Best Practices for Multi-Panel Plots

[sgdecker]

I was thinking of requesting an example be added that uses the declarative interface to make a four-panel plot using all the "tricks" available to make the plot look good with the minimum possible code. One such trick I have figured out is using copy() to eliminate code repetition, but I haven't figured out how to eliminate the huge margins in the resulting plot. It may be that this is a missing feature in PanelContainer.show() (maybe have it call plt.tight_layout()?), or maybe I just haven't figured out the trick yet. So this is a multi-stage discussion:

1. Determine the "best" way to eliminate huge margins in declarative plots, or add the functionality to enable this.
2. After (1) is done, add an example showing the creation of a four-panel plot (an example exists, but not using the declarative interface) using the copy trick, the elimination of giant margins, and any other improvements I haven't thought of.

Here's where I'm at so far:

from datetime import datetime

import xarray as xr
from xarray.backends import NetCDF4DataStore
from siphon.catalog import TDSCatalog
from metpy.plots import ContourPlot, MapPanel, PanelContainer
import metpy.calc as mpcalc

def get_nam211(init_time, valid_time):
    ymd = init_time.strftime('%Y%m%d')
    hr = init_time.strftime('%H')
    filename = f'{ymd}_{hr}00.grib1'
    cat_name = ('https://thredds.ucar.edu/thredds/catalog/grib/NCEP/NAM/'
                'CONUS_80km/NAM_CONUS_80km_' + filename +
                '/catalog.xml')
    ds_name = 'NAM_CONUS_80km_' + filename
    
    cat = TDSCatalog(cat_name)
    ds = cat.datasets[ds_name]
    ncss = ds.subset()
    query = ncss.query()
    query.time(valid_time).variables('all')
    nc = ncss.get_data(query)
    data = xr.open_dataset(NetCDF4DataStore(nc)).metpy.parse_cf()
    return data

init_time = datetime(2021, 9, 23, 12)
plot_time = datetime(2021, 9, 24, 0)

nam = get_nam211(init_time, plot_time)
tmpk = nam['Temperature_isobaric'][0,15,:,:]
tmpc = tmpk - 273.15
tmpc_gauss = mpcalc.smooth_gaussian(tmpc, 8)
tmpc_9pt = mpcalc.smooth_n_point(tmpc, 9)

raw = ContourPlot()
raw.data = tmpc
raw.contours = list(range(0,20))
raw.linecolor = 'green'
raw.clabels = True

gauss = raw.copy()
gauss.data = tmpc_gauss

ninept = raw.copy()
ninept.data = tmpc_9pt

kelvin = raw.copy()
kelvin.data = tmpk
kelvin.contours = list(range(0,300))

mp1 = MapPanel()
mp1.layout = (2, 2, 1)
mp1.area = [-87, -78, 40.5, 45]
mp1.projection = 'lcc'
mp1.layers = ['states', 'coastline', 'borders']
mp1.title = 'Raw Data'
mp1.plots = [raw]

mp2 = mp1.copy()
mp2.layout = (2, 2, 2)
mp2.title = 'Gaussian Filter'
mp2.plots = [gauss]

mp3 = mp1.copy()
mp3.layout = (2, 2, 3)
mp3.title = '9-pt Smoother'
mp3.plots = [ninept]

mp4 = mp1.copy()
mp4.layout = (2, 2, 4)
mp4.title = 'Kelvin'
mp4.plots = [kelvin]

pc = PanelContainer()
pc.size = (12,10)
pc.panels = [mp1, mp2, mp3, mp4]
pc.show()

I like the reduction in code repetition; I don't like the huge margins.

[dopplershift]

I wonder if the huge margins are Cartopy-related. I know in the past I've found multi-panel plots with Cartopy to be tricky to get to look how I want.

[kgoebber]

I often play with the figure size to adjust margins/spacing on the cheap. Also when you go to save the figure using the box_inches='tight' does a nice job of reducing the white space on the outside edge of the figure.

[sgdecker]

Thanks for the box_inches tip, although that triggers a MatplotlibDeprecationWarning, so there must be a "better" way to do this these days.

[sgdecker]

Ahh, bbox_inches is the correct (or at least currently correct) spelling of the argument.

[akrherz]

I ensure that adjustable="datalim" is set on the cartopy axes so that funky things are not done with the splines and my specified axes position is enforced. FWIW