Downscaling atmospheric circulation data for monsoon rainfall forecasting in India

The potential of a surface-based approach for rainfall downscaling has been explored using a statistical method. The methodology has been applied in a highly seasonal semiarid monsoon-dominated region of India. An automatic circulation pattern classification method is used in which daily rainfall occurrence has been conditioned on CP type using a fuzzy-rule based process and the rainfall is forecast as a stochastic process coupled to the circulation pattern. The model was calibrated for a 10-year period, 1985-1994 and also validated. Model parameters obtained from 1985-1994 daily time-series were used to predict the long-term monsoon rainfall during 1961-1994 for the Jhabua station. The predicted mean monthly rainfall shows a good fit with observed rainfall within a mean variability range of 5 to 15%. The results thus obtained may be used for predicting the arrival of the monsoon, agricultural and crop planning, farming risk assessment, rainfall-runoff modelling, water balance studies and the development of climate change scenarios.