A better way to caculate and handle cml coverage maps

[cchwala]

Idea:

• Calculate the covered pixels for a given grid for each cml
• Store this info along with the Comlink data
• Maybe store everything in a xarray dataset:
  • cml rainfall series (all with a common time index) indexed by cml_id
  • cml rainfall grid (lat, lon, time)
  • cml coverage (lat, lon, cml_id)
  • or maybe cml coverage (list_of_covered_x_y_indices, cml_id)

This way, once all coverage maps for the individual cml have been calculate, it would be easy to get the coverage for a given point in time, taking into account the varying NaNs of the CML rainfall time series.

Open questions:

• speed of calculating coverage for each CML, instead of building a convex hull for all of them in geopandas/shapely? (but, could be parallelized)
• size impact of coverage matrices?
  • matrices would only be of dtype bool
  • however, for large grids and a thousands of CMLs, that would still be a lot of bits...
  • maybe storing only the covered indices is better, just include a simple way to project these onto the grid