Revisiting the Cape Cod Bacteria Injection Experiment Using a Stochastic Modeling Approach
Bromide and resting-cell bacteria tracer tests carried out in a sand and gravel aquifer at the USGS Cape Cod site in 1987 were reinterpreted using a three-dimensional stochastic approach and Lagrangian particle tracking numerical methods. Bacteria transport was strongly coupled to colloid filtration through functional dependence of local-scale colloid transport parameters on hydraulic conductivity and seepage velocity in a stochastic advection-dispersion/attachment-detachment model. Information on geostatistical characterization of the hydraulic conductivity (K) field from a nearby plot was utilized as input that was unavailable when the original analysis was carried out. A finite difference model for groundwater flow and a particle-tracking model of conservative solute transport was calibrated to the bromide-tracer breakthrough data using the aforementioned geostatistical parameters. An optimization routine was utilized to adjust the mean and variance of the lnK field over 100 realizations such that a best fit of a simulated, average bromide breakthrough curve is achieved. Once the optimal bromide fit was accomplished (based on adjusting the lnK statistical parameters in unconditional simulations), a stochastic particle-tracking model for the bacteria was run without adjustments to the local-scale colloid transport parameters. Good predictions of the mean bacteria breakthrough data were achieved using several approaches for modeling components of the system. Simulations incorporating the recent Tufenkji and Elimelech [1] equation for estimating single collector efficiency were compared to those using the Rajagopalan and Tien [2] model. Both appeared to work equally well at predicting mean bacteria breakthrough using a constant mean bacteria diameter for this set of field conditions, with the Rajagopalan and Tien model yielding approximately a 30% lower peak concentration and less tailing than the Tufenkji and Elimelech formulation. Simulations using a distribution of bacterial cell diameters available from original field notes yielded a slight improvement in the model and data agreement compared to simulations using an average bacteria diameter; variable bacterial cell diameters lowered the modeled peak concentrations and more significantly diminished the tailing behavior, particularly for the Rajagopalan and Tien model of collision frequency. Spatial variability in detachment had little effect on the results. The Lagrangian particle transport model representing the non-idealities of the colloid transport process appears to be a robust, grid-free method for modeling field-scale distribution problems where incorporation of fine-scale heterogeneity would necessitate large numbers of computational cells. The stochastic approach based on estimates of local-scale parameters for the bacteria-transport process both captures the mean field behavior of bacteria transport and calculates an envelope of uncertainty that brackets the observations in most simulation cases.
- Research Organization:
- Lawrence Livermore National Lab. (LLNL), Livermore, CA (United States)
- Sponsoring Organization:
- USDOE
- DOE Contract Number:
- W-7405-ENG-48
- OSTI ID:
- 943825
- Report Number(s):
- UCRL-JRNL-226331; ESTHAG; TRN: US200902%%431
- Journal Information:
- Environmental Science and Technology, Vol. 41, Issue 16; ISSN 0013-936X
- Country of Publication:
- United States
- Language:
- English
Similar Records
Development of RWHet to Simulate Contaminant Transport in Fractured Porous Media
Geostatistical and Stochastic Study of Flow and Transport in the Unsaturated Zone at Yucca Mountain