, describing BTEX degradation mediated by multiple electron-accepting processes, represents one of the few attempts to date to quantify a complete sequence of natural attenuation processes with a detailed field data set. Because the case study is representative of many petroleum-product contaminated sites, the results and insights obtained from this study are of general interest and relevance to other fuel-hydrocarbon natural attenuation sites

Notes: While significant progress has been made in the theoretical development of the simulation/optimization (S/O) approach for ground water remediation design, its application to large, field-scale problems has remained limited. To demonstrate the applicability and usefulness of the S/O approach under real field conditions, an optimization demonstration project was conducted at the Massachusetts Military Reservation in Cape Cod, Massachusetts, involving the design of a pump-and-treat system for the containment and cleanup of a large trichloroethylene (TCE) plume. The optimization techniques used in this study are based on evolutionary algorithms coupled with a response function approach for greater computational efficiency. The S/O analysis was performed parallel to a conventional trial-and-error analysis based on simulation alone. The results of this study demonstrate that not only would it be possible to remove more TCE mass under the same amount of pumping assumed in the trial-and-error design, but also substantial cost savings could be achieved by reducing the number of wells needed and adapting dynamic pumping. In spite of the large model size of more than 500,000 nodes and a long planning horizon of 30 years, the optimization modeling was carried out successfully on desktop PCs. This field demonstration project clearly illustrates the potential benefits of applying optimization techniques in remediation system design.

Notes: Conceptual geological models based on geophysical data can elucidate aquifer architecture and heterogeneity at meter and smaller scales, which can lead to better predictions of preferential flow pathways. The macrodispersion experiment (MADE) site, with 〉2000 measurements of hydraulic conductivity obtained and three tracer tests conducted, serves as an ideal natural laboratory for examining relationships between subsurface flow characteristics and geophysical attributes in fluvial aquifers. The spatial variation of hydraulic conductivity measurements indicates a large degree of site heterogeneity. To evaluate the usefulness of geophysical methods for better delineating fluvial aquifer heterogeneities and distribution of preferential flow paths, a surface grid of two-dimensional ground penetrating radar (GPR) and direct current (DC) resistivity data were collected. A geological model was developed from these data that delineate four stratigraphic units with distinct electrical and radar properties including (from top to bottom) (1) a meandering fluvial system (MFS); (2) a braided fluvial system (BFS); (3) fine-grained sands; and (4) a clay-rich interval. A paleochannel, inferred by other authors to affect flow, was mapped in the MFS with both DC resistivity and GPR data. The channel is 2 to 4 m deep and, based on resistivity values, is predominantly filled with clay and silt. Comparing previously collected hydraulic conductivity measurements and tracer-plume migration patterns to the geological model indicates that flow primarily occurs in the BFS and that the channel mapped in the MFS has no influence on plume migration patterns.

Notes: Results are presented for numerical simulations of ground water flow and physical transport associated with a natural gradient tracer experiment conducted within a heterogeneous alluvial aquifer of the Natural Attenuation Study (NATS) site near Columbus, Mississippi. A principal goal of NATS is to evaluate biogeochemical models that predict the rate and extent of natural biodegradation under field conditions. This paper describes the initial phase in the model evaluation process, i.e., calibration of flow and physical transport models that simulate conservative bromide tracer plume evolution during NATS. An initial large-scale flow model (LSM) is developed encompassing the experimental site and surrounding region. This model is subsequently scaled down in telescopic fashion to an intermediate-scale ground water flow model (ISM) covering the tracer-monitoring network, followed by a small-scale transport model (SSM) focused on the small region of hydrocarbon plume migration observed during NATS. The LSM uses inferred depositional features of the site in conjunction with hydraulic conductivity (K) data from aquifer tests and borehole flowmeter tests to establish large-scale K and flow field trends in and around the experimental site. The subsequent ISM incorporates specified flux boundary conditions and large-scale K trends obtained from the calibrated LSM, while preserving small-scale K structure based on some 4000 flowmeter data for solute transport modeling. The configuration of the ISM-predicted potentiometric surface approximates that of the observed surface within a root mean squared error of 0.15 m. The SSM is based on the dual-domain mass-transfer approach. Despite the well-recognized difficulties in modeling solute transport in extremely heterogeneous media as found at the NATS site, the dual-domain model adequately reproduced the observed bromide concentration distributions. Differences in observed and predicted bromide concentration distributions are attributed to aquifer heterogeneity at the decimeter (dm) and smaller scales. The calibrated transport parameters for the SSM (i.e., 1:7 for the ratio of mobile-to-total porosity; 2.5 × 10−3 day−1 for the mass-transfer coefficient; 1 m for longitudinal dispersivity; and 0.1 m for transverse dispersiv-ity) are consistent with separate numerical simulations of two earlier tracer experiments at the site. The multiscale modeling approach adopted in this study permits the incorporation of both large-scale geologic features important for flow simulation and small-scale heterogeneities critical for transport simulation. In addition, the dual-domain transport model provides a foundation for multispecies reactive transport modeling studies of natural attenuation of hydrocarbons during NATS.