Assessment of Uncertainty in the Spatial Distribution of Rainfall Using Geostochastic Simulation

Abstract

The uncertainty in the spatial rainfall distribution and basin mean rainfall for flood discharge estimation has not been considered in hydrologic practice. The geostochastic simulation has potential for the assessment of uncertainty of the spatial rainfall distribution and basin mean rainfall. This study compares three geostochastic simulation methods, including the circulant embedding method (CEM), the sequential Gaussian simulation (SGS), and the turning bands method (TBM), for assessing the uncertainty in the spatial distribution of rainfall. These methods are found to be comparable in terms of spatial standard deviation, coefficient of variation, interquartile range, overall range, and the distribution of basin mean rainfall. Further, the estimates of basin mean rainfall by these methods are almost similar to those by conventional spatial interpolation methods. However, CEM and TBM are found to be superior to SGS in the simulation performance based on mean error (ME), mean percentage error (MPE), mean absolute error (MAE), and root-mean-square error (RMSE). The geostochastic simulation methods have potential for flood risk assessment.