MOC reversal in the Nordic Seas

[gustavo-marques]

The issue

During the CESM3 PI control procedure, we noticed a reversal in the AMOC circulation within the Nordic Seas, as summarized in the slide below:

Mixed Layer Eddies

We have identified that this reversal is being driven by the Mixed Layer Eddies (MLE) parameterization. There is positive feedback between the MLE overturning and convection, and eventually, the residual AMOC is dominated by the MLE overturning. We do not yet fully understand this feedback. However, we have implemented a fix by tapering BLD and MLD (h and H) used in the MLE's streamfunction, which seems to prevent this issue (i.e., stop the feedback).

Taper

H' = H / (1 + (H/D)^P)

H' is a saturation curve that leaves small values untouched and progressively compresses large ones.

Parameters

• D — depth: Sets where bending begins.
• P — power: Sets how hard it bends.

Example with D=900 and P=4:

Animations

Below are animations of the annual mean AMOC resulting from MLE across different simulations.

349 (no taper)

Reversal in the circulation within the Nordic Seas is present from year 1.
❌ AMOC reversal

349_amoc_mle.mp4

360a (D=1500, P=4)

✅ No reversal

360a_amoc_mle.mp4

365 (no taper initially, then added D=1500, P=4 @ year 19)

There was no MLE taper until year 19, and the reversal is present up to this year. Notice the big difference between years 18 and 19.
Before year 19 ❌ AMOC reversal
After year 19 ✅ No reversal

365_amoc_mle.mp4

366 (D=2000, P=4)

✅ No reversal

366_amoc_mle.mp4

367 (D=3000, P=4)

❌ AMOC reversal

367_amoc_mle.mp4

368 (D=4000, P=4)

❌ AMOC reversal

368_amoc_mle.mp4

[gustavo-marques]

Updating 366 up to year 68. Still showing no reversal.

366 (D=2000, P=4)

✅ No reversal

366_amoc_mle.mp4