Author(s): T. C. Pagano; Q. J. Wang; P. Hapuarachchi
Linked Author(s):
Keywords: Rainfall-runoff modelling; Forecast evaluation; Ensemble prediction
Abstract: Many hydrologic research studies assume perfect knowledge of observed precipitation. This assumption does not hold in forecasting situations in which future precipitation may be partially or entirely unknown. This study used the Ensemble Streamflow Prediction procedure to generate 15 day ahead streamflow forecasts for many catchments in southeast Australia. While simulation performance of the GR4J model was very high for many catchments, forecast skill dropped rapidly with lead-time. Skill at longer lead-times was mostly due to the model's ability to reproduce the hydrologic seasonal cycle. When seasonality is removed, almost none of the catchments had meaningful skill beyond 3 days lead-time.
Year: 2011