Positioning wind roses onto an existing x-y axis

[guidocioni]

I have a meteogram with many variables and I would like to add a "strip" with wind roses like this

I'm using gridspec and already made space for an axis (ax_winds). Following the example on the documentation I'm trying to create an inset ax based on this axis and just tried to plot a single wind rose at a certain point in time.

gs = gridspec.GridSpec(nrows=4, ncols=1, height_ratios=[1, 1, 0.3, 1])
ax_winds = plt.subplot(gs[2])
wrax_cham = inset_axes(ax_winds,
         width=1,                             # size in inches
         height=1,                            # size in inches
         loc='center',                        # center bbox at given position
         bbox_to_anchor=(dset_city['valid_time'][5], 0.5), # position of the axe
         bbox_transform=ax_winds.transData,    # use data coordinate (not axe coordinate)
         axes_class=WindroseAxes,    # specify the class of the axe
)
wrax_cham.bar(wdir[:,5], wspd[:,5], opening=1, edgecolor='white')

But I'm getting some errors

ValueError: 'bboxes' cannot be empty

which are probably related to how I'm passing the parent axis to inset_axes.

Does anyone have an example of positioning multiple wind roses over a define time axis?

Full code is here if someone is interested https://github.com/guidocioni/icon_d2_eps/blob/master/plotting/plot_meteogram.py

[ocefpaf]

What happens if you pass the projection in your subplot?

plt.subplot(gs[2],  projection"="windrose")

[guidocioni]

projection"="windrose"

Yep, that fixes the issue. However now I have two different wind roses appearing...

Any idea what's going on? For now I'm just plotting a single time step

ax_winds = plt.subplot(gs[2], projection="windrose")
wspd = wind_speed(dset_city['u10'], dset_city['v10']).metpy.dequantify()
wdir = wind_direction(dset_city['u10'], dset_city['v10']).metpy.dequantify()
wrax_cham = inset_axes(ax_winds,
         width=0.8,                             # size in inches
         height=0.8,                            # size in inches
         loc='center',                        # center bbox at given position
         bbox_to_anchor=(dset_city['valid_time'][5], 0.3), # position of the axe
         bbox_transform=ax_winds.transData,    # use data coordinate (not axe coordinate)
         axes_class=WindroseAxes,    # specify the class of the axe
)
wrax_cham.contourf(wdir[:,30], wspd[:,30])
wrax_cham.tick_params(labelleft=False, labelbottom=False)
ax_winds.xaxis.set_major_locator(locator)
ax_winds.xaxis.set_major_formatter(formatter)
ax_winds.tick_params(axis='y', which='major', labelleft=False)
ax_winds.tick_params(axis='x', which='both', labelbottom=False)

[ocefpaf]

Not sure. If you can make a self-contained example that one can easily run without external data I can try to debug. My guess is that mpl is drawing the polar axis too and not only just the windrose.

[guidocioni]

Not sure. If you can make a self-contained example that one can easily run without external data I can try to debug. My guess is that mpl is drawing the polar axis too and not only just the windrose.

I noticed that if I reduce the size of the axis then it seems to fill the background circle.
Anyhow, I took the time and packed a self-contained example with some test data (you need some data to plot...): it's really small, have a look.

https://drive.google.com/file/d/1SZo0rzTIO6nSDkHVywXg_NeZEX7ptg5L/view?usp=sharing
https://drive.google.com/file/d/14WfhMxCta5xHjkulaznmgxj6lUzt-Rcm/view?usp=sharing

I'm trying to reproduce something similar to this

of Course, it doesn't have to be exactly the same.