lfricAtlasMap not working as expected

[ctgh]

(I've assigned this issue to people based on their activity in this repository - hope that's OK.)

I have created an LFRic-JEDI LETKF test in this PR. When plotting the output I noticed that the locations of the increments do not match up with the locations of the observations, which indicates that something has gone wrong.

• The locations of the increments and observations match up correctly in the assimilation code, so it doesn't look like a problem either with the solver in oops or the geometry iterator in LFRic-JEDI.
• I tried two ways to plot the increments and they both gave the same output, so I don't think it's a plotting issue.

The issue appears to be in the way that monio saves the increments to netCDF (using the Atlas writer). Doing some digging I found that a mapping between the LFRic and Atlas coordinates is created in the createLfricAtlasMap routine. That seems to jumble up the locations and cause the effect I saw. If I replace the assignment to lfricAtlasMap as follows:

lfricAtlasMap = indices;

the plots appear exactly as expected. I have updated the PR linked above to reflect that.

I don't know precisely why the mapping doesn't work, but it seems I will need some kind of workaround. One quick but potentially ugly solution would be to pass another boolean into the createLfricAtlasMap routine which controls how the mapping is set up, and enable the user to configure that in the yaml. A better option would be to handle everything automatically, but having not yet looked in great detail I am uncertain how much work that would entail.

[mo-joshuacolclough]

As commented here: #54 (review)
It could be that the Atlas lon/lats are not being written to file in the correct order.

[ctgh]

Closing this as it appears the problem was in the lfric-jedi geometry iterator. So far testing indicates that fixing the iterator solves the problems. If there are any anomalous results I will take another look.