Assessment of SSA predictions of Earth temperature records

In this paper, the forecasts of Earth temperature records made by A. Pepelyshev and A. Zhigljavsky in 2009 are compared with the data actually observed during 2010–2014. It is demonstrated that the forecasts made in 2009 are quite accurate. In the second part, the SSA-based change-point detection algorithm proposed by Moskvina and Zhigljavsky in 2003 is applied to the same temperature records data. The results show that the data does not have essential structural breaks except perhaps a small rise of general level of temperatures at around 1998.

Keywords

singular spectrum analysis, stability of forecasts, long-horizon forecasting, Earth temperature, change-point detection

2010 Mathematics Subject Classification

62M20

Full Text (PDF format)

Published 27 September 2016