Abstract
An analysis of variance approach, for systematically studying and evaluating the interannual variability and predictability of seasonal mean fields, is demonstrated using an ensemble of six 50-year simulations of the 500 hPa geopotential height field from the Meteorological Research Institute-Japan Meteorological Agency (MRI-JMA) global atmosphere model forced by observed sea surface temperatures (SSTs). The model performance is analysed, for the seasons June-July-August (JJA) and December-January-February (DJF), and compared with NCEP reanalysis.
The magnitude and geographical distribution of the unpredictable weather noise variability (interannual variability of seasonal mean series arising from intraseasonal variability) are simulated well by the model. The spatial distribution of the potentially predictable variability (the interannual variability of seasonal means from which sampling error due to weather noise variability is removed) is simulated fairly well for DJF, but relatively poorly for JJA.
Despite this, however, several extratropical wintertime patterns, identified as the EOFs of the seasonal mean fields, are well simulated by the model. These include the two patterns forced by the El Niño-Southern Oscillation (ENSO): the Tropical Northern Hemisphere (for DJF), and the Tropical Southern Hemisphere (for JJA) patterns. In addition, the three patterns forced by the tropical Indian Ocean: Western Pacific (for DJF), Meridional Wavetrain (for JJA), and Southern Indian Ocean (for JJA) patterns, are also well simulated.
