Ncep Dynamical Seasonal Forecast System 2000

JULY 2002 AMERICAN METEOROLOGICAL SOCIETY | I n April 2000, a new dynamical seasonal prediction system was introduced at the National Centers for Environmental Prediction (NCEP; the acronyms used in this paper are summarized in the appendix). The transition to the new system was hastened by a computer fire in September 1999 and subsequent changeover from a Cray C90 to an IBM-SP computer system. This article will be a reference for people who are using the NCEP numerical seasonal forecast products. The first-generation dynamical seasonal prediction model was based on the notion that the seasonal predictability in the Northern Hemisphere extratropics mainly stems from equatorial tropical Pacific sea surface temperature anomalies (SSTA). This predictability motivated the NCEP Coupled Modeling Project to couple the atmospheric model with a Pacific basin ocean model (Ji et al. 1995). A major effort was made to build an ocean data assimilation system and a coupled prediction system to produce skillful SSTA forecasts over the tropical Pacific and to improve the physical parameterizations of the atmospheric model to better simulate the middle-latitude atmospheric response to anomalous tropical SST forcing. This approach led to successful wintertime seasonal predictions (Ji et al. 1995; Kumar et al. 1996; Behringer et al. 1998; Ji et al. 1998). The second-generation dynamical seasonal prediction is designed 1) to further improve the coupled modeling system to refine wintertime prediction, and 2) to take advantage of additional sources of predictability during other seasons. The latter requires incorporating new physical processes. In particular, in recent years the effect of soil moisture shows promise of enhancing summertime predictability (e.g., Koster et al. 2000). Forcing by snow and sea ice may also enhance predictability. Atmospheric initial conditions may also help increase prediction skill. Incorporation of these processes in a dynamical prediction system requires taking care with the initialization procedure. The dynamical seasonal prediction is still a basic research problem, because our understanding of the NCEP DYNAMICAL SEASONAL FORECAST SYSTEM 2000