The best possible accuracy of standard turbulence shortcuts
- Good it would do
- 2 / 5
- Can AI help
- 4 / 5
- How neglected
- 3 / 5
- Can we check it
- 4 / 5
Why we stopped
Rejected on prior art and decision value: the scout's optimisation is ill-posed by its own source, a funded CliMA group has already answered keep-tuning-or-replace online for 2D geophysical turbulence (tuning works except at coarse resolution, where backscatter is needed), and the remaining offline accuracy floor, though cheap and checkable on the public pyqg benchmark, changes no identified modelling decision because offline skill correlates weakly or negatively with online skill. The tractability checker and outside skeptic dissent.
Why it's stuck
Climate models lean on simplified turbulence formulas. Work on the mathematical limit of how good those formulas can ever be has only been done for simple 2D flows.
What we'd try
Extend the optimization to ocean and atmosphere-like flows to learn whether to keep tuning current formulas or replace them.
Source: https://arxiv.org/pdf/2106.03711
Think we got it wrong?
Ruled-out problems can come back. Tell us what we missed.
Members can read the full record for this problem, check its sources and comment. Open the member view →