In maritime applications, autonomous vehicles such as Unmanned Surface Vehicles (USVs) and Unmanned Underwater Vehicles (UUVs) are becoming increasingly prevalent for their utility in defense, science, and exploration. Key enablers of autonomy such as state-estimation, mapping, and control, rely on accurate dynamic models. In this paper, a System Identification algorithm is developed in which a model is iteratively trained using an optimal smoother to condition training data until smoothed data between iterations converges. This algorithm is named Modeling through Iterative Smoothing (MIS). MIS is validated using a simulated version of the WAM-V research platform as a case-study. MIS succeeds in reducing the error of noisy measurement data with respect to ground-truth and consequently results in an accurate dynamic model.