Innovative modelling techniques and approaches are key to gathering accurate and reliable information for the seasonal to decadal climate predictions. As such WP2 focuses on the application of advanced statistical and modeling techniques which improve predictions while maintaining a lower computational cost.
Overall, WP2 seeks to advance the state-of-the-art in climate prediction by addressing key challenges and leveraging cutting-edge techniques to enhance the accuracy and reliability of S2D predictions, ultimately contributing to more effective climate adaptation and resilience strategies.
Key highlights include:
- Understanding Model Errors and Signal-to-Noise Issues: Key to the success of using modelling is to avoid pre-existing model errors or signal-to-noise issues. Therefore at the onset of the project, we work to gain an in-depth understanding of the errors that exist. To do this, we will use existing multi-model simulations and employ machine learning, model output statistics (MOS), and clustering techniques. An emphasis will be placed on variables, seasons, and averages relevant to the I4C community.
- Development of Novel Filtering Techniques: Novel filtering techniques will be developed to efficiently identify predictable signals from large multi-model ensemble predictions. Process-based constraints will be applied to subsample ensemble members to increase forecast skill.
- Super Resolution Data Assimilation Technique: Novel data assimilation techniques such as super resolution data assimilation combines data assimilation with machine learning. This helps with model bias correction and ensures a more efficient use of the high-resolution observations in the data assimilation.
- Investigation of Ocean-Eddy Resolving Models: The project will assess the benefits of using ocean-eddy resolving models in climate predictions, particularly in predicting extreme oceanic and atmospheric events over Europe.
- Flux-correction and Supermodelling: innovative approaches such as flux-correction, super modelling, and super-resolution data assimilation help to keep the computation cost low.