Noise and discontinuities in WOA

[aekiss]

Just splitting this off as a separate issue from #74 to document it now that it's fallen off Slack.

World Ocean Atlas 2013 v2 has spurious discontinuities and noise in salinity and temperature (and possibly other fields), particularly in the 0.25deg product.

For example, SSS has >3psu discontinuities in the Kara Sea in the monthly climatology woa13_decav_s01_04v2.nc.

These move around and disappear in different months and extend down to 15m.

Discontinuities and noise also appear in near-surface temperature in the same pattern:

In addition to the discontinuities there are large areas with pixel-scale noise in both SSS and SST.

This is apparently an artefact of the interpolation method, as this region seems to be entirely unsupported by observations in January.

There are also large-scale differences between the 1deg and 0.25deg products in the Arctic, for example in the Beaufort Sea and around Severnaya Zemlya, and both also display differing spurious fonts (e.g. in the Beaufort Sea) - click on this image to enlarge:

I informed NOAA about this in January. As of late June it had not been investigated but they said they'd look into it. They are working on a release of WOA18 scheduled for September - hopefully it won't contain these flaws. In the meantime our WOA13v2-based restoring field will need smoothing - see #74.

[aekiss]

Unfortunately the same problems afflict WOA2018-prerelease salinity (https://www.nodc.noaa.gov/OC5/woa18/woa18data.html, downloaded here: /short/v45/aek156/woa18-prerelease).

[aekiss]

Simon suggested we check whether the PHC climatology is better in the Arctic: http://psc.apl.washington.edu/nonwp_projects/PHC/Climatology.html
The website says "PHC = WOA (Levitus) '98 everywhere except in our arctic domain, where we have blended in the AOA field (from EWG), and the BIO data to produce a more realistic arctic region."

This has global monthly T & S climatologies, but only at 1 degree resolution.

[aekiss]

However I gather that WOA13v2 is better than WOA98 in terms of Southern Ocean polynyas so perhaps that rules out PHC...?

[aekiss]

Noise and discontinuities are also present in WOA23 temperature (top) and salinity (bottom) in /g/data/ik11/observations/woa23:

[aekiss]

Also see COSIMA/access-om3#161 (comment)