Atlite now requires Python 3.6 or higher.
We changed the Atlite backend for storing cutout data. Existing cutouts either need to be migrated with the appropriate functions or (what we recommended) recreated.
The backend change also includes some changes to the API. Most notably:
The xarray for cutouts is now exposed as Cutout.data
The Cutout.meta attribute was deprecated in favour of Cutout.data.attrs
xarray and dask can now handle some data caching automatically. If you wish to preload some data before your calculation, you can now use Cutout.data.load() to load all of the cutouts data into memory. (Warning: Requires a large enough memory.)
- The order of coordinates (indices) for Cutouts changed: x and y (e.g. longitude and latitude) are now
both ascending (before: x ascending and y descending).
The argument show_progress of function atlite.convert.convert_and_aggregate does not take strings anymore.
The argument layout of function atlite.convert.convert_and_aggregate must be a xarray.DataArray.
- Due to the change of the order of coordinates in cutouts the order of coordinates in matrix passed to convert_* functions
changed likewise: x and y are both ascending now.
Due to the change of the order of coordinates in cutouts the order of elements returned by grid_coordinates() has changed.
Due to the change of the order of coordinates in cutouts the order of elements in the attribute grid_cells has changed.