Application of simple clustering on space‐time mapping of mean monthly rainfall pattern

Rain gauges in many watersheds are not enough to represent the spatial and temporal variations of precipitation. Many agricultural activities can be highly sensitive to these variations and therefore different monthly or annual precipitation statistics have been widely used in the literature for agricultural planning and management. For example, mean monthly rainfall has been used as one of the main criteria for classifying suitable regions for dryland farming. In this paper, a method for space‐time estimation of cumulative mean monthly rainfall (CMMR) is presented. The proposed method for estimating CMMR pattern is based on developing time‐varying variogram coupled with ordinary Kriging (OK). OK is selected as a spatial exploration method and all of its parameters are optimised using weighted least square method. K‐means clustering has also been applied to increase the model efficiency. Space‐time estimation of CMMRs in each cluster has been carried out based on the observed CMMRs of the rain gauges in each specific cluster. The proposed method is tested using 25 years of monthly rainfall records of 14 rain gauges in the Maroon watershed in south of Iran. The results of the case study have shown that both methods (estimation with and without clustering) are proper choices for space‐time modelling of rainfall pattern; however, the method in which the rain gauges are classified to different clusters has shown a better statistical performance in terms of normalised mean square error. Copyright © 2010 Royal Meteorological Society