Understanding the limits of two lumped hydrological models through divergences between daily and sub-daily projections

The present paper identifies four common reasons for divergences between projections of flow metrics simulated by lumped conceptual hydrological models at a daily time step, and a daily-averaged 3-h time step. These reasons are based on five case studies simulated by two hydrological models in North America in the context of climate change, and are related to: the filtering of subdaily discrete and cyclic processes; the time-step dependent solving of nonlinear equations involving logical conditions; the time-step dependent solving of nonlinear sequential equations; and the timing of the flood events within the daily window. The time-step dependent simulation of the corresponding internal processes requires compensatory mechanisms in other processes in order to achieve an equivalent performance of the hydrological models in calibration at both time steps. These compensatory mechanisms can be identified in the parameter sets of the models and often imply a time-step dependent seasonality of the internal variables. Divergences between hydrological projections at daily and subdaily time steps reveal significant future shifts in the hydrological regime or in the runoff-generating processes, outside the domain of validity of the calibrated parameter sets. Time-step dependent projections underscore the need for detailed insight before using projections for engineering applications. Furthermore, the comparison of hydrological simulations from different time steps is a valuable approach to assess the domain of validity of calibrated parameter sets, and to understand the behavior of conceptual hydrological models under non-stationary climate conditions.