The analysis of time-dependent data is often guided by the question of how dominant structures develop over time. It is important to understand how patterns or structures identified for one time step evolve over time, by changing or moving in the domain. To gain insight into such evolving structural change it is crucial to effectively compare different time steps. This paper proposes a comparison method for twodimensional flow fields. The method is based on a feature description using invariant moments. The specific strength of these moments is their invariance under scaling and rotation, thus facilitating an identification of features even if they occur at other positions, with changed orientation, and variation in size. In addition the moments themselves can beused to define a similarity measure. To evaluate the significance of this concept it has been applied to wind speed data from meteorological simulations.