Spatial Analysis of Monthly and Annual Precipitation Trends in Turkey

Global climate change could have important effects on various environmental variables in many countries around the world. Changes in precipitation regime directly affect water resources management. So that, it is important to analyze the changes in the spatial and temporal rainfall pattern in order to improve water resources management policies. For this reason, non-parametric Mann-Kendall rank correlation test is used in order to examine the existence of trends in annual and monthly rainfall distribution. To understand the regional differences of precipitation in Turkey, the detected trends are spatially interpolated using geostatistical techniques in a GIS environment. The main objective of this paper is to evaluate three interpolation methods, concerning their suitability for spatial prediction of temporal trends of Turkey's monthly and annual rainfall data. The study used a dense and homogeneous monthly precipitation database comprising 120 rain-gauge stations over a 32 years testing period of 1975-2009. The results conclusively show that significant positive trends are both infrequent and found only in outlying stations during March, April and October. In order to estimate and characterize the magnitude of observed changes at unmeasured locations, Ordinary Kriging, Inverse Distance Weighted and Completely Regularized Spline interpolation methods were employed and compared. A comparative analysis of interpolation techniques shows that Ordinary Kriging with having RMSE of 0.148 is the best choice. This is followed by Inverse Distance Weighted (RMSE 0.151), and Splines (RMSE 0.152). Cross validation of the results shows the largest over prediction at Kars rainfall station and largest under prediction at Burdur station. Upon for the examination of the cross-validation and spatial error clustering results, the Ordinary Kriging method was concluded to be the best algorithm in the interpolation process.