I was checking in the literature about a formal definition for hyperspectral sensors, and the most complete definition I found was from The Center of Space Research (@ University of Texas in Austin):

Hyperspectral data sets are generally composed of about 100 to 200 spectral bands of relatively narrow bandwidths (5-10 nm), whereas, multispectral data sets are usually composed of about 5 to 10 bands of relatively large bandwidths (70-400 nm).

Is there any consensus about it? Does exist an international body or forum to deal with remote sensing definitions?

  • I can't find any 'official' definition or distinction between the two by any sort of governing body. All of my sources cite the two specific distinctions as you mention in your question - number of and width of the individual bands. Some sources also suggest that hyperspectral is also contiguous bands. An example given is that a 20 band image could be considered hyper rather than multi if there were no gaps in the spectral coverage. – Chris W Jun 9 '14 at 20:03
  • Thanks Chris W, very interesting the contiguous aspect. I missed this notion when I was investigating the concept! – Gorgens Jun 12 '14 at 18:10

Not that I know. Sometimes the sensors are named and then the definition change (e.g. Advanced Very High Resolution Radiometer would not be called VHR anymore, but it was in 1978)

Your definition is quite practical but does not tell you what you have between the two ranges (e.g. 15 bands like MERIS, Rapideye red Edge is 40 nm... I would put those two in the mlti-spectral category, but as you can see this is out of range).

That being said, I would not rely too much on the number of bands for a definition of hyperspectral. What is more important to me is the method of acquisition. See the difference between radiometer (for multispectral) and spectrometer (for hyperspectral) below:

Radiometers. Radiometers are used to measure the amount of electromagnetic energy present within a specific wavelength range. The measurement is expressed in Watts (W) which is a unit of measurement for power. Radiometers are usually used to detect and measure the amount of energy outside the visible light spectrum and are used to measure ultraviolet (UV) light or infrared (IR). A typical use for a UV meter is in the museum lighting world where the presence of UV is can be very troublesome. UV energy hastens the ageing process due to its higher energy content so any energy below 400nm needs to be filtered out or eliminated. Another application for a radiometer is in the detection and measurement of infrared or IR. It is used to detect and measure heat on a surface. Technicians use them to safely detect and repair overheating motors or shorted out wiring. Radiometers can measure very quickly because they are simple meters that use only one sensor with a filter designed to just measure the wavelength range they were intended for.

Spectrometers. Spectrometers, like radiometers, are instruments that are also used to measure a specific wavelength range. The biggest difference is spectrometers use an optical grating or prism and multiple sensors to break down the incoming energy into different wavelengths or components. Spectrometers are not complete instruments and need to be paired with optics in order to work correctly. It can be used with a camera system to measure watts per square meter SR nm ( W / m2*SR*nm ) or with a cosine corrected head to measure irradiance and report watts per square meter nm ( W / m2 * nm ). Spectrometers can have up to 2048 sensors so they are highly analytical and can give very precise data and can measure very accurately. And since they are not complete systems, they can be adapted and used in multiple industries and applications.

So the bandwidth with a radiometer is fixed for a given purpose, while the spectrometer potentially provides a continuous spectrum.

  • +1 Great answer, which would be even better with a definitive peer-reviewed source. – Aaron Jun 9 '14 at 19:53

A high number of bands does not make a sensor "hyperspectral", although the hyper could suggest that "many" is the solution. However, the typical feature of hyperspectral data is that it results in a continuous curve of reflectance across the range of wavelengths measured. Therefore, the spectral interval has to be narrow, much narrower than for the multispectral sensors. Landsat 8, for example, averages across a broad spectral regions, such as the Blue Band which covers 0.452-0.512nm (=60nm!= The hyperspectral HyMap airborne sensor has a spectral interval of 15nm in the VIS.

A good example for a (semi)hyperspectral satellite is CHRIS-PROBA developed by ESA see here. In Mode 5 it only has 32 bands, but these still allow a more or less contiguous spectral curve. It´s not always size that matters :)

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.