Geography Reference
In-Depth Information
the possibility of examining other river attributes, such
as substrate composition or suspended sediment concen-
tration. Many airborne sensors and commercial satellites
provide multispectral data, most often consisting of blue,
green, red, and near-infrared bands, and more advanced
hyperspectral instruments measure radiance in a larger
number of narrower bands. We briefly examine the effects
of sensor spectral resolution in Section 3.4.3, and hyper-
spectral remote sensing is considered in greater detail
in Chapter 4 (Fonstad). Here, we focus on the spectral
characteristics of the river channels themselves.
(i.e., the near-infrared),
L
S
(
λ
) can comprise a large pro-
portion of
L
T
(
), overwhelming the other components
in Equation (3.14). Because
L
S
(
λ
) is unrelated to the
channel attributes of interest, this situation can compli-
cate, if not preclude, mapping of these attributes. In other
words, the noise associated with water surface reflectance
can drown out the signal associated with radiance that
has interacted with the water column and possibly the
streambed (Legleiter et al., 2009).
Images of rivers are often adversely affected by sun
glint due to strong reflections from the water surface.
This problem can be mitigated by carefully planning
data acquisition so as to avoid unfavorable combina-
tions of illumination and viewing geometry. Due to the
highly variable nature of water surface topography in
rivers, however, sun-dominated surface reflectance will
always be present to some extent. Because such reflectance
is spectrally flat, radiance measured in wavelengths for
which
L
W
(
λ
3.3.1 Reflectancefromthewater surface
For most applications, reflectance from the water sur-
face is an extraneous source of radiance that confounds
mapping of channel attributes such as depth, substrate
type, and water column optical properties. As discussed
above, the magnitude of the surface-reflected radiance
can be determined from Fresnel's Equation (3.9) for a
level water surface and remains small (0.02-0.03) as long
as the solar incidence angle is not too large. Radiative
transfer simulations indicate that more realistic, irregular
water surfaces have greater reflectance (Legleiter et al.,
2004, 2009). The spectral shape of the surface-reflected
radiance
L
s
(
λ
) also depends on surface roughness. For
a level water surface,
L
s
(
) is assumed negligible can be attributed
entirely to reflection from the water surface, along with
the ubiquitous contribution from the atmosphere. Sub-
tracting this radiance component across the spectrum
could thus serve to remove the contribution of
L
S
(
λ
λ
)
to
L
T
(
) and help to isolate the water-leaving signal of
interest. In fact, this procedure is used to correct for sun
glint in oceanographic remote sensing (Hooker et al.,
2002; Hochberg et al., 2003). Two issues that arise in a
fluvial context imply that such an approach is not directly
applicable to rivers, however. First, the texture of the
water surface, and hence the magnitude of
L
s
(
λ
λ
) consists of reflected diffuse
skylight and thus increases slightly at shorter wavelengths
that experience greater scattering within the atmosphere.
For the special case of a view direction opposite the
direct solar beam, specular reflection results in an
L
s
(
), is dic-
tated by local flow hydraulics and is thus highly spatially
variable, implying that a single surface-reflectance cor-
rection factor cannot be subtracted uniformly across an
image. Second,
L
W
(
λ
) cannot be assumed negligible, even
in the near-infrared, due to the shallow depths and highly
reflective substrates typical of rivers. Radiance observa-
tions from longer, shortwave-infraredwavelengths, where
L
W
(
λ
)
spectrum that literally mirrors the solar spectrum - pure
sun glint. For rougher water surfaces, the orientation of
individual surface facets becomes more variable, a greater
number of facets reflect light from the brighter, near-sun
portion of the sky, and
L
s
(
λ
) begins to resemble the solar
spectrum rather than the background sky.
An important property of reflection from the air-water
interface is that all wavelengths are affected to a similar
degree - that is, surface reflectance can be approximated
as spectrally flat.When
L
S
(
λ
) would be negligible even for very shallow waters,
could be used for this purpose, but the multispectral
sensors used in most stream studies lack such bands. For
these reasons, detection and removal of water-surface
reflectance remains difficult.
λ
) is normalisedby the incident
flux the resulting surface reflectance
R
S
(
λ
) is nearly equal
for all wavelengths,
R
S
(
λ
λ
≈
R
s
. The magnitude of
R
S
is
greater for rougher water surfaces, however. Although
surface reflectance can increase the radiance measured
above a river channel, this increase is purely additive
and does not modify the spectral shape of the radiance
signal because all wavelengths are affected approximately
equally. Importantly, for those portions of the spectrum
for which the water-leaving radiance
L
W
(
)
3.3.2 Opticallysignificant constituents
of thewater column
The optical properties of the water column are a crucial
consideration in any effort to derive information on the
wetted portion of a river channel from remotely sensed
data. Even for attributes such as channel morphology or
λ
)issma l
Search WWH ::
Custom Search