The Preferred Structure of the Interannual Indian Monsoon Variability

Abstract

The leading empirical orthogonal function (EOF) of the June-Sept. mean, rotational horizontal wind at 850 hPa and 200 hPa (over the region 12.5°S–42.5°N, 50°E–100°E) from 56 years (1948–2003) of reanalysis (from the National Centers for Environmental Prediction) shows strong anti-cyclonic circulation at upper levels, strong Indian Ocean cross-equatorial flow and on-shore flow over western India at lower levels . The associated principal component (PC) is correlated at the 0.75 level with the seasonal mean observed Indian Monsoon rainfall (IMR). Composite differences of vertically integrated divergence (surface to 800 hPa) and vorticity (surface to 500 hPa) between ``strong'' years (PC-1 exceeds one standard deviation σ) and ``weak'' years (PC-1 less than − σ) suggest increased rising motion and storminess over the Bay of Bengal and central India. Composite difference maps of station rainfall from the India Meteorological Department (IMD) between strong years and normal years (weak years and normal years) are statistically significant over central India, with strong (weak) years associated with increased (decreased) precipitation. In both cases the maps of rainfall anomalies are of one sign throughout India. The correlation of PC-1 with global seasonal mean SST is strong and negative over the eastern equatorial Pacific, but positive in a surrounding horse-shoe like region. Significant negative correlation occurs in the northwestern Indian Ocean. The lag/lead correlation between the NINO3 SST index and PC-1 is similar to but stronger than the NINO3/IMR correlation. Modest (but significant) negative correlation is seen when NINO3 leads PC-1 (or IMR) by one-two months. Strong negative correlation is seen when PC-1 (or IMR) leads NINO3. The projections of running five-day means of horizontal rotational winds at 850 and 200 hPa onto EOF-1 (after removing the seasonal mean for each year) were pooled for strong, normal and weak years. The strong and normal year probability distribution functions (pdfs) are nearly indistinguishable, but the weak year pdf has more weight for moderate negative values and in both extreme tails and shows some hint of bi-modality.