GIS-BASED SPATIAL INTERPOLATION OF RAINFALL ACROSS ARABIA

Accurate spatial interpolation of monthly and annual rainfall is required for validating satellite-based rainfall products and parameterizing spatially distributed climate and hydrologic models. Spatially distributed estimates of annual and monthly rainfall across Arabia were obtained from sparse station data using GIS-based univariate and multivariate spatial interpolation techniques. A high-resolution digital elevation model was introduced as an exhaustive ancillary variable to augment the accuracy of the spatial interpolators. Accuracy of the spatial interpolators is investigated using the cross-validation technique and expressed as root mean square error between observed and estimated rainfall. Both annual and monthly rainfall exhibited very strong spatial continuity with the spherical model providing the best fit for the observed variograms and cross- variograms. Based on cross-validation technique, colocated cokriging provided the most accurate spatial interpolation and thus it was used to grid both annual and monthly rainfall across Arabia. Mean annual and monthly rainfall surfaces obtained by colocated cokriging are not only the most accurate, but also the least smoothed ones and depict clearly the local spatial patterns and regional trends.