Prediction of Indian Monsoon Rainfall: Further Exploration

This study expands recent research into the predictability of Indian monsoon rainfall anomalies. In addition to the April latitude position of the 500 mb ridge over India, and Darwin pressure tendency, the May surface resultant wind speed in a strategic area of the jet axis over the western equatorial Indian Ocean (0–10°S, 45°–50°E) is used as predictor. Regression models developed on 20-yr portions of the 1939–83 record are employed to predict the summer monsoon rainfall anomalies of the 25 yr 1959–83. Correlation, root-mean-square error, bias, and absolute error are presented as measures of forecast performance on the “independent” dataset. It is found that the model constraint optimal for predictive purposes is to be ascertained empirically. “Updating” is not necessarily superior to the use of a fixed regression base period. Relationships between preseason indicators and monsoon rainfall were, strongest in the early and late portions of the 1939–83 record, and weakest in the 1950s and 1960s. Antecedent departures in the late-scale circulation setting allowed prediction of more than 60% of the interannual variance of Indian monsoon rainfall for the period 1959–83, and nearly 80% for 1969–83.