Regional Climate Downscaling by Artificial Neural Network
Wachiraporn Permpoonsinsup and Dusadee Sukawat*
Department of Mathematics, Faculty of Science, King Mongkut’s University of
Technology Thonburi, Bangkok 10140, Thailand
Abstract
Global models for climate change show that in the future temperature of the world is raising. However, regions of the world may experience different changes in temperature. Moreover, the regional climate changes are stronger than the global change. The main idea in this paper is to interrelate regional climate parameters to large-scale variables using an interpolation technique. Interpolation is used to downscale the output from global to regional climate models. Network is also used to train temperature data based on neural network technique. The temperature data from the National Center for Environmental Prediction (NCEP) reanalysis data are trained for temperature at Bangkok. In training phase, the error are minimized and artificial neural networks (ANNs) are adjusted for the connect weights. Accordingly, output data from the regional model are compared with observation data of the Thai Meteorological Department.
Keywords: Downscaling, Artificial Neural Network.
KMITL Sci. Tech. J. Vol. 12 No. 2 Jul. - Dec. 2012
