A seasonal calibration approach of conceptual hydrological models for improved real-time river forecasting

Conceptual hydrological models can play an important role in real-time river forecasting systems due to their limited calculation time and versatility. Nevertheless, their simplified structure, very often based on the water content in multiple storages, and constrained physical background hampers their applicability in seasonally influenced catchments. In particular, these models often show good forecasting performance in one season (e.g. for high discharges in wet seasons), but fail to capture events in other seasons (e.g. due to typical high intensity precipitation during dry periods). To overcome this issue, we propose a seasonal calibration approach for conceptual hydrological models, based on the results of a seasonal sensitivity analysis. The obtained seasonal models however induce an additional challenge within a continuous real-time river forecasting system: the transition from one seasonal model to another. The latter is of particular importance when the volume of the storages in the conceptual model changes between different seasons. An application with the conceptual NAM model for three catchments in Belgium will be used to illustrate the proposed calibration strategy and a number of possible solutions for the transition issue.