DEVELOPMENT AND CALIBRATION OF A THREE-DIMENSIONAL VARIABLE-DENSITY MULTICOMPONENT REACTIVE TRANSPORT MODEL TO A CHEMICAL OXIDATION FIELD TRIAL

A reactive transport computer code was developed to simulate geochemical reactions in variable-density flow systems. The objective of this work is to use this code to integrate field-measured data and test conceptual models for groundwater remediation of the chlorinated solvents TCE and PCE using the injection of permanganate.

Three-dimensional numerical models were created of a pilot-scale permanganate oxidation trial. In this experiment, the density of the permanganate solution was the primary mechanism driving the migration of the oxidant. The physical domain consisted of a sandy aquifer overlying a silty clay aquitard with an irregular contact between the units. Reaction by-products and geochemical parameters were monitored using depth-discrete nested sampling points over an eight month period. Chemical processes simulated in the models include DNAPL dissolution, permanganate oxidation of TCE and aquifer organic carbon, calcite dissolution, and precipitation of manganese oxides.

A large aqueous data set was used to calibrate the model. Initial calibration efforts focused on the temporal distribution of the oxidant, and the production of chloride. Of the aqueous components in the model, the chloride data provided the most direct indication of the occurrence and intensity of the oxidation reaction. Subsequent calibration efforts incorporated groundwater pH, alkalinity, calcium, and potassium data. The incorporation of multiple data sets required the development of weighting schemes, and statistical analysis of model residuals and the degree of correlation between the different types of data. Although several investigations of parameter optimization have been published in groundwater literature, parameter estimation in multicomponent reactive transport modeling represents an emerging field of study. As the complexity of the system may limit the effectiveness of trial and error calibration methods, the application of inverse modeling methods for parameter estimation is investigated.