Work package 2

Post-processing and tailoring

In the post-processing stage, the primary forecasts are treated in order to generate tailor-made final forecasts for specific applications. Based on the evaluation results, a decision will be made whether to use an approach in which the different forecasts (lagged and from different systems) are merged into one grand ensemble of forecasts, possibly through a weighting scheme (e.g. using Bayesian Model Averaging), or an approach based on parallel but separate forecasts.
This may involve a correction of the forecast bias or, alternatively, a calibration of the forecast (see e.g. Olsson and Lindström (2008) for an application to discharge forecasts). The primary forecasts will be translated into probability forecasts, e.g. the probability of different intensity levels for each time step in the forecast (1-36 h). 

The forecast for a given location and time period can be formulated either as a probability that the precipitation exceeds some critical threshold (e.g. 25% probability of rainfall exceeding 10 mm/hr) or as the most probable intensity (or range of intensities). Post-processing will also include further tailoring to be consistent with the input format required in subsequent hydrologic and hydraulic models.

MSB timetable