Water conservation again

[MartinDix]

Suites are currently using ancillaries from /scratch/tm70/kr4383/cylc-run/ancil-gen-22-04-2026/share/data/n96e_mom025_20260316 and
/scratch/tm70/kr4383/cylc-run/ancil-gen-22-04-2026/share/data/n96e_mom025_20260316/qrparm.landfrac.

The qrparm.rivseq files here have some blocked outflow points which is likely affecting water balance. They're also missing the Black Sea - Mediterranean connection

Unfortunately at the moment it's a manual correction for these. See also #48.

[anton-seaice]

Might be able to use remapping weights like

https://github.com/ACCESS-NRI/om3-scripts/blob/main/mesh_generation/generate_rof_weights.py

to ensure all runoff gets moved into the ocean (as long as the runoff reaching the mediator isn't lost to a land mask)

[kieranricardo]

@MartinDix the suites are still using your fixed river sequence from a while back: /g/data/vk83/prerelease/configurations/inputs/access-cm3/ancil/n96e_momO1_20201102/qrparm.rivseq.20241001, but all the other ancillaries from /scratch/tm70/kr4383/cylc-run/ancil-gen-22-04-2026 are being used.

I haven't checked whether this matches the new land mask (I'm not sure how the UM calculates the landfraction on the river grid), but we're not getting the massive water conservation errors we used to have.

@anton-seaice yeah I think your mapping will automatically fix this! I think the UM still outputs the river runoff if the outflow points are land, it just gets lost in cap/mediator.

[anton-seaice]

Yea I think so

[MartinDix]

Current runs show a large decrease in ocean mass, even when accounting for changes in sea ice and land water (note that snow accumulation on land is now handled by the iceberg scheme, so don't have to include this, c.f. #33).

There are minor problems with the river sequence file which means that some high N latitude rivers don't reach the ocean. However I think these are now handled by inland basin corrections so don't affect the overall water balance..

The main problem is a mismatch between the river outflow from the atmosphere diagnostics and the river inflow from the ocean diagnostics. If the difference between these is added to the ocean mass the balance looks good (apart from the mysterious small residual annual cycle).

[MartinDix]

@kieranricardo suggested the problem was a mismatch between the river spreading and the updated land sea mask. A run w/o the spreading removes the difference between the two river flow diagnostics and the sea level now looks good (still with the mystery annual cycle)

[MartinDix]

100 km run with CABLE and no spreading shows a very similar water balance, again with the residual annual cycle

[MartinDix]

@kieranricardo notes that the sea ice ponds have an annual cycle equivalent to about 3 mm/year.

Only the snow accumulation over the permanent ice points goes to the iceberg scheme. The snow elsewhere on land should be included in the total mass, but was missing in the above calculations. Including it decreases the residual substantially.

Terms are

• ocean mass
• sea ice
• snow on sea ice
• meltponds on sea ice
• soil water (+ groundwater for JULES)
• snow on land (excluding permanent ice points)
• atmospheric water
• water in rivers

[MartinDix]

With a correction to the river storage term to exclude outflow points