Temp / salt comparison with observations

[hrsdawson]

Compare against the Pathenet et al. 2021 climatology for upper ocean temperature and salinity, and against Schmidtko et al. 2014 for bottom shelf temperature.

Previous issue using old simulations is #63

[hrsdawson]

Here's a comparison of bottom temperature and salinity against Schmidtko et al. (2014) on the shelf and WOA in the abyss. The experiments are compared to the Control, not observations. Note: Will need to update this to use the new WOA23 observations when available on Gadi, however there is little change in biases when using WOA13 and WOA18, so I don't expect much change when updating.

In summary:

• Basal freshens the bottom shelf and warms the Amundsen Sea relative to Control.
• BasalGade has the opposite effect - mostly salinifies the bottom shelf and cools the Amundsen Sea relative to Control.
• BasalGadeBrine has a similar effect on bottom salinity, but the cooling in the Amundsen, relative to Control, is reduced/fixed.

[hrsdawson]

Here's the same figure but with all experiments compared to observations. It's a bit hard to tell whether BasalGadeBrine is is closer to reality than the Control. The temperature biases in the Amundsen are reduced slightly in BasalGadeBrine compared to Control, but the cool biases in the Bellingshausen Sea are larger. Similarly, the salinity biases in the Amundsen in BasalGadeBrine are slightly reduced compared to the Control, but we have a larger salinity bias in the Ross Sea.

I still need to compare the upper-ocean changes against the Pauthenet climatology.

[aekiss]

Hi @hrsdawson I've downloaded WOA23 to /g/data/ik11/observations/woa23, but I'm not sure if it will need further processing to be useable for your comparison. What WOA18 files did you use?

[hrsdawson]

Thanks @aekiss! I was interpolating the objectively analyzed climatology (annual mean from 1955 to 2017) onto the model grid. But perhaps I should be doing it the other way around. Conservatively interpolating the model onto the coarser obs grid?

[aekiss]

What directory on Gadi are you getting WOA18 from?

/g/data/ik11/observations/woa18 is climatologies for the 2005-2017 period (A5B7), not 1955-2017.

[hrsdawson]

I have the woa18_decav_t00_01.nc and woa18_decav_s00_01.nc files on my gdata here: /g/data/e14/hd4873/obs_data/woa/woa18/. Probably not very efficient to have them stored in my own directory. I'm updating the analyses to use the new WOA23 now.

[aekiss]

I was interpolating the objectively analyzed climatology (annual mean from 1955 to 2017) onto the model grid. But perhaps I should be doing it the other way around. Conservatively interpolating the model onto the coarser obs grid?

I'm not sure it matters too much whether you interpolate model to obs or vice versa, but I guess the fairest comparison is to interpolate to whichever is the coarser grid.

Conservative interpolation is appropriate for when you want to close a budget or maintain some integral/average, but 1st order conservative is blocky and 2nd-order can give spurious extrema, so in other cases smoother interpolation may be better.