Identifying Runoff Parameters and Predicting Future Runoff with GCM Data in a Mountainous Basin in Japan

In order to make relevant predictions of future runoff on the regional scale, which are useful for decision makers, analysis needs to be carried out on the basin scale. However, hydrological models still have uncertain parameters on which the accuracy of the model performance depends. In this study, we aimed to predict the future runoff on the regional scale using a hydrological model in which the runoff parameters are identified.

The study basin of Sameura Dam (472km²) is located in the central mountain region of Shikoku, in the west of Japan. The hydrological model used in this study consists of the infiltration model presented by Diskin and Nazimov (1995) and the storage–discharge relations originally described by Horton (1936). The storage–discharge equation for flood flow comprises two parameters, the exponent p and the constant k, which are different in each flood event. The flood runoff analysis is carried out 10 000 times for 22 flood events, which are selected from the calibration period of 1986–2000, while changing the values of the two parameters in the storage–discharge equation using a double-loop algorithm. The result for each flood event shown on the basis of the Nash–Sutcliffe criteria (NSE) is visualized and the visualized NSEs are superposed, then the optimum values of the parameters for the basin are identified. The optimized parameter set is applied to the 22 flood events selected from the validation period of 2001–2015. The NSE–based median of 22 flood events is 0.867, which is comparatively good.

The simulation for future runoff projections is carried out with MIROC5, provided based on the Inter-Sectoral Impact Model Intercomparison Project (ISIMIP), in accordance with the emmissions scenario RCP8.5. Bias–correction is carried out for MIROC5 outputs of 9 grids of data of 0.5-degree resolution surrounding the study basin using the observed data of the calibration period of 1986–2000. The runoff simulation is carried out for the period of 2080–2099 future projections, and the results show that the direct runoff increases between May and September.