[MITgcm-support] regional zooming method

Mon Sep 23 09:23:57 EDT 2013

Matt, if you can solve your problem with one-way nesting, via pkg/obcs, then you will not suffer the time step limitation that Martin mentions and we have considerably more experience with that both in polar (i.e., with sea ice) as well as non-polar seas. 

Dimitris

> On Sep 23, 2013, at 1:07 AM, "Martin Losch" <Martin.Losch at awi.de> wrote:
> 
> Hi Matt,
> 
> looking for a model with variable grid resolution? See here <http://www.awi.de/en/research/research_divisions/climate_science/climate_dynamics/news/>
> (Just kidding).
> 
> Without too much experience myself, the limitation of a zoomed mesh is obviously the aspect ratio of the largest to the smallest grid size that you can achieve with this, and the that the time step will be determined by the smallest grid cell (CFL) in your domain. (That's actually the issue that the people behind the above link have all the time. Great resolution where you need it, but also enormously small time steps where you don't want them)
> 
> As far as I know there is no perfect nesting, and again the ratio between the grid cells of coarse and fine mesh is very limited, maybe even more than in the zooming approach.
> 
> Martin
> 
>> On Sep 20, 2013, at 7:02 PM, Matthew Mazloff <mmazloff at ucsd.edu> wrote:
>> 
>> Hello
>> 
>> I would like to zoom in on a region of my domain and am looking for some advice on how to best do this.
>> 
>> I have used telescoping where DX and DY are vectors but would prefer something better as this is not ideal far away from the region where, e.g., DX is still small but DY becomes large 
>> 
>> So it seems my two choices are coming up with a mesh or nesting. 
>> 
>> For the mesh case I assume the model cannot handle two cells becoming one on a cell wall…is this correct? Has anyone been successful in coming up with a mesh of quadrilaterals that zooms in on a region. I know the mesh can be derived, by successful I mean that the model output looks physical. 
>> 
>> For nesting -- has anyone had success with 2-way nesting? Is nesting_sannino the place to start? Any pitfalls I should be aware of?
>> 
>> Thanks
>> Matt