1/20° topography

[AndyHoggANU]

It looks like we will begin a second attempt at a higher resolution (1/20°) PanAnt configuration. Last time we attempted this we simply sampled the old ACCESS-OM2-01 bathymetry to give the simplest possible 1/20° configuration. But that is probably not ideal in the longterm. Furthermore, @ChrisC28 is doing the same thing in mom6-eac (see COSIMA/mom6-eac#6), so perhaps we should combine methodologies.

Possible options seem to be:

• The block subsampling that we used last time;
• Interpolating the 0.1° bathymetry?
• Sourcing a new and updated bathymetry and starting from scratch.
  It would be nice to think that the 3rd option is possible, but I don't know how possible it is.

Any suggestions or comments? Any other pitfalls we need to think about, especially in matching low resolution boundary conditions?

[adele-morrison]

I vote for:

• Sourcing a new and updated bathymetry and starting from scratch.

Given these are regional domains, hopefully it will be only minimally time consuming if we need to manually fix any problem areas.

Possibly @pedrocol and I could work on this next week when he visits. Except I know nothing about what's needed. @russfiedler or @aekiss, do you have scripts / documentation to share for what was done for the ACCESS-OM2 bathymetry and advice on what we need to do differently for C-grid?

[ChrisC28]

I'm somewhat agnostic about the exact approach to generate the new grid and topo. However, as caught by @aidanheerdegen in MOM6-eac the grid has some odd values that looked like round-off/truncation from the interpolation. Not sure if that's a problem with the pan-Ant config but it should probably be checked.
As I discussed at length with @russfiedler, ideally the high-res grid should "nest" into the low-res grid, so that the T, u, v and q points of the low-res grid sit on top of certain high-res grid points.
I'd be leaning towards starting the topo from scratch. The EAC region has quite highly resolved topography as the various voyages up and down the coast has swathed it to death. This data should make it's way to GEBCO or ETOPO. There are a lot of features, such as submarine canyons, that aren't going to be resolved at 1/10th degrés grid spacing that I'd like to target with my planned higher res runs.

[aekiss]

It would be good to establish a reproducible workflow for the topo generation.
For ACCESS-OM2 the 0.25deg and 1deg topographies have repositories that fully document how they were generated, e.g. https://github.com/COSIMA/make_025deg_topo, and the exact commit used to generate them is recorded in the topog.nc history metadata.
Hand edits were needed for important sills and straits using https://github.com/COSIMA/topogtools/blob/master/editTopo.py but these edits were recorded in topog_edits.txt and form part of the workflow for reproducibility.
This reproducible approach has been very useful when we've needed to tweak the topography.
The topo workflows for ACCESS-OM2 contain checks and fixes to ensure that the topo will work properly on a B grid. It should be a bit more straightforward with MOM6 due to the C grid.

[aidanheerdegen]

Given it is a C-grid it probably makes sense to start from scratch rather than just using the ACCESS-OM2 grid. Is there an existing 0.1 degree C-grid bathymetry for MOM6?

As for generating a new bathymetry, a new Bedmap2 product has just been released

https://www.bas.ac.uk/project/bedmap-2/

It extends up to 60S and is 1km resolution.

Or Bedmachine Antarctica v2

https://nsidc.org/data/NSIDC-0756/versions/2

That is 500m resolution, but only extends to 70S. The Bedmachine data is utilised in GEBCO, so that would be the one to use if that was the preferred to Bedmap2. Using GEBCO means you also have a relatively seamless global version that is compatible with the higher resolution Pan-Antarctic.

Because Paul asked about interpolating the Bedmachine data on to the ACCESS-OM2 grid, I ended up writing up a blog post about using proj4 for this

https://climate-cms.org/posts/2021-11-02-pyproj-regrid.html

In case it is useful.

[ChrisC28]

Perhaps a naive question: what are the main challenges with generating a "good" topography? In MITGCM I interpolate ETOPO onto the relevant grid (T-points) and it's done. I can imagine modifying straits/internal seas/sills etc.... however for the generation of the basic topography, what are the issues we need to be aware of, as it doesn't appear to be a straightforward process (when compared with MITGCM, for example).

[AndyHoggANU]

There's no doubt that the C-grid is much more amenable to naive grid production than B-grid.
Maybe the best thing to do is to take one of these configurations, produce a basic grid at higher res, ensure the boundaries will be well-behaved and then give it a burl?

[aekiss]

Yep, keep it simple to begin with, and then tweak it via scripts on github to keep it reproducible.

@ChrisC28 https://github.com/COSIMA/make_025deg_topo/blob/master/make_topog.sh will give you an idea of what things needed to be done for that model, including

• float_vgrid ensures the vertical grid consists of double-precision values that also exist at single precision
• deseas removes lakes and unwanted seas
• do_partial_cells
• min_max_depth
• apply_mask
  etc

[pedrocol]

I would just add a smoothing prior to masking (it could be a Shapiro filter, I have a fortran code for this) and probably the coast will need hand editing (for extra smoothing and specific topographic features). Also, perhaps it is a good opportunity to extend the horizontal grid further to the south in order to fully include the cavities in a near future.

[micaeljtoliveira]

I would be happy to help with this, in particular with the scripts and making sure everything is reproducible.

[adele-morrison]

Fantastic @micaeljtoliveira! Sounds like we have enough info here to get us progressing with this. Let's work on it together next week when @pedrocol is in town.

[aidanheerdegen]

Also worth bearing in mind what your optimal dimensions (in cells) should be. Maybe it isn't such an issue in MOM6, but MOM5 grids were generally chosen to be divisible by as many factors as possible, to make decomposition as straightforward and flexible as possible.

[adele-morrison]

Here's a summary of the work flow needed for the new bathymetry:

0. Create horizontal grid.
1. Interpolate most recent GEBCO data onto new model grid. Existing script: gen_topo.f90
2. Remove isolated seas. Existing script: deseas.f90
3. Apply minimum and maximum allowed ocean depths. Existing script: min_max_depth.f90
4. Remove potholes. Existing scripts (but can remove code for fixing non-advective coastal cells): fix_nonadvective_mosaic.f90 and check_nonadvective_mosaic.f90
5. Generate the land mask from the topography. Existing script: topog2mask.py
6. Possibly optional: Smooth regional boundary to match forcing model grid.

Some notes on choices that need to be made in this process:

• Which bathymetry data product to use. GEBCO.
• Size of domain. For PanAntarctic we want to extend far enough south to include ice shelves later. Also consider the optimal dimensions for different decompositions of core layouts here.
• Minimum ocean depth. We'll be running z* initially, so aim for ~10 m.

Compared with the work flow for MOM5 bathymetry, we no longer need to do: coastal editing of non-advective cells, editing to ensure minimum thickness of partial cells, calculating single/double precision vertical grids.

[micaeljtoliveira]

As discussed during the meeting, here is the new git repository to collect the pieces and bits of code used for these tasks:

https://github.com/COSIMA/domain-tools

I will start by adding the Fortran tools mentioned in the work flow.

[micaeljtoliveira]

I've now added the Fortran tools (plus a simple CMake based build` system) to the new repository.

[aekiss]

The ACCESS-OM2 topography was based on GEBCO2014, but the latest is GEBCO2021
https://www.gebco.net/data_and_products/gridded_bathymetry_data/gebco_2021/
so if we use GEBCO I think we should use that version as there has been quite a bit of progress in mapping bathymetry since 2014: https://seabed2030.org/mapping-progress

[dpath2o]

Has anyone had a good and thorough look at GEBCO 2021? Others that I work with won't touch it for high resolution (less than a 1/10 of a degree) 3D ocean dynamic modelling. I have correspondence on a secure network that highlights these issues and it is something that I'd like to investigate myself, but it's down there on the priority list at the moment.

[aekiss]

Hm, thanks for the heads up - I haven't looked into it, and was naively assuming newer=better. Can you be more specific about the flaws?

[dpath2o]

Yes, of course. I'll get into my work email tomorrow and bring it into this conversation.

[yuhang559]

Hi, I am Yuhang, a UTAS Honours student. In my Honours project, we compare two high-resolution models (MITgcm) configured with two bathymetry data ( BedMachine Antarctica and SRTM15_PLUS) in the Denman region, East Antarctica. BedMachine is smoother and maybe lacks some small-scale features, whereas SRTM15 is rougher with more small-scale features. In the current stage, we can only see the two bathymetry datasets are different and make a significant impact on ocean circulation. As we haven't compared the two datasets with the observations (seal data), so we can't give a definite answer on which bathymetry data is better.
COSIMA_Yuhang Liu.pdf

[aekiss]

Thanks @yuhang559!

[aekiss]

Following Alex Fraser's talk at COSIMA meeting today, we could also put in grounded icebergs as land in order to get fast ice in the right places (if we have tensile strength in the sea ice).

[adele-morrison]

@adfraser maybe you can check if SIS2 has the required dynamics for us?

[adfraser]

Hi @aekiss and @adele157 -- it looks like Kate Hedstrom has investigated Arctic fast ice with SIS2: https://twitter.com/UAFOceans/status/1417624387008696320. It makes me think it supports fast ice out of the box, but I can't confirm quickly sorry.

This brings up an intersting possibility. If you have an easy way to get fast ice happening, there's a chance @dpath2o 's PhD could pivot to using this (rather than CICE6) to do at least some of his PhD work. Getting CICE6 running correctly is proving to be tricky and many of Dan's aims could be accomplished without waves. @willrhobbs @PaulSpence any thoughts?

[adele-morrison]

Great that SIS2 looks good for fast ice.

Sounds like a nice option for @dpath2o to use this model - shall we chat offline about what simulations / diagnostics are needed and if it would be suitable? If so, we could perhaps do a 5 year test of the model with a) the new bathymetry, and b) new bathymetry plus grounded icebergs?

[adfraser]

Thanks @adele157 - I'll have a chat just with @dpath2o, @willrhobbs and @PaulSpence too - could be really good!

BTW -- speaking of bathymetry, I note that IBCSO 2 was just released. I know that IBCSO 2 has the new multibeam near Cape Darnley from the Japanese program (and I'm not sure if the new GEBCO does or not). Could be another one to throw in the mix when thinking about new bathymetry products.
https://ibcso.org/ibcso-version-2-released/
@aekiss

[aekiss]

Thanks @adfraser. I'm not an expert on bathymetric datasets, but IBCSO 2 seems definitely worth considering. They describe it as "the most authoritative seafloor map of the area south of 50°S" and it also includes sub-ice shelf bathymetry. Hopefully it can be smoothly combined with some other dataset to cover north of 50°S.

[adfraser]

@aekiss Jodie Smith from Geoscience Australia will know more if you need - she's a contributor to IBCSO2.
I've always preferred IBCSO (1) on the continental shelf but I haven't done any comparisons in the last few years.

[aekiss]

GEBCO2022 has just been released, and includes sub-ice shelf data https://www.gebco.net/data_and_products/gridded_bathymetry_data/gebco_2022/
The GEBCO2022 resolution is 15 arc-second (ie 1/240 deg = 460m at equator and finer zonally nearer the poles), whereas IBCSO is coarser (500m x 500m).
They're both GEBCO projects but I don't know how GEBCO2022 is related to IBCSO 2 south of 50S (maybe they are the same, apart from resolution?)
I've contacted Jodie Smith for advice.

[aekiss]

Boris Dorschel (lead author of the IBCSO2 release paper) has confirmed that IBCSO2 has been merged into GEBCO2022. So GEBCO2022 sounds like a good choice, unless there are issues with it that @dpath2o knows about.

[dpath2o]

I've downloaded the full GEBCO2022 and I will incorporate into the CICE6 stand-alone that I'm currently standing up locally as I'll be using a cubic spline to reduce to the ACCESS-OM2 CICE tri-polar grid as a first step. However, with my other hat, I'm currently working with USN (FNMOC/NRL) and I'll see if they have any prickly bits that are cautionary for operational oceanography and advise as required.

[PaulSpence]

would be good to get Russ Fiedler to take a look https://github.com/russfiedler P

…

On Wed, Oct 5, 2022 at 2:19 PM Christina Schmidt ***@***.***> wrote: I created the bathymetry and land mask for the Southern Hemisphere at 1/20° resolution for 75 z* levels with the help from @micaeljtoliveira <https://github.com/micaeljtoliveira> . Overall, it looks pretty good to me! [image: image] <https://user-images.githubusercontent.com/38216156/193973100-a802ef88-9fb8-40ae-abdd-f42a51fc1921.png> The bathymetry on the Antarctic shelf looks very similar to the original GEBCO 2022 data: [image: image] <https://user-images.githubusercontent.com/38216156/193973139-2913d1c1-e47e-4d4a-918b-66406b259f39.png> The land mask for South America and Africa does not seem to be “blocky”: [image: image] <https://user-images.githubusercontent.com/38216156/193973307-5a42566b-7d85-4198-b4ad-ec5683689966.png> I haven't smoothed the northern boundary yet to match the bathymetry with the 1/10° forcing. Does that need to be done manually? Any other thoughts on this bathymetry or can I go ahead and use this for test runs with the panan-005? — Reply to this email directly, view it on GitHub <#12 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABSWJXHKDTIP5D4PA3CM3X3WBTXVNANCNFSM5WYGZDWQ> . You are receiving this because you were mentioned.Message ID: ***@***.***>

-- *I’m sending this message at a time that suits me. I don’t expect you to reply outside of your own work hours.* Paul Spence, Assoc. Prof. ARC Future Fellow Institute for Marine and Antarctic Studies <https://www.imas.utas.edu.au/> University of Tasmania, Hobart, Australia https://paulspence.github.io/

[russfiedler]

Looks to be quite a few problems remaining. e.g. New Zealand's North Island, inland seas in South America. Also, have a good think about what you are trying to model in areas like Patagonia. What detail is needed? In addition, check that you've got deeper channels/canyons/sills well represented.
It would be useful to put the topography file somewhere public so others can comment.

[schmidt-christina]

Thanks, for the comment @russfiedler . Here is a zoom into South America and NZ (top: land mask, bottom: bathymetry) and I don't see inland seas/other problems there, but correct me if I am wrong:

This will be the bathymetry for the PanAntarctic configuration at 1/20° resolution with a northern boundary at 37°S or the equator (still needs to be decided), so the main focus should be modelling Antarctic shelf processes and possibly ACC dynamics.

Can you provide me with a list of the deeper channels that you have in mind and which might be a problem?

Is there a public folder on gadi where I should put the file? Otherwise it is now here
/g/data/e14/cs6673/shared/panan-005_bathy/
(Coordinates and attributes are still missing in the files and it's on our list to fix that)

[russfiedler]

I think that some of the problems that I thought might be there are just artifacts of the plotting software.
You'll need to chat with some real oceanographers(!) with respect to resolving important pathways.

[aekiss]

Great work @schmidt-christina!

I'm not sure whether sills will be a problem at 1/20°. Sills are more of an issue at coarser resolution, where the mean depth doesn't capture the maximum at sills.

We did a lot of tweaking at 0.25°, but not in the southern hemisphere as far as I recall.
https://github.com/COSIMA/make_025deg_topo/blob/master/topog_edits.txt

At 1° there were also some edits in Patagonia and the ITF, but they're very unlikely to be needed at 1/20°
COSIMA/access-om2#158 (comment)

[aekiss]

It might be worth plotting the difference between your new topo and the one used in ACCESS-OM2-01 (with your topo regridded to the ACCESS-OM2-01 grid). Plot both the absolute and relative depth changes to see if anything looks weird.

[AndyHoggANU]

Hi @schmidt-christina - this looks good. I do have some slight worries about the number of channels and fjords, but we should remember that MOM6 is a very robust code. I think it may be worth just giving this a go and seeing what happens?

As it’s running, be sure to monitor sea level — if you have any enclosed lakes then they will show up as their level evolves.

I think the only outstanding question is whether you need to degrade the resolution on the northern boundary to match the original. I think not, but I would defer to @angus-g here …

[schmidt-christina]

I now compared the new topo with ACCESS-OM2-01 (using ht) @aekiss . There are some big differences between them with the new topo being often deeper (red shadings, unit is meter). It seems that often the differences are in regions of gradients in the topography. I am surprised by how big the differences are but I wouldn't think there is anything systematically wrong.

[adfraser]

That Western Weddell Sea difference is probably due to better observations in a newer bathymetry product. It must be very data-sparse there - often multiyear sea ice cover so hard to get a multibeam in there. But wow 400m of difference is wild.

[aekiss]

Hm, interesting. Those differences don't look too concerning to me - I guess the interpolation between the bathymetric observations is least accurate where the gradients are large, so the differences between successive versions of GEBCO are biggest there? That doesn't explain why the new topo is usually a bit deeper though.

Do they both have the same maximum depth, and do they use the same vertical grid and minimum partial cell height?

What do the relative depth differences look like?

2d histogram of old vs new topo can be useful to look for systematic differences, and a plot of partial vs full thickness can catch issues with partial cells - see
COSIMA/access-om2#158 (comment)
and
https://github.com/COSIMA/topogtools/blob/master/bathymetry.ipynb

[adfraser]

The large differences on the Antarctic continental shelf slope in general aren't surprising, given the attention paid to improving/getting better data in these areas (e.g., for better understanding of warm water access) - e.g. I was involved in this one which got new multibeam data offshore of the Totten in 2015: https://agupubs.onlinelibrary.wiley.com/doi/10.1002/2017GL074433 - Fig 3b gives a few transects across the shelf showing huge (400 m+) differences between products (Bedmap2 looks particularly bad in this region). I assume the new multibeam from our cruise is in the gebco2022 product.

[adfraser]

(Probably more concerning is that the shelf itself isn't improved! But what can you do?)

[schmidt-christina]

Yes, the maximum depth and vertical grid is the same (and I assume the minimum partial cell height as well as I haven't modified anything, but I am not really sure where to look that up).

This is the relative depth difference:

[adfraser]

@schmidt-christina ah - that's a good way of looking about it - and now I see that the shelf is changed in many regions! Good stuff.

[aekiss]

Thanks - interesting that there are a few big relative differences in the open ocean. I guess the blue one near 60E, 60S is on the Kerguelen plateau?

[adfraser]

No, it's south of Kerg. Kerg is 50S.
Edit: oops you said plateau. It looks like it's on the Elan Bank though.

[pedrocol]

I have a few comments, I don't know how useful they will be, but perhaps worthy to mention. I discussed this with Jean-Marc Molines, he did all the DRAKKAR bathymetries, and in particular, the bathymetry for the North Atlantic 1/60deg config (NATL60).

• For those who want to take a look to the last version of DRAKKAR 1/12deg global bathy: wget --no-check-certificate https://ige-meom-opendap.univ-grenoble-alpes.fr/thredds/fileServer/meomopendap/extract/eORCA12.L75/eORCA12.L75-I-JZ/eORCA12_bathymetry_v2.4.nc. This extended grid goes south until 85.7 (needed to include the cavities).
  Maybe interesting for comparison.

• He always perform changes manually at the coast, following a high resolution coast line (order 100m), and including very small bathymetric features (like islands, fiords, etc). This gives a very high realism look to the bathymetry, but he is sure that resolving all these bathymetric features has no impact on the large scale off-shore dynamics, and that maybe masking these places would had the same impact. These small features were not a problem when initializing the model in terms of stability, but the initial T/S data in those places is most likely not very precise.

• To finish, he also always performs a smoothing using a Shapiro filter, before introducing the mask. This is done to deal with some erratic values that may exist in the dataset and also considering that the dataset is created from different sources.

[aekiss]

Alistair Adcroft mentioned at the COSIMA workshop that several bugs were fixed recently in MOM6's handling of cavities so that "ceiling topography" is handled in the same way as bottom topography. So we should make sure we're using the latest MOM6 source code.

[adele-morrison]

Update on 1/20deg bathymetry from our meeting 20/1/23:

• Christina to update the bathymetry using the fraction parameter in domain-tools.
• Try running with this new bathymetry again (needs coupler mask). Also first check that we have runoff set to zero (and not constant runoff). @angus-g to check this with @schmidt-christina.

[schmidt-christina]

We still have issues with the bathymetry that @micaeljtoliveira is working on. We can't use the new bathymetry yet, especially because it has lakes again:

• depth in topog.nc also has depth values where the fraction is < 0.5 although I set it to 0.5
• ocean_mask considers the correct fraction but now ocean_mask and depth in topog.nc don’t agree
• after applying the fraction criterion new lakes occur (see one isolated yellow gridcell in ocean_mask)
• geolon_t and geolat_t are not the longitude and latitude, but again sea_area_fraction
• ocean_mask.nc doesn’t have coordinates
• NaNs should be something like -1e30

[micaeljtoliveira]

• geolon_t and geolat_t are not the longitude and latitude, but again sea_area_fraction
• ocean_mask.nc doesn’t have coordinates

These are now fixed.

Regarding the inconsistencies between the depth and the mask, after chatting with @aekiss about this, we agreed that we need to change the workflow and do things in a slightly different way as for ACCESS-OM2. In particular, all fixes to the topography, including setting the depth to zero for cells that have a fraction of ocean smaller than some value, should be done before generating the mask.

[aekiss]

At today's Antarctic Margins meeting Esmee Van Wijk commented that topography in the Denman region is quite different from datasets (e.g. bedmap), e.g. Denman trough is significantly deeper in reality, and some of the ridges might be over 100m too shallow in bedmap (and presumably GEBCO).

Steve Rintoul added that the thermocline is sharp and similar depth to ridges in this region, so can get critical transitions when water below the thermocline can make it over a ridge, so the combination of thermocline depth and bathymetry is important to get right in modelling. Much of the bathymetry data is from airborne gravimetry, which lacks spatial resolution, so smoothing reduces amplitude of extrema (troughs and ridges).

I may not have got all the details right - we could contact Esmee and Steve for more info.

[access-hive-bot]

This issue has been mentioned on ACCESS Hive Community Forum. There might be relevant details there:

https://forum.access-hive.org.au/t/bathymetry-for-ocean-model-at-any-resolution/462/8

[aekiss]

Just noting here a new bathymetric dataset (SYNBATH) that fills in the gaps between multibeam tracks with statistically reasonable high-resolution synthetic data https://doi.org/10.1029%2F2021ea002069