MODTRAN 4 for example allows for pixel adjacency effects, and is the basis of
the commercially available FLAASH (Fast Line-of-sight Atmospheric Analysis of
Spectral Hypercubes) plugin for the image processing software ENVI (AdlerGolden et al. 1999; Exelis VIS 2012). A number of research level atmospheric
codes that are used in the literature have limited availability or no support. Two
that occur regularly in the shallow water mapping literature are TAFKAA (used in:
Goodman and Ustin 2007; Mobley et al. 2005; Lesser and Mobley 2007) and cWOMBAT-c (used in Brando et al. 2009). TAFKAA is documented in Gao et al.
(2000) and Montes et al. (2001) whereas the basis of c-WOMBAT-c is described
in de Haan et al. (1997) and Brando and Dekker (2003). Freely available codes
such as the above mentioned SBDART and libRadtran can also be used to
parameterize a correction, as can the 6SV code which includes polarization
(Kotchenova et al. 2006; Kotchenova and Vermote 2007). It should be noted that
almost all use of radiative transfer models for atmospheric correction requires
some estimating of unknown parameters. For this reason, despite the demonstrable
success of model based approaches (Ferrier and Trahair 1995), highly accurate
atmospheric corrections are rarely achieved simply by ‘turning the handle’ on a
model based approach.
Vicarious calibration—Vicarious calibration refers to the process of taking
in situ above or below water reflectances for refining an atmospheric correction. In
principle it operates identically to the empirical line correction, but is typically
performed after a radiative transfer model based atmospheric correction and uses
actual above or below-water reflectances over the area of interest. Specific
instruments exist for collecting boat based in situ reflectances, such as the gimbal
mounted DALEC instrument (Slivkoff 2010). Given the expense of acquiring
airborne hyperspectral data and the difficulty of performing accurate corrections by
modeling alone, collection of vicarious calibration data is always recommended.
4.2.4 Cross Track Variation and Correction
Imagery from both satellite and airborne sensors can be affected by variations in
the view angle onto the Earth’s surface across the image. This is particularly true
for aquatic targets as the air–water interface can be highly reflective for certain
solar-view geometries. The extent of variation across an image is a function of the
altitude of the sensor and the width of imaged area. The satellite sensor IKONOS,
for example, orbits at around 700 km and has a 0.9° field of view imaging 11 km
on the surface. For IKONOS the variation in view angle is therefore extremely
small, less than 0.5° from nadir, and cross-image view angle effects are negligible.
Push-broom airborne sensors have a wider field of view to accommodate a sufficiently wide swath at low altitudes. For example the current CASI-550 has 40.4°
across track field of view (to the left and right of aircraft travel) (ITRES 2008)
imaging a swath of width approximately equal to the altitude of flight. Therefore
the variation in view angle from left to right side of the swath is ±20.2° off nadir,
4 Hyperspectral Applications
91
the commercially available FLAASH (Fast Line-of-sight Atmospheric Analysis of
Spectral Hypercubes) plugin for the image processing software ENVI (AdlerGolden et al. 1999; Exelis VIS 2012). A number of research level atmospheric
codes that are used in the literature have limited availability or no support. Two
that occur regularly in the shallow water mapping literature are TAFKAA (used in:
Goodman and Ustin 2007; Mobley et al. 2005; Lesser and Mobley 2007) and cWOMBAT-c (used in Brando et al. 2009). TAFKAA is documented in Gao et al.
(2000) and Montes et al. (2001) whereas the basis of c-WOMBAT-c is described
in de Haan et al. (1997) and Brando and Dekker (2003). Freely available codes
such as the above mentioned SBDART and libRadtran can also be used to
parameterize a correction, as can the 6SV code which includes polarization
(Kotchenova et al. 2006; Kotchenova and Vermote 2007). It should be noted that
almost all use of radiative transfer models for atmospheric correction requires
some estimating of unknown parameters. For this reason, despite the demonstrable
success of model based approaches (Ferrier and Trahair 1995), highly accurate
atmospheric corrections are rarely achieved simply by ‘turning the handle’ on a
model based approach.
Vicarious calibration—Vicarious calibration refers to the process of taking
in situ above or below water reflectances for refining an atmospheric correction. In
principle it operates identically to the empirical line correction, but is typically
performed after a radiative transfer model based atmospheric correction and uses
actual above or below-water reflectances over the area of interest. Specific
instruments exist for collecting boat based in situ reflectances, such as the gimbal
mounted DALEC instrument (Slivkoff 2010). Given the expense of acquiring
airborne hyperspectral data and the difficulty of performing accurate corrections by
modeling alone, collection of vicarious calibration data is always recommended.
4.2.4 Cross Track Variation and Correction
Imagery from both satellite and airborne sensors can be affected by variations in
the view angle onto the Earth’s surface across the image. This is particularly true
for aquatic targets as the air–water interface can be highly reflective for certain
solar-view geometries. The extent of variation across an image is a function of the
altitude of the sensor and the width of imaged area. The satellite sensor IKONOS,
for example, orbits at around 700 km and has a 0.9° field of view imaging 11 km
on the surface. For IKONOS the variation in view angle is therefore extremely
small, less than 0.5° from nadir, and cross-image view angle effects are negligible.
Push-broom airborne sensors have a wider field of view to accommodate a sufficiently wide swath at low altitudes. For example the current CASI-550 has 40.4°
across track field of view (to the left and right of aircraft travel) (ITRES 2008)
imaging a swath of width approximately equal to the altitude of flight. Therefore
the variation in view angle from left to right side of the swath is ±20.2° off nadir,
4 Hyperspectral Applications
91