Organizing and storing model outputs

[pkerins]

As we continue to move beyond the predefined study areas of the Atlas of Urban Expansion, we need a better way to store and organize map outputs. For example, a given area could be mapped multiple times by the same model (ie model applied to different imagery); in theory each of those multiple images could also be mapped by multiple models. Then there are composite maps to deal with, along with ancillary data like cloud masks, cloud scores, raw predictions values, etc. At the moment we don't have a good, intuitive, and above all navigable way of storing these different potential outputs.

[pkerins]

For the record, the current solution here isn't terrible: you can group and store outputs by place & model, or however you like. You just have to remember to use the map_id parameter in map_scenes / map_tile. (If you don't, there is risk of overwriting; at a minimum, the outputs will be stored by the very opaque scene_id. Maybe this should be a mandatory argument.) Those outputs will then be stored by scene within a dedicated director inside of .../maps/. Not bad, but there is probably still room for improvement.