Join the Band (Combinations)
Most satellite images have multiple bands. We can change the way an image is displayed to highlight certain features in the images.
With kudos to Little Feat’s Join the Band album, we’ve been looking at satellite images a lot recently and have seen how many satellite images are composed of multiple bands, or layers. When working with imagery, we can change the way they are displayed to highlight different features on the land surface. Let’s look at an example using a Landsat 8 image zoomed to an area around Christchurch. This image hosts a whopping 11 bands, designated as shown below:
We’re spoilt for choice here. Just a few highlights before we move on. The coastal aerosol band is a new one with Landsat 8 and highlights shallow water fine particles like dust and smoke in the atmosphere. Bands 2, 3 and 4 are the visible light bands while 5 in Near Infrared. SWIR stands for shortwave infrared which, among other things, allow us to distinguish between wet earth from dry earth, and for seeing differences between surface geology not apparent in visible wavelengths. The panchromatic band we can use for pansharpening – note the finer resolution. The cirrus band works only with high altitude cirrus clouds – this is a useful layer for removing clouds from images. The final two bands sense thermal heat from the ground over two wavelength ranges, and at a much coarser resolution.
(Just another quick aside – this image was downloaded from the USGS’s Earth Explorer website. The compressed file was originally just under a Gb in size. When uncompressed it ballooned out to 3Gb and included a separate image for each band as shown at right. I then used the composite bands tool to stack them all into one image layer with 11 bands. There are some extra files in here as well. LC815Dec2016.tif is the image we’ll be looking at today.)
Let’s add this to a map in Pro and see what we’ve got:
Looks a bit funky, but there’s a good reason why. In the legend we can see entries for Red, Green and Blue and the bands that Pro is using to display them. By default, it just takes the first three bands for RGB. We know from the band table above that these don’t match. In the Symbology pane at right, we can rearrange the bands and get a more natural colour display:
(In ArcMap we can do the same from the Symbology tab.) That’s a bit more like it – let’s zoom in closer to home:
Wow – glorious! We’ve got some clouds to contend with, which is common when working with imagery. Note the sediment coming out of the Waimak – might have been a Norwest day. Also check out the disturbed sediments around the northern bays of Banks Peninsula. This image was captured on 15 Dec 2016 – here’s what the weather was like.
(Ain’t the internet amazing?) We could go into the metadata to figure out exactly what time the image was captured but let’s not. Instead, now that we’ve seen we can rearrange the bands, let’s play around with this image. Below, I’ve reset the band combinations to 5, 4, 3 – this is the standard “false colour” combination you may have seen before:
Why “false”? We’re using the infrared band for red – our eyes can’t see these wavelengths but Pro simply uses the values in the layer as if they were the red values. This allows us to visualise the infrared and is useful for highlighting vegetation – it’s akin to an NDVI layer in some ways, only completely different.
The next combination is 7, 6, 4 which is suited to distinguishing urban areas and is less susceptible to haze (though we still have clouds):
This doesn’t do a spectacular job – perhaps Christchurch really is the “Garden City” so urban areas don’t stand out so well. By comparison, here’s one from LA where things stand out much better:
A 6, 5, 2 combination gives us a bit more contrast between different agricultural land covers:
This is a nice combination for picking out surface water features – all those darker areas north of the Waimakariri are reservoirs.
Hopefully, this helps highlight the point that, more than just creating pretty (and odd) pictures, these combinations are useful for differentiating between different features. If we’re wanting to pick out all the surface water features, the combination above might suit for further analysis and mapping. Thus far, our wetware has been doing a visual interpretation of what these combinations show. We’ve still got to go a few steps further if we want to extract these into a useful GIS layer. We’ll been slowing building up to this, but next stop: image classification.