Abstract
This paper proposes a gauge-radar rainfall composition method using Cokriging-type Sequential Gaussian Simulation (co-sGs). It estimates rainfall fields quantitatively consistent with ground gauged rainfall data keeping the spatial distribution pattern of radar rainfall. The method generates possible rainfall realizations with estimation uncertainty; the generated rainfall fields are to be used to evaluate the uncertainty of runoff prediction with a distributed rainfall-runoff model through a Monte Carlo simulation formulation. Application of co-sGs and its cross validation indicated that the generated rainfall has higher accuracy than one calibrated by Dynamic Window method currently used for operational purpose in Japan.
