Dynamical Downscaling for Seasonal Prediction of Indian Summer Monsoon Rainfall Using a Regional Climate Modeling System

Abstract

The importance of the Indian summer monsoon is quite well known by the people inhabiting the Indian sub-continent. Maximum rainfall is received during the JJAS months over India which is extremely important for the agricultural activities, hydro-power, industries, mining etc. Seasonal scale prediction of the monsoons and its rainfall puts huge expectation on the operational forecasting centers and demand for a skillful prediction of the monsoon rainfall is ever increasing by the common people as well as planning communities. Though the prediction skill of the monsoon rainfall has substantially increased over the past few decades, it is not quite reliable. This thesis aims at improving the skill of monsoon rainfall forecast by following the method of dynamical downscaling of the global model s forecast using a regional climate modeling system. The CFSv2 seasonal forecasts are used as the input conditions for two regional climate models namely RegCM and WRF. Following the objective of the thesis, seven state-of-the-art- GCMs are evaluated for their skill of predicting the monsoon rainfall over the hindcast period of 27 years from 1982-2008. The models ECMWF, CFSv2 possess good skill in simulating the mean seasonal rainfall. Intra-seasonal variability of monsoon rainfall is one of the most important characteristics of the monsoon rainfall which is not quite well captured by the ECMWF model. Comprehensive statistical analysis shows that the CFSv2 at IITM has better skill in reproducing the mean rainfall pattern as well as intra-seasonal variability. From the findings of this chapter, the CFSv2 prepared at IITM at T382 spectral resolution is used for driving the regional climate models (RMCs). Before performing the dynamical downscaling by the RCMs, sensitivity studies are carried out for the parametrization schemes in the RegCM and WRF for customizing the model for the Indian summer monsoon domain. The RCMs show significant discrepancies when the parametrization schemes are chosen differently.