Time series can be decomposed as the sum of several components, each one of them having a different frequency. Using this approach, we can see which are the frequencies that dominate: for example, there might be a yearly component and a weekly component.  

This spectral decomposition is useful for understanding cycles, and the dynamics of a time series. It can also be used as a preliminary tool for studying seasonality in detail. This knowledge can then be used to choose the right seasonality structure for removal from the series.