Spatial variation of Saudi Arabia’s annual rainfall using a regional prediction model

Abstract

In this study, Saudi Arabia is divided into five geographical regions: northwestern, southwestern, western, central, and finally, eastern with central. The study shows the possibility of constructing and formatting rainfall as a regional variable depending on the locality. Two techniques derived from the point cumulative semivariogram (PCSV) model are used to predict rainfall in the five regions. These are, namely, the trigonometric point cumulative semivariogram (TPCSV) method and the classical weighing average method (classical point cumulative semivariogram (CPCSV)). These methods are applied to data from 21 weather stations in the five regions for a 35-year period from 1970 to 2005 in the majority of stations. It is found that the absolute error between observed values and the predicted values using TPCSV and CPCSV does not exceed 5 and 13 %, respectively, except in the central region, where the absolute error is 18 % using both methods. Regarding the spatial rainfall variation in the region, the strength of correlation is calculated in terms of angles among the sites using the TPCSV technique, the smaller the angle the stronger the correlation. The standard weighing function (SWF) is used to find the weight of rainfall records at each site, which expresses the effect of locality on rainfall variation. The SWF approach reflects the homogeneity of rainfall around the pivot sites in all regions except the central. High-correlation angles (85, 90, and 88°) among the sites of the region explain the weakness of correlation among this set of station, where the error of prediction both method in this region exceeds 18 %. PCSV succeeds in explaining the spatial rainfall variation and finding the weight of rainfall phenomena for each site in all five regions, while successfully predicting rainfall.