A0593
Title: Common persistent cycles
Authors: Anthoulla Phella - University of Glasgow (United Kingdom) [presenting]
Vasco Gabriel - University of Victoria (Canada)
Luis Filipe Martins - ISCTE-IUL (Portugal)
Abstract: Paleoclimatic time series such as ice volume, CO2 emissions, and temperatures have very similar time series patterns during ice ages and inter-glacial periods, particularly common persistent long cycles. Their study is revisited using a dataset that ranges from the period -145,000 to the present (in 1000-year intervals), contributing to the empirical debate on how temperatures are determined by CO2 and orbital variables. A novel state-space approach is developed that offers a simple, reliable and robust characterization of common persistent cycles, which are present in two or more time series. The first contribution is methodological. To explore the properties of the model that capture the dynamics of a time series with persistent cycles fully, the single-observed state space representation is started. The variable modelled is decomposed into two components, each depending on one or more state variables: noise and cycle. The first moments of the model's variable, the associated Kalman filter, and the properties of the (Q)MLE estimator are derived. As an extension, the properties of a multivariate version of the previous model are studied where more than one variable common persistent cycles are shared and are affected distinctively by the short-term dynamics. The model is then empirically found to fit well the observed paleoclimatic data, providing a simple alternative to the methods developed in a recent study.
