Prewhitening of Climatological Time Series

Abstract

The most common procedures to prewhite a climatological time series are considered and compared with those based on Fourier analysis. In particular, advantages and shortcomings of the anomaly method and the seasonal differences are noted. Some modifications for the Fourier prewhitened synthesis are introduced to preserve the smoothness of the spectra. A simple recursion filter with prewhitening ability is presented and compared with the other filter using synthetic and a real time cries. Special attention is given to the case of short time series. The anomaly method and the recursion filter appear as suitable procedures to replace a Fourier technique giving both similar outputs but differing in the details. The seasonal differences, although very simple to apply, introduce substantial modifications in the output and should be avoided. When the data series contains a few times the cycle to be eliminated the recursion filter seems to be safer than the anomaly method.