Regionalization of Northwest Iran Based on Daily Rainfalls and Rain’s Time Intervals Using PCA, Ward and K-mean Methods

Document Type : Research Article

Authors

• B. Fallahi
• A. Fakheri Fard
• yaghoub dinpazhoh
• S. Darbandi

10.22067/jsw.v0i0.15302

Abstract

Having a correct view of the effective factors on climatic changes by explanation of a considerable part of the total variance in data with limited number of principal components the analytical methods of decreasing data dimensions, such as PCA are important tools in water resources planning. In this study PCA method as a projection tool for projecting the information space on the limited and specific axes, ward’s method as a hierarchical clustering and k-mean as partitioning clustering method has been applied in this research. Using this methods and application of daily precipitation data of 60 meteorological stations during a 35 years period (1970-2004), 4 types of delineated regions were come out on the basis of daily precipitations, distance-quantity index, time intervals and rainy days series. S statistic test algorithm was used for homogeneity test of the regions. Results showed the nature of the PCA method is such that projects the data space on the main axes and shows the real space. But in the hierarchical methods, clusters do not describe the real structure. Therefore we do expect that the resulting clusters of PCA would be more realistic than that of methods. But hierarchical methods have the advantage of containing the wider clustering information on the basis of homogeneity than the others.

Keywords

• regionalization
• Principal Components Analysis
• Cluster analysis
• Ward
• K-mean

---