Call for Matplotlib & Cartopy Suggestions/Feedback to Benefit NASA Missions

[battistowx]

Hey everyone!

We have some exciting news Thomas Caswell who is the project lead for Matplotlib over at the Brookhaven National Lab. They are on year 2 of a 5-year ROSES grant to work on improving matplotlib and cartopy to benefit NASA-funded projects, and have until the end of FY26 to gather suggestions and plan. This is an amazing opportunity to improve these two widely-supported libraries that essentially every geospatial Python coder is exposed to, and the community feedback from the Openscapes channel is a perfect place to start.

As you brainstorm, use the thread below to record any ideas or suggestions. Once we've got a good bulk of ideas, we can start to distill some goals and share them with the developers. Below, I've added some additional resources, including links to the current and previous ROSES grant submissions from Matplotlib.

Additional Resources

• Matplotlib Documentation
• Cartopy Documentation
• 2024-NASA-ROSES-F7 Proposal (Current)
• 2020-NASA-ROSES-E7 Proposal

[jules32]

@andypbarrett had this idea yesterday: create a built-in Projection class for NSIDC Polar Sterographic and EASE Grid vX projections.

[andypbarrett]

More details on adding Projections for NSIDC Polar Stereographic and EASE GRID v2.

These are two projections used widely in the cryo community for Arctic and Antarctic satellite data and derived products. The NSIDC Polar Stereographic has north and south polar projections. EASE Grid vX (not really a grid but that is legacy terminology) has north, south and global projections. There are two versions of the north and south polar stereographic projections; one based on a sphere (the "Hughes 1980 ellipsoid) and one based on WGS84. There are also two versions of EASE Grid; version 1 is based on a sphere, version 2 is on WGS84 ellipsoid.

It would be really nice to have these CRS/Projections as derived classes. In particular, plugging the projection parameters for the NSIDC Polar Stereographic CRS into cartopy's XXXXXPolarStereo class does not render correctly.

NSIDC Guide to Polar Stereoraphic projections
EPSG:3411
EPSG:3412
EPSG:3413
EPSG:3976

NSIDC Guide to Ease Grids
EPSG:3408
EPSG:3409
EPSG:3410
EPSG:6931
EPSG:6932
EPSG:6933

3411, 3412, 3408, 3409, 3410 are all marked as Superceded (but not deprecated) in the EPSG registry but are still used by some legacy products. The WGS84 versions are the preferred CRS for new datasets.

Note: I am using epsg.org which the canonical registry for EPSG codes and associated CRS. epsg.io is out dated and not maintained.

[betolink]

More:

It'll be great to have better integration with xarray and geospatial data, so it gets easier to go from an xarray DataArray with spatial coordinates and CRS information to a good map with minimal configuration. i.e. automatically detecting the CRS from CF/xarray metadata and using Cartopy appropriately for the projection, extent, and coordinate transforms. This could also build on the patterns established by tools like hvPlot, where plotting geospatial xarray data is much more seamless. (Not sure if this would be duplicating what GeoViews does)

The goal would be that the users shouldn't have to understand Cartopy internals to make a correct map from an xarray dataset.
Something like

import xarray as xr
import matplotlib.pyplot as plt

ds = xr.open_dataset("temperature.nc")

# Ideally, this would "just work":
ds.temperature.plot.map() # <-- hvplot and xarray plot have no "carto" awareness 
plt.show()

instead of something like

import cartopy.crs as ccrs
import matplotlib.pyplot as plt

ax = plt.axes(projection=ccrs.NorthPolarStereo())

da.plot(
    ax=ax,
    transform=ccrs.PlateCarree(),
)

ax.coastlines()