Potential Predictability of Sea Surface Temperature in a Coupled Ocean--Atmosphere GCM

Using the Flexible Global Ocean--Atmosphere--Land System model (FGOALS) version g1.11, a group of seasonal hindcasting experiments were carried out. In order to investigate the potential predictability of sea surface temperature (SST), singular value decomposition (SVD) analyses were applied to extract dominant coupled modes between observed and predicated SST from the hindcasting experiments in this study. The fields discussed are sea surface temperature anomalies over the tropical Pacific basin (20oS--20oN, 120oE--80oW), respectively starting in four seasons from 1982 to 2005. On the basis of SVD analysis, the simulated pattern was replaced with the corresponding observed pattern to reconstruct SST anomaly fields to improve the ability of the simulation. The predictive skill, anomaly correlation coefficients (ACC), after systematic error correction using the first five modes was regarded as potential predictability. Results showed that: 1) the statistical postprocessing approach was effective for systematic error correction; 2) model error sources mainly arose from mode 2 extracted from the SVD analysis---that is, during the transition phase of ENSO, the model encountered the spring predictability barrier; and 3) potential predictability (upper limits of predictability) could be high over most of the tropical Pacific basin, including the tropical western Pacific and an extra 10-degrees region of the mid and eastern Pacific.