RGB composite images

An RGB composite combines three separate data fields — one for each of the red, green and blue colour channels — into a single true-colour (or false-colour) image. This technique is widely used with satellite imagery to produce natural-looking scenes or to highlight specific atmospheric or surface features.

In this example we retrieve three visible channels from a Meteosat Second Generation (MSG) satellite swath via MARS and combine them into a composite image.

import earthkit.data as ekd

import earthkit.plots as ekp

Fetching the data

We request channels 1, 2 and 3 of MSG instrument 210 (SEVIRI) from a sample file, which returns a three-field FieldList - one field per channel - which we can index directly.

data = ekd.from_source("sample", "rgb-sample.grib").to_fieldlist()

High-level API: ekp.geo.rgb_composite

For a quick single-panel composite over a specific domain, use the high-level earthkit.plots.rgb_composite function. earthkit-plots normalises each channel to the range [0, 1] automatically before combining them, so you do not need to pre-process the data.

ekp.geo.rgb_composite(data, domain="Africa").show()

<earthkit.plots.components.figures.Figure at 0x1141c7510>

Visualising individual channels

Before compositing it can be helpful to look at the individual channels. Each channel captures slightly different wavelengths of visible light, so bright regions in one channel may be darker in another.

We create a 2×2 figure, plot each channel in its natural colour, then reserve the last panel for the composite.

import cartopy.crs as ccrs

COLORS = {
    "red": (1, 0, 0),
    "green": (0, 1, 0),
    "blue": (0, 0, 1),
}

figure = ekp.Figure(crs=ccrs.NearsidePerspective(), rows=2, columns=2)

for i, color in enumerate(COLORS):
    channel_plot = figure.add_map()
    channel_plot.grid_cells(data[i], colors=["black", COLORS[color]])
    channel_plot.title(f"{color} channel")

composite = figure.add_map()
composite.rgb_composite(data)
composite.title("RGB composite")

figure.title("RGB composite image at {time:%H:%M on %-d %B %Y}")
figure.show()