Grid Implementation Consistency w.r.t. Paper Formulation

[mikonvergence]

@geospatial-jeff has opened an issue on the EG own repo: https://github.com/earth-genome/majortom/issues/4 and it highlights how the official code is not exactly reflecting the formulation presented in the paper.

It's up to us to decide how to document this, but given the fact that quite a few datasets have been created with the original grid implementation from this repository, we should probably leave everything as is, and instead document better.

We could consider:

• Improving the current source code documentation to clearly explain the steps and their implications
• Indicate where the discrepancy lies in the source code in the form of a comment
• Indicate where the discrepancy lies in the manuscript (possible to update in arxiv)
• Release precomputed geometry of the most common Major TOM 10 km grid (and possibly a few others)

[mikonvergence]

Update (copied from the other issue):

I finally had time to look at what we put in the paper.

The ESA original implementation is in my view closer to that definition. We have even provided an example in text:

"So, row ‘317D’ is at latitude −317δ_lat."

The solutions are equivalent only if N_r is an even number. However, if it is odd, then one of the "spacing segments" of length δ_lat is positioned exactly on top of the equator, which means that it's reference latitude will be below equator.

The official implementation gets around this via circular shift of the +-90 degree point (end and beginning of the segment sequence) to the 0 degree line.

I don't suppose any change to the manuscript is needed, but we should probably clean up documentation of the ESA repository to explain this more carefully.

I will leave this open until we document the code to make a note of this