Approximation ("meta") models have been used in coupled optimization and simulation models to improve computational efficiency. In most instances, multiple simulation runs have been done before the optimization, which are used to fit an approximate model that is then used for the optimization. In this study, we propose a dynamic meta-modeling approach, in which artificial neural networks (ANN) and support vector machines (SVM) are embedded into a genetic algorithm (GA) optimization framework to replace time-consuming flow and contaminant transport models. Data produced from early generations of the GA are sampled to train the ANN and SVM and the numerical models are periodically called to dynamically update the ANN and SVM. This allows the meta model to adapt to the area in which the GA is searching and provide more accuracy. Preliminary results show that a well trained ANN or SVM can achieve satisfactory accuracy. Different approaches to dynamic training will be presented at the conference.