Attempts have been made by the weather scientists to forecast Indian monsoon since last century. H.F. Blanford, the first director general of IMD made tentative forecast about the monsoon from 1882-1885 based on the amount of snowfall over the Himalayas.
According to him excessive snowfall over the Himalayas which lowers the temperature difference between the Indian landmass and the Indian Ocean is generally characterised with poor monsoon. Sir John Elliot (1888) found that climatic factors in neighbouring regions, in particular Australian weather parameters, have their impact on raisin India. In 1904 Gilbert Walker revolutionised the forecasting method and included Southern Oscillation (see previous account) in forecasting.
He developed a multiple regression model using 4 parameters-accumulation of Himalayan snow at the end of May, pressure in South America during spring, pressure in Mauritius in May and rainfall in Zanzibar in April and May. Later he also added rainfall in Sri Lanka in May and SOI in spring. In 1979 V. Thapliyal introduced the ‘dynamic stochastic transfer model’ which improved the accuracy to 75 per cent.
Recently Vasant Gowariker, Secretary, Deptt of Science and Technology, New Delhi suggested that the process of forecasting could be split into two stages-a simple model easily understood by administrators to help in decision-making and the detailed forecast following at the end of May.
He proposed a ‘power regression’ model using 16 parameters. These parameters include : El Nino (same year), El Nino )previous year), north India temperature (March), central India temperature (May), east coast of India temperature (March), northern hemisphere temperature (January and February), 500 ridge (April), 50 rte (ridge trough extent (Jan. & Feb.), 10 (30 km) westerly wind (Jan.), northern hemisphere pressure, Southern Oscillation Index, Darwin Pressure (spring), Argentina pressure (April), Indian Ocean equatorial pressure (January-May), Himalayan snow cover (January-March), and Eurasian snow cover (previous December). Of the 16 parameters, 6 are temperature based, 3 wind based, 5 pressure based and 2 snow fall based.
Since the relationship of monsoon rainfall with individual predictors exhibits a non-linear relationship the improvement was made in this model (1989) by determining a curvilinear relationship by fitting the equation of different degrees. The monsoon rainfall fore cast provided in 1989 by using this new model came out to be correct. In 1992 F1 Nino and SOI weir unfavorable yet the prediction was for 92 per ceil normal rainfall and the actual was 93 percent (Roo Karnani, 1993, pp. 114-115).