EVALUATING THE USE OF DIFFERENT DISTANCE MEASURES IN STATISTICAL DOWNSCALING OF CLIMATE PARAMETERS USING THE K-NN METHOD

摘要

The k-nearest neighbor (k-nn) algorithm is one of the simplest and most utilized tools for statistical downscaling of large-scale General Circulation Model (GCM) outputs. The accuracy of this method relies on the selected distance measure for calculating the similarity between future and past events as well as the considered number of neighbors. In this study, seven distance metrics were used in conjunction with the Hadley Centre climate data to downscale monthly maximum and minimum temperature as well as average precipitation for the River Severn basin in the UK. The analysis of the results showed that although the predictions of average minimum and maximum temperature are insensitive to the number of neighbors and selected distance measure, the average monthly precipitation may vary by up to 40% depending on the choice of distance measure, but is less effected by the number of considered neighbors.