19. REMOTE SENSING OF ALBEDO USING THE BRDF IN
RELATION TO LAND SURFACE PROPERTIES
177
3.
DERIVING ALBEDO FROM MULTIANGULAR
REFLECTANCE OBSERVATIONS
Intrinsic land albedo may be derived in a form decoupled from
atmospheric conditions from integrating the BRDF over the hemispheres of
viewing and illumination angle, yielding directional-hemispherical (blacksky) or bihemispherical (white-sky) albedo, respectively. The BRDF can be
derived from atmospherically corrected multiangular remote sensing
observations of reflectance through the use of a BRDF model that
parameterizes the vegetation effects influencing the BRDF and albedo.
Operationally, semi-empirical BRDF models describing the BRDF with
three independent parameters are currently the most practical option at the
kilometer scale. Several feasible and validated models are available
(Roujean et al., 1992; Rahman et al., 1993; Wanner et al., 1995). Even
though semi-empirical BRDF models describe the physics of light scattering
in vegetation only very approximately or even mostly empirically, there is
some evidence linking model parameter values with general land surface
type (Engelsen, 1996; Hu et al., 1997) or vegetation properties (Hyman and
Wanner, 1997; Roujean et al., 1997; Disney and Lewis, 1998).
BRDF and albedo of vegetated surfaces are a function of the optical
properties and three-dimensional spatial structure of vegetation. In order to
recover these from an inversion, one approach to the modeling is to separate
leaf-level and crown-level effects. They reflect different although often
coupled aspects of the underlying scattering process, one being related
mainly to the LAI and one mainly to vegetation structure. This approach is
taken by the Ambrals BRDF model, scheduled to produce the MODIS
BRDF/albedo product globally with 1 kilometer spatial resolution in 7
spectral bands (spanning the visible and near-infrared), once every 16 days,
after the launch of MODIS in December 1999 (Wanner et al., 1997).
The rationale of the modeling, developed by Roujean et al. (1992)
specifically for remote sensing applications, is as follows. The reflectance of
a scene is described as consisting of a linear superposition of the following
three components.
An overall isotropic scene reflectance is given as a Lambertian constant.
Leaf-level volume scattering, dependent both on leaf optical properties
(reflectance and transmission) and their spatial density or leaf area index
(intra-crown gaps), is calculated from approximations to radiative
transfer theory by Ross (1981). It is given by the so-called RossThick
kernel function
(Roujean et al., 1992).
Crown-level surface scattering due to the three-dimensional structure of
individual plants, either in the form of protrusions or gaps in otherwise
RELATION TO LAND SURFACE PROPERTIES
177
3.
DERIVING ALBEDO FROM MULTIANGULAR
REFLECTANCE OBSERVATIONS
Intrinsic land albedo may be derived in a form decoupled from
atmospheric conditions from integrating the BRDF over the hemispheres of
viewing and illumination angle, yielding directional-hemispherical (blacksky) or bihemispherical (white-sky) albedo, respectively. The BRDF can be
derived from atmospherically corrected multiangular remote sensing
observations of reflectance through the use of a BRDF model that
parameterizes the vegetation effects influencing the BRDF and albedo.
Operationally, semi-empirical BRDF models describing the BRDF with
three independent parameters are currently the most practical option at the
kilometer scale. Several feasible and validated models are available
(Roujean et al., 1992; Rahman et al., 1993; Wanner et al., 1995). Even
though semi-empirical BRDF models describe the physics of light scattering
in vegetation only very approximately or even mostly empirically, there is
some evidence linking model parameter values with general land surface
type (Engelsen, 1996; Hu et al., 1997) or vegetation properties (Hyman and
Wanner, 1997; Roujean et al., 1997; Disney and Lewis, 1998).
BRDF and albedo of vegetated surfaces are a function of the optical
properties and three-dimensional spatial structure of vegetation. In order to
recover these from an inversion, one approach to the modeling is to separate
leaf-level and crown-level effects. They reflect different although often
coupled aspects of the underlying scattering process, one being related
mainly to the LAI and one mainly to vegetation structure. This approach is
taken by the Ambrals BRDF model, scheduled to produce the MODIS
BRDF/albedo product globally with 1 kilometer spatial resolution in 7
spectral bands (spanning the visible and near-infrared), once every 16 days,
after the launch of MODIS in December 1999 (Wanner et al., 1997).
The rationale of the modeling, developed by Roujean et al. (1992)
specifically for remote sensing applications, is as follows. The reflectance of
a scene is described as consisting of a linear superposition of the following
three components.
An overall isotropic scene reflectance is given as a Lambertian constant.
Leaf-level volume scattering, dependent both on leaf optical properties
(reflectance and transmission) and their spatial density or leaf area index
(intra-crown gaps), is calculated from approximations to radiative
transfer theory by Ross (1981). It is given by the so-called RossThick
kernel function
(Roujean et al., 1992).
Crown-level surface scattering due to the three-dimensional structure of
individual plants, either in the form of protrusions or gaps in otherwise