Long-Range Predictability in the Tropics. Part I: Monthly Averages

The sensitivity to initial and boundary conditions of monthly mean tropical long-range forecasts (1–14 weeks) during Northern Hemisphere winter is studied with a numerical model. Five predictability experiments with different combinations of initial conditions and prescribed ocean boundary conditions are conducted in order to investigate the temporal and spatial characteristics of the perfect model forecast skill. It is shown that initial conditions dominate a tropical forecast during the first 3 weeks and that they influence a forecast for at least 8 weeks. The initial condition effect is strongest over the Eastern Hemisphere and during years when the El Niño–Southern Oscillation (ENSO) phenomenon is weak. The relatively long sensitivity to initial conditions is related to a complex combination of dynamic and thermodynamic effects, and to positive internal feedbacks of large-scale convective anomalies. At lead times of more than 3 weeks, boundary forcing is the main contributor to tropical predictability. This effect is particularly strong over the Western Hemisphere and during ENSO. Using persisted instead of observed sea surface temperatures leads to useful forecast results only over the Western Hemisphere and during ENSO.