Use of Model Error as a Screening Tool for Selecting a Model Structures for a Lab-Scale NAPL Dissolution Experiment

Deficiencies in groundwater modeling can be attributed to two reasons: imperfect mathematical description of the physical process and uncertainty in the adopted values of model specific parameters. While the effect of parameter uncertainty on the model performance has been the subject of extensive research, little attention has been paid to the effects of the imperfect or mis-conceptualization of the model structure. The error resulting from the imperfect formulation of the model structure, referred here as model error, is not random but systematic and thus grows asymptotically with increase in forecast length. For a given problem, the availability of multiple model structures leads to the question-- which model structure is the best choice and why? To illustrate the problems related to the selection of model structure, an attempt has been made here to model a data set from a lab-scale NAPL dissolution experiment. The laboratory model setup is composed of a sand tank fitted with extraction and injection wells and multilevel observation ports. TCE NAPL was released into the system and breakthrough concentrations were observed. Laboratory measured values of the physical properties of the porous medium, such as hydraulic conductivity and porosity, were assumed to be the true known values. The observed concentration breakthroughs were assumed to represent true measurements from an unknown model response. Then the dissolution process of the TCE NAPL source was simulated by using steady-state and transient dissolution models. The interaction of contaminants with the porous media was simulated under equilibrium and rate-limited sorption models. The simulated result was compared to the observed breakthroughs to quantify the model error for a particular model structure. This work is a preliminary attempt to illustrate how laboratory observed concentration breakthrough data could be used as a screening tool to select a suitable model structure from a set of available competing model structures.