Stratospheric Predictability: Basic Characteristics in JMA 1-Month Hindcast Experiments for 1979–2009

This study investigates basic characteristics of stratospheric predictability in the Northern Hemisphere using 1-month hindcast (HC) experiment data from the Japan Meteorological Agency for 1979–2009. The author describes characteristics of forecast properties of spread, error (root-mean-square error), and anomaly correlation, contrasting the stratosphere and troposphere for different seasons and exploring the so-called spread–skill relationship for the winter stratosphere. The properties are defined for each HC set (ensemble forecasts initialized on the same day). The error and anomaly correlation are calculated with the ensemble mean as measures of forecast accuracy. The author also examines the role of stratospheric sudden warmings (SSWs) in variations in forecast accuracy. Results show that, for lead times shorter than about 10–15 days, the accuracy of the HC data is higher on average and more variable in the stratosphere than in the troposphere, especially for the northern winter. This is reflected in larger averages and variability in the predictable time limit, or the characteristic time scale of useful predictions, for the winter stratosphere. The author also reveals that the spread–skill relationship for the northern winter stratosphere is characterized by the existence of notable outliers from their expected linear distribution; the outliers have markedly large errors for given spreads. Most outliers are contributed by HC sets initialized before observed major SSWs. Such HC data fail to reproduce the strength and/or shape of the stratospheric polar vortex, including both onset and recovery phases of SSWs. The HC data tend to yield a too-strong vortex and shorter-than-average predictable limit.