Revising calibration of a lumped watershed model to simulate extreme streamflow

A present-day challenge in hydrologic modeling lies is yielding extremes in simulated streamflow. The missing signature of anthropogenic climate change in flood magnitudes and frequencies is often attributed to conventional model calibration, which does not separate catchment processes into average and extreme flows. Moreover, traditional model calibration typically considers a function of error variance as the objective function, which can yield the central tendencies of the observed streamflow but fails to furnish extreme flow characteristics. Addressing this issue, the current study aims to revise traditional hydrologic model calibration approaches for the purpose of improving extreme flow simulations. To achieve this, a comprehensive list of 20 objective functions for a lumped-conceptual hydrologic model calibration on two Indian peninsular river basins are considered. The results suggest that the Kling–Gupta efficiency for extreme thresholds and the mean squared error for the flow duration curve are the objective functions that can improve model simulation for the extremes. Furthermore, it is observed that a change in the objective function substantially changes two model parameters: the number of quick flow tanks and the height of the soil moisture tank. The revised calibration also improves the seasonality associated with the annual maximum flood. In addition, this study discusses the likely impact of parametric uncertainty on extreme flow simulation.