An Introduction to the Physical Basis for Deriving River Information by Optical Remote Sensing - Fluvial Remote Sensing for Science and Management - page 43

limited in several important respects. To develop such
correlations, field data must be collected at the same time
the image is acquired, a requirement that negates one of
the primary advantages of remote sensing. Ground-based
measurements alsomust be associated with specific image
pixels, placing a premium on accurate geo-referencing of
image data, which can be difficult for images with rela-
tively coarse spatial resolution. Moreover, the resulting
calibrations are scene and sensor-specific and cannot be
readily extended to other locations, nor to images of the
same river acquired at other times, due to spatial and
temporal variations in solar radiation and differences in
sensor characteristics. Notably, however, more recently
developed techniques could help to overcome these limi-
tations by incorporating hydraulic (Fonstad and Marcus,
2005) or photogrammetric (Lane et al., 2010) informa-
tion to calibrate image-derived depth estimates when field
data are not available.
Similarly, a more physics-based approach built upon
radiative transfer theory could also help to alleviate some
of these issues by providing greater generality, but such
techniques are not yet available to the fluvial commu-
nity. At present, remote sensing of rivers typically
involves a certain degree of empiricism. Nevertheless,
even regression-based methods must have some under-
lying physical basis, or meaningful correlations with
field data could not be established. The strength of these
correlations, and the extent to which accurate maps of
river attributes can be derived from them, thus depends
on the physics governing the interaction of light and
water, even if these processes are not explicitly considered
during image analysis. For this reason, we maintain that
regardless of position along the continuum from purely
empirical to fully physics-based, an understanding of
radiative transfer processes is essential.
This chapter provides an introduction to these pro-
cesses, and focuses specifically on passive optical remote
sensing of river bathymetry. Estimating water depth
from image data was one of the earliest applications
of remote sensing technology to aquatic environments
(e.g., Lyzenga, 1978), and is perhaps the most mature
use of such methods in a fluvial context. Moreover, in
the shallow, relatively clear-flowing streams of interest
to many scientists and managers, efforts to characterize
other channel attributes must account for the influence
of a water column of variable thickness, or depth. A
discussion of the physical basis for bathymetric mapping
thus encompassesmany of the principles underlying other
applications as well. Thematerial presented in this chapter
is a river-centric distillation of more general treatments of

remote sensing concepts (Schott, 1997), radiative transfer
theory and numerical modeling (Mobley, 1994), opti-
cal properties of various types of water bodies (Bukata
et al., 1995), and recent advances in our understanding of
the unique radiative transfer processes operating in shal-
low water (e.g., Mobley and Sundman, 2003; Dierssen
et al., 2003); the interested reader is referred to these
publications for additional detail. Similarly, more thor-
ough discussion of the topics covered herein can be
found in our earlier work (Legleiter et al., 2004; Legleiter
and Roberts, 2005; Legleiter et al., 2009; Legleiter and
Roberts, 2009).
By focusing our attention on passive optical remote
sensing of the active river channel, we have chosen to
neglect other, complementary technologies and approa-
ches. For example, an active formof remote sensing, Light
Detection And Ranging (LiDAR), has become the pre-
ferred method of topographic measurement for research
on surface processes (Slatton et al., 2007), and newly
developed, water-penetrating green LiDAR systems could
enable precise, simultaneous mapping of both subaerial
topography and wetted channel bathymetry (Bailly et al.,
Chapter 7; Kinzel et al., 2007; McKean et al., 2009). Nor
do we address thermal remote sensing, mapping of ripar-
ian vegetation, or efforts to quantify water quantity and
quality via remote sensing; coverage of these topics is
available elsewhere.
This chapter is organised to first provide some back-
ground on general remote sensing concepts and radiative
transfer theory and then focus more specifically on the
optical properties of the fluvial environment and the ways
in which channel characteristics can be inferred from
measurements of reflected solar energy. The following
sections:
1) introduce quantities used to characterise light fields
and provide an overviewof the radiative transfer processes
operating in shallow streams;
2) describe the spectral characteristics of river chan-
nels, including the water surface, optically significant
constituents of the water column, and bed and bank
materials;
3) illustrate how information on river morphology can
be derived from image data by separating the remotely
sensed signal into component parts and evaluating their
relative magnitudes; and
4) summarise the effects of sensor characteristics such
as spectral, spatial, and radiometric resolution on the
accuracy and precision of image-derived depth estimates.
The chapter concludes by evaluating the prospects for
more physics-based remote sensing of rivers, discussing