Example that uses restoring to the ECCO solution on a tripolar grid?

[glwagner]

Right now we have an example that runs a "near global" simulation, but we don't have an example that restores to the ECCO solution in polar regions on a tripolar grid.

[francispoulin]

The ACC example that @simone-silvestri and I are working on in #142 restores to ECCO solution but uses LatitudeLongitudeGrid.

Would there be interest in changing grids for this example? I have never used the tripolar grid so don't know much about it.

[simone-silvestri]

For that example we need to use the latitude-longitude grid because the tripolar grid supports only domains on a sphere that

• are periodic in longitude (between 0 and 360 degrees)
• reach 90 degrees latitude.

In my opinion, on regional domains, the latitude-longitude grid is much better also for visualization and postprocessing, the tripolar grid requires remapping onto a latitude-longitude grid for things like zonal averages which makes everything a little more clunky.

I am not sure about, for example, an arctic simulation. In that case, a tripolar grid has the advantage of being trivial to set up (probably the latitude-logitude grid is also easy to set up), but still the con of having a more complicated data structure to work with (2D coordinates which are a bit of a nightmare for interpolation).

[glwagner]

Hmm yea, it also seems easier to use a lat-lon grid. We just need an abstraction to rotate the "apparent pole" of the lat lon grid to some prescribed coordinate (eg somewhere on the equator).

[glwagner]

So in summary it will still be useful to have a tripolar grid example!