[MITgcm-support] Tracer instabilities when I reduce hFacMinDr

[C Spencer Jones]

Hi Kaitlin,
I had a similar problem, but with the Prather advection scheme, when the
maximum bathymetry depth exceeded the sum of the cell depths (these often
have pretty similar values). I found that reducing the maximum depth of my
bathymetry by a couple of meters solved the problem. Maybe that is worth a
try?
Thanks,
Spencer

On Tue, 30 Oct 2018 at 11:20, Naughten, Kaitlin A. <kaight at bas.ac.uk> wrote:

> Hi again,
>
>
> The three simulations I tried yesterday (decrease timestep even more,
> increase vertical viscosity, increase vertical diffusivity) all showed
> similar instabilities as before. Also, increasing the (horizontal) Leith
> viscosity doesn't fix the problem. I can't easily increase the horizontal
> diffusivities as these are set by GM.
>
>
> Could GM be part of the problem here? Or Leith? Clearly it's something to
> do with the thinner partial cells, but I think I've ruled out a simple CFL
> error...
>
>
> Please let me know if you can think of anything else!
>
>
> Many thanks,
>
> Kaitlin
> ------------------------------
> *From:* MITgcm-support <mitgcm-support-bounces at mitgcm.org> on behalf of
> Naughten, Kaitlin A. <kaight at bas.ac.uk>
> *Sent:* 29 October 2018 11:34:26
> *To:* MITgcm Support
> *Subject:* Re: [MITgcm-support] Tracer instabilities when I reduce
> hFacMinDr
>
>
> Hi Martin,
>
>
> Thanks so much for your ideas. I'm waiting on the results of some
> simulations, but I'll respond to what I can now:
>
>
>
>    1. advcfl_W_hf_max and advcfl_wvel_max are the same for every monitor
>    output. Is this normal/okay? For my simulations with hFacMinDr=20, they are
>    usually very slightly different.
>    2. The instabilities mostly happen on the side of a trough. It's not
>    the steepest side of the trough, and not as steep as many other parts of
>    the domain. So, I don't think there is any correlation with steep or rough
>    bathymetry. But, this trough is a region of strong sea ice formation, so
>    maybe that's causing strong downslope flows.
>    3. I also see no correlation with the transition from minimum 5m to
>    minimum >5m. The instability is well within the region with vertical
>    resolution 25m, so 5m is the smallest allowable partial cell.
>    4. I looked at the vertical structure and the instabilities seem to be
>    coming from the bottom partial cells, especially water columns where that
>    cell is 5m. They spread 1-2 cells in each direction (both vertical and
>    horizontal) but the worst values are clearly in the bottom. So, that seems
>    to suggest vertical CFL errors.
>
>
> I've started a few simulations as tests:
>
>
>    1. Decrease timestep by a factor of 10 compared to my original
>    simulations with hFacMinDr=20 (rather than a factor of 4)
>    2. Increase vertical viscosity (viscAr) by a factor of 10
>    3. Increase vertical diffusivities (diffKrT and diffKrS) by a factor
>    of 10
>
>
> Hopefully I will have some results tomorrow.
>
>
> All the best,
>
> Kaitlin
> ------------------------------
> *From:* MITgcm-support <mitgcm-support-bounces at mitgcm.org> on behalf of
> Martin Losch <Martin.Losch at awi.de>
> *Sent:* 28 October 2018 10:32:04
> *To:* MITgcm Support
> *Subject:* Re: [MITgcm-support] Tracer instabilities when I reduce
> hFacMinDr
>
> Hi Kailin,
>
> this looks like a difficult problem to solve. Here are my ad-hoc ideas:
>
> The unphysically negative temperatures are most likely just the
> consequence of your instability and then the flux-limiters (which are never
> perfect) can’t help. They end to mask problems in the initial state, which
> is not helpful either. Maybe you want to try a non-flux-limited scheme to
> see where the problem starts.
>
> When you checked the clf-numbers, did you also check advcfl_W_hf_max? In
> contrast to advcfl_wvel_max, this number takes into account the hFacC (so
> it will be larger for smaller hFacC, i.e. hFacMinDr)?
> Also are your instabilities correlated (in space) with rough or steep
> topography?
>
> The minimum cell thickness is computed like this (as you probably already
> know):
> MAX( hFacMin, MIN(hFacMinDr*recip_drF(k),oneRL) )
> so that for thin cells (10m) you’ll have a minimum thickness of 5m, but
> for thicker cells you’ll have drF*hFacMin, which is larger than 5m for
> drF>50m. Does this tansition have anything to do with your instability
> pattern?
>
> Did you have look at the vertical structure of your instabilities. Often
> not only the horizontal but also the vertical viscosity/diffusivity can
> cause cfl-like problems.
>
> What happens if you decrease your time step even further? Does the problem
> go away?
> Try larger viscosities (for momentum) …
>
> Martin
>
>
> > On 24. Oct 2018, at 12:47, Naughten, Kaitlin A. <kaight at bas.ac.uk>
> wrote:
> >
> > Hi everyone,
> >
> > I'm currently running an ice shelf configuration with hFacMin=0.1,
> hFacMinDr=20, and vertical resolution ranging from 10 m (top few layers) to
> 25 m (in the cavity) to 100+ m (deep ocean). I'm trying to decrease
> hFacMinDr so I can get a better representation of the ice shelf base, as
> the current value of 20 m combined with the 25 m resolution in the cavity
> doesn't allow the partial cells to do very much.
> >
> > I'm trying to get hFacMinDr=5 working. I expected to have to reduce the
> timestep to prevent vertical CFL errors, and at this point I have reduced
> the timestep by a factor of 4 (which should exactly compensate for the
> thinner cells, given the CFL criterion is linear, correct?) However, the
> simulation still looks very strange, in particular the bottom water
> temperature and salinity in a few patches on the continental shelf (not in
> the cavity): it looks like checkerboard instabilities or false extrema,
> even though I'm using a flux-limited advection scheme (33). In some cells,
> the temperature goes down to -7 C.
> >
> > I checked the CFL terms in the monitor output, and the vertical terms
> are on the order of 1e-2. This is similar to my previous simulations with
> hFacMinDr=20. and timestep 4 times larger. So, why am I still seeing these
> instabilities? Are they still CFL errors (despite what the monitor terms
> seem to suggest) or could they be something else?
> >
> > I am using the linear free surface until I find a stable timestep for
> hFacMinDr=5, so we can rule out any problems related to nonlinear free
> surface or r*.
> >
> > Many thanks,
> > Kaitlin Naughten
> > This message (and any attachments) is for the recipient only. NERC is
> subject to the Freedom of Information Act 2000 and the contents of this
> email and any reply you make may be disclosed by NERC unless it is exempt
> from release under the Act. Any material supplied to NERC may be stored in
> an electronic records management system.
> > _______________________________________________
> > MITgcm-support mailing list
> > MITgcm-support at mitgcm.org
> > http://mailman.mitgcm.org/mailman/listinfo/mitgcm-support
>
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mailman.mitgcm.org/mailman/listinfo/mitgcm-support
> ------------------------------
> This message (and any attachments) is for the recipient only. NERC is
> subject to the Freedom of Information Act 2000 and the contents of this
> email and any reply you make may be disclosed by NERC unless it is exempt
> from release under the Act. Any material supplied to NERC may be stored in
> an electronic records management system.
> ------------------------------
> ------------------------------
> This message (and any attachments) is for the recipient only. NERC is
> subject to the Freedom of Information Act 2000 and the contents of this
> email and any reply you make may be disclosed by NERC unless it is exempt
> from release under the Act. Any material supplied to NERC may be stored in
> an electronic records management system.
> ------------------------------
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mailman.mitgcm.org/mailman/listinfo/mitgcm-support
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.mitgcm.org/pipermail/mitgcm-support/attachments/20181030/5c86caa7/attachment-0001.html>