In this talk I introduce the concept of an optimal persistence pattern. An optimal persistence pattern (OPP) is a component of a time-varying field that remains auto-correlated for the longest time-lags. Techniques for extracting OPPs provide efficient methods not only for isolating persistent patterns in stationary time series, but also for detecting trends, discontinuities, and other low-frequency signals in nonstationary time series. Examples of OPPs drawn from meteorological and oceanographic data sets will be presented and interpreted. The use of OPPs as a basis for extending statistical forecast schemes will also be discussed. A major problem with the technique, which arises in all multivariate regression analyses, is that it yields artificially large decorrelation times for small sample sizes. The question of what constitutes a sufficient basis set for extracting reliable information remains to be answered.