The most predictable patterns and prediction skills of subseasonal prediction of rainfall over the Indo-Pacific regions by the NCEP Climate Forecast System

Abstract

The most predictable patterns and prediction skills of subseasonal rainfall prediction for the Indo-Pacific regions are investigated using the daily hindcast product of the NCEP Climate Forecast System version 2. A maximum signal-to-noise empirical orthogonal function analysis indicates that on 30–60-day time scale the most predictable patterns exhibit a zonal dipole over the tropical Indian Ocean and the surrounding oceanic areas of the Maritime Continent in boreal winter and spring, associated with the eastward propagation of the Madden–Julian Oscillation. Differently, the most predictable patterns in boreal summer and autumn present a meridional dipole between a tropical belt (from the Arabian Sea, the Bay of Bengal, and the South China Sea to the western Pacific) and an equatorial belt (from the equatorial Indian Ocean to the equatorial Maritime Continent), related to the northward propagation of intraseasonal oscillation. Both the 30–60-day and 10–20-day patterns can be predicted at lead times around one half of their respective longest life cycles, with worst skills in summer. For the winter–spring 30–60-day patterns, the atmospheric drive on oceans plays an important role, and the corresponding winds at both upper and lower levels are consistent with the Gill response, suggesting the importance of the evaporation-wind feedback mechanism. During summer–autumn, cold sea surface temperature anomalies cause a change in land-sea thermal contrast, which is important for monsoon variations.