Illustration of the automated model selection procedure. Up to twenty harmonics were computed in a brute force approach, depending on data availability. The statistical significance of increasing the model’s complexity was assessed using both F-test (p < 0.05) and BIC (ΔBIC < 0) criteria. Violations of these criteria were accepted if the two subsequent, more complex, models both indicated significant fit improvements. The improvement χ²/N_DOF of the fit is shown in green.