Grid-based parameter regionalization of distributed hydrologic models based on geospatial dataset for flash flood predictions in heterogeneous catchments in Japan

Abstract

National-scale parameter regionalization of a distributed rainfall-runoff model (1K-DHM) is promoted for flash flood prediction in river basins in Japan. Representative model parameter sets of 1K-DHM for 7 geospatial characteristics including land-use and soil properties were identified through calibration in 53 donor catchments with 1 dominant geospatial characteristic and a leave-one-out cross-verification within each parameter group. This resulted in a parameter map of 1K-DHM for the 30-second grid cells with a national-scale coverage in Japan. The identified parameter (IP) sets yielded noticeable differences in the depth-discharge relationship in 1K-DHM, which explains the difference in runoff characteristics among geospatial categories. The transferability of IP to heterogeneous catchments was validated against individual optimized parameter sets (OP) for 70 receptor basins evaluated using Nash-Sutcliffe efficiency (NSE) and the normalized peak discharge error (PDE). The results show that IP (median NSE: 0.74, median PDE: 0.028) demonstrated comparable performance to OP (median NSE, 0.76; median PDE, 0.014), which indicates that a distributed rainfall-runoff model with model parameters determined by land use and soil properties can predict floods in ungauged basins.