An Introduction to the Physical Basis for Deriving River Information by Optical Remote Sensing - Fluvial Remote Sensing for Science and Management - page 61

between the streambed and water column, as well as the
other factors included within the constant A .
The only quantity in Equation (3.16) that varies on a
pixel-by-pixel basis within an image is the one of primary
interest, the flow depth d . For a reach without tribu-
tary inputs or other sources of suspended sediment, the
inherent optical properties of the water column can be
assumed homogeneous, and these properties determine
the values of K (

height; 2) spectral resolution , which refers to the number
of wavelength bands in which radiance measurements are
made, as well as the location andwidth of these bands; and
3) radiometric resolution , which describes the sensor's
ability to detect small changes in radiance and depends on
the manner in which the continuous upwelling radiance
signal is converted to discrete, digital image data. In the
following paragraphs, we examine each of these sensor
characteristics and discuss their influence on the utility of
remotely sensed data for specific applications.
Perhaps the most obvious decision made in planning
the acquisition of remotely sensed data, or identifying
appropriate existing data sets, is that of spatial resolu-
tion, or pixel size. In essence, the image data must be
sufficiently detailed to allow the user to 'see' the river
features of interest. The ratio of mean channel width
to image pixel size thus represents a fundamental con-
straint. If this ratio is on the order of one, even detecting
the channel could be problematic; this is often the case
for small- to medium-sized rivers less than 50m wide
m and moderate-resolution satellite systems (e.g., Land-
sat's 30m pixels). A number of commercial satellites now
provide much greater spatial resolution, often with a sub-
meter panchromatic band and multispectral bands with
pixel sizes of 2mor less; current examples include Ikonos,
QuickBird, GeoEye, and WorldView-2.
Airborne systems provide greater flexibility because the
flying height can be adjusted so as to achieve a desired
pixel size, and some low-altitude deployments achieve
resolutions on the order of a few cm (e.g., Carbonneau
et al., 2004; Lejot et al., 2007). Acquiring such high
resolution data can mitigate the problem of mixed pixels
along the channel banks, although some degree of mixing
is inevitable. With smaller pixels, the area over which
such mixtures occur will be reduced, as will the area
subject to negative depth estimates along shallow channel
margins. Within the channel proper, ambiguity due to
sub-pixel variations in depth and bottom reflectance
can largely be avoided with high spatial resolution data.
These detailed images also enable more traditional photo
interpretation, and features such as woody debris will
often be clearly visible. Such high resolution data are not
without drawbacks, however. Other factors to consider
include the practicality of flying at low altitudes, the small
area covered by individual images along a flight line, the
massive volume of data involved, and the effort required
to geo-reference and analyse large numbers of images.
For many applications, then, a somewhat coarser reso-
lution, with pixels on the order of 1-4m, might be more
appropriate for large-scale mapping and monitoring.

). Substrate composition
might vary within a reach, if not on a pixel scale, but
the ratio R B (

λ )and R C ( λ

λ 2 ) remains approximately constant
across bottom types. These other factors therefore have
relatively little influence on X , implying that this image-
derived quantity primarily depends on depth and is thus
useful for bathymetric mapping. In practice, depth maps
are produced by regressing field measurements of depth
against X values for the corresponding image pixels and
applying the resulting calibration equation throughout
the image. The radiative transfer simulations, field spec-
tra, and image processing reported by Legleiter et al.
2009 provided both theoretical and empirical validation
of the critical assumptions underlying this ratio-based
approach.

λ 1 ) / R B (

3.4.3 The roleof sensor characteristics

In addition to the physical processes that govern the
interaction of light and water and thus dictate the optical
properties of river channels, the characteristics of the sen-
sors themselves are also an important consideration. The
technical specifications of an imaging system play a key
role in determining which river attributes can be remotely
mapped, and with what degree of accuracy and precision.
For the most part, the physics that control the upwelling
spectral radiance must simply be accepted, although the
timing of data acquisition can be specified so as to avoid
unfavourable circumstances such as high turbidity or
strong sun glint. Faced with these largely immutable
physical constraints, prospective users of remotely sensed
data can improve their chances of deriving useful river
information by selecting instrumentation appropriate for
a particular project's objectives. Certain types of sensors
will be better able to provide image data from which the
attributes of interest can be reliably inferred than other
kinds of instruments. Financial and logistical considera-
tions always play a role, of course, but inplanning a remote
sensing mission one would like to specify three principal
sensor characteristics: 1) spatial resolution , which is typ-
ically equated with the edge dimension of an image pixel
on the ground and depends on sensor optics and flying