Point cloud plots

A point cloud plot renders each data value as a coloured scatter point at its geographic location. There is no interpolation — each source point appears as an individual dot whose colour encodes its value.

This makes point_cloud a good choice for:

• unstructured grids such as HEALPix or reduced Gaussian, where you want to see the raw cell distribution before any regridding,

• sparse observation data where points are irregularly spaced,

• quickly checking the coverage and values of a dataset.

For a filled-surface rendering of the same data, use grid_cells (which preserves cell boundaries) or contourf (which interpolates between values).

Example: HEALPix 2 m temperature

We will use a HEALPix GRIB file at H128 resolution. The HEALPix grid distributes points roughly uniformly over the sphere, so a point cloud gives a good visual impression of the native grid density.

import earthkit.data as ekd

import earthkit.plots as ekp

data = ekd.from_source("sample", "healpix-h128-nested-2t.grib")

chart = ekp.Map(domain="Europe")

chart.point_cloud(data, units="celsius")

chart.coastlines()
chart.gridlines()
chart.legend()
chart.title()

chart.show()

Controlling point size

The s keyword argument (passed through to matplotlib’s scatter) controls the point size in points². Smaller values reduce overlap on dense grids; larger values make individual points easier to see when the data is sparse.

figure = ekp.Figure(rows=1, columns=2, domain="Europe")

ax = figure.add_map()
ax.point_cloud(data, units="celsius", s=2)
ax.title("s=2 (small points)")

ax = figure.add_map()
ax.point_cloud(data, units="celsius", s=20)
ax.title("s=20 (large points)")

figure.coastlines()
figure.legend(location="right")

figure.show()

Applying a style

You can pass a Style object to control the colour map, contour levels and units — the same way as with other plotting methods.

style = ekp.styles.Style(
    colors="Spectral_r",
    levels=range(-10, 35, 5),
    units="celsius",
    extend="both",
)

chart = ekp.Map(domain="Europe")

chart.point_cloud(data, style=style, s=5)

chart.coastlines()
chart.gridlines()
chart.legend()
chart.title()

chart.show()

Point cloud vs. grid cells

The key difference between point_cloud and grid_cells is how each grid cell is drawn. point_cloud places a single dot at the cell centroid, while grid_cells shades the full area of the cell.

figure = ekp.Figure(rows=1, columns=2, domain=["France", "Spain"])

ax = figure.add_map()
ax.point_cloud(data, style=style, s=10)
ax.title("point_cloud")

ax = figure.add_map()
ax.grid_cells(data, style=style)
ax.title("grid_cells")

figure.coastlines()
figure.borders()
figure.legend(location="right")

figure.show()