Zorder for metpy.BarbPlot

[kylejgillett]

Hey folks,

I am plotting some data onto a CartoPy map but when adding some of the CartoPy features such as lakes, in particular, sit on top of the wind barbs. I have looked through the documentation and have tried setting a zorder, but it doesn't seem to be a feature for metpy.BarbPlot?

Any thoughts?

[dopplershift]

Well that's problematic. Yeah, we don't expose zorder (that's pretty nitty gritty) in the declarative, but that's not great here. Here's a work-around:

import xarray as xr

from metpy.cbook import get_test_data
from metpy.plots import ContourPlot, ImagePlot, MapPanel, PanelContainer, BarbPlot
from metpy.units import units

# Use sample NARR data for plotting
narr = xr.open_dataset(get_test_data('narr_example.nc', as_file_obj=False))

contour = ContourPlot()
contour.data = narr
contour.field = 'Temperature'
contour.level = 850 * units.hPa
contour.linecolor = 'red'
contour.contours = 15

barb = BarbPlot()
barb.data = narr
barb.field = ('u_wind', 'v_wind')
barb.level = 850 * units.hPa
barb.skip = (10, 10)

panel = MapPanel()
panel.area = 'us'
panel.layers = ['coastline', 'borders', 'states', 'rivers', 'ocean', 'land', 'lakes']
panel.title = 'NARR Example'
panel.plots = [contour, barb]

pc = PanelContainer()
pc.size = (10, 8)
pc.panels = [panel]

pc.save('temp.png')  # Triggers creation of objects
barb.handle.set_zorder(10)
pc.show()

The .handle attribute hold the matplotlib object that gets created by plotting, but it's only available after something force the plot to "materialize". I do that here by forcing the save.

@kgoebber Thoughts? I'm wondering if for BarbPlot we should set a better default.

[dopplershift]

I'll also add that apparently we have no example of BarbPlot, it's just in the tutorial.

[kylejgillett]

I actually tried this, but not by forcing the img generation first, so this does work now and is a decent temporary fix. The other thing I was going to add to this post was that an example of using and manipulating BarbPlot would be useful. Thanks again, Ryan!

[kgoebber]

Probably adding a default zorder for the barbs would be the best, rather than exposing yet another plotting trait. The gallery has plenty with a BarbPlot() example, but it would probably be better to get at least one example in the docs. It would be good to capture an example or two that expose how to access the some of the raw matplotlib elements to change things about how the axes are being plotted including this zorder trait.

What's interesting is that according to Cartopy, the zorder for adding LAND and OCEAN is -1, so that barbs don't plot over them is interesting...

[kylejgillett]

Another, kind of related question. In the Docs for FilledContourPlot(), there are limited controls for colorbar, essentially you can only set its orientation and label font size (from what I can see). I have built my plot using MapPanel() inside a PanelContainer() and I accessed the raw figure functionality to add figure.text wherever I want - which I think is why the colorbar is messed up.

I would like to manually adjust the scale and position of the colorbar, etc. Are there any simular work arounds for this as well?

[dopplershift]

@kgoebber The problem was with LAKES, not the others. I think it looks like Cartopy assigns a zorder of 1.5 in this case.

@kylejgillett Looks like we don't store the colorbar object anywhere. This is problematic, since this seems to be one of the least accessible objects in Matplotlib. The best way I can get a reference to it is: fig.axes[-1]._colorbar. I'm not sure what customization options you had in mind, but @kgoebber might have some thoughts. I'll just say in general, this is where we end up in tension between simplicity and functionality.