Atmospheric Seasonal Predictability Experiments by the JMA AGCM

Abstract

Atmospheric seasonal predictability is investigated using the Japan Meteorological Agency (JMA) Atmospheric General Circulation Model (AGCM) which is a global spectral model of T63 resolution used for former operational one-month farecasts. Four-month ensemble integrations were performed from nine consecutive days of initial condition preceding the target season. All four seasons in the 15-year period from 1979 to 1993 are chosen as the target seasons. The model was forced with observed sea surface temperature (SST) during the time integrations. Verification by the anomaly correlation of seasonal averaged 500 hPa height are summarized as follows. (1) In terms of a 15-year mean, ensmble average forecasts tend to have higher skill than means of individual forecasts, persistence forecasts and climatological forecasts over the whole globe. (2) In the Northern Hemisphere, skill is high in spring and winter, and low in summer and autumn. (3) Skill is higher over East Asia and North America, than over Europe and Atlantic regions. In winter, the relatively higher skill over East Asia and North America can be attributed to reasonable reproducibility of Western Pacific (WP) and Pacific/North American (PNA) teleconnection patterns by the model. (4) A larger ensemble size improves the skill of the model both for extratropical regions and the tropics, and for all four seasons. (5) The skill of one-month lead time forecasts is lower than that of no lead time forecasts for all regions and all seasons. Focusing on the Northern Hemisphere, the relatively large skill degradation with respect to lead time is striking in spring, which suggests that the model is more sensitive to initial conditions in spring than in summer and autumn. (6) Skill enhancement for strong El Niño and the Southern Oscillation (ENSO) years is generally evident over the whole globe, but is not so striking in the Northern Hemisphere. Skill over North America increases for ENSO years, but there is no clear skill enhancement over the Europe and Atlantic regions. The model’s ability to reproduce the interannual variability of precipitation was also investigated. Geographycal distributions of interannual temporal correlation coefficients between observed precipitation, and model ensemble average forecast precipitation show that correlations are generally higher in the tropics than in the extratropics for all seasons. In the tropics, skill is relatively higher in the equatorial eastern Pacific ocean and lower over India. The model has difficulty in simulating the interannual variability of precipitation for the Indian summer monsoon season and for the East Asian rainy season in summer, which may originate from the model’s poor climatology over these regions in summer.