Probabilistic Forecasting Course
Forecasting may reduce but cannot eliminate inherent uncertainties about the future. One approach to managing this is to estimate remaining uncertainties in a probabilistic fashion and thus arrive at a probabilistic forecast. This estimation can be done using ensemble techniques, statistical post-processing or a combination of these.
The quality of resulting forecasts can be assessed through a process called forecast verification. Verification of probabilistic forecasts uses different techniques than verification of deterministic forecasting. Within the framework of a forecast – decision – response system, effective forecasting requires that attention is given to additional issues such as forecast visualization, communication, decision-making and training.
The present course comprises an introduction to these topics. It is aimed at both practitioners and scientists.
- Introduction to uncertainty, risk and probability
- Techniques for estimating predictive hydrological uncertainty: ensembles and post-processing
- Verification: how good is my (probabilistic) forecast?
- Forecasting applications: (i) storm surge forecasting for the North Sea coast; (ii) fluvial forecasting in Rhine, Meuse and the EFAS system
- Serious game: making forecast sensitive decisions
- Using probabilistic forecasts in operational practice.
This course was part of the Delft Software Days – Edition 2018 (DSD-INT).