Impact of thinner upper ocean layer on SST variance

[islasimpson]

Periodically we look at plots like the ones below and question why we have more variability now in our SSTs. This is interannual variability of DJF seasonal means. In CESM3 we have more greens over the South Pacific and the mid-latitude oceans as well as red around the Kuroshio and the Gulf Stream and the south of Greenland.

Below is a comparison of DJF variance in TS on the atmosphere side, the upper layer of MOM, and the average of the upper 4 layers in MOM between 247 and CESM2 piControl and ERSSTv5. I used the "native" MOM history files and simply averaged the top 4 layers (I think that's OK because each of these layers look like they represent 2.5 m depth).

In panels (a) we can see the bias in the variance of TS relative to ERSSTv5 which reproduces the fact that we see more variance in CESM3 compared to CESM2. Panel (d) then shows the variance in the first layer of MOM. Note that this is NOT the same as TS on the atmosphere side e.g., see the region to the west of the UK - there are pretty obvious differences between panels (d) and (a). For the impacts of the thinner upper-most layer, panel (e) shows the variance in the average over the first 4 MOM layers (10m). (f) is the difference between the variance in the upper most layer and the variance in the average over the upper 4 layers. The difference is very small. With a finer contour interval you can see that there are differences (g) but they're really small, and this doesn't explain the difference between CESM3 and CESM2. So this difference in SST variance between CESM3 and CESM2 is a real difference and doesn't seem to be related to the fact that SST is now defined based on a thinner surface layer.

Note that I think the contouring in the CVDP plots might make it looks worse for CESM3 than it really is. CESM2 is biased low in many places while CESM3 is biased high. Biased low in the first plot above as a lighter blue than the dark blue of the obs whereas bias high the contouring changes to green, which I feel makes it look much worse for CESM3 than it does for CESM2. CESM3 is more biased in the South Pacific, but it is less biased in other places e.g., parts of the North Pacific.

[islasimpson]

Here is the difference between the variance in TS and the variance in the top ocean layer. It would be good to make sure that we expect this. I'm using potential temperature in the ocean.

@gustavo-marques - do you know if we expect the temperature of the top layer of the ocean to be equal to the surface temperature on the atmosphere side or are they different? I think TS on the atmospheric side is backed out from the radiative flux, so maybe it's expected.

[islasimpson]

@swrneale I'm a bit confused about why MPAS is coming into this. The comparison above is between the upper layer ocean temperature and the TS (radiative temperature). Isn't that the same comparison as you're looking for except within CESM? Would it be helpful to just see the climatological version of this as opposed to the standard deviation to address your question?

[swrneale]

Oh sorry, I meant MOM. Too many models.

[islasimpson]

Oh ok. That's what the above is (I think). I took the top layer temperature from MOM, regridded onto the CAM grid and compared with TS. Do you want to see the mean climatology version of this or am I still misunderstanding?

[swrneale]

Oh right sorry, didn't quite appreciate that. This implies that CAM must know something about ocean emissivity in order to recover the same TS as MOM from the lowg-wave up. I'll check.

[islasimpson]

Just to clarify - the top layer ocean temperature and CAM's TS are not the same in the above plots. But the differences are very small.