| ||Area classification|
Vegetation comparison for 1998 and 1999
Satellite-based mapping of land use necessitates the ability to separate water from bare ground, built-up areas, hardwood forest, softwood forest, agricultural areas etc. If the spectral signature of a given surface can be differentiated in the sensor channels, then it is possible to let the computer make an area classification. However, classes may not be separable at a single point in time, but only when seen at different times in the growing season. This is called multi-temporal classification.
During image classification it is possible to identify a specific area type on the screen (and use it as a training area), determine the spectral signature, and then let the computer identify all the pixels that have the same spectral signature. In this way, large regions can be mapped very quickly and easily.
There are still, however, several unresolved problems. It is especially difficult to distinguish between different types of vegetation as their spectral signatures can be so much alike. Furthermore, the same type of vegetation has different signatures, depending on which stage it's at in the growth season and on other
factors, such as soil humidity and atmospheric conditions.
Research is therefore concentrated on the potential refinement of area classification based on satellite data. One way is to try to optimise the spread of sensors covering specific channels in the 'visible' and 'near infrared' ranges of the spectrum.
A satellite with many narrow channels is said to have a high degree of spectral resolution. In the future, satellites with high spectral resolution may make it possible to map the changes in vegetation provoked by the stress of pollution and drought. Remote sensing is expected to become an increasingly important tool for environmental mapping.