14
1: Richard Lucas, Aled Rowlands, Olaf Niemann, Ray Merton
1.3
Hyperspectral Systems
In recognising the limitations of multispectral systems, an improved technology based on spectroscopy was developed whereby the limited number
of discrete spectral bands was enhanced by a series (nominally> 50 bands)
of narrow (Full-Width-Half-Maximum; FWHM 2-20 nm) contiguous bands
(Aspinall et al. 2002) over the VNIR and SWIR wavelength regions. Similar
advances were also made in the thermal infrared (TIR) regions. Matrices of
spectral samples were then designed to be built up on a line-by-line basis to
form a two dimensional image (x- and y-axes), with a third dimension (zaxis) holding the spectral data for each sample (pixel). These new imaging
spectrometers were subsequently able to retrieve near-laboratory quality reflectance spectra such that the data associated with each pixel approximated
the true spectral signature of a target material, with sufficiently high signalto-noise ratio (SNR) across the full contiguous wavelength range (nominally
400-2500 nm) to ultimately form a 3-dimensional datacube.
Within these hyperspectral images, molecular absorption and particle scattering signatures of materials could be detected to unambiguously identify
and quantify the abundance of surface constituents (Buckingham et al. 2002).
Specific reflectance or absorption features could also be associated with different minerals or even biological and chemical processes, thereby providing an
opportunity to better characterise surface environments and dynamics. With
this advancement, the era of hyperspectral imaging was developed with airborne sensors deployed initially albeit largely for research purposes. Several
spaceborne hyperspectral sensors have also been deployed although, to date,
the data have been used largely for technology demonstration and research.
The following sections chronicle the development of these airborne and
spaceborne hyperspectral sensors and subsequently consider the characteristics of several, in the context of temporal, spatial, spectral and radiometric
resolutions, which render them unique for environmental applications.
1.3.1
Airborne sensors
Airborne hyperspectral remote sensing has been available since the early 1980s
(Table l.2). Early developments were characterized by small, purpose-built
sensors. Amongst the earliest of the scanning imaging spectrometers was a onedimensional profiler developed by the Geophysical Environmental Research
Company (GER) in 1981. This sensor gathered data in 576 channels over the
400-2500 nm wavelength range. In the same year, the Shuttle Multi-spectral
Infrared Radiometer (SMIRR) became operational and Canada's Department
of Fisheries and Oceans introduced the Fluorescence Line Imager (FLI). In 1983,
the Airborne Imaging Spectrometer (AIS) was first flown following a three year
period of development at NASA's Jet Propulsion Laboratory (Goetz 1995). The
principal driving force behind many of these initial developments came from
geological disciplines.
1: Richard Lucas, Aled Rowlands, Olaf Niemann, Ray Merton
1.3
Hyperspectral Systems
In recognising the limitations of multispectral systems, an improved technology based on spectroscopy was developed whereby the limited number
of discrete spectral bands was enhanced by a series (nominally> 50 bands)
of narrow (Full-Width-Half-Maximum; FWHM 2-20 nm) contiguous bands
(Aspinall et al. 2002) over the VNIR and SWIR wavelength regions. Similar
advances were also made in the thermal infrared (TIR) regions. Matrices of
spectral samples were then designed to be built up on a line-by-line basis to
form a two dimensional image (x- and y-axes), with a third dimension (zaxis) holding the spectral data for each sample (pixel). These new imaging
spectrometers were subsequently able to retrieve near-laboratory quality reflectance spectra such that the data associated with each pixel approximated
the true spectral signature of a target material, with sufficiently high signalto-noise ratio (SNR) across the full contiguous wavelength range (nominally
400-2500 nm) to ultimately form a 3-dimensional datacube.
Within these hyperspectral images, molecular absorption and particle scattering signatures of materials could be detected to unambiguously identify
and quantify the abundance of surface constituents (Buckingham et al. 2002).
Specific reflectance or absorption features could also be associated with different minerals or even biological and chemical processes, thereby providing an
opportunity to better characterise surface environments and dynamics. With
this advancement, the era of hyperspectral imaging was developed with airborne sensors deployed initially albeit largely for research purposes. Several
spaceborne hyperspectral sensors have also been deployed although, to date,
the data have been used largely for technology demonstration and research.
The following sections chronicle the development of these airborne and
spaceborne hyperspectral sensors and subsequently consider the characteristics of several, in the context of temporal, spatial, spectral and radiometric
resolutions, which render them unique for environmental applications.
1.3.1
Airborne sensors
Airborne hyperspectral remote sensing has been available since the early 1980s
(Table l.2). Early developments were characterized by small, purpose-built
sensors. Amongst the earliest of the scanning imaging spectrometers was a onedimensional profiler developed by the Geophysical Environmental Research
Company (GER) in 1981. This sensor gathered data in 576 channels over the
400-2500 nm wavelength range. In the same year, the Shuttle Multi-spectral
Infrared Radiometer (SMIRR) became operational and Canada's Department
of Fisheries and Oceans introduced the Fluorescence Line Imager (FLI). In 1983,
the Airborne Imaging Spectrometer (AIS) was first flown following a three year
period of development at NASA's Jet Propulsion Laboratory (Goetz 1995). The
principal driving force behind many of these initial developments came from
geological disciplines.