文摘
Bromide and resting-cell bacteria tracer tests conductedin a sandy aquifer at the U.S. Geological Survey CapeCod site in 1987 were reinterpreted using a three-dimensionalstochastic approach. Bacteria transport was coupled tocolloid filtration theory through functional dependence of local-scale colloid transport parameters upon hydraulicconductivity and seepage velocity in a stochastic advection-dispersion/attachment-detachment model. Geostatisticalinformation on the hydraulic conductivity (K) field that wasunavailable at the time of the original test was utilizedas input. Using geostatistical parameters, a groundwaterflow and particle-tracking model of conservative solutetransport was calibrated to the bromide-tracer breakthroughdata. An optimization routine was employed over 100realizations to adjust the mean and variance of the natural-logarithm of hydraulic conductivity (lnK) field to achievebest fit of a simulated, average bromide breakthrough curve.A stochastic particle-tracking model for the bacteriawas run without adjustments to the local-scale colloidtransport parameters. Good predictions of mean bacteriabreakthrough were achieved using several approaches formodeling components of the system. Simulations incorporating the recent Tufenkji and Elimelech (Environ. Sci.Technol. 2004, 38, 529-536) correlation equation forestimating single collector efficiency were compared tothose using the older Rajagopalan and Tien (AIChE J. 1976,22, 523-533) model. Both appeared to work equally wellat predicting mean bacteria breakthrough using a constantmean bacteria diameter for this set of field conditions.Simulations using a distribution of bacterial cell diametersavailable from original field notes yielded a slightimprovement in the model and data agreement comparedto simulations using an average bacterial diameter. Thestochastic approach based on estimates of local-scale parameters for the bacteria-transport processreasonably captured the mean bacteria transport behaviorand calculated an envelope of uncertainty that bracketedthe observations in most simulation cases.