We've been looking at adding support for reading in MPAS mesh files in UXarray, with some internal discussion about it here.
Internally, we represent unstructured grids using the UGRID conventions by encoding the input grid at the data loading step. We'd like to support directly converting from MPAS to UGRID.
One workaround I've explored is using the mpas_tools.scrip.from_mpas.scrip_from_mpas() function to first convert from MPAS to SCRIP and then from SCRIP to UGRID, since UXarray currently support converting from SCRIP to UGRID. However, there are a couple comments I'd like to discuss:
NetCDF
The current implementation of mpas_tools.scrip.from_mpas.scrip_from_mpas() is written with NetCDF and operates using file paths, meaning that it doesn't plug in nicely with our Xarray-styled code and workflows
We are interested in working directly with Xarray.Datasets without writing anything to disk
I've converted mpas_tools.scrip.from_mpas.scrip_from_mpas() to use Xarray instead of NetCDF as shown here with some more examples here
Would having an equivalent Xarray implementation of these conversion functions be something worth looking more into, especially with the dominance of Xarray in many scientific python workflows.
MPAS to UGRID
The solution to reading in MPAS files in UXarray works, but having a direct conversion from MPAS to UGRID would be more ideal
Are there any thoughts about implementing a ugrid_from_mpas() or similar function?
We are open to learning more about how we could best leverage the tools here and for any possible collaboration.
We've been looking at adding support for reading in MPAS mesh files in UXarray, with some internal discussion about it here.
Internally, we represent unstructured grids using the UGRID conventions by encoding the input grid at the data loading step. We'd like to support directly converting from MPAS to UGRID.
One workaround I've explored is using the
mpas_tools.scrip.from_mpas.scrip_from_mpas()function to first convert from MPAS to SCRIP and then from SCRIP to UGRID, since UXarray currently support converting from SCRIP to UGRID. However, there are a couple comments I'd like to discuss:NetCDF
mpas_tools.scrip.from_mpas.scrip_from_mpas()is written withNetCDFand operates using file paths, meaning that it doesn't plug in nicely with our Xarray-styled code and workflowsXarray.Datasetswithout writing anything to diskmpas_tools.scrip.from_mpas.scrip_from_mpas()to useXarrayinstead ofNetCDFas shown here with some more examples hereXarrayimplementation of these conversion functions be something worth looking more into, especially with the dominance ofXarrayin many scientific python workflows.MPAS to UGRID
ugrid_from_mpas()or similar function?We are open to learning more about how we could best leverage the tools here and for any possible collaboration.
Pinging @erogluorhan and @anissa111