Streamflow maps for run-of-river hydropower developments in Japan

Japan recently introduced a feed-in tariff for small-scale hydropower plants, promoting the development of run-of-river hydropower plants in small-sized basins; however, appropriate implementation requires gauging station streamflow data at substantial costs and time (i.e., more than several years). Thus, in this study, we generated streamflow maps for small-sized basins (∼10 km²) throughout Japan using artificial neural networks (ANNs). Modeled output streamflow characteristics relied upon the input variables obtained from 176 basin characteristics and consisted of mean annual streamflow (Q_MEAN), daily streamflow indices in a flow duration curve (Q_D), and a water volume index for run-of-river hydropower energy production (W_D95). We preliminarily investigated the impacts of selecting the input variables obtained from 176 basin characteristics on performances of the ANNs, which indicated that the ANNs showed high robustness for disinformative input variables and multicollinearity between input variables. Although Q_MEAN, high Q_D, and W_D95 performed well, low Q_D were inadequate, possibly due to snowmelt contributions and small catchment sizes obstructing the detection of geological impacts. To accurately estimate the streamflow characteristics throughout Japan, we emphasize the importance of developing robust methods for correcting wind-induced precipitation undercatch and a spatial interpolation for precipitation in high-montane areas. Nevertheless, the ANNs for Japan proposed herein significantly outperformed a previous study exhibiting excellent global-scale ability. A map expressing run-of-river hydropower potential in small-sized basins was generated and closely corresponded to the spatial distribution and electrical output of existing hydropower plants. Furthermore, we demonstrated that the hydropower potential map reproduces the hydropower developments corresponding to the history of the electric power systems in Japan, which reflects its high reliability. Therefore, the hydropower potential map can greatly aid the exploration of optimal sites for hydropower developments.