Multiple Regression Model for Long Range Forecasting of South-West Monsoon Rainfall for Pune, Maharashtra S.S. Chinchorkar 1 , V.B. Vaidya 2 and Vyas Pandey 2 * Anand Agricultural University, Anand 388 110, India 1 Polytechnic in Agricultural Engineering, Anand Agricultural University, Dahod xxx xxx, India 2 Department of Agricultural Meteorology, BACA, Anand Agricultural University, Anand xxx xxx, India Received: July 2011 Abstract: The large spatial variability in monsoon rainfall over India demands for regional model to predict the seasonal rainfall. Hence, a local model was developed for predicting seasonal (June-September) rainfall of Pune using multiple regression technique. The monthly weather data of 36 years (1970-2005) was used for model development and data for next five years (2006-2010) were used for validation purpose. The model explained 81% variability in seasonal rainfall with model error of 2.52%. During the validation period (2006-2010), the performance of model was quite satisfactory with model error of -1.94% only. This model was used to predict the rainfall for 2011 season. Results suggested that the rainfall during 2011 would be higher (51.2%) than the normal rainfall in Pune. Key words: Multiple regression, rainfall forecasting, rainfall analysis, statistical model. A major portion of annual rainfall over India is received during the south-west monsoon season (June-September). There are known vagaries of the monsoon as regards the onset as quantum of monsoon rains and its distribution in different parts of the country. Although the regional rainfall has large year to year fluctuations, the south-west monsoon rainfall over the country as a whole varies by about 10% of the mean rainfall (Rajeevan et al., 2004). However, this small fluctuation in the seasonal rainfall can have devastating impacts on India’s economy. Therefore, prediction of south-west monsoon rainfall deserves high priority in India. The India Meteorological Department (IMD) has been issuing long range forecasts of the south-west monsoon rainfall since 1886. During the period 1988-2002, IMD’s operational forecasts were based on the 16 parameter power regression and parametric models (Gowariker et al., 1989 and 1991). Two new models (8 and 10 parameter models) were introduced in 2003, with which the seasonal Indian Summer Monsoon Rainfall for the country as whole are issued in mid April and an update or second stage forecast is issued by the end of June for different homogenous regions of India. Although at many a time the SW-monsoon rainfall of the country as a whole had been normal, but there have been quite large variation in regional rainfall distribution e.g., in 2006 country received 878.6 mm (normal rainfall), but Gujarat received 151% of the normal rainfall. Hence, regional or location-specific models are needed to be developed. Varshneya et al. (2010) developed models for different regions of Gujarat State using regression techniques in which the best models were selected based on higher R 2 and lower model error. These models explained 74 to 93% variability in seasonal rainfall with models error ranging between -2.5 to 5.1%. Sable et al. (2007) developed models for various locations of Maharashtra State, in which the group matching technique was used. From the matching group weighted mean was calculated to predict the seasonal rainfall of different locations of Maharashtra State. Pune is situated on leeward side of Western Ghats located in a basin surrounded by uplands and hills and situated near the periphery of the Deccan Plateau. Local topography plays a major role in the rainfall receipt at Pune. There is no model available for predicting rainfall of Pune. Hence an attempt has been made to predict the south-west monsoon rainfall for Pune station. Materials and Methods Weekly meteorological data of Pune (18º 31’ N and 73º 52’ E) for forty one years (1970- Annals of Arid Zone 51(1): 1-3, 2012 *E-mail: pandey04@yahoo.com