Matplotlib plots

[csteele2]

I have a couple of questions that maybe others have come across. While the gallery of the uxarray documentation shows figures with different projections, when I use the same code, I get different looking figures, presumably because the image returned seems to have a regular axes, not a geoaxes.

My other issue is for real plots. Interactive plots in a notebook are nice, but what if I wanted to make scripted png plots? As far as I have tried, I can not extract a matplotlib figure or axes, as only an image gets returned. Is there a way to plot the data directly in matplotlib without using the built-in .plot? I would like to have more, easier, control over the axes and figure than seems possible using the built-in functions and not have the holoviews,etc. dependecies on a headless server. Additionally, a contour function would very much be needed, but I don't see much on that. There are some tangential threads on this already, but in order to really use MPAS seriously, I need to be able to make similar plots as a WRF workflow. I don't consider interpolating the data back to a regular grid to be more than a stopgap, as that has high potential to introduce error (especially with any sort of station verification). Happy to add any portion of this to the prioritizations thread if appropriate.

[philipc2]

Hi @csteele2

Our direct matplotlib support is very minimal as of right now. We have some support for creating PolyCollection and LineCollection objects that can be used in within matplotlib workflows, however the performance is much less than ideal and we haven't spend much time recently looking into it. However, this is something I can take a look at this upcoming iteration, since it does seem like someone the community would benefit from.

We can track the progress with #840 and #430

This page might be a good read from HoloViews.. All of our .plot() methods suppot a backend='matplotlib' parameter, which uses matplotlib as the renderer through HoloViews.

[csteele2]

Okay, that looks like a decent band aid. So, the only plot method that will return a geoaxes is points? when trying like polygons, I notice it doesn't make a geoaxis, and even the rendered image is just a simple lat, lon image.

[csteele2]

Okay, I was able to pull the matplotlib elements out of holoviews and use them, albeit with some difficulties remaining when a geoaxes isn't returned.

[philipc2]

@csteele2

We've added better support for directly using Matplotlib in our most recent release. We've added a user guide section describing how to generate plots, including projection and geographic feature support

https://uxarray.readthedocs.io/en/latest/user-guide/mpl.html

Also pinging @rljacob @brianpm @ahijevyc if you have any input on it!

[ahijevyc]

I love the new matplotlib support--especially cartopy. I ran the examples in the user guide with my own data.