Research papersLong-term spatio-temporal drought variability in Turkey

•An innovative method is suggested for principle component number determination.•Low variance explained score cause instability on spatio-temporal drought pattern.•SO and drought phases are not completely coherent in four drought homogeneous regions.

The spatio-temporal variability of drought is presented by evaluating homogeneously distributed 250 station records from 1931 to 2010 for 80 years' duration in Turkey. The drought analysis is implemented using Standardized Precipitation Index (SPI) in terms of SPI-1, SPI-3, SPI-6, SPI-6AS (SPI-6 April to September) and SPI-12. The principle component analysis (PCA) is applied to SPI time series to identify spatial and temporal drought patterns. SPI time series are classified into two groups (1st group: SPI-1, SPI-3, SPI-6AS; and 2nd group: SPI-6 and SPI-12) according to the similarity in spatial drought patterns. SPI-3 and SPI-12 are selected as representative members of each group for spatio-temporal analysis. A relationship among correlation area (An), correlation coefficient (CC), principle component numbers (Fn) and total variances explained (Vexp) are investigated for identifying four well-defined drought vulnerable homogeneous regions over Turkey mainland. Mean percentages of extreme, severe, and moderate drought areas are calculated as 3.13% (2.81%), 3.75% (4.06%) and 7.19% (7.50%) for SPI-3 (SPI-12) based on 80 years in all drought vulnerable regions. Spectral characteristics of drought are also investigated based on fast Fourier transform (FFT) method. It is observed that while southeastern and western parts of Turkey are more stable due to the highly-correlated variances of spatial patterns; central parts and few pockets in northern areas of Turkey are less stable regions because of the low-correlated variance scores (below 10%). Furthermore, the impact of extreme phases of the ENSO (El Nino/La Nina) on droughts in four drought regions over Turkey is discussed.