ROLE OF PREDICTORS IN STATISTICAL DOWNSCALING SURFACE TEMPERATURE FOR MALAPRABHA BASIN

Nagraj S. Patil, Soumya S. Bankapur

Abstract: A change in the statistical dissemination of weather patterns or arrangements which lasts for a prolonged time is called climate change. Researchers vigorously work to figure out past and future climate by using theoretical and observations models. Based on the physical sciences GCMs, are generally used in theoretical approaches to match past climate data, projection of future data, and to associate with the causes and consequences in climate change. As Global Climate Models are only accessible or available at coarse resolution the downscaling technique has been acknowledged as an essential component for the assessment of climate change impacts. In this paper, a surface temperature data is considered as a predictand and has been downscaled for the Malaprabha basin.The purpose of this study is to performance different methods of sensitivity analysis for selecting appropriate predictors influencing the predictand and to study the predictor-predictand relationship. Artificial Neural Network(ANN) methodology is utilized to downscale CanCM4 GCM surface temperature to local scale on monthly time series. The sensitivity analysis method PCA provides the most probable predictors strongly influencing the performance of the ANN model. Hence, this approach shows the importance of sensitivity analysis which helps in the performance of model for downscaling

Keywords: Climate Change; Statistical Downscaling; Artificial Neutral Network; Global Climate Model; CanCM4 GCM.

DOI: https://doi.org/10.15623/ijret.2016.0530014