[aspect-devel] Velocity boundary conditions - lithosphere ext/compression

Wed Jul 5 05:23:10 PDT 2017

Hi John,

I'm getting some stability now through an error in my implementation of asthenosphere rheology properties. Simple error!

I'll have a play around for a while and see what errors pop up - hopefully can work this into a cookbook as it is a simple extension of the continental_extension.prm

Thanks for your help, will let you know if I run into any other problems.

Phil

________________________________
From: Aspect-devel <aspect-devel-bounces at geodynamics.org> on behalf of HERON, PHILIP J. <philip.j.heron at durham.ac.uk>
Sent: 04 July 2017 23:26:20
To: aspect-devel at geodynamics.org
Subject: Re: [aspect-devel] Velocity boundary conditions - lithosphere ext/compression

Hi John,

Thanks for the reply - I will look at implementing some of these suggestions. I've already tried playing around with the velocities, to no success.

I will try to use a different seed to localize the deformation. An idea was to simply expand the continental extension model, however, this seems like it may require a bit more tweaking.

I am interested in continental collision, intraplate deformation (benchmarking my own work with other codes) and indeed continental extension. I have been using the numerical code SOPALE for a while, so it would be interesting to me to be able to reproduce certain established papers of lithosphere deformation that use such a code (and similar). For instance, to be able to produce Huismans and Beaumont (Nature, 2011) paper, or the work you did with Susanne in 2015.

I can dig out what benchmarks for such endeavours would be required.

I'll apply those changes and get back to you!

Cheers,

Phil

________________________________
From: Aspect-devel <aspect-devel-bounces at geodynamics.org> on behalf of John Naliboff <jbnaliboff at ucdavis.edu>
Sent: 04 July 2017 21:03:37
To: aspect-devel at geodynamics.org
Subject: Re: [aspect-devel] Velocity boundary conditions - lithosphere ext/compression

Hi Philip,

I ran the model and the velocities certainly are showing interesting behavior. Over time the velocity magnitudes increase, but the flow pattern is basically reversing at each visualized output (1 Myr intervals).

While the velocity boundary conditions are constant, this change in flow pattern appears to reflect that different portions of the lithosphere (sides vs interior) are alternating between moving and up and down. These vertical motions are swamping out the signal from the applied velocity boundary conditions along the side walls.

In some ways this kind of behavior looks similar to "drunken seaman" behavior (instability associated with free surface), but I don't think that is the case. Rather, I think the model is having trouble to localize large-scale deformation in the lithosphere and this is part of the response. Significantly, I noticed that in the parameter file the velocities may be a bit lower than intended. The variable "cm" is set to "0.001". With the other specified values this gives a convergence velocity of 0.1 cm/year applied on each side in the upper 120 km. This value is quite low for continental convergence and I think if you increase this value to 1 cm/year it may eliminate some of the anomalous behavior. I would also try decreasing the CFL number to 0.2 or 0.1.

Aside from trying those fixes, may I ask if there is a specific end goal of these models? In other words, is this experiment designed to produce a model of long-term continental shortening? If so, different initial conditions may help.

If the goal is indeed to model some form of long-term convergence (continental collision, subduction, ...), is there a particular prior study (or studies) you have in mind? Always happy to help reproduce prior work, which can then be added as benchmarks cases to the ASPECT repository. A number of us are currently working on subduction problems, so lots of information, ideas, tips and parameter files to share.

A note to all new or potential ASPECT users - the offer to help with reproducing prior studies or discussing ideas for new setups is open to everyone.

Cheers,
John

*************************************************
John Naliboff
Assistant Project Scientist, CIG
Earth & Planetary Sciences Dept., UC Davis

On 07/04/2017 07:36 AM, HERON, PHILIP J. wrote:

Hi John,

thanks for the reply - I've attached the log.txt for the prm file I sent, along with the error output.

I have tried and played around with a great number of parameters to try and get this to work, but I think it is good to get some outside perspective.

Point 1 - thanks for that, I've applied that change to the models.

Point 2 - the vertical sides seems to be immune to the fluctuating velocity, the interior of the model has the blow up in velocity.

Point 3 - I'm just running a fixed mesh model now to see where it gets to. Will keep you posted.

Furthermore, I'm going to run a series of tests on reference viscosity after noticing this post:

https://github.com/geodynamics/aspect/issues/1814

Issues with linear solver convergence due to choice of Reference viscosity · Issue #1814 · geodynamics/aspect<https://github.com/geodynamics/aspect/issues/1814>
github.com
I've had some issues with choosing the "right" reference viscosity (in models with varying viscosity) in order to get convergence of the linear system and was wondering if there might be some way o...

Philip J. Heron
Junior Research Fellow
Dept. of Earth Sciences
Durham University
web: http://philheron.com<http://philheron.com/>

________________________________
From: Aspect-devel <aspect-devel-bounces at geodynamics.org><mailto:aspect-devel-bounces at geodynamics.org> on behalf of John Naliboff <jbnaliboff at ucdavis.edu><mailto:jbnaliboff at ucdavis.edu>
Sent: 04 July 2017 01:22:47
To: aspect-devel at geodynamics.org<mailto:aspect-devel at geodynamics.org>
Subject: Re: [aspect-devel] Velocity boundary conditions - lithosphere ext/compression

Hi Philip,

Yes, certainly happy to help on this front! A few things to discuss.

1. Unrelated to your issue, I completely forgot that there is a small bug with the heat production term in this cookbook that Anne Glerum pointed out to me a week or two ago. This will cause the crust to cool more rapidly than it should over time, but it is not readily apparent over 10's of Myr with advection on. This is unrelated to the anomalous velocities in your model, but I will push a fix for this this evening or tomorrow morning. Apologies to all who are currently using the cookbook example. The quick fix is to change the "Initial concentrations crust" parameter from 1 to 1e6. Alternatively, the "fix" I'll push tomorrow will allow you to set different constant heating rates for each compositional layer (see "compositional heating" method).

2. Onto your issue. I've run a similar extension model (1000 x 600 km) with similar boundary conditions along the sides (balanced inflow outflow) and the velocity field was stable over time. Are your horizontal velocities fluctuating throughout the model or near the vertical sides?

3. In similar models I used a fixed mesh (no AMR). Have you tried the model with a fixed mesh? Very likely to be unrelated, but might be worth using a fixed mesh to see if the error occurs at a similar time and place in the model.

3. I'm running your parameter file now, but in the meantime can you send the log.txt file from the output folder?

Regardless of the root issue, we can get this worked out quickly. If the issue is not readily apparent, we can start from a similar working input file and compare the differences.

Thanks for sending in this question to the "devel" email list. Really helpful for others to see where issues are cropping up!

Cheers,
John

*************************************************
John Naliboff
Assistant Project Scientist, CIG
Earth & Planetary Sciences Dept., UC Davis

On 07/03/2017 08:14 AM, HERON, PHILIP J. wrote:

Hello all,

I was just extending the 'continental_extension.prm' cookbook to include sub-lithosphere mantle, and have ran into issues of convergence. I wonder if anyone on the group could help!

I've attached the extended cont extension model, I was wondering if anyone could point me in the right directions of how to get it to converge. The issue is that after 10-15 timesteps velocities within the model start to blow up.

I feel that the temperature profile initially is fine. Although applying a variable horizontal velocity to the boundary, the initial horizontal velocity field is smooth. However, almost immediately the horizontal velocity starts to fluctuate.

Are there better techniques to be able to get this  model to converge? Am I missing something simple? I've played around with a few different things but with little luck.

Thanks in advance!

Phil

Philip J. Heron
Junior Research Fellow
Dept. of Earth Sciences
Durham University
web: http://philheron.com<http://philheron.com/>

_______________________________________________
Aspect-devel mailing list
Aspect-devel at geodynamics.org<mailto:Aspect-devel at geodynamics.org>
http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel

_______________________________________________
Aspect-devel mailing list
Aspect-devel at geodynamics.org<mailto:Aspect-devel at geodynamics.org>
http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.geodynamics.org/pipermail/aspect-devel/attachments/20170705/964ee984/attachment-0001.html>