DOWNSCALING OF CLIMATE VARIABLES USING SUPPORT VECTOR MACHINE (SVM)

Abstract

World's climate which had been observed over the past several decades is consistently associated with changes in a number of components of the hydrological cycle and hydrological systems such as: changing precipitation patterns, intensity and extremes; widespread melting of snow and ice; increasing atmospheric water vapour; increasing evaporation; and changes in soil moisture and runoff. The Intergovernmental Panel on Climate Change (IPCC), an authoritative international body predicts that global temperatures will rise by 1.1 to 6.4° C by the end of 21 century and this will head to negative influence to the nature in some cases. Thus it is, very important that scientist make prediction of the future climate as the first step of mitigation planning and adaptation. Global Climate Models (GCM) are considered to be the best tool to predict future climate with resolutions of hundreds of kilometers whilst the spatial resolution of regional climate models (RCM) which can give input data required for hydrological application as finer spatial resolution of the order of tens of kilometers. Further, many impact applications require the equivalent of point scale climate variations that are parameterized in coarse-scale-models. In view of the above, the output from a GCM has to be downscaled to obtain the information relevant to hydrologic studies. Statistical downscaling method is based on the view that the regional climate is conditioned by two factors: the large scale climate state and regional/local physiography. The large-scale output of GCM simulation is fed into this stastical model to estimate the corresponding local and regional climate characteristics. In this study, Multi Linear Regression (MLR) and Support Vector Machine for Regression (SVR) method approach were applied for statistical downscaling for precipitation and temperature variables in Roorkee area.