Flood scaling under nonstationarity in Daqinghe River basin, China

Flood scaling issue is usually studied under stationary conditions. However, in recent decades, climate change and anthropogenic activities have changed hydrological processes, and stationary assumption has been questioned. To test the flood scaling invariance (simple scaling or multiscaling) and analyze the influence of environmental change on flood scaling parameter, in this study, eight mesoscale sub-watersheds in Daqinghe River basin were selected as the study area, and the trend and change point of annual maximum flood peak (AMFP) series were detected, respectively. All the AMFP series had downward trend, and the change point was around 1979. Therefore, the AMFP series are nonstationary. To analyze the flood scaling issue in the Daqinghe River basin, the AMFP series were reconstructed under the environmental conditions before and after the change point, respectively. Then, flood quantiles were calculated using the reconstructed stationary series. We also used GAMLSS (Generalized Additive Model in Location, Scale and Shape) to calculate flood quantiles based on the observed nonstationary AMFP series. According to the flood quantiles calculated by the above methods, the relationship of the drainage areas of the sub-watersheds and the flood quantiles was fitted with power function. Flood quantiles of the reconstructed stationary and observed nonstationary series showed obvious flood multiscaling. The increase in rainfall depth causes the increase in flood scaling exponents with the increase in return period, and different change ratios of land use before and after change point resulted in the flood scaling exponents of reconstructed series before 1979 were smaller than those after the change point at same return period.