Sub-seasonal streamflow forecasts provide useful information for a range of water resource management and planning applications. This work has focused on improving forecasts for one such application: the management of water available in an open channel drainage network to maximise environmental and social outcomes in a region in southern Australia. Conceptual rainfall-runoff models with a postprocessor error model for uncertainty analysis were applied to provide forecasts of monthly streamflow. Two aspects were considered to improve the accuracy of the forecasts: 1) state updating to force the models to match observations from the start of the forecast period, and 2) selection of a calibration period representative of the forecast period. Five metrics were used to assess forecast performance, representing the reliability, precision, bias and skill of the forecasts produced, using both observed and forecast climate data. The results indicate that assimilating observed streamflow data into the model, by updating the storage level at the start of a forecast period, improved the performance of the forecasts across the metrics when compared to an approach that “warmed up” the storage levels using historical climate data. The shorter calibration period improved the performance of the forecasts, particularly for a catchment that was expected to have experienced a change in the rainfall-runoff relationship in the past. The results highlight the importance of identifying a calibration record representative of the expected forecast conditions, and if this step is ignored degradation of predictive performance can result.
展开▼
