Prediction of summer monsoon rainfall over India and its homogeneous regions

Abstract

ABSTRACTThe coherent regions for various fields such as sea level pressure, temperature, geopotential height and zonal wind anomalies at the surface, 850, 500 and 200 hPa levels in pre‐monsoon months (January through May) and seasons (winter, spring) have been identified by applying the shared nearest neighbour algorithm. The fields over the corresponding cluster regions could be possible predictors for Indian summer monsoon rainfall as well as the rainfall over various homogeneous regions of India. The time series have been constructed by averaging the parameters over the respective clusters. The relationship between these time series and the summer monsoon rainfall over India and its well‐defined homogeneous regions over India, (northwest India, central northeast India, northeast India, west central India and peninsular India), has been examined during the positive and negative phases of effective strength index tendency using the simple technique of multiple regression. Along with the linear relationship, the non‐linear relationships between the cluster parameters and the seasonal rainfall have also been considered. Independent cluster parameters have been selected by cross‐validation procedure and the performance of each predictive model is tested. The extreme yearly rainfall departures over India are qualitatively well predicted by the model. Also, the unprecedented droughts over India in 2002 and 2009, where all earlier models have failed to forecast, are well predicted by the present model. The performances of models for summer monsoon rainfall prediction over homogeneous regions of India are convincing.