Downscaling Climate Variables to River Basin Scale in India for IPCC SRES Scenarios using Support Vector Machine

摘要

Realistic assessments of the local impacts of natural climate variability and projected climate change in the future are important to make independent judgments about actions required to mitigate and manage natural disasters; manage the natural environment and their water resources in a sustainable manner. A river basin which integrates some of the important systems like ecological and socioeconomic systems can be ideal to study the impact of climate change on the water cycle at a local scale. General circulation models (GCMs) are among the most advanced tools to simulate climatic conditions on earth hundreds of years into the future. The GCMs are generally run at coarser scale to cover the whole globe and as a result they are inherently unable to represent local scale features. Consequently, there is a continuing need for new and improved techniques for obtaining effective projections of hydrological and meteorological variables at the river basin scale. Downscaling is one such technique, which is gaining popularity in estimating these variables at regional and local scales by translating information simulated by GCMs at global scale. This paper emphasizes the importance of downscaling to a river basin scale and presents a methodology to downscale monthly climate output from GCM to this scale using Support Vector Machine (SVM). Implementation of the methodology is demonstrated by downscaling maximum temperature to Malaprabha reservoir catchment in India (which is considered to be a climatically sensitive region), using simulations from the third generation Canadian Global Climate Model (CGCM3) for IPCC SRES scenarios A1B, A2, B1 and COMMIT.