NASA GEOS-S2S (Version 2) model for sub-seasonal to seasonal predictions:The major teleconnections, troical cyclone activity, and ENSO

Ability of the new NASA GEOS-S2S (Version 2) seasonal prediction model, which is a member of the North American Multi-Model Ensemble (NMME) system, is validated by investigating its sub-seasonal to seasonal predictive skill for the major large-scale teleconnections, tropical cyclone (TC) activity, and ENSO. The major teleconnections include the North Atlantic Oscillation (NAO), the Arctic Oscillation (AO), and the Pacific North American (PNA). Predictive skill of TC activity is assessed by TC activity indices that consider genesis frequency and intensity of the TCs. ENSO predictive skill is examined by long-lead prediction of the Niño3.4 SST.

Comparison in the teleconnection patterns between observation and 36-year hindcasts indicates that the model successfully captures the spatial structure of the major teleconnections over the Northern Hemisphere. Geographical locations of the positive/negative anomalies seen in the observation are quite realistically reproduced by the model, though some underestimation of the observed magnitude of the anomalies is found from NAO simulation. Interannual variation of the January/February averaged teleconnection indices (initialized in late December) demonstrates the encouraging predictive skill that exceeds correlations of 0.5 for all three teleconnections.

Seasonal predictability of the TC activity for the 1982-2017 TC seasons shows that the Genesis Potential Index (GPI) averaged over June through September (initialized in late May) predicted at the beginning of the TC season has correlations greater than 0.5 with observation over the North Atlantic and 0.8 over the Western Pacific basin.

Long lead predictability of the Niño 3.4 SST indicates that the DJF SST (ENSO maturity period) predicted in spring (MAM, 9 month lead) has correlation greater than 0.8 with observation. However, three categorical predictability (El Niño, La Niña, and Neutral) indicates that the predicted categories for the 1982-2017 DJFs are 75% consistent with observations at 6 month lead.