Regionalization of precipitation in Iran using principal components and cluster analysis

Principal components and cluster analysis were used for grouping the monthly precipitation over Iran into somehomogeneous regions. Such grouping is considered as an essential and primary step for pre-process in downscaling of the general circulation models outputs. The monthly precipitation data of 42 stations for the period of 1967-2003 (37 years) forming a 42×444 matrix was used as input file of the principal components analysis. Using PCA, this matrix was reduced into a 42×33 matrix that accounted for about 96% of total variance in the observational time series. The reduced matrix was used as the input file of the Cluster analysis. The results regionalized the country into six different regions.