In many cases where operational modal analysis is of interest, harmonic components are disturbing the estimation of modal parameters. Sometimes the harmonics can be included in the parameter estimation process and subsequently removed (because the damping values become unrealistically low). However, this is difficult if a natural frequency is very close to, or coincides with, a harmonic. In this paper we investigate a recently proposed method for removing harmonics when estimating spectral densities for operational modal analysis. The method is essentially based on removing the harmonics in a periodogram, prior to estimating the spectral density by smoothing the same periodogram. This makes it a very time efficient method, as the harmonic removal is integrated into the spectral analysis process. To produce data for testing the method, a finite element dynamic model representing a wind turbine was created. The MATLAB-based finite element code CALFEM was used to achieve a flexible description of the tower. An aerodynamic force representation excited the rotating structure. An unbalance in one of the blades was introduced to produce a harmonic at the rotation speed of the turbine. To validate the efficiency of the harmonic removal, modal parameters were estimated using simulated data prior to, and after, removing the harmonics. The results show that the harmonic removal with the proposed method is efficient.