H43H-01 INVITED 13:40h
Effects of Aquifer Heterogeneity on Mass Flux Emission From DNAPL Entrapment Zones
Organic waste products that are in the form of dense non-aqueous phase liquids (DNAPLs) exhibit complex flow and entrapment behavior in the subsurface. Heterogeneity at different scales from pore to macro-scale in combination with flow instabilities controls the fluid flow behavior and hence the final entrapment architecture. Two primary mechanisms contribute to mass flux emission from source zones where DNAPLs are entrapped. The first is the result of the soluble constituents of the DNAPL dissolving into the flowing groundwater through mass transfer that occurs at DNAPL-water interfaces in residual zones and pools. The relationships of mass flux generation to entrapped mass from residual zones where NAPLs exit in the from of ganglia or residuals and from pools where the saturation varies from high to low values are different due to the differences in the contact morphologies between the flowing aqueous phase and entrapped DNAPL. The second process that contributes to emission is a result of rebounding of dissolved mass that has diffused into the low permeability zones of the aquifer during active DNAPL dissolution. Aquifer heterogeneity that controls the NAPL entrapment and groundwater flow plays a critical role in the generation of mass flux through both these mechanisms. The relationships between the DNAPL mass in the source zone and dissolved mass emission need to be understood and quantified to evaluate the benefits of partial mass removal from source zones during remediation. Methodologies for up-scaling the mass transfer from laboratory to field scale that take into consideration effects of heterogeneity are in their early stages of development. Experimental investigations conducted in intermediate-scale test tanks packed to represent various degrees of heterogeneity that produce different entrapment morphologies and architecture are used to obtain an insight into mass transfer processes. An up-scaling methodology for mass transfer that uses stochastic parameters of the heterogeneity field and parameters that define the entrapment architecture is developed. The data generated in the intermediate-scale test tanks are used to validate this up-scaling methodology.
http://cesep.mines.edu
H43H-02 13:55h
Stochastic fusion of information to characterize DNAPL source zones: conceptual framework and preliminary results from theoretical, computational, and experimental studies
Dense Nonaqueous Phase Liquids (DNAPLs) are prevalent at a large number of sites throughout the world. The variable release history and geologic heterogeneity make the spatial distribution of DNAPLs in the source zone complex. This causes difficulties in cleanup and can contribute to long-term groundwater contamination for decades to centuries. Therefore, the spatial distribution, mass, and composition of DNAPLs present in the source zone need to be characterized in great detail so that efficient remediation schemes can be designed. Development of a cost-effective technology that provides detailed images of DNAPL distributions in the source zones thus is essential. We present here the conceptual framework and preliminary results of a technology that delineates DNAPL source zones in three-dimensions without extensive invasive sampling. This new technology based on the stochastic fusion of information methodology, assimilates results of hydraulic, conservative, and partitioning tracer tomography surveys to derive the best estimate of the DNAPL residual distribution and its uncertainty. Specifically, it first analyzes the information derived from hydraulic tomography to identify the three-dimensional hydraulic heterogeneity of the aquifer. It then improves the estimate of the heterogeneity by incorporating new information acquired from the conservative tracer tomography. Afterward, the improved estimate of heterogeneity is used to simulate the hydraulic tomography such that more detailed information about the response of the subsurface becomes available. This new information again is fed back to the technique to update the estimate of the heterogeneity. The iterative process continues until the available information and measurements are fully utilized to identify the heterogeneity that controls the spatial distribution of DNAPLs. The newly derived knowledge of heterogeneity is then used to design partitioning tracer tomography tests to accurately depict the spatial distribution of DNAPL residual saturation in the source zone. The proposed techniques are being tested and validated through numerical experiments and two sandboxes.
H43H-03 14:10h
DNAPL Source Zone Morphology and Mass Flux: the Role of Aquifer Heterogeneity
To design and implement successful remedial actions, it is necessary to understand the dissolution behaviour of the DNAPL. Remediation cannot be assessed purely in terms of mass or volume removed, but must focus on the mass flux leaving the source zone on completion of cleanup. This study considers the significance of DNAPL spill morphology on source zone dissolution and subsequent mass flux using a combination of numerical simulations and 2-D laboratory experiments. Using the Turning Bands method and assumption of a log-normal distribution of natural aquifer permeability values, a number of stochastic aquifer realizations have been generated using a range of mean, variance and correlation scale parameters. In each test aquifer, a 500L release of PCE was simulated followed by a 3000 day dissolution period using a multi-component, multi-phase simulator. Temporal variations in mass flux leaving the source zone were recorded, and the vertical PCE distribution within the source zone was found to be the dominant controlling factor. This value was then used in a new method for normalising mass flux values. These simulations were repeated for aquifer realisations with different variance and correlation scale values, and corresponding normalised flux values also followed a similar trend. Any exceptions could be explained by examining the PCE spill morphology, which could then be directly related to the aquifer heterogeneity. Six heterogeneous aquifer fields were generated with the same Turning Bands technique and used in a series of laboratory tank experiments. The experiments involved multiple surfactant flushes and a suite of partitioning tracer tests. Following periods of surfactant flushing, vertical PCE mobilization occurred by reduced interfacial tensions and aqueous PCE concentrations in the tank effluent had increased. The measured PCE mass flux values at different stages of remediation were normalised by the observed vertical DNAPL distributions, and the resulting data showed close similarity with the trends seen in the numerical simulations.
H43H-04 INVITED 14:25h
Detecting NAPLs Heterogeneously Distributed in the Subsurface
A particularly difficult task facing engineers and managers concerned with subsurface spills of nonaqueous phase liquids (NAPLs) is determining where the NAPL is and how much is there. Borrowing from past work in petroleum reservoir engineering, partitioning interwell tracer tests (PITT) were developed for characterizing the NAPL source zone and assessing the performance of remediation technologies. PITTs have been used to determine domain-average NAPL saturations as well as the spatial distribution of the NAPL. While these tracer tests work well when the NAPL is distributed uniformly throughout the domain, if NAPL is located nonuniformly, either as millimeter-scale ganglia or pools that are centimeter-scale and larger, the flow paths of the injected tracer solution may bypass NAPL-contaminated zones. In this case, the transfer of tracer mass from the main flow paths to the NAPL may be slow, resulting in extensive tailing of tracer breakthrough curves and underestimation of NAPL mass. In this work we examined the influence of nonuniform NAPL distribution and local-scale mass transfer resistance on the accuracy of measured NAPL saturations using PITTs. Two mathematical models were used along with laboratory column experiments to explore the influence of tracer partition coefficient, tracer detection limit, and injected tracer mass on NAPL measurement when the NAPL was distributed nonuniformly. When dimensionless mass transfer coefficients were small, NAPL measurement errors decreased with decreasing tracer partition coefficient, decreasing tracer detection limit, and increasing injected tracer mass. Extrapolating breakthrough curves exponentially reduced but did not eliminate systematic errors in NAPL measurement. Although transport in a single stream tube was used in the mathematical models and laboratory experiments, the results from this simplified domain were supported by data taken from a three-dimensional computational experiment, where the NAPL resided as large pool. Based on these results, we suggest guidelines for interpreting tracer breakthrough data to ascertain the importance of mass transfer limitations on NAPL measurements.
H43H-05 14:40h
Using Tracers to Describe NAPL Heterogeneity
Tracers are frequently used to estimate both the average travel time for water flow through the tracer swept volume and NAPL saturation. The same data can be used develop a statistical distribution describing the hydraulic conductivity in the swept volume and a possible distribution of the NAPL in the heterogeneous flow field. To determine the statistical distributions we use a simple genetic algorithm to evaluate the parameters of an advective dominated stochastic Lagrangian model. Eight possible NAPL architectures are considered in an attempt to describe the NAPL heterogeneity. The method is appropriate when transverse dispersion is small but would not be appropriate where the formation is dominated by interbeded sands and clays. The approach is illustrated using tracer data from two different sites. Once the model parameters are determined the effects of the site specific NAPL heterogeneity on pump-and-treat remediation can be observed. Change in contaminant mass flux and fraction of contaminant mass removal are projected as a function of the number of pore volumes extracted by the pump-and-treat system.
H43H-06 14:55h
Stochastic Analysis of Partial DNAPL Mass Reduction in a Highly Heterogeneous Glaciofluvial Aquifer
Predictions of the impact of partial mass removal at DNAPL contaminated sites in achieving regulatory compliance goals, as measured in terms of aqueous-phase mass fluxes and contaminant concentrations at a down-gradient boundary, have largely been made using analytical models that are based on a number of simplifying assumptions [e.g., Sale and McWhorter, 2001; Rao et al., 2001; Soga et al., 2004; Falta, 2004]. In this paper, we will make use of a multiphase compositional model to explore the relative times predicted for complete depletion of the DNAPL source due to natural dissolution in a highly heterogeneous aquifer, and if significant environmental benefits can be achieved through DNAPL-zone source removal via enhanced remedial technologies. To gain insight into the implications of various representations of the local-scale kinetic DNAPL-dissolution process, aquifer heterogeneity and the complex architecture of a DNAPL source zone, the aqueous-phase contaminant concentrations and mass fluxes arriving at a down-gradient compliance boundary will be analyzed in a conditional stochastic framework. The hydrogeologic setting upon which the high-resolution simulations are based is a heterogeneous fluvial aquifer in South West Germany, referred to as the aquifer-analog dataset. The aquifer was intensively characterized in three dimensions by the researchers at the University of Tübingen, Germany, for hydrogeological parameters that include permeability, effective porosity, grain size, mineralogy and sorption coefficients.
H43H-07 15:10h
Stochastic Analysis Of Water-Oil Phase Flow In Heterogeneous Media By Combining Karhunen-Loeve Expansion And Perturbation Method
We present a novel approach to modeling stochastic multiphase flow problems, for example NAPL flow, in a heterogeneous subsurface medium with random soil properties, in particular, with randomly heterogeneous intrinsic permeability and soil grain size. A stochastic model for steady state water-oil flow in two-dimensional random field is developed using the Karhunen-Loeve Moment Equation (KLME) approach and is numerically implemented. An exponential model is adopted to define the constitutive relationship between phase relative permeability and capillary pressure. The log-transformed intrinsic permeability Y(x) and soil pore size distribution $\beta$(x) are assumed to be Gaussian random functions with a separable exponential covariance function. The perturbation part of these two log-transformed soil properties is then decomposed into an infinite series based on a set of orthogonal normal random variables . The phase pressure, capillary pressure and phase mobility are decomposed by polynomial expansions and perturbation method. Combining these expansions of Y(x), $\beta$(x) and dependent pressures, the steady state water-oil flow equations and corresponding boundary conditions are reformulated as a series of differential equations up to 2nd order. These differential equations are solved numerically and the solutions are directly used to construct moments of phase pressure and capillary pressure. We demonstrate the validity of the proposed KLME model by favorably comparing 1st and 2nd order approximations to Monte Carlo simulations. The significant computational efficiency of the KLME approach over Monte Carlo simulation is also illustrated.
H43H-08 INVITED 15:25h
DNAPL Migration in Heterogeneous Porous Media
This talk provides an overview of field studies, laboratory experiments, and numerical modelling that has been conducted in the past 15 years to examine dense, non-aqueous phase liquid migration in heterogeneous porous media. Particular emphasis will be placed on relating what has been learned under controlled conditions (i.e., the laboratory, numerical models) to what is observed at actual field sites.