GCM hindcasts of SST forced climate variability over agriculturally intensive regions

The NASA/Goddard Institute for Space Studies (GISS) climate model is forced with globally observed sea-surface temperatures (SST) in five simulations, 1969-1991, with individual runs beginning from altered initial atmospheric conditions. The interannual variability of modeled anomalies of the Southern Oscillation Index, mid-tropospheric temperatures, 850 mb zonal winds and Outgoing Longwave Radiation over the tropical Pacific Ocean, which has the largest SST anomaly forcing, are strongly correlated with observed trends which reflect ENSO cycles. The model's rainfall variability over three agriculturally intensive regions, two tropical and one mid-latitude, is investigated in order to evaluate the potential usefulness of GCM predictions for agricultural planning. The correct sign of Zimbabwe seasonal precipitation anomalies was hindcast within a useful range of consensus only for select seasons corresponding to extreme ENSO events for which anomalous circulation patterns were rather realistically simulated. The correlation between hindcasts of Nordeste monthly precipitation and observations increases with time smoothing, reaching 0.64 for 5-month running means. Consensus between individual runs is directly proportional to the absolute value of Nino3 SST so that during El Nino and La Nina years most simulations agree on the sign of predicted Nordeste rainfall anomalies. We show that during selected seasons the upper tropospheric divergent circulation and near surface meridional displacements of the ITCZ are realistically represented by the ensemble mean of the simulations. This realistic simulation of both the synoptic mechanisms and the resulting precipitation changes increases confidence in the GCM's potential for seasonal climate prediction.