Calculation of Transversal Mercator Projection for glacier coordinates #68
Conversation
|
And thank you @albangossard for suggesting using CoordRefSystems instead of Geodesy: it is much more easy and intuitive to use, althought I spend quite some hours understanding how Mercator projections work... all this mess when the Earth is flat... |
There was a problem hiding this comment.
Hey @facusapienza21, thank you for that work! Computing the projections with CoordRefSystems looks pretty straightforward. I like the idea of having the coordinates in a global format, this will make the post-processing easier to plot and to interface with external data sources.
| latitudes = map(x -> x.lat.val, transform.(Ref(mean(easting)), northing)) | ||
| longitudes = map(x -> x.lon.val, transform.(easting, Ref(mean(northing)))) |
There was a problem hiding this comment.
I have the feeling that these averages are a bit touchy. From what I understand this is to compute the central point. Is the grid always uniform? Otherwise the projection might be incorrect in some situations. At the minimum we could check that the spacing is uniform, don't you think?
|
Following on @albangossard point about the uniform grid points: I have the feeling that these averages are a bit touchy. From what I understand this is to compute the central point. Is the grid always uniform? Otherwise the projection might be incorrect in some situations. At the minimum we could check that the spacing is uniform, don't you think? Yes, I thought about this as I was doing it... It is true that the glacier won't fix exactly in a square grid, but for mountain glaciers this discrepancy is minimal (even smaller than with usual projections, since the UTM projection is done with a custom longitude centered at the glacier). This discrepancies are already present in the dataset since the glacier products are reported in northing and easting, so grinding here has already been made. The conversion to latitude and longitude can introduce some small discrepancies, but it is rather minimal I will say. This differences are more considerate in high latitudes. I added a warning message in the new version of the code to report to be careful with this when the latitude is near polar regions (~80 degrees ref). |
|
Notice that I also added one more I am not sure how the files for the glacier construct tests where created, but maybe we need to update them? |
Thank you for this complete explanation!
I double checked the content of the jld2 files when creating them but as |
|
I updated the files but the errors don't pass in CI. I had run the tests in local and they work perfectly well, so I am not sure the source of the issue. I noticed there are some comments regarding tests in CI in Sleipnir, so maybe this is the issue? Let me know if you can fix this on your side. I am going to be offline until next Tuesday, so feel free to merge the PR and we figure this out later. |
After having updated the jld2 files in #69, the tests finally passed both locally and in the CI. I'm not sure to understand why. Anyway it's working now! |
@albangossard @JordiBolibar I worked a bit more on the solution of #67
I am still not sure the computed projections are 100% correct, but this is a first step forward in having actual latitude/longitude grid. Please take a look at the code and let me know if you have any suggestions!
Please don't accept this PR yet, since I will probably try to polish a few pieces in the code soon. But at least you can have a flavor of how this would look like.