Hyperspectral Sensors and Applications
25
(in terms of width and location) such that observations can be made in up to 15
moveable bands with these selected through telecommands. This increases the
utility of the instrument as selected portions of the spectrum with contiguous
channels can be viewed. The ASTER sensor also has channels centred on
selected absorption features. Specifically, band 6 (SWIR) is centred on a clay
absorption feature (associated with hydrothermal alteration), whilst bands 8
(SWIR) and 14 (TIR) are centred on a carbonate feature, thereby allowing
discrimination oflimestones and dolomites. Bands 10 to 12 (TIR) are designed
to detect sulphate and silica spectral features, whilst band 10 (with 6) allows
discrimination of common minerals, including alunite and anhydrite (Ellis
1999).
Only the Hyperion and CHRIS systems provide a continuous spectral profile
for each pixel value (Table 1.2). The Hyperion observes in 220 bands extending
from 400 to 2500 nm with a FWHM of 10 nm. The CHRIS observes in the
410-1050 nm range in 63 bands with bandwidths ranging variably from 1-3 nm
at the blue end of the spectrum to about 12 nm at the NIR end. The red-edge
region (~690-7 40 nm) is sampled at a bandwidth of approximately 7 nm.
Although a continuous spectral profile is useful, it can be argued that for
many applications, the information required can be easily extracted from
a few specific spectral bands, particularly since many are highly correlated
and often redundant. Feature extraction and selection therefore becomes an
important image processing issue under such circumstances. Aspects of over
sampling are discussed in more detail in Chapters 8 and 9.
A particular benefit of CHRIS is that multi-angle viewing can be achieved
in the five along track view angles. For example, using PROBA's agile steering
capabilities in both along and across track directions, observations of targets
outside the nominal field of view of 1.3° can be obtained. From CHRIS, surface
biophysical parameters are being estimated using a number of techniques
ranging from traditional vegetation indices to more advanced techniques such
as BRDF model inversion. MERIS also provides directional viewing capability.
1.3.2.2
Spatial Resolution and Coverage
The spatial resolution of spaceborne hyperspectral sensors is coarser compared
to most airborne sensors, due largely to their greater altitude. MODIS and
MERIS sensors provide data at spatial resolutions ranging from 250-300 m to
1000 m. Hyperion observes at 30 m spatial resolution whilst CHRIS acquires
data at spatial resolutions of 36 m and 19 m with 63 and 18 spectral bands
respectively. MODIS provides data of the same location at variable spatial
resolutions of 250 m-1000 m (Table 1.3). As suggested earlier, coarse spatial
resolution can limit the ability to extract pure spectral signatures associated
with different surfaces. Orbview 4 (launch failure) was to include a finer (8 m)
spatial resolution 200 band hyperspectral sensor.
The swath width, and hence the area of coverage, also varies between satellite sensors. MODIS and MERIS have swath widths of 2330 km and 1150 km
respectively, allowing wide area coverage within single scenes. Such data are
25
(in terms of width and location) such that observations can be made in up to 15
moveable bands with these selected through telecommands. This increases the
utility of the instrument as selected portions of the spectrum with contiguous
channels can be viewed. The ASTER sensor also has channels centred on
selected absorption features. Specifically, band 6 (SWIR) is centred on a clay
absorption feature (associated with hydrothermal alteration), whilst bands 8
(SWIR) and 14 (TIR) are centred on a carbonate feature, thereby allowing
discrimination oflimestones and dolomites. Bands 10 to 12 (TIR) are designed
to detect sulphate and silica spectral features, whilst band 10 (with 6) allows
discrimination of common minerals, including alunite and anhydrite (Ellis
1999).
Only the Hyperion and CHRIS systems provide a continuous spectral profile
for each pixel value (Table 1.2). The Hyperion observes in 220 bands extending
from 400 to 2500 nm with a FWHM of 10 nm. The CHRIS observes in the
410-1050 nm range in 63 bands with bandwidths ranging variably from 1-3 nm
at the blue end of the spectrum to about 12 nm at the NIR end. The red-edge
region (~690-7 40 nm) is sampled at a bandwidth of approximately 7 nm.
Although a continuous spectral profile is useful, it can be argued that for
many applications, the information required can be easily extracted from
a few specific spectral bands, particularly since many are highly correlated
and often redundant. Feature extraction and selection therefore becomes an
important image processing issue under such circumstances. Aspects of over
sampling are discussed in more detail in Chapters 8 and 9.
A particular benefit of CHRIS is that multi-angle viewing can be achieved
in the five along track view angles. For example, using PROBA's agile steering
capabilities in both along and across track directions, observations of targets
outside the nominal field of view of 1.3° can be obtained. From CHRIS, surface
biophysical parameters are being estimated using a number of techniques
ranging from traditional vegetation indices to more advanced techniques such
as BRDF model inversion. MERIS also provides directional viewing capability.
1.3.2.2
Spatial Resolution and Coverage
The spatial resolution of spaceborne hyperspectral sensors is coarser compared
to most airborne sensors, due largely to their greater altitude. MODIS and
MERIS sensors provide data at spatial resolutions ranging from 250-300 m to
1000 m. Hyperion observes at 30 m spatial resolution whilst CHRIS acquires
data at spatial resolutions of 36 m and 19 m with 63 and 18 spectral bands
respectively. MODIS provides data of the same location at variable spatial
resolutions of 250 m-1000 m (Table 1.3). As suggested earlier, coarse spatial
resolution can limit the ability to extract pure spectral signatures associated
with different surfaces. Orbview 4 (launch failure) was to include a finer (8 m)
spatial resolution 200 band hyperspectral sensor.
The swath width, and hence the area of coverage, also varies between satellite sensors. MODIS and MERIS have swath widths of 2330 km and 1150 km
respectively, allowing wide area coverage within single scenes. Such data are