Help us verify the rewritten WOMBAT

WOMBAT has been rewritten to enable it’s use in both MOM5 and MOM6. Before we release the rewritten version of WOMBAT, ACCESS-NRI needs your help to confirm that it’s behaving as expected.

ACCESS-NRI have run two 1-deg repeat-year-forced ACCESS-OM2 experiments that differ only in the version of WOMBAT they use:

  • om2_1deg_jra55_ryf_bgc - uses the old version of WOMBAT (as used in prior ACCESS-OM2-BGC simulations). This experiment was run using this configuration.
  • om2_1deg_jra55_ryf_wombatlite - uses the rewritten version of WOMBAT. This experiment was run using this configuration (which is the same as above except for changes required to use the new WOMBAT).

Both experiments were run for 100 years and have sea-ice BGC is disabled.

Where are the data?

Outputs from these experiments can be found on Gadi at:

/g/data/zv30/model_evaluation/ACCESS-OM2/generic_WOMBATlite_v1/om2_1deg_jra55_ryf_bgc
/g/data/zv30/model_evaluation/ACCESS-OM2/generic_WOMBATlite_v1/om2_1deg_jra55_ryf_wombatlite

Please join the zv30 project on NCI to access these data.

The outputs include all WOMBAT diagnostics at monthly frequency.

Intake-esm datastores for the data are available in the above directories to simplify loading the data. This notebook demonstrates using the intake-esm datastores to compare the climatological means of every WOMBAT diagnostic between the two experiments.

What can you do?

We need BGC-inclined folk to help us confirm that the rewritten WOMBAT is behaving as expected. For example:

  • Take a look at the plots in the notebook linked above and tell us if something doesn’t look right
  • Suggest additional diagnostics you’d like to see
  • Dive into the data yourself and check your favourite WOMBAT diagnostics
  • Anything else - it’s all helpful!

Changes to be aware of in the rewritten WOMBAT

Some changes to WOMBAT were made during the rewrite out of necessity or convenience:

  • The units of many diagnostics are different (including the prognostic tracers which are in mol/kg).
  • Surface flux diagnostics are named <tracer>_stf (e.g. dic_stf) instead of stf0##
  • There are some unavoidable changes that mean the rewritten WOMBAT can’t exactly reproduce the original:
    • New sinking scheme (now a fully implicit upwind advection scheme)
    • New available light scheme (SW penetration decays with depth based on the ocean opacity)
    • New order of operations
3 Likes

Hi everyone, Dougie,

Looks great. The comparison shows minor differences between hard-coded and generic tracer WOMBAT in the primary production and phytoplankton-related fields, likely because of the unavoidable differences in the sinking schemes.

For me, this is an excellent place to launch from. We now can easily move into MOM6 generic tracers.

Thanks for all your hard work!

pearse

1 Like

Had another look, this time making my own figures and looking at deep properties after 100 years. I compared all the major tracers.

“generic” = generic tracer version of WOMBAT
“old” = previous hard-coded version of WOMBAT

Generic WOMBAT outputs tracers in mol/kg
Old WOMBAT outputs tracers in mmol/m3

I converted generic WOMBAT outputs to mmol/m3 by multiplying them by 1000 x 1000 x 1.026

I then plotted Generic minus Old WOMBAT (see below).

Summary:

  • There is less DIC and Alkalinity in the generic WOMBAT at all levels
  • There is an accumulation of O2 in the mid-depth (1000 m) North Pacific and loss of O2 elsewhere
  • There is a loss of NO3 where increases in O2 are most intense, otherwise there appears to be a redistribution of NO3 from water masses at 1000 metres to water masses closer to the surface (250 metres). This matches what happens to O2.
  • There is a greater utilisation of Fe at the surface, an increase in detrital, phytoplankton and zooplankton carbon in the upper ocean, meaning an increase in productivity of ecosystems.

My intuition:
I think the slight changes that were inevitable in porting to the generic tracers has caused a weaker transfer of organics to deep water masses, increasing the nutrient contents held at the surface, which has increased primary production. The C storage potential of the ocean is often inverse of the rate of primary productivity, which explains the decrease in DIC. The decrease in Alkalinity, may also be a consequence of this, since less CaCO3 is being transferred to depth and adding less alkalinity back in the abyss. Just an educated guess at this stage though.

2 Likes

Thanks for taking a deep look @pearseb! I guess to properly convert between the two units requires the spatio-temporally varying density. I don’t know how much difference this makes. If it’s the case that users will always prefer units per volume rather than per mass, we could convert the units in generic WOMBAT using the ocean densities prior to outputting the diagnostics?

For those interested, I’ve also run a 1-deg repeat-year-forced ACCESS-OM3 experiment that uses the rewritten version of WOMBAT. This experiment uses this configuration and uses the same grid and forcing as the ACCESS-OM2 experiments. Please be aware that ACCESS-OM3 and related configurations are still under active development.

Where are the data

Outputs from the 100-year ACCESS-OM3 experiment can be found on Gadi at:

/g/data/zv30/model_evaluation/ACCESS-OM3/generic_WOMBATlite_v1/om3_1deg_jra55do_ryf_wombatlite

As above, an Intake-esm datastore for the data is included in this directory. The notebook linked above has been extended to include the ACCESS-OM3 data.

2 Likes

Note, one of the directories in the above paths has been renamed and the data now reside at:

/g/data/zv30/non-cmip/ACCESS-OM2/generic_WOMBATlite_v1/om2_1deg_jra55_ryf_bgc
/g/data/zv30/non-cmip/ACCESS-OM2/generic_WOMBATlite_v1/om2_1deg_jra55_ryf_wombatlite
/g/data/zv30/non-cmip/ACCESS-OM3/generic_WOMBATlite_v1/om3_1deg_jra55do_ryf_wombatlite

The Intake-esm datastores and example notebook have been updated accordingly.

I just had a look and the density used in the conversion factor does seem to make a difference. As an example, if I use density 1035 like was in the notebook you provided the new wombat lite in OM2 has more surface DIC than the old version in some regions, whereas if I use the 1026 like Pearse then as he showed, surface DIC is decreased globally in the new version. I couldn’t find the density field for the runs to try with the variable density field, maybe that is just me not knowing how to use the new intake catalogue yet!

2 Likes

Thanks for taking a look @Lizzie! I’m guessing we would want the MOM5 diagnostic rho which unfortunately is not saved as part of the standard suite of diagnostics for ACCESS-OM2 (and so is not in these runs). Even if we had it, I think the conversion would be inexact using temporally-averaged quantities.

As above, if it’s the case that users will always prefer units per volume rather than per mass, we could convert the units in WOMBATlite prior to outputting the diagnostics?

As for feedback on these runs, we’re most interested in any differences in the output from the rewritten WOMBAT that might indicate an issue or bug.