Workshop description and aim
Ensemble forecasting -- which uses a set of predictions to improve on a single-model output -- have been very successful in improving operational weather forecasting and are also used in many other fields such as data science and economics. Ensemble techniques are even used in state-of-the-art machine learning competitions to improve performance. Their use in Space Weather (SW) forecasting could not only improve forecast accuracy but also provide simple model uncertainties that are crucial for improving end-user understanding of the products available.
In the past few years the use of ensembles in SW forecasting has grown. However, in comparison to terrestrial weather forecasting, their potential hasn’t been fully used yet.
The main goal of this workshop is to make concrete steps towards improving the SW forecasting capabilities by implementing ensemble techniques that have been successful in other forecasting fields, especially terrestrial weather. In order to achieve this main goal, this workshop aims to shed light into questions such as:
1. Usefulness of Ensembles. Given the limitations of input data and/or models used for foresting, can ensemble techniques contribute significantly to improve forecasting of SW events? Are the ensemble methods implemented in other fields transferable to SW forecasting?
2. Type of Ensembles. Multi-model input ensembles versus initial conditions perturbation single-model ensembles: when and how to apply either case.
3. Combination of Forecasts. Are simple combinations (e.g. averages) of forecasts more successful than complex combinations (e.g. weighting schemes) in SW forecasting?
This workshop aims to bring together researchers from academic institutions, operational forecasters across all subfields of SW, experts in terrestrial weather forecasting, and experts in financial/economical forecasting -- among others.