Aggregation to a polygon

[BSchilperoort]

While it is not really regridding, aggregation to a shapefile (polygons) should also be possible as an extension to the current capabilities. This can aid, e.g., hydrologists in aggregation of precipitation to a basin.

PyShp can read shapefiles and is implemented in pure python.
Shapely has binary wheels on PyPI so it does not require compilation or conda.

In a quick test I have managed to get a 2D array with overlap fractions between a polygon from a shapefile and a grid.

1. Load in shapefile, grab first polygon
2. Turn grid coordinates into polygons using shapely.box
3. Compute overlap between every grid box and the shapefile polygon
4. Massage back into 2D array

This 2D array can be converted into weights to transform the gridded data into mean values. The 2D array can be (very) sparse.

Additional optimization can occur by making off part of the grid that is out of bounds of the shapefile polygon. This will speed up computation significantly.

Quick demo here: https://gist.github.com/BSchilperoort/db955c4183b3721493875688d7417026

[BSchilperoort]

This post by @rabernat actually goes through the entire workflow, in an efficient manner: https://discourse.pangeo.io/t/conservative-region-aggregation-with-xarray-geopandas-and-sparse/2715

Instead of geopandas, we can make use of shapely and pyshp.