Band Combination Mixer Introduction to GIS: multispectral images
Welcome to the Band Combination Mixer
In this activity you will work with a synthetic multispectral image of a farming landscape. The landscape contains healthy crops, crops that are growing poorly, bare fields, a river, a lake, a forest and a small town. You will learn how a GIS turns image bands into colors on the screen, and you will use that skill to find crops that are growing poorly.
Learning objectives
After this activity you will be able to:
- Explain that a multispectral image stores several bands for the same area, and that each band records a different part of the electromagnetic spectrum.
- Assign bands to the red, green and blue display channels to make true-color and false-color composites.
- Interpret a color-infrared composite, in which healthy vegetation appears red.
- Display a single band in grayscale or with a pseudocolor ramp, and adjust its contrast stretch.
- Explain spectral resolution as the number and width of the bands a sensor records.
What to do first
Select Start below. Then work through the steps in order, using the Next step button at the bottom of each step or the step buttons at the top of the screen. Before some actions you will be asked to predict what will happen. Make your best guess; predictions are not marked.
Open Key terms at any time to look up a word. The activity takes about 30 minutes.
Note: your work is not saved. If you reload or close the page, your answers are lost. At the end, a summary screen lets you print or screenshot your evidence of completion.
Use the arrow keys to move the selection. Hold Shift with an arrow key to move eight pixels at a time. Press Enter or Space to select the pixel or grid square.
Step 1. One image, five bands
A multispectral image stores several bands for the same area. Each band records how much sunlight the ground reflects in one range of wavelengths of the electromagnetic spectrum. A band is a raster: a grid of pixels with one number in each pixel.
The map shows a true-color view of the image. Select a band below to display that band alone. Bright pixels reflect a lot of energy in that band; dark pixels reflect little.
Where the bands sit in the spectrum
Simplification. This app uses generic band names (Blue, Green, Red, NIR, SWIR) instead of band numbers, because band numbering differs between sensors. For example, the red band is band 3 on some sensors and band 4 on others. The wavelength ranges are approximate values typical of moderate-resolution sensors. Many real sensors record two or more SWIR bands; this app uses one.
Spectral resolution describes how many bands a sensor records and how narrow each band is. A sensor with more, narrower bands has finer spectral resolution. This image has five fairly broad bands. A hyperspectral sensor can record hundreds of narrow bands, so its spectral resolution is much finer.
Step 2. Mix bands into a color composite
A screen makes every color by mixing red, green and blue light. These are the three display channels. A color composite assigns one band to each channel. The value of that band at each pixel controls how bright that channel is.
This step starts with a true-color composite: the Red band in the red channel, the Green band in the green channel and the Blue band in the blue channel. The caption under the map always lists the current assignment.
Band-to-channel assignment
Simplification. In composites, every band uses a fixed linear stretch: the visible bands (Blue, Green, Red) run from 0 % to 30 % reflectance and the infrared bands (NIR, SWIR) from 0 % to 50 %. A GIS usually calculates a stretch for each image. A fixed stretch keeps the composites comparable with each other.
Step 3. Display one band
A single band holds one value per pixel, so a GIS can display it as a grayscale image: low values are dark and high values are bright. A GIS can also apply a pseudocolor ramp, which assigns a color to each value. Pseudocolors do not show the real color of the ground. They make differences between values easier to see.
This step starts with the Red band in grayscale.
Histogram and contrast stretch
The histogram shows how many pixels have each value in the band. The stretch decides how values become display colors. Values at or below the minimum get the first color of the ramp, and values at or above the maximum get the last color. Drag the minimum and maximum lines, or use the sliders and number boxes.
Bar heights use a square-root scale so that small groups of pixels, such as the town, remain visible. The strip under the bars shows the display color for each value.
Step 4. Compare spectral profiles
A spectral profile is a line chart of one pixel's values in every band. Different surfaces reflect sunlight differently in each band, so their profiles have different shapes. You can use these shapes to identify what is on the ground.
Collect profiles
Click a pixel on the map to draw its profile. With a keyboard, focus the map, move the crosshair with the arrow keys and press Enter. Collect at least one profile each for vegetation (crops or forest), water and bare soil. Up to four profiles are kept; the oldest is replaced. Change the display to help you find each surface:
The bands are spaced evenly along the horizontal axis. They are not drawn to wavelength scale.
Step 5. Field task: find the crops that are growing poorly
The request. A farm manager writes: "Find the parts of the fields where crops are growing poorly, so that we can inspect them."
Mark the areas
A grid now covers the map. Each grid square is 8 by 8 pixels (240 m by 240 m). Switch between the displays below, then mark every square where crops appear to be growing poorly. Click or drag on the map to mark or unmark squares. With a keyboard, focus the map, use the arrow keys to move the selected square, and press Space to mark or unmark it.
Marked squares: 0
Step 6. Concept check
Answer the five questions below. If an answer is not correct, read the feedback and try again. Your summary records how many questions you answered correctly on the first attempt.
Summary: Band Combination Mixer
This page is your evidence of completion. Enter your name, then print the page or take a screenshot.