题目：Nonlinear time sereis analysis: from scaling behavior to prediction tools
袁乃明，2012年获得博士学位,中国科学院大气物理研究所东亚区域-环境重点实验室副研究员,中科院百人计划C类候选人。主要研究方向:气候预测理论,气候系统的长期记忆性,非线性时间序列分析等.近年来发表SCI论文近30篇。其中在J.Climate, JGR, Sci.Rep等著名期刊发表研究论文8篇。2016年曾获世界气象组织Mariolopoulos教授信托基金奖。目前主持国家自然科学基金青年项目1项，面上项目1项。
In this talk, long-term memory (LTM) in climate variability is introduced and a new perspective on climate prediction is proposed. Using a recently developed model, Fractional Integral Statistical Model (FISM), any given climatic time series with LTM can be decomposed into two components: historical memory component M(t) and weather scale excitations . M(t) represents the long-lasting influences accumulated from the past history, which can be considered as a kind of “inertia” in climate system. , however, is not a simple white noise, but contains dynamical information which triggers the current climate regime to change. While for , it is still a big challenge to make reliable estimations. In this talk, several potential approaches for the estimation o f are introduced, among which, we provide more details on a recently proposed method, Detrended Partial-Cross-Correlation Analysis. It is argued that may be better estimated by establishing a hierarchical prediction model, but more efforts are needed in the future.