Seasonal Prediction of North Pacific SSTs and PDO in the NCEP CFS Hindcasts

Abstract

Abstract Seasonal prediction skill of North Pacific sea surface temperature anomalies (SSTAs) and the Pacific decadal oscillation (PDO) in the NCEP Climate Forecast System (CFS) retrospective forecasts is assessed. The SST forecasts exhibit significant skills over much of the North Pacific for two seasons in advance and outperform persistence over much of the North Pacific except near the Kuroshio–Oyashia Extension. Similar to the “spring barrier” feature in the El Niño–Southern Oscillation forecasts, the central North Pacific SST experiences a faster drop in prediction skill for forecasts initialized from November to February than those from May to August. Forecasts for the PDO displayed a constant phase shift from the observation with respect to lead time. The PDO skill has a clear seasonality with highest skill for forecasts initialized in boreal spring. The impact of ENSO on the PDO and North Pacific SST prediction was investigated. The analysis revealed that seasonal prediction skill in the central North Pacific mainly results from the skillful prediction of ENSO. As a result, the PDO is more skillful than persistence at all lead times during ENSO years. On the other hand, persistence is superior to the CFS forecast during ENSO-neutral conditions owing to errors in initial conditions and deficiencies in model physics. Examination of seasonal variance and predictability (signal-to-noise ratio) further articulates the influence of ENSO on the PDO skill. The results suggest that improvement of ENSO prediction as well as reduction in model biases in the western North Pacific will lead to improvements in the PDO and North Pacific SST predictions.