Vegetation canopies are inherently 3-D, with various levels of clumping
occuring at different scales: the within-crown, the tree and the canopy
level. As such their reflectance anisotropy in the optical domain is not
accurately simulated by adopting a purely turbid medium representation.
For example, the local reflection maximum in the retro-reflection
direction - known as the 'hot spot' - is precisely due to the finite
size of the scatterers in the canopy layer, as well as the spectral
contrast with the underlying background. Structurally homogeneous canopy
representations (i.e., plane-parallel vegetation layers containing
uniformly distributed finite-sized scatterers with specified
distributions of their orientation: 1-D') thus constitute the simplest
canopy representation capable of matching multi-spectral space borne
reflectance measurements under almost any view and illumination
geometry. In fact, at large spatial resolutions, 1-D' canopy
repesentations may generate reflectance fields that are
undistinguishable from those generated over 3-D surface types. However,
when the spatial resolution of the observing sensor becomes finer, then
horizontal photon migration in conjunction with multi-directional canopy
sampling issues start to affect the aptitude of 1-D' canopy
repesentations to match multi-angular and multi-spectral reflectance
observations over 3-D target surfaces.
In any case, the interpretation of the state variables of such
radiatively equivalent 1-D' models is far from simple, and the canopy
reflectance modelling community is only just beginning to address these
issues. The
quantitative documentation of horizontal and vertical transport
phenomena at high spatial resolutions requires 1) the availability of
structural (and to a lesser extent also spectral) information on 3-D
vegetation canopies, 2) a sophisticated 3-D RT model capable of
representing the local radiation transfer within such a heterogeneous
environment, and 3) the usage of suitable variance reduction and/or
software parallelisation techniques in order to keep the computation
times within acceptable levels. Recently, the implementation of the
'photon spreading' technique within a parallelised Monte Carlo
raytracing model has allowed to perform the forward simulation of fluxes
and bi-directional reflectance fields over structurally heterogeneous
scenes at very high spatial resolutions. Initial results of these
efforts will be presented, and a number of issues and caveats related to
these forward and inverse modeling problems will be documented.
Copyright. All Rights Reserved.