Sensitivity of dynamical seasonal forecasts to ocean initial conditions

The impact of ocean data assimilation on dynamical El Nino-Southern Oscillation forecasting is studied by looking at forecasts started from ocean initial conditions produced with and without data assimilation. Sensitivity is further examined by comparing coupled forecasts started from initial conditions obtained using different ocean-forcing fields. A total of four ocean reanalyses for the period 1990-97, two analyses using subsurface data assimilation and two using only wind and sea-surface-temperature (SST) information are considered. Different wind-stress forcing produces significantly different analyses when subsurface data are not assimilated, but these differences are much smaller when subsurface temperatures are assimilated. The statistics of the forecast errors show that, with data assimilation, the forecasts of the NINO3 area (5S-5degreesN. 150degreesW-90degreesW) SST anomaly clearly beat persistence over all lead times (1-6 months), which is not the case when subsurface ocean observations are not used. The forecasts from the analyses using different wind forcing and no data assimilation differ considerably. The forecasts started from the analyses with data assimilation are generally similar irrespective of which wind product is used, suggesting that data assimilation is effective at reducing the impact of errors in the wind. For both wind forcings, data assimilation improves the forecast skill.