Abstract
Two sets of experiments were performed. The first set, denoted SSTA, consisted of 90-day forecasts with sea surface temperature anomalies updated with observed values daily during the entire integration. For the summers 1987 and 1988, three SSTA experiments were made using three different initial conditions centered on 1 June of that year, separated by 1 day. The second set of experiments, denoted CSST, used the same initial conditions as the first set, but the integrations were performed using climatological SSTs. All numerical experiments were done using the NMC T80 spectral model of 1990, which is the same model used in making operational medium-range forecasts.
Simulated seasonal ensemble-mean rainfall was compared with satellite estimates of precipitation and observed station rainfall data. Overall agreement between them is good. Two centers of maximum rainfall, over the Arabian Sea and the Bay of Bengal, are captured by the model, but it fails to capture the movement of the rainfall associated with the Indian monsoon. The model is able to simulate the interannual variability of rain in India and over the Sahel, although the simulated convection in the central Pacific associated with the 1987 warm episode is not realistic.
When the model is able to simulate the convection associated with the SSTAs, then the updated SSTs have a large positive impact on tropical impact seasonal forecasts. The impact on the extratropical forecasts is, in general, positive but small.
