Fix ssh incrementation in VVL code (Met Office)

[drew]

Context

Code improvements and revisions required for FOAM assimilation system in GO6

Analysis

Initial attempts to increment ssh directly as a change to the surface eta resulted in the introduction of fresh-water blotches into the ocean, presumably due to increased volume added with no heat/content change.

Fix

Chris Harris has devised a scheme to implement the ssh change though out the water column through changes to e3t. Associated changes to heat and salt content have also been added.

[drew]

Here is a Met Office intranet accessible html formatted version of an early email detailing the problem

• ​http://www-hc/~haddp/FreshWaterBlotches.html

[drew]

Under the original ssh incrementation code

• https://forge.ipsl.jussieu.fr/nemo/browser/trunk/NEMOGCM/NEMO/OPA_SRC/DYN/sshwzv.F90?rev=5518 (ll 126--132)

freshwater blotches developed in the southern ocean -- and probably elsewhere. This is manifest in my mi-am200/1/2 set of runs for which I will now show some evidence from mi-am201.

[drew]

SLA increment for 2nd day

[drew]

Difference in SSS when adding ssh increment

[drew]

Difference in SST when adding SLA increment

[drew]

Here is the sea level anomaly increment on the 2nd day of one of the offending runs

While here is the corresponding difference in SSS when adding that increment

There are also changes to the temperature, but it is harder to associate that directly to the sea level anomaly increment

[drew]

Some fairly obvious freshening when volume is added through the ssh. Likely associated with no change in salt content. Not as clear what is happening with heat content. Freshening also was obvious in mean profile bias. Shown below are the mean and rmse bias for the new VVL code followed by bias in GO5.

[drew]

Hovmuller stats for run with VVL causing freshwater blotches.

[drew]

Hovmuller stats for GO5 run

[drew]

Hovmuller stats for run with VVL causing freshwater blotches.

[drew]

In an attempt to counteract this, salt and heat content changes were introduced to balance the volume change.

• [6525]

While the freshwater blotches disappeared -- or at least were reduced in this revision.  The SSH mean bias in the North Pacific was greatly reduced in this revision -- which was subsequently associated with an inability of the model to retain the sea level increment. 

[drew]

A Met Office intranet review of what happen when implemented in mi-am700/1/2 set of runs is documented here

• ​http://www-hc/~haddp/PAPERS/FOAMSLA.pdf

which I will summarize below.

[drew]

Binned obs minus background stats for 1995/03 in mi-am200.

[drew]

Increments for March 16 of mi-am700 run

[drew]

Model Tendency for 16 March of mi-am700 run.

[drew]

Dan Lea observed that the obs-bkg for analysis runs had large error in the Pacific -- here for mi-am700 run in March 1995.

Further, plotting the increments and model tendency for 16 March

showed that the correction to this error was present in the increments plots, but did not carry forward into the model tendency for the day (i.e. the increment was not retained).

It was thought this may be due to the creation of spurious model convergence/divergence in the process of pushing the increment -- applied at the surface -- down through the water column through the adjustment of e3t in the VVL code. Therefore, Chris Harris further revised the sshwzv.F90 code to recast the ssh incrementation as a increment on the full column e3t in a manner similar to what is done for the addition of freshwater.

• https://forge.ipsl.jussieu.fr/nemo/changeset?reponame=&new=6608%40branches%2FUKMO%2Fdev_r5518_sshinc_with_VVL%2FNEMOGCM%2FNEMO%2FOPA_SRC%2FDYN%2Fsshwzv.F90&old=6526%40branches%2FUKMO%2Fdev_r5518_sshinc_with_VVL%2FNEMOGCM%2FNEMO%2FOPA_SRC%2FDYN%2Fsshwzv.F90

• through [6589], [6605], and [6608]

• The code is functional at

• source:branches/UKMO/dev_r5518_sshinc_with_VVL/NEMOGCM/NEMO/OPA_SRC/DYN/sshwzv.F90@6605​

• source:branches/UKMO/dev_r5518_sshinc_with_VVL/NEMOGCM/NEMO/OPA_SRC/DYN/sshwzv.F90@6608​

[drew]

Additional changes to trasbc.F90 to properly conserve heat and salt content where added correctly at

• source:branches/UKMO/dev_r5518_sshinc_with_VVL/NEMOGCM/NEMO/OPA_SRC/TRA/trasbc.F90@6660​ [6620] [6660]

[drew]

These changes have been subsequentedly added to the NEMOVAR community branch

• ​http://fcm3/projects/NEMO.xm/log/branches/UKMO/dev_r5518_v3.4_asm_nemovar_community
• source:branches/UKMO/dev_r5518_v3.4_asm_nemovar_community [6754]

[drew]

To date, there seems to be little difference between adding the trasbc changes or not. Perhaps Ana has more comprehensive evidence.

Salinity profile obs minus background stats for run with trasbc changes [6608] and without [6660].
Note: A transition was made in mid-run between [6620] and [6660] of run with trasbc (black). Will be replaced with cleaner comparison.

[drew]

Profile stats for run with trasbc changes (black) and without (blue)

[nemo]

Looks like an internal issue within the Met Office.

[nemo]

Reassigned to UKMO officer as the current owner does no longer have the permissions to close tickets.

[mathiot]

Has been included in the 2017 merge @ revision 9023