Akademik Veri Yönetim Sistemi
Theoretical and Applied Climatology, cilt.106, sa.3-4, ss.499-510, 2011 (SCI-Expanded)
Rainfall over Turkey portrays highly variable character both spatially and temporally. The aim of this study is to redefine main rainfall clusters of Turkey by using k-means methodology and investigate spatial shifts in the redefined rainfall clusters in subsequent periods with respect to North Atlantic Oscillation (NAO) patterns. Initially, monthly rainfall totals are subjected to k-means clustering by taking into consideration 148 stations covering the 1977-2006 period. Considering the maximum silhouette value and lowest negative silhouette number, six rainfall clusters are determined as optimum classifications for this climate period. The results indicate that Aegean-Marmara and Eastern Anatolia-Central Anatolia geographic regions are characterized as single rainfall cluster contrary to the conventional geographical regions. The Mediterranean region is characterized with two separate sub-regions indicating highly variable rainfall distribution characters over the region. The study further adapts a similar classification for 10-year sub-periods to determine spatial shifts of the redefined rainfall clusters for the last 30 years. From one decade to another, temporally drier and wetter clusters are observed with underlying shifting causes in relation to NAO patterns. Parallel to other studies in the literature, NAO is found to be partially useful in explaining the temporally dry trends while less useful in justifying wet periods. On the other hand, coefficient of variation (COV) is introduced in order to explain the temporal shifts in the clusters. Strong relations are obtained between the regions with the higher COV numbers and highest cluster shifts, while smaller COV numbers are associated with the most stable clusters. © 2011 Springer-Verlag.