Now I See (What's in a Multispectral Image)
I’m often asked by friends and family what it is that I do for work.The truth is that it that I sometimes have a hard time explaining remote sensing, and most people that ask are probably just making small talk. Do they really want a lesson on the electromagnetic spectrum and how sensors on spaceborne satellites can detect emissions outside of the visible portion of it? Probably not. But when I’m feeling up to the task and think I am talking to someone who would be at least half-interested in the subject, I sometimes attempt to delve into a basic explanation of remote sensing.
I have found the hardest concept for a lot of people to grasp is how multispectral images differ from the kinds of pictures we take with our smartphones and digital cameras. I have discovered if you really want someone to be interested in this topic all you have to do is show them a picture. I think someone important once said, “a picture is worth a thousand words.” So, I thought today we could take a look at some pictures and see how a true color image (like one we could take with our smartphone or digital camera) might differ from one that takes advantage of wavelengths outside of the visible portion of the electromagnetic spectrum.
I’ve been paying fairly close attention to the Rim Fire that is currently burning in California. This fire has been a beast to deal with and has burned nearly 400 square miles of land. An image of this fire was captured by Landsat 8 on August 31, 2013. Landsat 8 is capable of capturing vivid and crisp true color images. But, of the 11 ranges of wavelengths (A.K.A. bands) that Landsat 8 is capable of capturing, only five of them sense visible light. The other six bands capture information ranging from the near infrared through thermal infrared to shortwave infrared portions of the electromagnetic spectrum.
When we load a multispectral image into image analysis software, we can choose which band we want to display as red, which band we want to display as green, and which band we want to display as blue. Depending on how we choose to visualize an image, we can get a very different sense of what is going on in the sky, on the land, and even in the water. Some bands can see through clouds and smoke, some can pierce through water, and some make healthy vegetation appear very bright. Depending on what you are looking for, you might choose to visualize the image differently. The great news is that when we perform analysis on multispectral imagery, we can take advantage of all of these bands to extract the information we are looking for. Here are three different visualizations of the same image of the Rim Fire.
True Color Image of California Rim Fire captured by Landsat 8, August 31, 2013. Note how you can see smoke and clouds, but it’s difficult to see exactly where the fire has burned or where it is currently burning.
SWIR 2 (Red), Near Infrared (Green), Blue (Blue) Image of California Rim Fire captured by Landsat 8, August 31, 2013. With this band combination the smoke appears blue, the ground is more visible through both smoke and clouds, and the fire scar appears as bright pink.
Thermal Infrared (1) Image of California Rim Fire captured by Landsat 8, August 31, 2013. For this image, only one band is displayed and a red color table has been applied. The very bright pixels in this image depict the hottest areas in the image and represent the current perimeter of the Rim Fire.