POAMA 2 Seasonal SST Index Skill

The plots on this page compare the skill of the POAMA-2 forecasts with that from other international models for SST index regions in the tropical Pacific and Indian Oceans

Temporal correlation coefficients of Nino3 and Nino4 indices

Fig 1a: Temporal correlation coefficients between the monthly SST anomalies in the Niño3 region (averaged over 210-270oE, 5oS-5oN) of the seasonal forecast systems and the corresponding observations for all initial months during the period 1982.1 - 2006.12. The 3-month mean along the forecast lead months has been applied on all indices data prior to the calculation. X axis is the lead time from 0-6 months. The ensemble hindcasts in this figure were selected from the ECMWF seasonal forecast System3 (ECMWF-S3), the Frontier Research Centre model (SINTEX-F), two versions of the NCEP Climate Forecast System (CFSV1 and CFSV2), the POAMA v1.5b (P15b) and the POAMA-2 multi-model ensemble (PMTMD), respectively.

Fig 1b: As in Fig 1a, except for the Niño4 region (averaged over 190-240oE, 5oS-5oN).

Temporal correlation coefficients of WIO and EIO indices

Fig 2a: As in Fig 1a, except for the WIO region (averaged over 50-70oE, 10oS-10oN).

Fig 2b: As in Fig 1a, except for the EIO region (averaged over 90-110oE, 10oS-0oN).

Temporal correlation coefficients of IOD indices

Fig 3: As in Fig 1a, except for the IOD indices (the differences between the WIO and EIO) of the seasonal prediction systems and the statistic model (dot-dashed purple curve; following the Dommenget and Jansen, 2009) for the verification months September-October-November (SON) during the period 1982 - 2006. The 1-2-1 filter has been applied.