Geostatistical modeling to simulate daily rainfall variability in Iran

Abstract Rainfall variability is among the main challenges confronted when simulating the spatial patterns of climatic changes under different environmental conditions, particularly in countries with arid and semiarid climates such as Iran. Climate changes simulation, through geostatistical modeling, have made possible to develop the understanding of spatial variability, e.g. daily rainfall. This article presents some spatial variability simulations of average values of the daily rainfall for Iran from 170 stations and 39,042 rainfall points by comparing geostatistical techniques based on the prediction errors. For the spatial variability simulation of average values of the daily rainfall, rainfall data series of 1975–2014 was used to analyze the accuracy of geostatistical models. Four statistical error assessment measures, mean absolute deviation prediction errors, mean square prediction errors, root mean square prediction error (RMSPE), and coefficient of determination (R2), were used to assess and compare the interpolation techniques. Tetraspherical Ordinary Kriging, Exponential Kernel Smoothing, Order 5 Polynomial Kernel Smoothing, and Quartic Kernel Smoothing were selected as the best spatial models for simulating daily rainfall variability, in the order of their performance. The RMSPE varied between 0.042 and 2.639 were predicted by employed models for average values of the daily rainfall.