Regional boreal summer intraseasonal oscillation over Indian Ocean and Western Pacific: comparison and predictability study

Abstract

The boreal summer intraseasonal oscillation (BSISO) has two major activity centers, the northern Indian Ocean and tropical Western North Pacific, which dominate the monsoon intraseasonal variability over South Asia and East Asia, respectively. The spatial–temporal structures of BSISO over the Indian Ocean (10°S–30°N, 60°–105°E) (IOISO) and Western Pacific (10°S–30°N, 105°–150°E) (WPISO) are examined by corresponding the leading modes of daily OLR and 850-hPa zonal wind (U850). The IOISO features a northeastward propagation with a 30–45 days energy peak and the first principal component (PC1) has maximum variance in May, while the WPISO propagates northward with a broad spectral peak on 10–60 days and the PC1 has maximum variance in August. Because of the large regional differences, two regional indices, the IOISO index and WPISO index, are defined by their corresponding first two leading PCs. The combined IOISO–WPISO index captures about 30 % (10 %) of U850 (OLR) daily variance over the entire IO–WP region (10°S–30°N, 60°–150°E), which doubles that captured by the Madden–Julian Oscillation (MJO) index (Wheeler and Hendon 2004) and is 50 % higher than that captured by the BSISO index (Lee et al. 2013). The combined index also shows superior performance in representing biweekly and pentad-mean variations in the Asian-Pacific summer monsoon region (north of 10°N). The predictability/prediction skill and simulated principal modes of two regional BSISO indices are explored by using data derived from the Intraseasonal Variability Hindcast Experiment project. The major regional modes are reasonably well captured, but the forecasted fractional variances of the leading modes and variability center’s locations exhibit significant deficiencies. The multi-model mean estimate of the predictability is 40–45 days for the IOISO index, whereas 33–37 days for the WPISO index. The less predictable WPISO is likely due to the existence of its significant biweekly component. The multi-model mean prediction skill is significantly higher with large initial amplitude (~20 days for two indices) than that with small initial amplitude (~11 days), suggesting that the prediction for development of BSISO is much more difficult than the prediction for mature BSISO disturbances’ propagation.