Efficiency analysis of two spatial interpolation methods od precipitation on the Kolubara river basin

Abstract

The efficiency comparison has been analyzed in this paper for the two spatial interpolation methods that have been most commonly used in the practice, for averaging precipitation data on the experimental catchment of the Kolubara river upstream of the Peštan river confluence, in the Republic of Serbia. The analyzed methods are most commonly used in engineering practice: the Thiessen polygon method and the Linear interpolation method based on the TIN network. The prediction efficiency of the average annual precipitation for a certain location, performed using the cross-validation method, showed that both analyzed methods have low efficiency of point data prediction and that both methods are very sensitive to the exclusion of at least one rain-gauge station. Sensitivity analysis of dependence spatially averaged value of annual precipitation for a given area and the number of precipitation stations, performed for four scenarios: S1 - only precipitation stations in the basin, S2 - precipitation stations up to 10 km from the basin boundary, S3 - precipitation stations up to 20 km from the basin boundary and S4 - precipitation station up to 30 km from the basin boundary, showed that for adequate application of both methods it is necessary to include the surrounding stations up to 20 km from the basin boundary.