Parameter sensitivity of a watershed-scale flood forecasting model as a function of modelling time-step

Despite significant developments, the simple, lumped, conceptual, rainfall-runoff model is still widely used for flood forecasting. What may not be appreciated is that, while such models can often be calibrated to give reasonable forecasts of flood flows, both parameter values and the fluxes of water through individual model components change significantly with the time step used. This means that such models should be used with caution for studies which require “internal” information, such as hydrograph separation or water quality studies that depend on knowing the fluxes through individual flow routes through the model and in studies which try to relate parameter values to physical features of the catchment. To demonstrate this time-scale limitation, a parameter sensitivity analysis was performed on a typical lumped conceptual model (SMARG) applied to a small rural catchment on the Irish East Coast for a number of different time-steps, flow regimes and evaluation metrics. A global sensitivity analysis method (GUI-HDMR, is applied to calculate sensitivity indices which varied greatly with time-step and evaluation metric used. The sensitivity of parameters also differed for different flow regimes. Care should be taken in using internal information and calibrated parameter in conceptual models because of the strong dependence on time-step