The session on “Advancing Climate Modelling through High-Resolution Simulations and CMIP–CORDEX Integration” offered a balanced view of where the high resolution and downscaling model communities stands as we move towards CMIP7. Higher-resolution models are clearly opening new scientific avenues, allowing scientists to assess for example the impacts of ocean eddies on climate and climate change. The high-resolution models can represent observed climate change patterns that have been missed by low resolution models. While this is a great advance, the results are not uniformly transformative in all areas. Instead, a more complex picture is revealed for example when global high resolution is shown to not necessarily be better than regional downscaling over small islands.
Where high resolution makes a difference
Several presentations showed that finer spatial resolution—especially in the ocean—does improve key processes:
- Yushi Morioka demonstrated that eddy-permitting simulations better capture Atlantic Niño variability, boundary currents, and Antarctic sea ice variability, long-standing weaknesses in coarser models.
- Gokhan Danabasoglu presented results from CESM-HR, where resolving eddies improves Southern Ocean heat transport and leads to a better representation of the Southern Ocean climate change trend, as well as being able to capture features like Pacific cold tongue intensification, which is absent in low-resolution models. The AMOC appears more stable in high resolution simulations, which has implications for tipping point studies that typically employ low resolution models.
These examples highlight that high resolution can change simulated climate, not just refine it, particularly through improved representation of mesoscale dynamics.
But improvements are uneven
At the same time, the session made clear that higher resolution is not a universal fix for all our model problems:
- Thomas Jung (EERIE project) showed that even eddy-rich models offer opportunities to study eddy activity, but still miss processes like tidal forcing, leading to biases especially near coastlines and shelves.
- High resoltuion models are also expensive to run and especially to spin up. Anna Denvil-Sommer showed that deep-ocean nudging does not shorten spin-up time, but instead exposes underlying model biases. This positions high resolution as a diagnostic tool as much as a solution.
In other words, resolving smaller scales improves some aspects of the system, but other missing physics still limit overall performance.
Global high resolution vs regional downscaling: no clear winner
The relationship between global high-resolution models and regional downscaling (such as done under the CORDEX protocol) remains unresolved:
- Hiroaki Kawase showed that no single model performs best everywhere, motivating ensemble approaches and km-scale downscaling (to ~5 km) for improved regional detail, especially for precipitation.
- Francisco Doblas-Reyes (for Kat Grayson) highlighted that high resolution improves orographic representation for small island climates, but it remains unclear whether global high-resolution models outperform CORDEX downscaling in practice.
This suggests that integration of approaches is the way forward, rather than replacing for example downscaling with high resolution modelling entirely.
Practical advances alongside scientific uncertainty
Beyond the science, there are important technical developments:
- Large high-resolution ensembles (e.g., CESM-HR, as part of the MESACLIP project) are now available, enabling broader community analysis.
- The EERIE project provides open, cloud-based access to eddy-rich simulations and explores AI-assisted spin-up, helping manage computational costs.
These advances are making high-resolution modelling more usable by the broader community, even when we do not have the resources to run high resolution models ourselves.
Takeaway
This session demonstrated the value of high-resolution models, but also moved beyond the simple narrative that “higher resolution is better.” It demonstrated:
- Clear gains in representing specific processes (especially ocean dynamics and extremes)
- Persistent and newly exposed biases, even at high resolution
- Mixed evidence on whether resolution alone reduces uncertainty
- No definitive advantage yet over regional downscaling approaches
High-resolution modelling is powerful, but not a silver bullet. Its real value may lie as much in revealing model limitations and guiding improvements as in delivering immediately better projections.
