Potential Predictability of Seasonal Means Based on Monthly Time Series of Meteorological Variables

Abstract

Based only on monthly mean data, an analysis of variance method is proposed for decomposing the interannual atmospheric variability in seasonal-mean time series into components related to ?weather noise? and to slowly varying boundary forcing and low-frequency internal dynamics. The ?potential predictability? is then defined as the fraction of the total interannual variance accounted for by the latter two components. A study using synthetic data showed that the method proposed here is comparable in performance to conventional methods requiring daily data. The technique was applied to gridded global data of monthly surface temperature, 500-hPa height, and 300-hPa wind in order to examine the geographical and seasonal dependencies of their potential predictability. For all the variables, the highest potential predictability tends to be found in the Tropics, where seasonal anomalies in the atmosphere are strongly coupled with the underlying sea surface temperature anomalies and the weather noise component is relatively weak. In contrast, the predictability is generally low over the extratropics. Surface temperature, however, exhibits relatively high predictability over the subtropical and midlatitude oceans, particularly over the midlatitude North Pacific in winter, where the El Niño?Southern Oscillation events exert strong influences through atmospheric teleconnection. These results appear physically reasonable and consistent with our current understanding based on previous observational and model-based analyses.