Distributions of predi ction capacity of T63L16 model for mediumrange forecast at different spatial scales

The 500 hPa geopotential height fields of NCEP (National Centers for Enviornment al Prediction) reanalysis data from 1970 to 2003 and hindcasting results from 1982 to 2002 derived from Monthly Dynamic Extended Range Forecast System of Nation al Climate Center (T63L16 Model) were analyzed by the spherical harmonic expansi on and variance analysis to the spherical harmonic coefficients. Spatial scale dependence of daytoday forecast capacity of T63L16 model and predicable stab le components for midiumrange forecast were studied. The results suggest that capacity of the model prediction is rather isotropic in total wave numbern, and the main errors happen at total wave number from 5 to 10, or synoptic scale wave. Based on analyzing the evolution of the ratio of T63L16 model inte rnal variance to climatic external variance, an effective scaledependent ratio threshold was defined which is the average of ratio for 26-40 day prediction to quantitatively decide the predictable limits of T63L16 model for a variety of s patial scales. Furthermore, a set of rational weight coefficients which represen t wave energy spectrum were introduced to study the relationship between pred ictable limits of the model and latitudinal wave number m, as well as total wave number n. The results indicate that the daytoday predictable limits of the T63L16 model are relevant to the zonal and total wave number, as well as the seasons. It was shown that predicable limits, as a whole, gradually shorten with spatial scale reduction, but the variation is not completely monotonic: th e limits at latitudinal wave number of 2 are shorter than that between 3 and 5. In addition, the seasonal averaged predicable stable components to midiumrange forecast were examined. As far as global mean is concerned, for above 6 day for ecast, the predictable stable components are zonal wave number smaller than 12/7 (or total wave number smaller than 17/13) in summer/other seasons, while for 11 -15 day forecast, predictable stable components are zonal wave number smaller th an 5, 3, and 2 (or total wave number smaller than 13, 8, and 7) in both winter and summer, spring, and fall, respectively. The understanding of predictable com ponents for mediumrange forecast could provide the basis for developing new pr ediction strategies and methods, as well as for improving forecasting skills.