Supermodelling is a novel approach to combine different models dynamically. The models exchange information during the simulation which leads to new synchronized solutions, a supermodel. New learning methods are developed to train this combination of models using observational data. If the models complement each other, the supermodel potentially outperforms the imperfect individual models.
Current projects
November 2022 – October 2026: EU Horizon Europe Impetus4Change: Improving near-term climate predictions for societal transformation (I4C). In I4C, supermodelling is further developed as part of innovative new methods to mitigate prediction model deficiencies and close the gap between current skill and potential predictability, with a particular focus on seasonal-to-decadal timescales for the North Atlantic sector. The aim is to deliver climate predictions with improved quality at regional scales in particular for Europe and for variables of high societal relevance such as temperature, precipitation and their extremes. To help to address these goals, we will combine the atmosphere and ocean connected supermodels as developed in TOSCP and train this combination, resulting in a full atmosphere-ocean supermodel.
January 2022 – June 2024: ERC PoC Towards Operational Supermodel Climate Prediction (TOSCP). In TOSCP we configured the first version of the STERCP ocean supermodel for seasonal predictions. We tested and assessed the skill of the supermodel compared to toe earth system models NorESM, CESM and MPIESM. A preliminary analysis shows that the supermodel improves the prediction compared to the individual models for January and April start across the May predictability barrier.
We continued the work in the NSF project on Coherent Precipitation Extremes by setting up a very efficient version of an atmosphere-connected supermodel based on CAM version 5 and 6. This state-of-the-art atmosphere connected supermodel is comparable in quality to the individual CAM models. It is expected that a trained version could improve upon this first version and outperform the individual CAM models.
May 2021 – June 2024: NFR Increasing the impact of supermodelling climate research in Norway (NorSuper). Funded by the Norwegian Research Council under the FORSTERKNING programme for increasing the impact of Norway’s participation in EU framework programmes for R&D projects, NorSuper’s primary objective was to give Norwegian civil society, academia, and public and private sectors access to important findings on climate supermodelling from the ERC-STERCP project. Our secondary objectives were to (1) encourage a greater involvement of Norwegian actors (students and other researchers) into the ERC-STERCP project and follow-up research (such as through the EU funded projects TOSCP and I4C), and (2) enhance awareness of the potential of climate supermodelling at the international level.
Past projects
September 2021 – November 2023: NSF Project 2015618—Coherent Precipitation Extremes in a Supermodel of Future Climate. This project started the development of a state-of-the-art atmosphere supermodel, based on different versions of the Community Atmosphere Model (CAM) to improve the representation of large scale coherent structures that impact extreme precipitation.
September 2015 – August 2021: ERC CoG Synchronisation To Enhance Reliability of Climate Predictions (STERCP). This project was an international collaboration between partners in both Norway and the Netherlands. It was funded by the European Union’s ERC under the Horizon 2020 program. STERCP is a continuation of the SUMO: Super Modeling by combining imperfect models project, which was funded under Framework Program 7.
The STERCP project was divided into four work packages. The first work package aimed to develop methods to reduce the systematic error of models. The second, third and fourth work packages investigated the predictability of the supermodel, whether the supermodel is able to reflect the interannual, decadal and climate variability.