Regridding Design Document

Introduction (minimally needed)

This package will provide geospatial regridding/remapping functionality on xarray data objects

Background

Geospatial gridded datasets (climate models, remote sensing images, etc.) frequency exist on dissimilar grids. Existing tools (CDO, NCL, NCO, ESMF, etc.) have addressed this with both stand-alone command line tools and python libraries. While existing python libraries have some of the desired functionality, they do not currently interface with xarray.

Requirements

This is a list of requirements:

• Interface directly with xarray, utilizing coordinates defined on DataArray objects.
• Provide a remap-like interface (da.remap.remap_like(da_target, how='bilinear'))
• Provide a remap-to interface (da.remap.remap_to([('lat', 0.5), ('lon', 0.75)], how='nearest'))
• Provide access to mapping tables
• A series of resampling/remapping algorithms
  • linear/bilinear interpolation
  • distance weighted
  • nearest neighbor -Nearest neighbor for 1D source coordinates will probably make it into xarray as a special case for my vectorized indexing rewrite (in reindex/reindex_like)
  • conservative area (1st/2nd order)
  • largest area fraction
• Use dask for out of core / scaling
• or just wrap existing remapping lib

Design

My original thought here would be to xarray’s accessor namespace on top of the DataArray. In my examples above, this would be the remap attribute.

To be efficient, many of these regridding techniques will need to make use of a tree datastructure for multi-dimensional lookups. It would nice could be integrated a bit more deeply into xarray (e.g., KDTreeIndex)

Application programing interfaces (API)

Often it is useful to specify function signatures, class methods / attributes.

Testing

This lists tests that should be used to verify that the code is working correctly, especially with respect to meeting requirements. Ideally, this would be a list of unit tests demonstrating that requirements have been met.

link to hackpad