Despite significant advances in the understanding of chlorinated solvent source zones and the maturation of several in situ remediation technologies (e.g., bioremediation), the ability to provide a priori predictions of the performance of remediation technologies in the field remains severely limited. This is attributed, in large part, to the inability to accurately quantify source zone mass and its spatial distribution, as well as to accurately estimate effective in situ mass transfer and transformation rates. The overarching goal of this research project was to develop and demonstrate a remediation design and performance assessment protocol that can efficiently assess the suitability of a remediation technology and predict remedial performance (e.g., mass removal/destruction) and the uncertainty associated with such predictions. This protocol couples careful characterization of the contaminant source with down-hole treatability testing and mathematical modeling.

Technical Approach

This project focused on a representative trichloroethene (TCE)-contaminated site (Commerce Street Superfund Site, Williston, VT) that supported the development, refinement and testing of the protocols and software tools. The project was implemented in three phases: Phase I involved the development of protocols and software tools for efficient estimation of key source characteristics (mass distribution metrics) governing remediation technology selection, design, and performance. In Phase II, work focused on laboratory-scale batch and aquifer cell testing to (a) support source zone characterization, (b) provide kinetic data for testing and refinement of upscaled mathematical models, and (c) guide the design and implementation of field-scale reactivity tests. A down-hole treatability (DHT) test was then conducted in Phase III to estimate (confirm) the in situ rate parameters needed for subsequent site-specific protocol application. Upscaled models and uncertainty analysis tools were incorporated into the widely-used solute transport modeling platform, Modular Three Dimensional Transport Simulator (MT3DMS), to facilitate adoption by practitioners and site managers.


Phase I: A novel statistical approach was developed and implemented for the reconstruction of source zone mass distributions and quantification of source zone metrics and associated uncertainty in heterogeneous subsurface formations. The approach employs trained discriminative random field (DRF) models, in conjunction with Monte-Carlo sampling methods, to generate contaminant mass realizations, conditioned on measured borehole data. Post-processing of these realizations yields estimates of source zone metrics (e.g., pool fraction, total mass) and associated uncertainty. These metrics and approximations of uncertainty can be used to predict source zone longevity, mass recovery behavior, and remedial performance and to inform further sampling for characterization and remediation. DRF model performance was evaluated through comparisons of predicted metrics with those obtained from ‘true’ mass distributions, generated with validated flow and transport models. Comparisons demonstrate that the trained DRF model can reconstruct realistic saturation and concentration fields for a range of non-aqueous phase liquids spill scenarios, significantly outperforming traditional kriging approaches.

Phase II: A matrix of laboratory microcosm experiments, assembled from site soil and groundwater samples, was undertaken to estimate batch TCE dechlorination rates in unaugmented and SiREM KB-1®-amended systems. Results demonstrated the need for bioaugmentation at the Commerce St. site. An aquifer cell system was constructed with field site soil, pre-loaded with TCE, bioaugmented with KB-1®, and operated to mimic the field-scale downhole treatability test. Effluent and side port measurements of volatile fatty acid, chlorinated ethene and ethene, and biomass concentrations were used, in conjunction with an enhanced version of the transport simulator MT3DMS with batch-measured rates, to explore effective bio-reaction rates (e.g., maximum substrate utilization rate, /!"#) under representative heterogeneous subsurface conditions. Batch-measured rates provided a good prediction of aquifer cell behavior (average relative error of 19%) when heterogeneity was explicitly modeled and TCE and dichloroethene isomers (cis-DCE) inhibition of vinyl chloride (VC) transformation was neglected. However, ethene production was significantly under-predicted when a uniform permeability was assumed, demonstrating the influence of local heterogeneity on dechlorination prediction accuracy. Ethene concentrations varied spatially within the domain, primarily associated with low permeability layers. To investigate the effect of the residence time on dechlorination, the flow rate was reduced by 50%. Under these conditions, the proportion of ethene (molar basis) increased from 26% to 54%. In this phase of the project, correlations for effective mass transfer coefficients were also developed to describe back-diffusion/desorption under a range of heterogeneous formation conditions.

Phase III: Employing aquifer cell-calibrated rate parameters adjusted for temperature effects, transport modeling of the field DHT test resulted in an over-prediction of ethene production by a factor of 2. Model sensitivity analyses suggest this discrepancy, observed despite the comparable sizes of the aquifer cell and pilot test treatment zone, was associated with unmodeled heterogeneity. Similar to the cell experiment, when the flow rate in the test zone was reduced by 50%, the observed proportion of ethene increased from 17% to 78% at the end of the treatment zone.

Adjoint sensitivity analysis was employed, in conjunction with a first-order second-moment (FOSM) uncertainty analysis method, to optimize borehole sampling for prediction of down-gradient flux-averaged concentration evolution at a contaminated site. The approach was implemented in MT3DMS, and initial source zone conditions were generated by averaging realizations of the DRF model. Results reveal that optimal sampling locations vary with the prediction time window. Comparison of predictions associated with the optimized versus a uniform sampling approach reveals that the FOSM model yields better estimates of down-gradient flux averaged concentration, associated with a significant reduction in variance.

Project results were integrated into a source zone remediation feasibility framework to guide practitioners on the use of the developed modeling methodologies. This framework provides an efficient method to perform site characterization and obtain screening-level forecasts of site behavior, with and without implementation of treatment remedies. Application of the framework to a realistic synthetic field scenario demonstrated its feasibility and potential benefits during conceptual site model refinement and remedial site management.


This research provides site managers, regulatory officials and the scientific community with protocols and software tools to (a) efficiently characterize source zone mass metrics and associated uncertainty, (b) estimate relevant mass transfer and reaction rates for use in upscaled models, and (c) predict remedial performance (or evolution of down gradient plume) and associated uncertainty. The methodologies and tools, although developed for microbial reductive dechlorination, are designed with sufficient flexibility to allow for implementation with other remediation technologies or combinations of remedies. In addition, the research team anticipate that other types of remedial performance data, such as isotopic analysis and proteomics, could be incorporated into the developed models to allow for refined predictions of remedial performance and optimization of remediation-specific site characterization in near real-time.


Boroumand, A. and L. M. Abriola. 2015. On the Upscaling of Mass Transfer Rate Expressions for Interpretation of Source Zone Partitioning Tracer Tests, Water Resources Research, 51:832–847. doi: 10.1002/2014WR015767.

Cápiro, N.L., F.E. Löffler, and K.D. Pennell. 2015. Spatial and Temporal Dynamics of Organohalide-Respiring Bacteria in a Heterogeneous PCE-DNAPL Source Zone. Journal of Contaminant Hydrology, 182:78-90.

De Paolis Kaluza, M. C., E. L. Miller, and L.M. Abriola. 2015. Markov Random Field Models for Quantifying Uncertainty in Subsurface Remediation. IEEE International Geoscience and Remote Sensing Symposium, 26-31 July 2015:4296-4299.

Phelan, T. J., L.M. Abriola, J.L. Gibson, K.M. Smits, and J.A. Christ. 2015. Development and Application of a Screening Model for Evaluating Bioenhanced Dissolution in DNAPL Source Zones. Journal of Contaminant Hydrology, 183:1-15.

Yang, L., X. Wang, I. Mendoza-Sanchez, and L.M. Abriola. 2018. Modeling the Influence of Coupled Mass Transfer Processes on Chlorinated Solvent Plume Persistence in Heterogeneous Source Zones. Journal of Contaminant Hydrology, 211:1-14.

Zhang, H., I. Mendoza-Sanchez, E.L. Miller, and L.M. Abriola. 2016. Manifold Regression Framework for Characterizing Source Zone Architecture. IEEE Transactions on Geoscience and Remote Sensing, 54(1):3-17.