Making composites#

Plotting individual bands is nice but we usually want to make some composite images, both RGB and false-color, to visualize information from multiple bands at once. We provide the xlandsat.composite function to make this process easier and produce composites that are compatible with both xarray and matplotlib.

As an example, let’s load our example scene from Manaus, Brazil, which is where the Solimões (brown water) and Negro (black water) rivers merge to form the Amazon river:

Plotting composites#

Composites can be plotted using the xarray.DataArray.plot.imshow method:

rgb.plot.imshow()

<matplotlib.image.AxesImage at 0x7fbfa1ae1060>

With no arguments, xarray takes care of creating the new figure and adding a lot of the different plot elements, like axis labels and units. If we want more control over the plot, we must create a matplotlib figure and axes separately and tell imshow to plot on those instead:

fig, ax = plt.subplots(1, 1, figsize=(10, 6))

rgb.plot.imshow(ax=ax)

# The "long_name" of the composite is the band combination
ax.set_title(rgb.attrs["long_name"].title())

# Make sure pixels are square when plotting to avoid distortions
ax.set_aspect("equal")

plt.show()

Well, this looks bad because some very bright pixels in the city are making the majority of the other pixels have only a small share of the full range of available values. This can be mitigated by rescaling the intensity of the image to a smaller range of reflectance values.

Note

Using xarray.DataArray.plot instead of xarray.DataArray.plot.imshow won’t work and will display histograms instead.

Rescaling intensity (AKA contrast stretching)#

We rescale the intensities of a composite to a given reflectance range by setting the rescale_to parameter when creating a composite. It takes a list of the min and max reflectance values allowed. For this image, we can arrive at the following values by trial and error until it looks nice:

rgb = xls.composite(scene, rescale_to=[0.01, 0.2])

# Pretty much the same plotting code
fig, ax = plt.subplots(1, 1, figsize=(10, 6))
rgb.plot.imshow(ax=ax)
ax.set_title(f"Rescaled {rgb.attrs['long_name'].title()}")
ax.set_aspect("equal")
plt.show()

Notice that we can more clearly see the different colors of the forest and the rivers. However, it can still be a bit hard to distinguish between some of the water bodies and the forest in the right side of the scene. Other band combinations can generate composites that better highlight these features.

Note

The rescaling has to be done when creating the composite so that we can use min/max values in reflectance units. After a composite is created, the original range of values is lost and we’d have to specify the min/max between 0 and 255 instead.

Color infrared composites#

Another common type of composite is the color infrared (CIR) composites, which uses the NIR, red, and green bands as the red, green, blue channels. The added information of the NIR band helps highlight vegetation, which can help us distinguish between the water and forest on the right. Let’s make one by specifying this band combination to xlandsat.composite to see what happens:

cir = xls.composite(
    scene, rescale_to=[0, 0.4], bands=("nir", "red", "green"),
)

fig, ax = plt.subplots(1, 1, figsize=(10, 6))
cir.plot.imshow(ax=ax)
ax.set_title(cir.attrs["long_name"].title())
ax.set_aspect("equal")
plt.show()

In this composite, the contrast between the forest and water bodies on the right are much clearer!

Other composites#

You can make pretty much any composite you’d like by passing the correct band combination to xlandsat.composite. For example, let’s make one with NIR as red, SWIR1 as green, and red as blue:

custom = xls.composite(
    scene, rescale_to=[0, 0.4], bands=("nir", "swir1", "red"),
)

fig, ax = plt.subplots(1, 1, figsize=(10, 6))
custom.plot.imshow(ax=ax)
ax.set_title(custom.attrs["long_name"].title())
ax.set_aspect("equal")
plt.show()

This particular composite is great at distinguishing between built structures in the city and along the canals (light green), the water ways (dark blue and purple), and the forest (orange).

Making composites

Contents

Making composites#

RGB composites#

Plotting composites#

Rescaling intensity (AKA contrast stretching)#

Color infrared composites#

Other composites#