River and iceberg runoff mapped to incorrect locations

[aekiss]

The river and iceberg fluxes are getting mapped to the wrong locations (in the middle of oceans instead of coastlines). Some of these locations are common between rivers and icebergs, and some differ. In both cases the patterns seem constant in time (which is odd).

Here's the liquid (river) runoff friver from a 1deg_jra55do_ryf experiment

and the solid (iceberg) runoff ficeberg

I'm guessing this affects all configurations, but I haven't checked. Presumably something to do with DROF or the mediator.

[aekiss]

somewhat related #37

[aekiss]

All the fields in drof.streams.xml and datm.streams.xml use the same

   <meshfile>./INPUT/JRA55do-ESMFmesh.nc</meshfile>

but RYF.friver.1990_1991.nc and RYF.licalvf.1990_1991.nc have lat = 720, lon = 1440, whereas all the other inputs in /g/data/vk83/experiments/inputs/JRA-55/RYF/v1-4/data/ have lat = 320, lon = 640.

So I guess we need a different meshfile for friver and licalvf?

[aekiss]

I used generate_mesh.py to make a new meshfile with

generate_mesh.py --grid-type=latlon --grid-filename=/g/data/ik11/inputs/JRA-55/RYF/v1-4/RYF.friver.1990_1991.nc --mesh-filename=/g/data/v45/aek156/github/COSIMA/om3-scripts/mesh_generation/JRA55do-drof-ESMFmesh.nc

This does a much better job (almost all the flux is at coastlines), but on a nonlinear, detailed colour scale (different from pictures above) there are still some artefacts visible, typically <1e-7kg/m^2/s. [edit - see below]

[aekiss]

The artefacts are not present in the original JRA55do files

[aekiss]

False alarm - the new JRA55do-drof-ESMFmesh.nc actually fixes the problem. The artefacts were lingering in the first month (is this to be expected with <tintalgo>upper</tintalgo> time interpolation?) and are absent in the subsequent month.

[dougiesquire]

Great catch @aekiss! I'm pretty surprised this ran at all with a mesh of the wrong size...

Will you submit PRs with the new meshes? If so it would be good to also update the filename of the old mesh also to include datm in the name.

[anton-seaice]

Good work Andrew - I am also surprised this ran!

I think we could add mesh checks to the driver ? i.e. to compare between the mesh and the data file. Similar to those in the MOM6/CICE6 drivers

[anton-seaice]

Is all the run-off entering the ocean around the coastline the expected behaviour? Or should we disperse it away from the coastline?

[aekiss]

Good idea @anton-seaice, a check in the driver would be good if it's easy to implement.

[aekiss]

Good question re. runoff redistribution, I was thinking the same thing. I'm not sure whether/how that is handled in ACCESS-OM3.

ACCESS-OM2 uses a runoff flux limiter that conservatively spreads the runoff over a sufficiently large area to keep the flux below a prescribed cap. All resolutions have a global cap of 0.03kg/m^2/s, and 0.1° also has a local cap of 0.003kg/m^2/s at some Arctic rivers to avoid excessively low salinity there. These caps are below the ~0.1kg/m^2/s we get at the Amazon mouth in MOM6-CICE6 1deg_jra55do_ryf (see above) so we should check whether that large flux is causing any problems.

Vertical distribution is another thing to consider. ACCESS-OM2 uses river_insertion_thickness = 40.0m rather than enhancing diffusivity at river mouths. MOM6 has some rivermix options. It's not clear to me whether we're using them.

[aekiss]

I'm also not sure what's done if the JRA55do runoff is located on land on the model grid. Is it teleported to the nearest wet cell?

[aekiss]

We need to check whether the area-integrated friver and ficeberg MOM6 diagnostics match the area-integrated fluxes in the RYF.friver.1990_1991.nc and RYF.licalvf.1990_1991.nc JRA55do source files.

[aekiss]

We also need to switch from <mapalgo>bilinear</mapalgo> to <mapalgo>conserve</mapalgo> for all fluxes (i.e. not temperature, humidity, velocity etc) in drof.streams.xml and datm.streams.xml - see #37. The area-integrated fluxes won't match the inputs until we do this.

I think the area-integrated flux check requires that we retain the current time interpolation setting <tintalgo>upper</tintalgo> too (rather than switching to <tintalgo>linear</tintalgo>).

[ofa001]

Probably premature to think about how icebergs will be handled in ACCESS CM3 here but ESM1.6 is implementing a version of the iceberg scheme we had in ACCESS-CM2, Also at last weeks ACCESS-NRI ESM meeting there was a discussion of the CMIP7 plan for additional freshwater/icebergs approach thats being suggested for the Fast track 'historical' runs so its an active topic of discussion and may well be at the ACCESS-NRI meeting.

[anton-seaice]

The COSIMA Spack and NRI Spack are building from the same version of the model source - it shouldn't matter which order the changes are implemented in. Do you have a preference on which order its done ?

[aekiss]

no preference. I'll struggle to find bandwidth for it this week, so don't let me hold things up

[anton-seaice]

Closed through linked PRs

noting #231 is still open to look at conserving run off correctly.