Data transmitted by Earth observation satellites vary a great deal depending on the sensors used. What are known as panchromatic data originate in the recordings of radiation reflected by the Earth’s surface at an interval that includes all the visible wavelengths. On the other hand, the richer multispectral data are obtained from the simultaneous recording of the same scene in 3 to 10 spectral bands that are not necessarily contiguous and usually located in the visible range, in the near infrared, medium infrared and thermal infrared. “Hyperspectral” sensors represent a major quantitative advance on both the above methods. These are able to record data simultaneously in hundreds of often contiguous and much narrower spectral bands (in the region of a few nm), thereby revealing in much greater detail the spectral properties of the elements observed for large acquisition windows. In these targeted areas of the electromagnetic spectrum, the continuous signature of each element is available rather than discrete measurements or rough averages. This fine spectral signature better reflects the spectral behaviour of the elements observed and permits a more precise identification and discrimination of objects.

An unprecedented precision

Each pixel of a hyperspectral image contains a continuum of information on the spectral response of the objects observed in large areas of the electro- magnetic spectrum. A considerable quantity of information is therefore available to estimate with unprecedented precision a series of bio-geophysical and bio- chemical variables. This makes hyper-spectral imaging a useful tool for a better understanding of atmospheric, hydro- logical and geological phenomena or for a better monitoring of vegetation or soil composition. The numerous applications, which some are already operational, include geological and mining prospecting, estimating the pollution levels and eutrophication of surface waters, monitoring sediment build-up and movement in estuaries and port areas, monitoring of the composition of the vegetative cover, evaluating the health of trees, identifying a deficiency or excess of nitrogen in major crops, and estimating organic matter content in soils or their susceptibility to erosion.

Making the most of this new method
By their very nature, hyperspectral data result in vast quantities of information that must be stored and processed. This requires a much increased computing power to classify raw data and brings the need to review and optimise the whole processing and analysis process. This field of research represents a challenge for both fundamental and applied research. Supported by the STEREO programme and by ESA’s PRODEX programme, this field of investigation has taken concrete shape since 2002 in the organization of four hyperspectral acquisition campaigns using various airborne instruments. Alongside these flight campaigns, the STEREO programme has also financed small-scale scientific projects enabling Belgian researchers working with international partners to investigate the technique and to acquire a certain know-how. The Belgian Science Policy Office has entrusted the organization of the airborne campaigns as well as the pre-processing, distribution and storage of data to the Vlaamse Instelling voor Technologisch Onderzoek (VITO). This research institute is also a partner in a Belgo-Swiss consortium that is developing the APEX imaging spectrometer for ESA (see box).
http://cvblocal.vgt.vito.be/