用天气变量时间序列估计天气的可预报性

ESTIMATING WEATHER PREDICTABILITY FROM THE TIME SERIES OF WEATHER VARIABLES

本文从非线性系统的吸引子概念出发,用单个气象时间序列重构维数较高的相空间并嵌入天气吸引子,根据相轨道上初始时刻紧邻的点随时间的演化来估计吸引子的维数和天气的可预报性。用500hPa亚洲环流指数和北京冬季气温的逐日资料计算表明,天气吸引子的维数分别为3.8和5.4;可预报时间尺度约6—14天,考虑相空间e指数膨胀因素后为4—9天。

Based on the concept of attractors of nonlinear system, the phase space with higher dimension is reconstructed by using observed single meteorological time series and then the weather attractor is embedded in it. The dimension of weather attractor and the weather predictability can be estimated from the time evolution of initially close pieces of trajectories. Computation results used daily data sets of the general circulation index at 500 hPa in Asia and Beijing temperature in wintertime show the fractal dimensions of 3.8 and 5.4 for these two attractors, respectively; and for which predictability time scale of 6-14 days, while weather predictability time scale of 4-9 days resulted from the e-folding expansion of trajectories in phase space.