Water conservation

[MartinDix]

Monthly total water mass and equivalent sea level change from cm3-run-18-09-2024-no-zap-no-iceberg

Even after accounting for snow accumulation (missing iceberg flux) there's a sea level decrease of ~ 3 mm/year.

[MartinDix]

From cm3-run-24-09-2024-fix-wind-stress

The seaice term now includes snow over sea ice Land water includes soil water and river storage

The river storage increases over the first year, but even with this included the decrease in the first year is larger.

[MartinDix]

River outflow from the UM matches river inflow in MOM so there's no error from the scaling by ocean fraction.

Rate of water loss is ~ 4e7 kg/s which is ~ 2.5% of the total river term.

UM has l_inland=.true. which means that the river flow which ends up in inland basins should be compensated.

[MartinDix]

UM is using ancillaries from /g/data/tm70/mrd599/ancil/n96e_momO1_20201102, created by suite u-cp883. The river sequence file qrparm.rivseq puts ocean outflow points in the Black and Caspian Seas which are 100% land in the 1 degree ocean model. Therefore the river outflow to these points is lost. These points were fixed in the CM2 river sequence file.

In these plots river outflow points are shown in blue, inland basin in brown and blocked outflow points in red. Original on left and fixed on right. Changed points highlighted with thicker arrows.

Outflow to these blocked points was ~ 2.5e7 kg/s, so perhaps not the total solution.

/g/data/tm70/mrd599/ancil/n96e_momO1_20201102/qrparm.rivseq is now the corrected version so any new runs will pick up this change.

[kieranricardo]

@MartinDix nice catch!

The new ancillary seems to generating NaN in the river outflow

[MartinDix]

Something went wrong in handling of masked points in the updated file. Corrected updated version is /g/data/tm70/mrd599/ancil/n96e_momO1_20201102/qrparm.rivseq.20241001. qrparm.rivseq is the original file.

[kieranricardo]

@MartinDix having a look at cm3-run-15-10-2024-revert-zapping and cm3-run-15-10-2024-river-seq-update, the run off has definitely improved. But for some reason the total water budget barely improves, and the sea level change actually gets worse (driven by a decrease in meltwater).

	cm3-run-15-10-2024-revert-zapping	cm3-run-15-10-2024-river-seq-update	$\Delta$
Runoff	114.9 mm/yr	118.9 mm/yr	4.0 mm/yr
Sea level change	-12.1 mm/yr	-16.7 mm/yr	-4.6 mm/yr
Meltwater to ocean	16.4 mm/yr	11.2 mm/yr	5.2 mm/yr
Total water change	-10.25 mm/yr	-9.5 mm/yr	0.75 mm/yr

[MartinDix]

Theriv-seq-update run looks a bit better in the second year. The larger rate of decrease in the first year suggests we're still missing some sort of storage term.

[MartinDix]

Equivalent sea level change over the second year shows a clearer benefit from the change to the river sequence.

However there's still something odd with the atmospheric budget even with the river fix. Calculating net flux to atmosphere as ocean evaporation + sublimation - ocean precipitation - river outflow and total water as soil water + snow + river water + atmospheric water shows an imbalance. Total water calculated from the instantaneous fields in the restart files, so difference between months should match the monthly average fluxes.

This difference is equivalent to 1.7 mm/year, so close to the sea level change.

[MartinDix]

The JULES land surface scheme includes a 3 m thick groundwater layer below the usual 4 soil layers. This was missing from my calculation of the soil water content. For some reason the diagnostic variable for this is the saturation fraction rather than the actual water amount, STASH variables

0 |  281 |SATURATION FRAC IN DEEP LAYER 
8 |  250  | SATURATION FRAC IN DEEP LAYER   

With this included the imbalance over year 2 is 1.9e6 kg/s = 0.12 mm/year.

The same calculation in a 40 year atmosphere only run with JULES gives a difference of 0.14 mm/year.

This is still larger than I'd expect.