An Evaluation of Deterministic and Geostatistics Interpolation Models to Precipitation Regionalization in Ardebil Province

Two deterministic methods [Radial Basis Functions (RBF) and Inverse Distance Weighting (IDW)], and two stochastic interpolation methods [Kriging and Co-Kriging] have been evaluated to estimate the annual precipitation in Ardebil Province. For this purpose, we used data taken from various stations across the province [4 synoptic stations, 1 climatological station and 24 rain gauges], over a decade (2005-1995). The performance of the above model was examined using cross validation and the indices of Mean Bias Error (MBE), Mean Absolute Error (MAE) and Root Mean Square Error (RMSE).Finally, to identify under-estimated and over-estimated areas of each method at elevations above the spatial distribution of stations, the output of each model was examined. The results suggested that Kriging model had lower errors than other methods and was more compatible with the elevation surfaces. Therefore, this method was identified as the best option for estimating f precipitation in areas without adequate data and intense topographic gradients in Ardebil Province.