A1032
Title: Correlation change detection for prediction in climate science
Authors: Zhaoyuan Li - The Chinese University of Hong Kong, Shenzhen (China) [presenting]
Josef Ludescher - Potsdam Institute for Climate Impact Research (Germany)
Jie Gao - Chinese university of Hong Kong(Shenzhen) (China)
Abstract: Reliable predictions of climate events remain a great challenge. The aim is to propose a new prediction method based on the correlation structure of multiple climate time series. The strength of the cross-correlation between climate variables or between different regions changes constantly, especially before and after some critical climate events, which is the basis for prediction by the proposed correlation change detection method. Algorithmic and computational properties are discussed, and simulations are presented to support the conclusions. This new approach can improve prediction and understanding of regional and global climate processes. It is applied to forecast Indian Ocean dipole events and La Nina episodes, respectively.
