Add VectorField/WindBarbs project operation

[philippjfr]

• Fixes VectorField does not reproject angle #295

[poplarShift]

As usual, terrific response time!

I believe you made a mistake in the calculation of us, vs from the angle, see my commit: project_vectors...poplarShift:project_vectors

Your fix seems fine otherwise. It does change the values of the angles though that get fed into the color mapping, so it's not ideal, but a good step in the right direction.

Geoviews:

Cartopy:

[philippjfr]

Thanks for the fixes. I think the coloring issue is to do with the magnitudes being transformed but the color range being computed before the transformation. I'll think about how best to fix that.

[ahuang11]

import numpy as np
import geoviews as gv
import matplotlib.pyplot as plt
import cartopy.crs as ccrs
import panel as pn
from metpy.calc import wind_components, wind_direction
from metpy.units import units

gv.extension("matplotlib")
pn.extension()

x = np.linspace(10, 50, 3)
X, Y = np.meshgrid(x, x)
U, V = 10 * X, 5 * Y
angle = np.pi / 2 - np.arctan2(-V, -U)
magnitude = np.hypot(U, V)

fig = plt.figure()
ax = plt.axes(projection=ccrs.Orthographic())

ax.quiver(X, Y, U, V, transform=ccrs.PlateCarree())
ax.set_global()
ax.coastlines()

vectorfield = (
    gv.VectorField(
        (X, Y, angle, magnitude), label="Vector Field", crs=ccrs.PlateCarree()
    ).opts(
        global_extent=True,
        projection=ccrs.Orthographic(),
        magnitude="Magnitude",
        show_legend=False,
    )
    * gv.feature.coastline()
)
pn.Row(
    pn.pane.Matplotlib(fig, dpi=180, width=400, height=400),
    pn.panel(vectorfield, width=300, height=300),
)

fig = plt.figure(figsize=(6, 6))
ax = plt.axes(projection=ccrs.Orthographic())
ax.barbs(
    X,
    Y,
    U,
    V,
    transform=ccrs.PlateCarree(),
    linewidth=0.5,
    sizes=dict(emptybarb=0.25, spacing=0.2, height=0.5),
)
ax.set_global()
ax.coastlines()

windbarbs = (
    gv.WindBarbs((X, Y, angle, magnitude), crs=ccrs.PlateCarree()).opts(
        global_extent=True,
        projection=ccrs.Orthographic(),
        show_legend=False,
        linewidth=0.5,
        sizes=dict(emptybarb=0.25, spacing=0.2, height=0.5),
        fig_size=200,
    )
    * gv.feature.coastline()
)

windbarbs_uv = gv.WindBarbs.from_uv((X, Y, U, V), crs=ccrs.PlateCarree()).opts(
    global_extent=True,
    projection=ccrs.Orthographic(),
    show_legend=False,
    linewidth=0.5,
    sizes=dict(emptybarb=0.25, spacing=0.2, height=0.5),
    fig_size=200,
    )

pn.Row(
    pn.pane.Matplotlib(fig, dpi=300, width=400, height=400),
    pn.panel(windbarbs, width=400, height=400),
    pn.panel(windbarbs_uv, width=400, height=400),
)

[ahuang11]

I think this is ready for review.

[hoxbro]

When I look at the figures, there seem to be differences in Matplotlibs and this implementation as we move closer to the border of the globe, see:

Do you know why?

[ahuang11]

When I look at the figures, there seem to be differences in Matplotlibs and this implementation as we move closer to the border of the globe, see:

Do you know why?

I suspect it's due to:

pivot{'tail', 'mid', 'middle', 'tip'}, default: 'tail'
The part of the arrow that is anchored to the X, Y grid. The arrow rotates about this point.

'mid' is a synonym for 'middle'.

[ahuang11]

Setting pivot='tail'

vectorfield = (
    gv.VectorField(
        (X, Y, angle, magnitude), label="Vector Field", crs=ccrs.PlateCarree()
    ).opts(
        global_extent=True,
        projection=ccrs.Orthographic(),
        magnitude="Magnitude",
        show_legend=False,
        pivot="tail",
    )
    * gv.feature.coastline()
)
pn.Row(
    pn.pane.Matplotlib(fig, dpi=180, width=400, height=400),
    pn.panel(vectorfield, width=300, height=300),
)

Results in the same plot:

[ahuang11]

All set.