Non-ideal transport of reactive solutes in heterogeneous porous media: 3. Model testing and data analysis using calibration versus prediction

The transport of reactive solutes is often influenced by multiple factors and processes. Complex mathematical models accounting for these multiple factors and processes are required to simulate accurately the transport of reactive solutes. There are numerous concerns and constraints associated with the application of mathematical models to the transport of reactive solutes, especially for field-scale problems. The specific issue of using calibration (fitting a model to measured data) for model evaluation and data analysis was examined in this paper. Selected laboratory data, as well as the results of the well-known Borden natural-gradient field experiment were used as case studies to illustrate the potential pitfalls of using model calibration for the analysis of reactive solute transport. It is shown that erroneous parameter values can be obtained from calibration if all significant transport factors are not properly represented. Perhaps of greater significance, the (mis)use of calibration can lead to a mistaken belief that the model accurately represents the physical system and, hence, result in a misinterpretation of the factors controlling solute transport.