When the issue of resolution is discussed, true remote sensing geeks will ask “which type of resolution?” You see, there are typically four types of resolution when talking about remotely sensed data: spatial, spectral, temporal, and radiometric. Spatial resolution is what most people think of when they hear the term resolution, and this has been discussed in many past blogs (“How Low Can You Go?” “Lidar – Accuracy Versus Resolution,” “Spatial Detail vs. Map Accuracy,” Resolution vs Minimum Mapping Unit). I’ll try and explain all these terms and why they may be important to understand.
As I mentioned before, spatial resolution is what most mean when talking about the resolution of their data. For remotely sensed imagery, it refers to the smallest ground object that can be resolved in the image, i.e., the pixel size. For example, the National Agriculture Imagery Program (NAIP) collects data with a 1m resolution, while the WorldView-2 satellite has a 1.85m pixel. The below images show the differences between a 1m, 10m, and 30m pixel (left to right) for the same area. Determining what is appropriate for your needs is covered in the “How Low Can You Go?” blog, but two major topics include cost and processing.
Waquiot Bay, MA NAIP aerial imagery in its native 1m format (left) compared to a 10m (middle) and 30m (right) version of the same data. While visual differences are easy to see, the file sizes also change from 2.5MB to 39kB to 23kB.
Higher spatial resolution may look much better, but it typically costs MUCH more. Also, keep in mind that amount of data (number of pixels) to cover an area is related to spatial resolution through a power factor. That is to say, if you cut the pixel size by two, you square the number of pixels (e.g., it takes four 10m pixels to cover the same area as one 20m pixel). Going from a 30m pixel to a 1m pixel results in 900 times the number of pixels!
The spectral resolution of a sensor refers to the number/location of spectral bands the sensor collects data in AND how wide those bands are. A good analogy may be, how many crayons are in your coloring box? Do you have 8 or 64 colors? If you only have 8 crayons and want the color teal, you can’t easily do it, but if you have the big box of colors, you may have teal at your fingertips. Digital sensors are similar, in that they are set up to look at specific regions of the electromagnetic spectrum. How many regions and how wide the regions are are related to spectral resolution. The USGS has a really cool tool which will show you the spectral response of different sensors for a variety of ground targets.
This graphic compares the spectral resolutions of Landsat 7 (L7) and the soon-to-be Landsat 8 (LDCM). The number, location, and width of the spectral bands differ between the two sensors. Graphic created by L.Rocchio & J.Barsi.
This is a pretty easy type of resolution to understand. Basically, it refers to the how often an area can be imaged by a sensor. For example, Landsat 8 has a repeat interval of 16 days, while WorldView-2 can typically revisit any place on earth in 1.1 days. For longer term monitoring, the Landsat data may be great, but to detect crop green-up or fire progression, something with much higher temporal resolution is needed.
This type of resolution brings in math. It typically refers to how many levels of brightness (contrast for example) a sensor can record, often called bit-depth. For example, an 8-bit image can record 256 levels of brightness (28) compared to a 16-bit image which can record 65,536 levels of brightness (216). Think back to the crayon analogy- how many shades of green (or red, or yellow…) are in your crayon box? With more shades of green (greater bit-depth), you can get see more detail in your object. Radiometric resolution can refer to how the sensor is “tuned” as well. Water bodies tend to reflect much lower amounts of energy than land, thus an ocean observing sensor will be tuned to be sensitive to a different dynamic range than a sensor tuned for land.
The image on the left displays the full 16-bit depth available with the WorldView-2 sensor. The image on the right has been resampled to 4-bit depth. While they may look similar, note the lack of detail in the runway and shallow water areas in the 4-bit image.
So Now You Know
This blog just scratches the surface of these topics, but should give you just enough to help you on your way to asking more intelligent questions about resolution. At the very least, when someone brags about their high-resolution data, you can respond with, “What type of resolution?”