Articles | Volume 13, issue 9
https://doi.org/10.1007/s00585-995-0995-x
https://doi.org/10.1007/s00585-995-0995-x
30 Sep 1995
30 Sep 1995

On forecasting abnormal climatic events in the tropical Atlantic Ocean

J. Servain and S. Arnault

Abstract. Modelling and observational evidence indicate that interannual variabilities of dynamic height and sea surface temperature (SST) in the eastern part of the tropical Atlantic Ocean (Gulf of Guinea) are largely induced by preceding fluctuations in wind stress, mainly in the western equatorial basin. A wind-driven linear ocean model is used here to test the possibility of forecasting the abnormal dynamic heights. A control run of the model, forced by 1964–1993 wind stress monthly means, is first conducted. Yearly test runs (1964–1994) are subsequently performed from January to August by forcing the model with observed winds from January to May, and then by forcing with the May wind assumed to persist from June to August. During the last three decades the largest deviations of dynamic height simulated by the control run in the Gulf of Guinea in boreal summer would have been correctly forecast from wind data related only to conditions in May of each year. However, for weak climatic anomalies, the model may forecast overestimated values. For the most part (about 20 times during the last 30 years), the sign of the observed SST anomaly in the centre of the Gulf of Guinea during the boreal summer is identical to the sign of simulated anomalies of dynamic height deduced from both control and test runs. Along the eastern equatorial waveguide, the sea level forecasting skill slowly decreases from the first 2 weeks of June until the second 2 weeks of August, but remains high on both sides of the equator throughout boreal summer, as is expected from the adjustment in a linear ocean model. It is established that throughout the year in the Gulf of Guinea the accuracy of the 1-month forecast dynamic height anomaly provided by the simple linear method is greater than that of the 1-month forecast assuming persistence.