An Introduction to the Physical Basis for Deriving River Information by Optical Remote Sensing - Fluvial Remote Sensing for Science and Management - page 65

For example, high spatial resolution corresponds to small
spatial bins, implying that the bins must encompass a
broader range of wavelengths in order for them to be
filled to the depth set by the sensor's radiometric res-
olution. Conversely, subdividing photons into narrow
spectral bands typically implies an increase in pixel size.
As a result, satellite or airborne sensors with very high
spatial resolution tend to be multispectral, with only
three or four bands, whereas hyperspectral instruments
tend to have larger pixel sizes. Radiometric resolution
determines bin depth and thus represents a key con-
straint, often dictated by the signal to noise properties
of the system electronics. In any case, these sensor char-
acteristics are a crucial consideration in any application
of remote sensing to rivers. Prospective users of this
technology must understand what levels of spatial, spec-
tral, and radiometric resolution are required for different
types of studies. For example, an investigation of channel
change might benefit from a highly sensitive detector
with 12-bit radiometric resolution that enables very pre-
cise depth estimates, whereas an effort to map different
algal species might prefer greater spectral resolution that
allows distinctive chlorophyll absorption features to be
distinguished. To make effective use of remote sensing
techniques in river management, project objectives must
be clearly defined and prioritised so that instrumentation
and/or existing data sources can be selected accordingly.

Whereas traditional techniques that rely upon statis-
tical correlations between ground-based measurements
and image pixel values produce scene-specific results,
a more physics-based approach could provide greater
generality and reduce the need for in situ observations.
Significant progress toward this end has already been
made in coastal and shallow marine environments. For
example, Mobley et al. (2005) used a numerical radia-
tive transfer model to create lookup tables containing
spectra for a range of depths, bottom types, and water
column optical properties. By matching these spectra to
hyperspectral image data, Lesser and Mobley (2007) were
able to map these attributes over a broad area and to
a high degree of accuracy, without extensive field data.
Remote sensing of rivers has not yet reached such an
advanced stage, but additional effort toward this goal is
justified by the advantages of a physics-based approach:
(1) the amount of field data required for calibration
could be greatly diminished, with a smaller number of
ground-based measurements used primarily for valida-
tion; (2) more flexible, generic mapping algorithms could
be developed and applied to larger spatial extents and/or
archival image data to examine river dynamics over an
expanded range of scales; and (3) the accuracy and preci-
sion of image-derived river information can be assessed
a priori within a forward image modeling framework
that represents physical processes along the image chain
(Legleiter and Roberts, 2009). Ultimately, we hope to see
continued progress toward more physics-based remote
mapping of fluvial systems.
In the interim, there is potential to leverage existing
monitoring programs to obtain the ground reference data
needed to implement traditional, empirical approaches to
remote sensing of rivers. For example, repeat cross-section
surveys conducted regularly for geomorphic monitor-
ing purposes might be coordinated with the acquisition
of remotely sensed data and thus used to calibrate
image-derived depth estimates. Exploiting the synergy
between current, ground-based data collection activities
and remote sensing campaigns could serve as a bridge
for extending spectrally-based bathymetric mapping over
longer stream segments and on a more routine basis. Such
an effort could facilitate various management objectives
and help to build confidence in the use of remote sens-
ing techniques in future monitoring programs. In any
case, we believe that even for practitioners using empir-
ical approaches, some appreciation of the underlying
radiative transfer processes is critical.
An improved understanding of these processes could
enable a number of innovative applications, beyond the

3.5 Conclusion

This chapter has provided an overview of the physical
basis for remote sensing of rivers. Various radiometric
quantities were defined and used to describe the radiative
transfer processes by which light and water interact in
shallow stream channels. We employed an image chain
analogy to link these processes to one another, discuss
the influence of the Earth's atmosphere, and partition
the radiance signal recorded by a remote detector into
a set of components. A particular application, estima-
tion of water depth, was emphasised throughout the
chapter and used to illustrate how river information can
be derived from remotely sensed data. Finally, the effects
of sensor characteristics - spatial, spectral, and radio-
metric resolution - on the nature and reliability of this
information was discussed. In keeping with the chapter's
principal objective, we have stressed the importance of
understanding the physical processes underlying remotely
sensed data, even when these data are primarily analysed
using empirical approaches.