Investigate and Compare Area Averaging in xarray and CDAT

[pochedls]

Specific details to investigate include:

• Generation of weights (particularly on non-rectangular grids)
• Ability to define regions (e.g., tropics or Niño 3.4 region)
• Handling of masked data
• Handling of 3D data (e.g., if you average a variable x[time, height, lat, lon], does it return xa[time, height]?)

[pochedls]

In explore_spatial_averages.ipynb I compare spatial averaging in xarray and CDAT. xarray has limited built-in functionality for generating weights (e.g., for regional averages), which does matter a little bit in comparisons with CDAT. It is able to subset regions using the .select() functionality, but if the native grid and target region don't have boundaries that match the weighting will not be exactly right (even on simple rectilinear grids). There is a need to extend xarray with better functionality in this regard.