Open wwieder opened 5 years ago
We've been working with xESMF lately. We'll probably need to think about some simple wrapper to handle the multi-model ensemble but that should be doable.
I have been using CDO operators for remapping. It seems possible that we can call CDO functions in Python scripts. I have not tested yet, but these operators would be useful for zonal statistics, vertical interpolation, etc. calculations.
https://github.com/Try2Code/cdo-bindings https://code.mpimet.mpg.de/projects/cdo/wiki/Cdo%7Brbpy%7D
CDO is great for this and so many purposes (subsetting, reformatting, GRIB-NetCDF conversions, etc.) related to the kinds of data we all generate and use - fast and powerful. If it will work, why reinvent the wheel?
A good example for our gallery could be a bias plot (model compared to obs), where something needs to be regridded in the process. We could pull examples from ILAMB for this.
GeoCAT also has an interpolation function where they migrated the linint2 function from NCL https://geocat-comp.readthedocs.io/en/latest/user_api/generated/geocat.comp.linint2.html https://www.ncl.ucar.edu/Document/Functions/Built-in/linint2.shtml
I used xESMF as suggested by @jhamman, but the current approach lost information around the coasts.
Can we use existing tools for regridding multi-model ensembles (e.g. LUMIP)