Spatial Structure, Forecast Errors, and Predictability of the South Asian Monsoon in CFS Monthly Retrospective Forecasts

Abstract

Abstract The spatial structure of the boreal summer South Asian monsoon in the ensemble mean of monthly retrospective forecasts by the Climate Forecast System of the National Centers for Environmental Prediction is examined. The forecast errors and predictability of the model are assessed. Systematic errors in the forecasts consist of deficient rainfall over India, excess rainfall over the Arabian Sea, and a dipole structure over the equatorial Indian Ocean. On interannual time scale during 1981–2003, two different characteristics of the monsoon are recognized—both in observation and forecasts. One feature seems to indicate that the monsoon is regionally controlled, while the other shows a strong relation with El Niño–Southern Oscillation (ENSO). The spatial structure of the regional monsoon can be characterized by the dominant rainfall between the latitudes of 15°N and 5°S in the western Indian Ocean. The maximum precipitation anomalies in the northern Arabian Sea are associated with the cyclonic circulation, while the precipitation anomalies in the equatorial western Indian Ocean accompany the easterlies over the equatorial Indian Ocean. In the ENSO-related monsoon, strong positive precipitation anomalies prevail from the equatorial eastern Indian Ocean to the western Pacific, inducing westerlies over the equatorial Indian Ocean. The spatial structure of the forecast error shows that the model is inclined to predict the ENSO-related feature more accurately than the regional feature. The predictability is found to be lower over certain areas in the northern and equatorial eastern Indian Ocean. The predictability errors in the northern Indian Ocean diminish for longer forecast leads, presumably because the impact of different initial conditions dissipates with time. On the other hand, predictability errors over the equatorial eastern Indian Ocean grow as the forecast lead increases.