Prediction of Indian summer monsoon rainfall (ISMR) using canonical correlation analysis of global circulation model products

Abstract

The Canonical Correlation Analysis (CCA) method has been used in this study for improving General Circulation Model (GCM) predicted rainfall over India during the southwest monsoon season. Hindcast runs for 27 years (1982–2008) from six GCM outputs are used. This statistical technique relates the pattern of multivariate predictor field (model rainfall) to the pattern of predictand fields (observed rainfall). It is found that the CCA method improves the skill of three of the GCMs at the all-India level. A noticeable improvement is also observed in the composite prediction with CCA as compared to the simple mean of raw GCM products. The skill of the composite prediction after applying CCA is higher compared to the simple mean of raw model products in several homogeneous zones such as the hilly areas, west central area and over some parts of northwest India. The possible reason for the improvement in the skill of some of the GCMs may be the similarity between the loading patterns of model predictions and the observed rainfall.