184
Chapter 19
For a given scene, expected maximal and minimal values of the LAI and of
plant density could be used to empirically determine the factors
and
from the observed extremes of the parameters Using these constants,
the BRDF model parameters would be nonlinearly rescaled to reflect more
physical parameters.
In order to establish whether such a scheme would work in practice, the
following requirements would have to be met:
The spatial distribution of the model parameters would have to show a
definite relation to the spatial distribution of surface properties.
Coupling between model parameters due to partial non-linearity between
kernels should be discounted as a source of the relationships seen, and
Since LAI and surface structure are both wavelength-independent
quantities, application of the method to the red and near-infrared band
should produce similar results.
There is some evidence in the literature that a partial interpretation of
kernel-based BRDF model parameters in this way may be possible (Barnsley
et al., 1997; Hyman and Wanner, 1997; Roujean et al., 1997; Disney and
Lewis, 1998), but the potential and limitation are still not fully understood.
In all likelihood, we will have to wait for more advanced evaluation of
POLDER data, and for the data stream from MODIS and MISR, before
being able to give a definitive answer. For the New England data set shown
here, volume scattering is found to be consistently low and geometric
scattering high in the urban areas, the opposite holding for the dense forests,
as expected and consistent with this proposal. However, GOES does not
provide a near-infrared channel for an independent check of these results,
and the question of kernel coupling under conditions of sparse angular
sampling still has to be explored further. For this reason, we refrain from
presenting these results here, but would like to suggest that this line of
thought may warrant further investigation from an improved and perhaps
more diverse data set providing at least two spectral bands.
6.
CONCLUSIONS
It was the purpose of this paper to consider whether the biospherically
and climatologically important quantity of land-surface albedo may be
derived from multiangular remote sensing at 1 km nominal spatial resolution
and how it may reflect land surface properties. We have given an example
from our more extended analysis showing that the retrieved albedo
magnitude and solar zenith angle dependence, both associated with the
BRDF of the surface pixels viewed, follow land surface properties as
identified by the NDVI of the region studied, a snow-based obscuration/gap-
Chapter 19
For a given scene, expected maximal and minimal values of the LAI and of
plant density could be used to empirically determine the factors
and
from the observed extremes of the parameters Using these constants,
the BRDF model parameters would be nonlinearly rescaled to reflect more
physical parameters.
In order to establish whether such a scheme would work in practice, the
following requirements would have to be met:
The spatial distribution of the model parameters would have to show a
definite relation to the spatial distribution of surface properties.
Coupling between model parameters due to partial non-linearity between
kernels should be discounted as a source of the relationships seen, and
Since LAI and surface structure are both wavelength-independent
quantities, application of the method to the red and near-infrared band
should produce similar results.
There is some evidence in the literature that a partial interpretation of
kernel-based BRDF model parameters in this way may be possible (Barnsley
et al., 1997; Hyman and Wanner, 1997; Roujean et al., 1997; Disney and
Lewis, 1998), but the potential and limitation are still not fully understood.
In all likelihood, we will have to wait for more advanced evaluation of
POLDER data, and for the data stream from MODIS and MISR, before
being able to give a definitive answer. For the New England data set shown
here, volume scattering is found to be consistently low and geometric
scattering high in the urban areas, the opposite holding for the dense forests,
as expected and consistent with this proposal. However, GOES does not
provide a near-infrared channel for an independent check of these results,
and the question of kernel coupling under conditions of sparse angular
sampling still has to be explored further. For this reason, we refrain from
presenting these results here, but would like to suggest that this line of
thought may warrant further investigation from an improved and perhaps
more diverse data set providing at least two spectral bands.
6.
CONCLUSIONS
It was the purpose of this paper to consider whether the biospherically
and climatologically important quantity of land-surface albedo may be
derived from multiangular remote sensing at 1 km nominal spatial resolution
and how it may reflect land surface properties. We have given an example
from our more extended analysis showing that the retrieved albedo
magnitude and solar zenith angle dependence, both associated with the
BRDF of the surface pixels viewed, follow land surface properties as
identified by the NDVI of the region studied, a snow-based obscuration/gap-