Prediction of Tropical Climate on Intraseasonal Timescale using Phase Space Reconstruction

Although considerable success has been achieved in weather prediction on the order of about ten days lead time, the prediction of climate variability on intraseasonal and seasonal timescales is still in developmental stage. The optimism for climate prediction comes from the realization that climate variability, especially in the tropics, is influenced mainly by slowly varying components of the climate system. Applying multichannel singular spectrum analysis (MSSA) to daily values of climate variables, such as outgoing longwave radiation (OLR) and low-level winds, the tropical climate variability is found to consist of nonlinear oscillations on intraseasonal time scales and large-scale seasonally persisting patterns. The nonlinear oscillations are found to be manifestations of the South Asian monsoon’s active-break cycles and the well-known Madden Julian Oscillation over the Indian and Pacific Oceans. Exploiting the coherent and more regularly varying nature of these nonlinear MSSA modes, this study has constructed a dynamical model for the prediction of tropical climate on intraseasonal time scale. The prediction model is constructed from the time series of the MSSA modes using time-delay embedding technique for the reconstruction of phase space. The predictions are expressed in a probabilistic manner by providing ensemble forecasts.