Assessment of interpolation methods for annual and seasonal precipitation in Mashhad plain

Rainfall is one of the most important components of the hydrologic cycle, and its spatial distribution plays an important role in water resources management. Choosing an appropriate interpolation method is one of the main challenges in optimally estimating the values at those locations where no samples or measurements were taken. The purpose of this research is to determine the best interpolation method for precipitation in Mashhad plain as the main agricultural areas in Khorasan Razavi Province. First, rainfall data was collected during the period of 2004 to 2013. A total of 63 stations were selected. Then, five interpolation methods, namely Kriging, co-Kriging, Regression, regression Kriging and Inverse Weighted Distance were used for estimating spatial distribution of precipitation. The Root Mean Square Error and Mean Bias Error was considered to select the best interpolation method. The interpolation approaches were evaluated using a cross-validation method. Results revealed that the most accurate interpolation method is based on the spherical model as the theoretical semivariogram model. The error of method showed that the regression kriging and three-variable regression (x,y,z) methods are the most accurate models with RMSE=5.8541 and MBE=0.3004 and RMSE=7.5888 and MBE=0.3498, respectively to interpolate annual precipitation over the study area. It was also deduced that for seasonal rainfall data, due to poor data correlation, it is better to use the Kriging method to interpolation method. The co-Kriging method was also recognized as the weakest method with least accuracy for rainfall interpolation.