Gulf stream late separation in 0.1deg IAF

[aekiss]

Gulf Stream in the 0.1deg IAF run used in the paper looks good at first, separating consistently at Cape Hatteras (35N): but starting about 2010 it starts to occasionally cling to the shelf for some distance before separating, and from about 2012 onwards it is separating several degrees too late more often than not: (these examples show the PV at about 300m depth)

[aekiss]

from p36 of the submitted paper:

The mean Gulf Stream separation latitude in ACCESS-OM2-01 is correct, but it is more variable than observed, occasionally separating 2–3◦ too far north; this appears as excessive SSH variability shortly after separation and broadening of the separated jet in Fig. 22c relative to observations. In addition, the transport is characterized by an overly strong anti-cyclonic recirculation near the separation (35◦ N, 70◦ W), whereas observations indicate a more uniform flow towards the Grand Banks. None of these models adequately captures the north flow of the North Atlantic Current along the east side of the Grand Banks, resulting in significant biases in the sea surface temperature and salinity (Figs. 10 and 11). Large biases in North Atlantic tracer distributions and air-sea fluxes are commonly attributed to the misrepresentation of these flows (Bryan et al., 2007). ACCESS-OM2-025 and ACCESS-OM2-01 also simulate the Loop Current (although with lower variability than the altimetric estimate), which flows through the Florida Straits to join the Florida Current and Gulf Stream. In contrast, the Caribbean circulation is incorrect in ACCESS-OM2, with the gyre circulation closed primarily via the Bahamas rather than the Florida Straits. Transport between Florida and Grand Bahama Island at 27◦ N is 23.4 Sv in ACCESS-OM2-025 and 20.4 Sv in ACCESS-OM2-01, significantly weaker than the observed 32.1 Sv (Meinen et al., 2010). This reduced inertia may contribute to the poor Gulf Stream separation, as seen in idealised experiments by Özgökmen et al. (1997), but more investigation is required.

[aekiss]

Some more plots (from these notebooks):

First up, here's the mean surface current and barotropic streamfunction over successive 4-year intervals. The separated Gulf Stream starts off looking OK (top row) compared to obs (bottom right) but then contracts and clings to the shelf before separating several degrees too late. The mean over all intervals is shown in bottom left. The excessive variability after separation blurs out the separated current in the mean compared to obs.

This Hovmoller shows the Gulf Stream migrating northward at 72W

The SSH variability starts off looking relatively OK (top row) compared to obs (bottom right) but then contracts and greatly weakens. In the final 4-year intervals the variability is weak post-separation, indicating the erroneous late separation path is relatively stable. In contrast, the variability over the full interval (bottom left) is far too strong near the coast, reflecting variation on timescales longer than 4 years.

The Florida Current transport (between Florida and Grand Bahama Island at 26.7 N) starts well below the observed 32.1 Sv (Meinen et al., 2010) and gradually declines further:

This reduced inertia may contribute to the poor Gulf Stream separation, as seen in idealised experiments by Özgökmen et al. (1997), e.g. their fig 5a, 9 which differ only in the strength of Ekman pumping:

I suspect the wind forcing is too weak, possibly due to calculating the stress using relative wind with the JRA55-do wind product which has already been adjusted to match time-mean scatterometer and radiometer winds, which are relative to the ocean surface current.

[StephenGriffies]

Great analysis @aekiss .

One thought that comes to mind is that the GS separation is degrading over time due to a degradation of the deep western boundary current, which is ultimately related to a degradation of the Nordic Sea overflows. Studies from Rong Zhang show that in these sorts of models, a strong DWBC is necessary for maintaining a good GS separation. See in particular this paper

https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2011JC007240

One suggestion is to examine the speed of the circulation at a few depths, such as 2000m, to see how the DWBC evolves. As per this paper, a weak DWBC can be caused by a poor Nordic Sea overflow. Her paper focuses on MOM025 resolution. Even so, we found that in CM2.6 (MOM01 horizontal resolution), the overflows are weak, thus contributing ultimately to a poor GS separation.

One question: is this MOM01 simulation using the 50level or 75level configuration? I wonder if the improved AABW in the 75 level configuration translates into a better Nordic Sea overflow...

[aekiss]

Thanks for the suggestions @StephenGriffies, I'm following them up.

Here are zonal transects of meridional velocity at 37N:

and at 35N:

The undercurrent does weaken, as you suspected.

[aekiss]

Sorry @StephenGriffies, I meant to say: it's the 75-level configuration

[aekiss]

The meridional velocity at 2000m shows a weakening of the undercurrent:

[StephenGriffies]

Thanks for this new plan-view images. I suggest showing into the subpolar gyre to see, as I suspect, that the overflows are weakening in concert with weakening DWBC.

[XihanZh]

Thank you @StephenGriffies and @aekiss. I did a few plots following your analysis. The four-year mean of meridional velocity at 3000m shows the overflows are weakening: I then plotted the transects of meridional velocity at 45N from 50W to 35W and the strength decreases throughout the water column. In addition, the AMOC also decreases with time in this run as it is shown in Fig.8 (a) in the submitted paper, and these processes do seem to be consistent with the mechanism in Zhang et al,2011.
I also looked at the transport of the Gulf Stream. Here is the Hovmeller of vertically integrated velocity at 75W. Picking the latitude from 33N to 36N to plot the time series of Gulf Stream transport: There is a clear downward trend of the transport.