Work Package 5 will build on efforts from the H2020 EUCP Project which have advanced predictive information by combining data from various sources, including uninitialised and initialised forecasts. EUCP worked to constrain climate projections with decadal predictions, aligning internal variability phases in ensembles to provide seamless predictive information across different timescales These methods will be tailored in I4C to blend subseasonal/seasonal predictions with seasonal/decadal predictions.
Some of the key methods to be explored in WP5 include:
- Storylines: WP5 will use ‘Storylines’ to classify climate projections into plausible, self-consistent potential storylines based on future evolution of critical climate drivers.
- Probabilistic Consistency: Simply merging different prediction ensembles at their merge point can lead to discrepancies, particularly affecting tail distributions and extreme events. Calibration methods and state-space modeling will be used to achieve probabilistic consistency and alleviate inter-ensemble discrepancies.
- Bias Correction and Optimization: I4C will develop a sophisticated blending approach based on state-space modeling and low-dimensional representations. To do this, I4C will use bias-corrected forecast data to avoid structural inconsistencies. These methods will be optimized for the North Atlantic and Europe, focusing on large-scale oceanic and atmospheric drivers and relevant extreme events in the Demonstrator cities.
- Collaboration is key: Effective interactions with other work packages are essential to ensure the success of the use of these forecasts. WP5 will work with WP3 and WP6 to ensure that blended outputs are tailored to their specific needs, maximizing impact and relevance.