Application of Weighted Semivariogram Model (WSVM) based on fitness to experimental semivariogram on estimation of rainfall amount

Abstract. This study presents a weighted semivariogram model (WSVM) which is intended to reduce the uncertainties in the selection of the best-fit semivariogram model and associated parameters. The proposed WSVM is based on the combined forecast method, providing the weighted average semivariogram by summing up the product of estimated semivariograms and weighted factors, which are related with the inverse of the objective function value associated with the optimal parameters of theoretical semivariogram models (TSVMs). A WSVM can save computation time in the estimation of rainfall amount without the identification of the best-fit TSVM commonly carried out by the cross-validation. Ten rainstorm events recorded at fourteen rain-gauges in North Taiwan's Shinmen reservoir catchment are used to develop and validate this model by comparing the estimated rainfall amount by the Kriging method with the WSVM and TSVM, respectively. The results of the model validation indicate that the proposed WSVM not only reduces the uncertainty of failing to select the best-fit TSVM, but also effectively provides more the accurate and reliable estimated rainfall amount than TSVM.