Performance of rainfall–runoff models in reproducing hydrological extremes: a case of the River Malaba sub-catchment

Abstract Daily River Malaba flows recorded from 1999 to 2016 were modelled using seven lumped conceptual rainfall–runoff models including AWBM, SACRAMENTO, TANK, IHACRES, SIMHYD, SMAR and HMSV. Optimal parameters of each model obtained an automatic calibration strategy. Mismatches between observed assessed a total nine “goodness-of-fit” metrics. Capacity the reproduce historical hydrological extremes was through comparison amplitude–duration–frequency (ADF) relationships or curves constructed based on flow quantiles. Generally, most performed better for high than low flows. ADF both various return periods 5 100 years well reproduced by SAC, TANK poorly SIMHYD SMAR, respectively. Overall, AWBM slightly other if are be considered simultaneously. The deviations these larger periods. It found that choice metric affects how performance can judged. Results this study also show when focusing extremes, uncertainty due particular should taken into consideration. Insights provide relevant information planning risk-based water resources applications.