Before collecting spectra, the instrument has to be warmed up for at least 30 min so
as to reduce the sensitivity drift (often called a “step”). The latter is a sudden change in
the reflectance curve of a target in a wavelength where the strands in the fiber-optic
cable of the spectroradiometer overlap. The sensitivity drift can be corrected directly
on the field by holding the fiber-optic cable away from the target and allowing the field
of views (FoVs) of the strands to overlap [Analytical Spectral Devices (ASD), 2009].
However, sensitivity drift is often saved in the spectral data and it has to be corrected
(Beal and Eamon, 1999; Salisbury, 1998).
Implementation of the spectrum averaging additionally ensures sufficient quantity
of usable information obtained expressed as the relative amount of noise contained in
the signal (i.e., signal-to-noise ratio, SNR). The higher the number of spectra
averaged, the higher the SNR value, and consequently the higher the amount of
usable information (ASD, 2009; Milton et al., 2009).
15.2.1.2 Environmental Factors The environmental factors that affect the reflectance measurements can be divided into atmospheric gases and illumination. As
outlined earlier, the complex reactions and absorption of the solar energy by the
atmosphere constituent gases, such as H 2 O, O 2 , O 3 , CH 4 , and NO, cause rapid energy
drop at longer wavelengths, especially in the SWIR spectrum. The spectroradiometer
production companies compensate for this natural effect by reduction of the spectral
resolution of the spectroradiometer at NIR and SWIR wavelengths (Salisbury, 1998).
Most remote sensing instruments operate in atmospheric windows by tuning the
detectors to specific frequencies that pass through the atmosphere with minimum
disturbances, that is, random noise. Such noise is in particular induced by the
atmospheric water vapor which exhibits strong absorption of the incident and
reflected radiation by water molecules. The optical spectrum is thus commonly
partitioned into four distinct wavelength regions: VIS, NIR, and SWIR-1 (1300–
1900 nm) and SWIR-2 (1900–2500 nm) (Figure 15.1).
In order to decrease the complexity that arises from the sun angle, the sun azimuthview angle relationship (the BRDF), spectral measurements on a field scale are often
assumed as isotropic, and the smallest nadir position of the sensor above the target
with the sun in its zenith is commonly used (Manakos et al., 2010; Pfitzner et al.,
2006). The parameters in the BRDF equation that describe the relative position of the
sensor to the sun are omitted. Thus, the sensor view on the ground FoV and the
orientation of the sun azimuth relative to any preferred orientations of the target are
the only significant geometrics. Since the FoV of high-altitude sensors is usually at
nadir, with a predetermined flight path, there is very little that can be done to
accommodate for the vegetation land cover variability that might affect the BRDF
response on the field. Spectral measurements are performed in close temporal
proximity between solar noon at high sun angle between 11:00 AM and 2:00 PM,
with sun height variation of less than 30°. At higher latitudes, directional measurements with off-nadir angles and wider FoV are implemented (Eklundh et al., 2011).
Measurements should be performed in such a way in terms of sensor position that the
BRDF variations are assumed as negligible. Additional minor variations can be
further minimized with spectra averaging (Manevski et al., 2012).
290
HYPERSPECTRAL REMOTE SENSING WITH EMPHASIS ON LAND COVER MAPPING
as to reduce the sensitivity drift (often called a “step”). The latter is a sudden change in
the reflectance curve of a target in a wavelength where the strands in the fiber-optic
cable of the spectroradiometer overlap. The sensitivity drift can be corrected directly
on the field by holding the fiber-optic cable away from the target and allowing the field
of views (FoVs) of the strands to overlap [Analytical Spectral Devices (ASD), 2009].
However, sensitivity drift is often saved in the spectral data and it has to be corrected
(Beal and Eamon, 1999; Salisbury, 1998).
Implementation of the spectrum averaging additionally ensures sufficient quantity
of usable information obtained expressed as the relative amount of noise contained in
the signal (i.e., signal-to-noise ratio, SNR). The higher the number of spectra
averaged, the higher the SNR value, and consequently the higher the amount of
usable information (ASD, 2009; Milton et al., 2009).
15.2.1.2 Environmental Factors The environmental factors that affect the reflectance measurements can be divided into atmospheric gases and illumination. As
outlined earlier, the complex reactions and absorption of the solar energy by the
atmosphere constituent gases, such as H 2 O, O 2 , O 3 , CH 4 , and NO, cause rapid energy
drop at longer wavelengths, especially in the SWIR spectrum. The spectroradiometer
production companies compensate for this natural effect by reduction of the spectral
resolution of the spectroradiometer at NIR and SWIR wavelengths (Salisbury, 1998).
Most remote sensing instruments operate in atmospheric windows by tuning the
detectors to specific frequencies that pass through the atmosphere with minimum
disturbances, that is, random noise. Such noise is in particular induced by the
atmospheric water vapor which exhibits strong absorption of the incident and
reflected radiation by water molecules. The optical spectrum is thus commonly
partitioned into four distinct wavelength regions: VIS, NIR, and SWIR-1 (1300–
1900 nm) and SWIR-2 (1900–2500 nm) (Figure 15.1).
In order to decrease the complexity that arises from the sun angle, the sun azimuthview angle relationship (the BRDF), spectral measurements on a field scale are often
assumed as isotropic, and the smallest nadir position of the sensor above the target
with the sun in its zenith is commonly used (Manakos et al., 2010; Pfitzner et al.,
2006). The parameters in the BRDF equation that describe the relative position of the
sensor to the sun are omitted. Thus, the sensor view on the ground FoV and the
orientation of the sun azimuth relative to any preferred orientations of the target are
the only significant geometrics. Since the FoV of high-altitude sensors is usually at
nadir, with a predetermined flight path, there is very little that can be done to
accommodate for the vegetation land cover variability that might affect the BRDF
response on the field. Spectral measurements are performed in close temporal
proximity between solar noon at high sun angle between 11:00 AM and 2:00 PM,
with sun height variation of less than 30°. At higher latitudes, directional measurements with off-nadir angles and wider FoV are implemented (Eklundh et al., 2011).
Measurements should be performed in such a way in terms of sensor position that the
BRDF variations are assumed as negligible. Additional minor variations can be
further minimized with spectra averaging (Manevski et al., 2012).
290
HYPERSPECTRAL REMOTE SENSING WITH EMPHASIS ON LAND COVER MAPPING