Role of Predictors in Statistical Downscaling of Surface Temperature for Malaprabha Basin
The earth’s surface temperature is an important parameter for several atmospheric phenomena and process. It is also considered as an indicator of environmental degradation and climate change. This study aims to first evaluate different methods of sensitivity analysis for selecting potential climate predictors influencing the predictand (surface temperature) to establish predictor-predictand relationship. Another important objective of the study is to explore the Artificial Neural Network (ANN) method to downscale surface temperature to local scale using CanCM4-Global Circular Model to local scale on monthly time series.
The work focused on exploring various sensitivity analysis techniques, drawing comparisons between them and determining the evaluation criteria to assess the performance of a downscaling model, to understand the factor affecting the climate and causing climate variation of the Malaprabha basin for the future scenarios. Pearson`s correlation coefficient, Spearman`s rank correlation coefficient, Principal component analysis (PCA) and One-way analysis of variance (ANOVA) methods have been compared and contrasted for predictor selection. PCA screened the most potential predictors and these sensitive predictors are used as an input to ANN model trained for the time period of 1971-1995 and validated using the observed Indian Meteorological Department (IMD) data for the period of 1996-2005. Trained ANN model downscales the future projections of surface temperature for the period 2006-2035.