Propagation of uncertainty through geochemical code calculations

Geochemical codes are used to identify the processes (speciation, solubility, and adsorption) controlling both natural water and waste leachate compositions. The uncertainties in geochemical calculations are important to environmental assessments, because decision-makers must be alerted to the potential error associated with predictions derived from the calculations. In this study, overall uncertainty is determined by propagating Monte Carlo-generated analytical and thermodynamic uncertainties through a geochemical code, and generating probabilistic distributions of the calculated output. Then Generalized Sensitivity Analysis (GSA) is used to separate the model responses into two classes or groups based on specified performance criteria. The relative contribution of the uncertainty associated with each input parameter to the output uncertainty is determined by comparing the cumulative distribution functions of the parameters in the two classes. The combined use of the Monte Carlo method with GSA is used to examine the significance of analytical and thermodynamic uncertainties in predicted pH values for a simple sodium bicarbonate buffer solution and for calcite saturation indices in the system Ca-CO3-Na-Cl-H2O. Statistical results concur with a geochemically based interpretation of each system. This agreement suggests that in the study of complex environmental systems that are not well known, GSA may be used both to focus the investigator's attention on the key parameters and uncertainties in the system, and to distinguish between the parameters that require more precise measurement and those that are sufficiently well known for model interpretation.