IEEE Geoscience and Remote Sensing Magazine - December 2021 - 232

affect each material in a pixel differently than the others
since the behavior of the reflectance as a function of the
angle is different for each material. This indicates that
each EM/material in a pixel can be variously affected by
topographic effects. Furthermore, the nontrivial relationship
between geometry and spectral signatures leads
to a more complex variation than single scaling of each
EM for high-albedo materials [51], [52]. Even small topographic
variations can significantly affect the ground
reflectance. For instance, in [53], experimental studies
found that even small slopes (of fewer than 10º) originating
from irregularities in the tree canopy can lead to
appreciable (enough to influence the results of subsequent
tasks) changes in the measured reflectance of vegetation
spectra.
INTRINSIC SPECTRAL VARIABILITY
Another important source of spectral variability is the intrinsic
variation pertaining the definition of a material, which
is also called intrinsic variability. The characterization of this
type of variability has been prominently studied in the area
of vegetation monitoring, where it poses a huge challenge
to identifying tree species from spectral measurements [54],
[55] and to the characterization of soil and mineral spectra.
Vegetation's spectral signature can change due to many factors,
including microclimates, soil characteristics, precipitation,
the presence of heavy metals, drought, foliage age, and
colonization by leaf pathogens [54]. The spectral signature
of soil is also heavily affected by variations in its composition
and moisture content [56]. Furthermore, intrinsic
spectral variability is common in mineral spectra, due to
differences in the grain size distribution and the presence of
variable amounts of impurities [57], [58]. Moreover, it also
depends on what level of detail is adopted to represent a
given material (e.g., a tree EM may be split into trunk and
leaf EMs), which is generally application dependent [59].
Although there is a large impact on EM spectral signatures,
the dependence of intrinsic spectral variability on physicochemical
parameters, which are usually unknown, makes
this area very hard to tackle.
One characteristic consistently observed in experimental
studies is the smoothness of the observed spectra
(i.e., the reflectance varies slowly between spectral
bands). This behavior can be taken into account when
designing SU algorithms. Moreover, unlike spectral
changes caused by illumination and topography effects,
intrinsic spectral variability frequently presents a considerable
dependence of the variability amplitude on
the spectral wavelength. For instance, the signatures of
different instances of minerals in the United States Geological
Survey (USGS) library, presented in Figure 7, show
complex dependence between the reflectance variation
and the wavelength. The samples from alunite and muscovite
have a variability that is far from uniform across
the spectrum. Moreover, different instances from pyrite
display complex variation, which is not consistent
across all samples, occurring independently in different
regions of the spectra. This behavior has been verified in
similar experimental studies, and it poses a great challenge
for differentiating mineral classes based on their
spectral signatures [60].
These characteristics are even more prominent in the
spectral variation of vegetation reflectance, which shows
significant dependence on the wavelength and behaves
very differently in visible, near-infrared, and short-waveinfrared
ranges [61]. This means that a simple scaling of
a reference spectral signature is usually not sufficient to
account for variations within tree species [54]. Extensive
experimental studies support this claim. In [54], the author
found that the variation of spectral reflectance in the
visible and near-infrared regions can occur independently
when measuring the tropical forest canopy in Brazil. Similar
inhomogeneity in spectral variation was also observed
in other studies with tropical tree species [62] and in many
distinctive environments, including conifer [63] and boreal
tree forests [64]. Similar nonuniform variation trends
are also consistently observed in seasonal changes, as indicated
by many experiments, including in salt marshes [65],
semiarid environments [66], and boreal tree species [67].
Furthermore, nonuniform spectral variations have also
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.5 1 1.5 2 2.5
(a)
Wavelength (µm)
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.5 1 1.5 2 2.5
(b)
Wavelength (µm)
FIGURE 7. Spectra variation samples from the USGS library. (a) Alunite. (b) Muscovite. (c) Pyrite.
232
IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE DECEMBER 2021
0.18
0.16
0.14
0.12
0.1
0.08
0.06
0.02
0.04
0.5 1 1.5 2 2.5
(c)
Wavelength (µm)
Reflectance
Reflectance
Reflectance

IEEE Geoscience and Remote Sensing Magazine - December 2021

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - December 2021