Interactive Genetic Algorithms for Inverse Groundwater Modeling: Issues with Human Fatigue and Prediction Models

This paper presents an interactive multi-objective evolutionary optimization based approach to solve the inverse problem of estimating heterogeneous aquifer parameters (in this case hydraulic conductivity) for a groundwater flow model. A hypothetical aquifer, for which the ‘true’ parameter values are known, is used as a test case to demonstrate the usefulness of this method. It is shown that using automated calibration techniques without using expert interaction leads to parameter values that are not consistent with site knowledge. In such cases, it is desirable to incorporate expert knowledge in the estimation process to generate more reasonable estimates. An interactive approach is proposed within a multi-objective framework that allows the user to evaluate trade-offs between the expert knowledge and other measures of numerical errors. For the hypothetical aquifer, this type of expert interaction is shown to produce more plausible estimates. A major issue with interactive approaches is ‘human fatigue’. One way of dealing with human fatigue is to use machine learning to model user preferences. This work presents some initial results that show that machine learning models can be used to augment user interaction, allowing the IGA to find good solutions with much less user effort.