The Climate Science2Policy workshop, organised by the European Commission’s EASME (Executive Agency for Small and Medium-sized Enterprises), took place virtually on the 17th and 18th of November 2020, bringing together a cluster of EU climate modelling projects, including EUCP, CRESCENDO, PRIMAVERA, APPLICATE, CONSTRAIN, BLUE ACTION, 4C, etc. Held online because of the COVID-19 pandemic and with the aim to ensure a continuation of climate science dialogue at the European level, around 180 participants attended the 2-day event.
Science presentations showcasing the latest results from the projects involved in the workshop brought our attention to policy-relevant conclusions and research gaps relevant to climate projections. Our main takeaways particularly focus on advances in projections and evaluation, as well as on model constraint.
Advances in climate projections and evaluation
The workshop focused on three major areas relating to advancing climate projections and their evaluation, with many speakers and sessions featuring assessments of the performance of our current selection of climate models:
- The ESMValTool: The first talk of the day discussed the ESMValTool which has already demonstrated significant improvements in the simulations of rainfall levels, the carbon cycle and the climatological effects of clouds in CMIP6 models over their older predecessors. This standardised assessment method is being used in the next Assessment Report from the IPCC.
- Convection permitting models (CPMs): Another popular topic in the workshop was pushing the boundaries of our climate projections using higher-resolution climate models. For example, convection permitting models (CPMs), a class of high-resolution models able to simulate atmospheric convection, were demonstrated to be much better at simulating extreme rainfall than regional-scale models, with up to 10 times less inherent bias in their projections. CPMs carry significant internal variability which can be addressed using a large ensemble of simulations, but their great computational cost renders this impractical. Attendees suggested either trading large ensembles for a better understanding of their uncertainties, or very specific tailoring towards user needs, reducing unnecessary simulation size. Indeed, many speakers touched on the demand from users of climate information for regional and local-scale climate projections of great relevance to local communities. Producing tailored simulations, as well as merging predictions and projections on various spatial and temporal scales to exploit their strengths, will bring important benefits for these users.
- Decadal climate predictions: Significant improvements were presented in the skilfulness of climate predictions on a decadal timescale, with some of their limitations being addressed using novel constraint techniques. These models have now begun to see significant utilisation by end-users, such as in the agricultural sector. They offer exciting opportunities to contribute to the Horizon Europe Mission on adaptation to climate change and the European Green Deal, with relevance to both the public and private sectors. Further work on developing these models, including creating large ensembles of initialised models which use current observations to fine-tune their simulations, have the potential to greatly improve the quality and credibility of their predictions.
Despite these advances, several presentations discussed areas where our current climate models are lacking, particularly in the polar regions and around climate tipping points, critical thresholds beyond which rapid climatic change may occur. Our current climate models are not suited to handling tipping points, and additional fundamental development is needed in our modelling capability. In the meantime, our focus should be on observations and close monitoring of potential tipping point indicators, some of which are predictable to an extent, to help inform mitigation policy.
Model constraint
Another important component of the workshop concerned advances in constraining the output of climate model ensembles. Climate models are improving, but still give a big range of results, therefore methods to reduce this range are more important than ever. Several speakers reported advances in the use of machine learning techniques to improve model constraint, while better methods of model weighting were presented that account for model performance. This approach can reduce the level of global warming seen in CMIP6 model projections. Another novel constraint method involved using decadal-scale predictions, which can make use of contemporary data, to constrain longer projections out to 50 years in the future.
Funded by the European Union under Horizon 2020.