Extracting a Smooth Trend from a Time Series: A Modification of Singular Spectrum Analysis

Abstract Singular spectrum analysis is commonly used in climatology to extract a trend from a noisy time series. Implicit in this method is the association of trends with high variance. In many cases, it may be more natural to associate trends with smoothness. This paper describes how singular spectrum analysis can be modified to incorporate this idea. The modified approach is illustrated using the annual central England temperature series.