Seasonal to Decadal Predictability of the ENSO, PDO, and AMO: Perspectives from the CMIP5 Decadal Hindcast Experiments

Abstract:

This study explores possibility for seamless seasonal to decadal prediction of surface air temperature (TAS), El Niño-Southern Oscillation (ENSO), Pacific Decadal Oscillation (PDO), and Atlantic Multi-decadal Oscillation (AMO) using decadal hindcast/prediction experiments of the Coupled Model Inter-comparison Project phase 5 (CMIP5). Long-lead prediction skill is evaluated by computing the anomaly correlation coefficient (ACC) and mean squared skill score (MSSS) of multi-model ensemble with respect to the observations and reanalysis data over the period of 1981-2012. The MSSS of global-mean TAS shows a prediction skill for a lead time of up to one year. This is much shorter than the one estimated from the ACC, indicating that the predicted global-mean TAS has a significant conditional bias. It is also found that prediction skill is highly sensitive to the season considered. The lowest prediction skill is found during the boreal winter when the long-term trend and inter-annual variability of TAS are not well reproduced especially over the Eurasian continent. This contrasts to the enhanced prediction skill over the oceans for a year. In most models, wintertime ENSO index is predictable at least one year in advance. Both PDO and AMO indices are also highly predictable on decadal time scales. Although their prediction skills vary in seasons, prediction lead time is still up to 6 years for PDO index and 9 years for AMO index in cold seasons. This result suggests that the two dominant low-frequency climate variabilities in the Northern Hemisphere are predictable on decadal time scales at least in their signs. With this finding, it is predicted by the CMIP5 decadal prediction experiments initialized in 2009 that the PDO and AMO will stay in their present phases for the next few years.