brianpm
commented
2 years ago @juliecaron I was messing around with vector plots and I think there are a few decisions that have to be made while designing these plots.
There are at least two distinct options for making the "mean" plots: quiver and streamplot. Playing around with them, I see pros and cons for each. They can both look good, but the quiver option with a contourf to show magnitude might be more information dense. I'll label that one as "Candidate 1" in this example, and the streamplot as "Candidate 2". For the difference panel, I'd also lean toward a quiver plot with a contour. In the example, I boosted the opacity parameter for the contour colors (alpha) to make the vectors more visible.
In both quiver plots, the density of the vectors needs to be reduced. I did that by striding through the data, I'm not sure if there is a better way. The main problem there is that how big the strides should be is dependent on the size of the data. Maybe an option is to make a guess for how many vectors should fit within the plot and then set the stride based on that and the grid size of the data.
Here's my example code and resulting plot:
proj = ccrs.PlateCarree()
lons, lats = np.meshgrid(uplev['lon'], uplev['lat'])
fig = plt.figure(figsize=(20,12))
# LAYOUT WITH GRIDSPEC
gs = gridspec.GridSpec(2, 4)
gs.update(wspace=0.25)
ax1 = plt.subplot(gs[0, :2], projection=proj)
ax2 = plt.subplot(gs[0, 2:], projection=proj)
ax3 = plt.subplot(gs[1, 1:3], projection=proj)
# too many vectors to see well, so prune by striding through data:
skip=(slice(None,None,5),slice(None,None,8))
# Candidate 1:
# - contourf to show magnitude w/ colorbar
# - vectors (colored or not) to show flow --> subjective (?) choice for how to thin out vectors to be legible
img1 = ax1.contourf(lons, lats, np.sqrt(uplev**2 + vplev**2), cmap='Greys', transform=ccrs.PlateCarree())
ax1.quiver(lons[skip], lats[skip], uplev[skip], vplev[skip], np.sqrt(uplev**2 + vplev**2).values[skip], transform=ccrs.PlateCarree(), cmap='Reds')
ax1.set_title(f"CAM5 Wind [{plev} hPa]")
## Add colorbar to Candidate 1:
cb_c1_ax = inset_axes(ax1,
width="3%", # width = 5% of parent_bbox width
height="90%", # height : 50%
loc='lower left',
bbox_to_anchor=(1.01, 0.05, 1, 1),
bbox_transform=ax1.transAxes,
borderpad=0,
)
fig.colorbar(img1, cax=cb_c1_ax)
# Candidate 2:
# - streamplot that shows the flow, colored to show magnitude (with colorbar)
strm = ax2.streamplot(lons, lats, u2plev.values, v2plev.values, density=2, color=np.sqrt(u2plev**2 + v2plev**2).values, cmap='Reds', transform=ccrs.PlateCarree())
ax2.set_title(f"CAM4 [{plev} hPa]")
cb_c2_ax = inset_axes(ax2,
width="3%", # width = 5% of parent_bbox width
height="90%", # height : 50%
loc='lower left',
bbox_to_anchor=(1.01, 0.05, 1, 1),
bbox_transform=ax2.transAxes,
borderpad=0,
)
fig.colorbar(strm.lines, cax=cb_c2_ax)
# Showing the difference is probably easiest with a contour plot,
# but then the anomalous circulation is hard to see, so could add vectors on top
udif = u2plev - uplev
vdif = v2plev - vplev
img3 = ax3.contourf(lons, lats, np.sqrt(uplev**2 + vplev**2)-np.sqrt(u2plev**2 + v2plev**2), transform=ccrs.PlateCarree(), norm=mpl.colors.TwoSlopeNorm(vmin=-5, vcenter=0.0, vmax=5), cmap='PuOr', alpha=0.5)
ax3.quiver(lons[skip], lats[skip], udif[skip], vdif[skip], transform=ccrs.PlateCarree())
# img3 = ax3.streamplot(lons, lats, udif.values, vdif.values, transform=ccrs.PlateCarree())
# img3 = ax3.barbs(lons[skip], lats[skip], udif.values[skip], vdif.values[skip], transform=ccrs.PlateCarree()) # does not look good for global data
ax3.set_title(f"Difference []", loc='left')
cb_d_ax = inset_axes(ax3,
width="3%", # width = 5% of parent_bbox width
height="90%", # height : 50%
loc='lower left',
bbox_to_anchor=(1.01, 0.05, 1, 1),
bbox_transform=ax3.transAxes,
borderpad=0
)
fig.colorbar(img3, cax=cb_d_ax)
[a.coastlines() for a in [ax1,ax2,ax3]]
juliecaron
andrewgettelman
New feature type
New plot and/or plot enhancement
What is this new feature?
we have lat/lon plots of scalar fields, but i'm adding the vector field option
Assistance required?
Yes, I will need some assistance before making a PR
Extra info
wish me luck!!