Spatial variability of rainfall trends in Iran

The main objective of this paper is to analyze the spatial variability of rainfall trends using the spatial variability methods of rainfall trend patterns in Iran. The study represents a method on the effectiveness of spatial variability for predicting rainfall trend patterns variations. In rainfall trend analysis and spatial variability methods, seven techniques were used: Mann–Kendall test, Sen’s slope method, geostatistical tools as a global polynomial interpolation and the spatial autocorrelation (Global Moran’s I), high/low clustering (Getis-Ord General G), precipitation concentration index, generate spatial weights matrix tool, and activation functions of semiliner, sigmoid, bipolar sigmoid, and hyperbolic tangent in the artificial neural network technique .For the spatial variability of monthly rainfall trends, trend tests were used in 140 stations of spatial variability of rainfall trends in the 1975–2014 period. We analyzed the long and short scale spatial variability of rainfall series in Iran. Spatial variability distribution of rainfall series was depicted using geostatistical methods (ordinary kriging). Relative nugget effect (RNE) predicted from variograms which showed weak, moderate, and strong spatial variability for seasonal and annual rainfall series. Moreover, the rainfall trends at each station were examined using the trend tests at a significance level of 0.05. The results show that temporal and spatial trend patterns are different in Iran and the monthly rainfall had a downward (decreasing) trend in most stations, and the trend was statistically significant for most of the series (73.5% of the stations demonstrated a decreasing trend with 0.5 significance level). Rainfall downward trends are generally temporal-spatial patterns in Iran. The monthly variations of rainfall decreased significantly throughout eastern and central Iran, but they increased in the west and north of Iran during the studied interval. The variability patterns of monthly rainfall were statistically significant and spatially random. Activation functions in the artificial neural network models, in annual time scale, had spatially dispersed distribution with other clustering patterns. The results of this study confirm that variability of rainfall revealing diverse patterns over Iran should be controlled mainly by trend patterns in the west and north parts and by random and dispersed patterns in the central, southern, and eastern parts.