Item

Abstract

Runoff, which is a crucial hydrologic variable to predict streamflow, is generally estimated by hydrologic models because it is really difficult to observe. Recently, runoff data is well distributed as a part of output from general circulation models. Some previous studies used runoff data to estimate future changes of streamflow. Although evaluating runoff data is basically impossible because of difficulties in its measurement, it is becoming effective to apply runoff data to estimate streamflow. In this study, runoff data calculated by land surface model was utilized to predict streamflow for both flood and drought events. Target basins were located in Kyusyu island, Japan, where snowfall is rarely observed. Atmospheric reanalysis data was used to carry out a land surface model, SiBUC (Simple Biosphere including Urban Canopy). To estimate streamflow, calculated runoff data by SiBUC was given to a flow routing model 1K-FRM, which is based on one dimensional Kinematic wave theory. Estimated streamflow was evaluated by river discharge at gauging station. The Nash-Sutcliffe index in annual daily discharge is generally higher than 0.6. The relative error in annual maximum daily discharge was 8% on average. In addition, regarding drought, the 355th largest daily discharge in each year represents the observed value with a 27% relative error on average. These results suggest a high applicability of runoff data to estimate streamflow for both flood and drought.