GLOBAL TEMPERATURE AS A REGULATOR OF CLIMATE PREDICTABILITY

The possibility of changes in the dynamical behavior of our climate sy stem as the global temperature fluctuates is investigated by consideri ng the large climate variation of the Pacific ocean-atmosphere system associated with Southern Oscillation Index (SOI) and modem methods of time series analysis. First the nonlinear structure of SOI is probed b y estimating local Lyapunov exponents. We obtain the sum of ail positi ve Lyapunov exponents (Kolmogorov entropy) as a function of time for t he period between June 1882 and May 1987. We then demonstrate that thi s time series and global temperature records are coherent over the Nyq uist frequency band from 0.0 to 0.25 cycles/year thus establishing lin ear relationship between many of their oscillatory components. Since t he inverse of the Kolmogorov entropy is a measure of predictability th is result indicates that global temperature may affect predictability with higher temperatures corresponding to lower predictability. We dis cuss how this result may eventually lead to the development of strateg ies for improved long-term predictions.