Downscaling of Climatic Variable Using Multiple Linear Regression

Climate change has been emerging as one of the major challenges in the global environment. Changes in climate would lead to greater negative impacts on natural and human systems. Statistical downscaling is an approach to establish statistical relationship between large scale climate variables called predictors and local-scale hydrologic variables called Predictands. The present study has been taken up to predict the future temperature and precipitation scenarios adopting statistical downscaling approach such as Multiple Linear Regression (MLR) for Upper Bhima Basin, India using the input of coarse variable from GCM.In this study for calibrating MLR model, the inputs are NCEP variables and the outputs are IMD variables like maximum temperature (Tmax), minimum temperature (Tmin), precipitation (Prec). To predict the local climatic variable like maximum temperature (Tmax),minimum temperature (Tmin), precipitation (Prec), the inputs are from global climate model (GCM) given to the MLR. Probable predictors for the study selected from the National Center for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) reanalysis data set for the base period 1969-2005. The future temperature and precipitation are based upon result of how close the value of Coefficient of determination (R2) in MLR. In this study, it is coming as 0.924, which is very, close to one and considered as best.