Lawrence Berkeley National Laboratory

iTOUGH2 provides inverse modeling capabilities for the TOUGH2 family of nonisothermal multiphase flow simulators. It can be used for a formalized sensitivity analysis, parameter estimation by automatic model calibration, and uncertainty propagation analyses. While iTOUGH2 has been successfully applied for the estimation of a variety of parameters based on different data types, it is recognized that errors in the conceptual model have a great impact on both the estimated parameters and the subsequent model predictions. Identification of the most suitable model structure is therefore one of the most important and most difficult tasks. Within the iTOUGH2 framework, model identification can be partly addressed through appropriate parameterization of alternative conceptual-model elements. In addition, statistical measures are provided that help rank the performance of different conceptual models. We present a number of features added to the code that allow for a better parameterization of conceptual model elements, specifically heterogeneity. We discuss how these new features can be used to support the identification of key model structure elements and their impact on model predictions.