Extreme sensitivity to initial values is an intrinsic character of chaotic systems. The evolution of a chaotic system has a spatiotemporal structure containing quasi-periodic changes of different spatiotemporal scales. This paper uses an empirical mode decomposition （EMD） method to decompose and compare the evolution of the time-dependent evolutions of the x-component of the Lorenz system. The results indicate that the sensitivity of intrinsic mode function （IMF） component is dependent on initial values, which provides some scientific evidence for the possibility of long-range climatic prediction.
Because atmosphere itself is a nonlinear system and there exist some problems using the linearized equations to study the initial error growth, in this paper we try to use the error nonlinear growth theory to discuss its evolution, based on which we first put forward a new concept: nonlinear local Lyapunov exponent. It is quite different from the classic Lyapunov exponent because it may characterize the finite time error local average growth and its value depends on the initial condition, initial error, variables, evolution time, temporal and spatial scales. Based on its definition and the at-mospheric features, we provide a reasonable algorithm to the exponent for the experimental data, obtain the atmospheric initial error growth in finite time and gain the maximal prediction time. Lastly, taking 500 hPa height field as example, we discuss the application of the nonlinear local Lyapunov exponent in the study of atmospheric predictability and get some reliable results: atmospheric predictability has a distinct spatial structure. Overall, predictability shows a zonal distribution. Prediction time achieves the maximum over tropics, the second near the regions of Antarctic, it is also longer next to the Arctic and in subtropics and the mid-latitude the predictability is lowest. Particularly speaking, the average prediction time near the equation is 12 days and the maximum is located in the tropical Indian, Indonesia and the neighborhood, tropical eastern Pacific Ocean, on these regions the prediction time is about two weeks. Antarctic has a higher predictability than the neighboring latitudes and the prediction time is about 9 days. This feature is more obvious on Southern Hemispheric summer. In Arctic, the predictability is also higher than the one over mid-high latitudes but it is not pronounced as in Antarctic. Mid-high latitude of both Hemispheres (30°S―60°S, 30°―60°N) have the lowest predictability and the mean prediction time is just 3―4 d. In addition, predictability varies with the seasons. Most re
CHEN Baohua1, LI Jianping1,2 & DING Ruiqiang2 1. College of Atmospheric Sciences, Lanzhou University, Lanzhou 730000, China
Recent advances in the study of nonlinear atmospheric and climate dynamics in China （2003 2006） are briefly reviewed. Major achievements in the following eight areas are covered： nonlinear error dynamics and predictability; nonlinear analysis of observational data; eddy-forced envelope Rossby soliton theory; sensitivity and stability of the ocean＇s thermohaline circulation; nonlinear wave dynamics; nonlinear analysis on fluctuations in the atmospheric boundary layer; the basic structures of atmospheric motions; some applications of variational methods.
