Predictability of Northeast Monsoon Rainfall over Southern India using Global Pressure Oscillations

Main focus of this hind cast study is to develop an algorithm for predicting northeast monsoon rainfall (NEMR) using global atmospheric pressure oscillations over south India. Analysis of 129–years datasets indicates that the Southern Oscillation Index (SOI) of April and May (AM) is inversely related (-0.56) with NEMR during 1959 to 2004; similarly the North Atlantic Oscillation (NAO) of January and February (JF) also exhibits similar relationship (-0.36) with predictant for the above study period. Simple linear equations are developed with above predictors to predict rainfall and they are tested for the succeeding six years, 1999-2004. To further confirm the consistency of above relationships, 20-year sliding window test is performed which is statistically significant. Secondly for extreme cases of rainfall events MannWhitney rank statistical test is performed and it repeats the same relationship. Above two predictors are statistically independent of each other and a multiple correlation coefficient (MCC) among two predictors and predictant is 0.66 for the common period 1959-1998, which is significant at the 0.1% level. Above MCC suggests a multiple regression equation for predicting NEMR, which is tested for succeeding six years (1999-2004). Leave-One-Out (L-O-O) cross validation test is applied for the estimated rainfall (significant at the 5% level), while root mean square error is 89.8 mm. Finally, observational evidence for variations of Hadley circulation, which is integral portion of northeast monsoon region is provided using NCEP/NCAR reanalysis data sets for the period 1959-2004 and the circulation features are very contrasting in extreme years of atmospheric oscillations.