METHODOLOGY FOR AUTOMATICALLY ESTIMATING STABILITY CONSTANTS FROM EXPERIMENTAL DATA

Uncertainties associated with geochemical models are challenging to quantify. The need to calibrate groundwater flow models has been recognized and several applications have been developed. In principle, there is no reason why the same approach can not be used for estimating thermodynamic constants from experimental data. UCODE_2005 is a universal parameter estimation program that, when used together with PHREEQC, can estimate not only parameter values but also how uncertain optimized values are. We demonstrate a methodology for estimating conditional stability constants for metal complexation by a remediation agent, carboxymethyl-b-cyclodextrin (CMCD).

The main uncertainty in estimated stability constants is, in our experience, associated with the conceptual model. For instance, the complexation ratio between metal and complexing agent, the number and types of complexation sites, and the most appropriate activity model are unknown in our example problem. Varying these parameters affects both optimized parameter values and how well the model describes data. A viable option is to develop more than one conceptual model in PHREEQC and estimate unknown parameters for each model. These models will inherently behave differently under various conditions and the most appropriate model will best mimic field conditions. UCODE_2005 provides several statistical tools that can aide in developing the conceptual models and can be used to evaluate and compare several models. These tools include 95% confidence intervals on estimated parameters, measures of importance of observations towards unknown parameter values, need for information to estimate unknown parameters, and, if more than one parameter is estimated, the correlation between these unknown parameters. In addition, shortcomings in the model may be recognized by identifying trends in differences between estimated and observed values.