Data from the Landsat series of satellites will be almost exclusively used for the remaining chapters of this book for two very sound reasons: the archive of this data type is the longest, and its spatial resolution is mostly adequate for the task of change detection. It will be demonstrated that much of the landcover change investigated would be very poorly captured using the coarse spatial resolution (~1 km2) of NOAA AVHRR.

There are five steps that must be completed before two Landsat scenes can be compared to detect and interpret landcover change. These steps are

• Preprocessing
• Registration
• Normalisation
• Differencing
• Display.

The most important refinements are the spatial Registration - which ensures that the 1972 and the present pixels are coincident - and the spectral or radiometric Normalisation that removes or minimises the influence of changes in the atmosphere and in the MSS sensor calibration between 1972 and the present. These two procedures ensure that the change detectable in the difference image is real. The change displayed in the image corresponds to real changes in landcover. All other influences have been removed or minimised.

To convince the reader of the capacity of satellite data to detect significant landcover change and the ease with which this change can be interpreted, we will work through several examples from one Landsat scene.

The demonstration scene is Perth, Western Australia, chosen because it shows extensive and varied landcover change that can usefully be related to several landuses. Two Landsat MSS scenes, January 1974 and December 1988, have passed through the five processing steps and the resultant 3 FCC images are displayed with the test sites indicated.

Use your browser to open each image in a new window to compare them.

The Difference FCC image contains all the landcover change information. As you can see, there is a substantial amount of landcover change spread across the entire Perth scene. The spatial patterns of this change are not difficult to interpret because they represent mostly human-made boundaries with a scattering of natural boundaries. But how to interpret the spectral information? What can we make of the colours?