Deal with data/no-data boundaries

[cortadocodes]

Background

The GLO-30 dataset provides data in 1° x 1° lat/lng tiles that contain land, not over areas that are just sea. Two things happen when we reach the sea:

• It pads coastlines with 0m elevations for some distance (unknown to me at this time, but probably just whichever parts of the sea are within its land-containing tiles)
• For areas that are purely in the ocean, it provides no corresponding files within the S3 bucket

Problem

We want to provide elevations corresonding to H3 cells, which is a hierarchical coordinate system. We have two problems for cells spanning land and sea:

1. For cells small enough to contain only the parts of the sea that are recorded as 0m elevations, are averages of the maximum resolution cell elevations meaningful?
2. What do we do for cells large enough to contain parts of the sea for which there is no data at all?

Temporary solution

Set any minimum resolution cells with no data to 0m elevation

• We can add a field to elevation nodes to differentiate them from cells that do have data that is 0m
• This makes hierarchical elevation calculations much easier

Proposed long term solution

Create a classifier that decides whether a cell is more on land than on sea

• This would be calculated by checking if a majority of the cell's children have a non-zero elevation, ie 4/7 cells are non-zero
• Record whether the cell is considered land or sea
• If a cell is more on sea than land, we can return null or 0m for it (or do something else)