京都大学防災研究所年報 第 48 号 B 平成 17 年 4 月 Annuals of Disas. Prev. Res. Inst., Kyoto Univ., No. 48 B, 2005 Use of Disaggregated Rainfall Data for Distributed Hydrological Modeling in Yodo River Basin Roshan K. SHRESTHA * , Takahiro SAYAMA, Yasuto TACHIKAWA, Kaoru TAKARA *Graduate School of Urban and Environmental Engineering, Kyoto University Synopsis This study tested the disaggregated rainfall field as the input field for hydrological simulation of various sub-catchments of the Yodo River. The multiplicative random cascade (RC) method (based on the beta lognormal model), the multiplicative random cascade HSA (RCHSA) method, and the space-time rainfall modeling (STRaM) method are employed to disaggregate the rainfall field from coarse (48-km, 100-min) to fine (3-km, 10- min) resolution. The Yodo River model, an OHyMoS assisted distributed hydrological model with saturated-unsaturated surface-subsurface flow mechanism, is used for simulating the runoff at Ootori, Ieno, Kamo and Inooka having catchments areas of 156 km 2 , 476 km 2 , 1469 km 2 and 1589 km 2 respectively. The simulated discharge using the disaggregated rainfall with STRaM is quite similar to the one obtained from using the radar observed rainfall. Keywords: downscaling, rainfall field, yodo river model, RCHSA, STRaM method 1. Introduction A fully distributed hydrological modeling, which includes distributed process descriptions, takes spatially and temporally distributed input fields, and employs fully distributed parameterization, often finds difficulty in producing entirely satisfactory outcome. Most of the distributed hydrological models are reported being unable to make accurate hydrological simulation and/or prediction (Reed et al. 2004). There are multiple reasonings that cause an advanced distributed hydrological model perform inferior than a simple lumped parameter model. Existence of scale gap between meteorological studies and hydrological studies is often cited as the dominant cause failing the distributed hydrological modeling. Most meteorological models perform well at the scale of several hundred kilometers in space and monthly scale in time. But the hydrological analyses require daily scale in time and few kilometers in space or even a finer scale. Numerical weather simulators of much finer scales in space and time are feasible but they are suffered heavily by a larger degree of uncertainty caused by problems in parameterization and lack of knowledge of meteorological processes at much finer scales (Bates et al., 1998; Chen et al., 1996; Giorgi and Mearns, 1991; Houze, 1997). This weakness comes to the core of the problem area because the hydrometeorological data are the one of key-role player in the hydrological studies involving a distributed hydrological model (Shrestha et al., 2004a). Recent advances in distributed hydrological modeling are continuously at odds with the use of inappropriate scale at various levels of the modeling (Shrestha, 2005). There is continuous interest in the study of scale dependent features and scale effects in hydrological modeling to understand the current limitations. A poor outcome is associated with the uncertainty coming not only from the heterogeneity of the parameters, processes and input field but also from their scale of representation (Shrestha et al., 2005). An intuitive solution to overcome the scale mismatch between the input field and the process description is to employing an appropriate disaggregation of the input field. Predictably, the basic tenet of this solution is the accuracy of the hydrological simulation involving disaggregated rainfall field, whose original source is of a much coarser scale equivalent to the one we usually obtain from a regional scale meteorological model. While such a claim at first glance seems unexpected, it is examined in this study involving multiple sets of disaggregated rainfall in multiple catchments of different sizes. The rainfall is disaggregated into finer resolution using three different disaggregation