A0291
Title: Modeling and forecasting the long memory of cyclical trends in paleoclimate data
Authors: Philipp Sibbertsen - University of Hannover (Germany) [presenting]
Tomas del Barrio Castro - University of the Balearic Islands (Spain)
Alvaro Escribano - Universidad Carlos III Madrid (Spain)
Abstract: The relevant cycles are identified and estimated in paleoclimate data of earth temperature, ice volume and CO2. Cyclical cointegration analysis is used to connect these cycles to the earth's eccentricity and obliquity and to see that the earth's surface temperature and ice volume are closely connected. These findings are used to build a forecasting model that includes the cyclical component as well as the relevant earth and climate variables, which outperforms models by ignoring the cyclical behavior of the data. The turning points can be especially predicted accurately using the proposed approach. Out-of-sample forecasts for the turning points of earth temperature, ice volume and CO2 are derived.
