unequal axis spacing

[milankl]

I don't think an non-equi-distant axis spacing is correctly visualised atm. This gif shows the same simulation but run on different grids, neither of them are equi-distant in latitude

At lower resolution the latitudes of these grids are almost equi-distant but not quite

julia> RingGrids.get_latd(OctaminimalGaussianGrid, 4)
8-element Vector{Float64}:
  73.79921362856322
  52.81294318999426
  31.704091745007943
  10.569882312576098
 -10.569882312576098
 -31.704091745007943
 -52.81294318999426
 -73.79921362856324

julia> RingGrids.get_latd(OctaHEALPixGrid, 4)
7-element Vector{Float64}:
  69.63586519368219
  48.59037789072914
  25.944479772370016
   0.0
 -25.944479772370016
 -48.59037789072914
 -69.63586519368219

or visualised

So Gaussian latitudes actually approach to being equidistant except near the poles, but the OctaHEALPix grid has latitudes some 25% further spread near the equator but 10% or so less near the poles

[lazarusA]

we haven't yet started working on different Grid projections. So, I don't know what are you looking at 😄 (something seems to work). Jokes aside, it will be good to have a MWE script (typical) for your use cases. That way we can generate the files in our end and possibly iterate faster.

Edit: ohh, I see, you are saving the matrix values and corresponding Lon,lat from the projections. ok, ok.

using SpeedyWeather
# generate different grid output systems

as for providing axis with unequal spaces, I don't think the current shader handles that, @TheJeran ? new feature?

[TheJeran]

At the moment visualization is equidistant. This is not planned for the final version but it does make things very easy for iteration at the moment.

I want to incorporate the functions for map projections https://pubs.usgs.gov/publication/pp1395 into shaders so you can reproject your data live. And when I do this I will need to include proper pixel scaling. So I will definitely return to this in the future.

@milankl Just a little fyi development has slowed significatly for a bit while I finish other essential projects. So it will be awhile. But it is something I strongly want to incorporate.

[milankl]

It's not super urgent, I just didn't know, so you could also just throw a warning instead? Just thought it might be helpful to list these possibly silent issues!

[milankl]

As you say Lazaro, "something seems to work" indeed 🤣

[TheJeran]

@milankl Hey Milan, I have reprojection working and can handle this now. But I need examples of datasets with these kind of grids to try to figure out a way to either autodetect or allow users to set-up the coordinate grid. Do you happen to have any datasets like this I can use for testing?

[milankl]

Oh that's awesome! I just sent you some netcdf via email but I realised you can also upload files here:

land_sea_mask_grids.zip

[milankl]

This is data using the HEALPix latitudes for example

these should align if the latitude spacing is correctly applied

[milankl]

In contrast for the full clenshaw grid which uses equiangle latitudes everything is already aligned well

[milankl]

And these are OctaHEALPix latitudes with everything further south than it should be (HEALPix is visualised further north than it should be)