Polar stereo stipple() fails for data on curvilinear/2d grid

[dlrhodson]

Hi Sadie!

Subject of the issue

Attempting to stipple() over a polar stereo contour plot from data that had 2d lon and lat grid (e.g lon(i,j)) fails.

Steps to reproduce

See attached code

Your environment

The version of cf-plot and its dependencies in which you are encountering an issue.
cf.environment(paths=False) Platform: Linux-5.14.0-503.38.1.el9_5.x86_64-x86_64-with-glibc2.34 HDF5 library: 1.14.2 netcdf library: 4.9.4-development udunits2 library: libudunits2.so.0 esmpy/ESMF: not available Python: 3.10.16 dask: 2025.3.0 netCDF4: 1.7.2 h5netcdf: 1.6.3 h5py: 3.14.0 s3fs: 2025.5.1 psutil: 7.0.0 packaging: 24.2 numpy: 2.2.4 scipy: 1.15.3 matplotlib: 3.10.3 cftime: 1.6.4.post1 cfunits: 3.3.7 cfplot: 3.4.0 cfdm: 1.12.1.0 cf: 3.17.0

Expected behaviour

A stippled plot should be produced

Actual behaviour

Tell us what happens instead. If you see an error please provide the traceback, ideally in full.

`Traceback (most recent call last):
File "/apps/jasmin/jaspy/miniforge_envs/jaspy3.11/mf3-23.11.0-0/envs/jaspy3.11-mf3-23.11.0-0-v20240815/lib/python3.11/site-packages/cfplot/cfplot.py", line 9579, in add_cyclic
field, lons = cartopy_util.add_cyclic_point(field, lons)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/apps/jasmin/jaspy/miniforge_envs/jaspy3.11/mf3-23.11.0-0/envs/jaspy3.11-mf3-23.11.0-0-v20240815/lib/python3.11/site-packages/cartopy/util.py", line 69, in add_cyclic_point
raise ValueError('The coordinate must be 1-dimensional.')
ValueError: The coordinate must be 1-dimensional.

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "/gws/nopw/j04/canari/users/dlrhodso/spring2025/analysis/FWC/polar_plot_example/./polar_test_2d_coord_data.py", line 10, in
cfp.stipple(f=data,min=28.0,max=40.0)
File "/apps/jasmin/jaspy/miniforge_envs/jaspy3.11/mf3-23.11.0-0/envs/jaspy3.11-mf3-23.11.0-0-v20240815/lib/python3.11/site-packages/cfplot/cfplot.py", line 5073, in stipple
field, xpts = add_cyclic(field, xpts)
^^^^^^^^^^^^^^^^^^^^^^^
File "/apps/jasmin/jaspy/miniforge_envs/jaspy3.11/mf3-23.11.0-0/envs/jaspy3.11-mf3-23.11.0-0-v20240815/lib/python3.11/site-packages/cfplot/cfplot.py", line 9583, in add_cyclic
field, lons = cartopy_util.add_cyclic_point(field, lons)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/apps/jasmin/jaspy/miniforge_envs/jaspy3.11/mf3-23.11.0-0/envs/jaspy3.11-mf3-23.11.0-0-v20240815/lib/python3.11/site-packages/cartopy/util.py", line 69, in add_cyclic_point
raise ValueError('The coordinate must be 1-dimensional.')
ValueError: The coordinate must be 1-dimensional.
`

Relevant data

The attached tar file contains polar_test_2d_coord_data.py which attempts to plot a stipple over data on a curvilinear grid (and fails). polar_test_1d_coord_data.py attempts to do the same, but after regridding the data to a standard lon-lat grid (ie 1d coordinate arrays). This works.
The proposed fix is stipple_fix.py adn regrid_fix.py which only makes minor changes to stipple() - to avoid applying cyclic corrections to the longitude coordinates for a 2d longitude array, but also re-writes regrid() to allow the regridding of data on a curvilinear grid (disclosure: I asked co-pilot to write this for me!) - this seems to work!

polar_plot_stipple_examples.tar.gz

[dlrhodson]

I modified the original comment as I had forgotten that regrid() also needs to change to make this work.

[dlrhodson]

One final comment - although regridding to a lon-lat grid is a work-around here, this is not optimal over the poles as the converging longitude lines mean we need a high resolution grid to resolve the details contained on the curvilinear grid over the poles.

[sadielbartholomew]

@dlrhodson Hi Dan! I am sorry, somehow I missed this post when it went up - I suspect I was on leave and it got hidden and lost amongst many other GitHub notifications when I returned. Anyway have now spotted this in the Issue Tracker, better late than never...

I'll add this to my list of issues to investigate. I understand you have a workaround but that it isn't ideal so I'll try to have something sorted soon, but I don't think I'll be able to fit in a new release of cf-plot for at least a month, so if you want to use a fixed version we may need to find a way to patch a local copy of the codebase pulled from GitHub (or a similar solution) with the fix in the meantime.

I will look at your proposed fix. If it seems a sensible way to address the issue - and I will report back here to let you know - then it would be ideal to credit you for your contribution of the fix, so either I can set your username as the commit authors in a PR I do, or you'd be welcome to do a PR of your own if you have the bandwidth and would like to.

Will report back by early next week at the latest!

[sadielbartholomew]

Sorry Dan just to add I may have to push back looking into this until next week (hosting a summer student this week so my schedule is pretty full), but I am aware of it now and will address it as soon as I have bandwidth to.

[dlrhodson]

Absolutely no rush on this one! I should probably get my act together and do a PR!

[sadielbartholomew]

Hi @dlrhodson,

I've taken a look at this now. Your solution certainly works, though really the code should be handling cyclic axes in a much more robust way than Andy's code approach (which you've made a small edit within, effectively) of inequality against 360 values - notably using cf-python field and axes querying methods such as field.cyclic, so eventually I will go in and sort all handling of cyclicity which would overwrite your changes.

But for now, this is certainly sufficient to solve the issue until I have time to overhaul the cyclicity logic (I am pushed for time at the moment sadly). I will add it in as a PR including a test to isolate the underlying bug you found.

Thanks for your work! To acknowledge your contribution, unless you state otherwise, I'll set your GitHub username/handle as the author of the fix commit, since you wrote it so it's best you get formally recognised for it, and I'll add your name to our list of contributors (you've contributed a lot before by way of bug reports at least so it is overdue to have your name on there I think!).

[sadielbartholomew]

OK I have initiated the fix PR for this in a branch at https://github.com/sadielbartholomew/cf-plot/tree/stipple-2d-curvi-grid but am about to start a small block of leave until the end of the bank holiday. Upon my return and by the middle of next week I can get it merged to close this Issue and then I'll aim to do a new release in September so it will work for you in a pip/conda-installable cf-plot.

But you've saved me the work and created your own workaround which I trust will work in the meantime! Thanks Dan!