| Based on the geo-potential height at 500hPa provided by NCEP/NCAR for period 1970～2003 and hindcasting results derived from National Climate Center Monthly Dynamic Extended Range Forecast System (T63L16 Model) for period of 1982～2002, spatial scale-dependence of predictability of day-to-day forecast of T63L16 Model and predicable stable components for mid-range forecast have been studied with the spherical harmonic expansion and variance analysis.Firstly, inter-annual variance of the climatic atmosphere and interior variance of T63L16 model have been analyzed in detail. The results show that the predictability depends to a great extent on the total wave-numbers in spatial scale. When the forecast time prolongs, the distribution of interior variance is close to that of exterior variance gradually. Predictability of day-to-day forecast of T63L16 Model is isotropic in total wave-numbers n. Furthermore, characteristics of distribution and of evolution of variance ratio, Ratio, in two-dimensional wave-number domain have been analyzed. The results indicate that Ratio increases as the wave-number increases in the first several days of forecasting, and the predictability of large-scale movement is better than small-scale movement from the start. Not only the time when the values of Ratio of different wave-numbers reach saturation at are different, but also the saturated values of Ratio are different. The error of small-scale movement grows faster than that of large-scale movement. After the forecast time prolonging for some days, Ratio reaches a stable state, i.e. the saturation state.On this basis, an index to determinate the predictability limits is given. Then the spatial scale-dependence of day-to-day predictability of T63L16 Model are analyzed. The results show that predictability limits of this model is associated with the zonal wave number m and total wave number n and is season-dependent. On the whole, predictability limits reduce as the wave-numbers increase, but the reduction tendency is not monotone simply. For zonal wave-numbers m=0, spring predictability limit of 23d is the longest in the four seasons. For long-wave and synoptic wave, summer predictability limits, with values of 14.1d and 13.7d respectively, are the longest, these of winter are shorter, and these of both spring and fall are shortest. For mesoscale wave, the longest predictability limit, with a value of 8.3d, occurs in summer and it is about two times of the predictability limits of other seasons. For the whole wave, the longest predictable limit of 17.2d occurs in winter.Subsequently, mid-range forecast of 6-15 days, which is beyond present forecast reliability, and the predictable stable components of T63L16 Model have been discussed. As far as global mean is concerned, for above 6 days' forecast, predictable stable components are the waves with zonal wave-numbers smaller than 12 (or total wave-numbers smaller than 17) in summer and with zonal wave-numbers smaller than 7 (or total wave-numbers smaller than 13 ) in other seasons, while for 11-15days' forecast, predictable stable components are the waves with zonal wave-numbers smaller than 5 (or total wave numbers smaller than 13) both in winter and summer, with zonal wave-numbers smaller than 3 (or total wave numbers smaller than 8) in spring and with zonal wave-numbers smaller than 2 (or total wave numbers smaller than 7) in fall.This paper has mainly explored the predicable stable components of the atmosphere and numerical model in 6-15 days, and the results could provide reference for improving forecast effect. So it favors researches on the development of multiple initial-value optimal ensemble method.
|
PDF Full Text Request