From spectral to ecological information

The interpretation of spectral information is at the very core of remote sensing data analysis. However, the spectral signal carries much more information about the land surface than just what is readily accessible to the human eye. In order to harness this information the original bands are often transformed into new synthetic bands, so-called spectral indices, by mathematical operations combining multiple bands. These spectral indices combine several advantages over just using the original reflectance. Firstly, they can dramatically enhance the separability of certain specific land cover types in visual or automated image interpretation, e.g. vegetation versus open soil. Secondly, many indices involve mathematical division of bands which has a normalizing effect on illumination variability within a single scene and also between scenes. This can reduce terrain or cloud-induced illumination effects, improve multi-temporal comparability in a time-series and may even reduce the need for precise atmospheric correction. Thirdly, spectral indices are usually geared towards describing actual physical measures of the land surface, such as the degree of vegetation cover or water stress within vegetation. These are measures which allow interpretation in an ecological context, while it is hard to reason what e.g. a reflectance of 20% in the green band actually means