Hyperspectral Imagery in Fluvial Environments - Fluvial Remote Sensing for Science and Management - page 80

advances of this type in the remote sensing of rivers; it
would be extremely useful for river managers to know the
likely quality levels of their data before they are contracted.

river habitats, whether these classifications are supervised
or unsupervised. For those larger projects that have the
money and need for a great deal of remotely-collected
river data, hyperspectral data are second to none, and are
likely quite cost-effective when compared with ground
measurements.
The future of hyperspectral imaging of river environ-
ments is very bright. Most importantly, the number of
hyperspectral sensors and organisations flying these sen-
sors is growing very rapidly. This has the long-term effect
of reducing costs. Computational power and data stor-
age, once major problems with using large hyperspectral
imagery, are quickly becoming non-issues. Hyperspectral
data can be used to test the usefulness and veracity of more
modest instruments (such as three-band aerial imagery)
to extract river instruments. The work by the ocean optics
community and a handful of river researchers are quickly
determining the uses and limitations of hyperspectral
imagery in detecting various water environments, and
doing so in a quantitative, replicable way that provides a
level of quality control necessary in the profession uses of
such imagery. Hyperspectral remote sensing of rivers is
beginning to be combined with other data types, such as
with LiDAR data (Hall et al., 2009). The combination of
hyperspectral, LiDAR, and high-resolution aerial imagery
to provide synoptic views riverscape constituents should
become feasible in the near future. Whether riparian or
fluvial, hyperspectral remote sensing of riverscapes is one
of the crowning jewels of river mapping.

4.6 Conclusions

Hyperspectral data are not a panacea for remote river
studies. At their best, they are still quite expensive and
logistically complex. At their worst, the effects of the
atmosphere and other factors can make interpretation
very difficult andmisleading. For these reasons alone, cau-
tion needs to be exercised when considering hyperspectral
imaging for river monitoring. The apparent realism and
flexibility of hyperspectral tools is considerable, and expe-
rience is needed to know how to analyse these data to
provide real river information rather than apparently real
yet false information.
There are, unfortunately, no hyperspectral instruments
in orbit that have the ground resolution needed to map
most of the world's small to medium-sized river environ-
ments. While some airborne hyperspectral imaging can
yield very high resolution imagery, they cannot approach
the cm-scale resolution that some digital cameras can
provide; such resolutions are necessary for some river
analyses such as extracting particle size information.
Low-flying remote control aircraft with the ability to
carry hyperspectral sensors and be used by civilians over
most river areas do not seem to become likely in the
near future. Most of all, hyperspectral imaging is only
available for the last few years at best, and it usually
is only available once an investigator pays for a tasked
mission. As such, long-term decadal change detection
via hyperspectral methods is not yet a reality. There are
not yet agreed-upon standards for using hyperspectral
data in river analysis, for example the many methods for
radiometric calibration of these images for rivers have
not been tested against each other systematically.
The advantages that come with well-used hyperspec-
tral imagery are, however, immense. The large number of
narrowly-defined image bands provides the perfect data
source for modern image processing algorithms. Specific
river variables, such as water depth, are strongly related
to particular wavelengths, and these can be singled out
from hyperspectral data and used to provide optimum
information extraction. Modern optical methods, cou-
pled with hyperspectral data and spectral libraries, can in
principle yield quantitative river environment informa-
tion without the need for in situ data measurement. The
multitude of layers allows for realistic classification of

Acknowledgments

I would like to acknowledge the efforts of AndrewMarcus
and Carl Legleiter, who contributed images, valuable
discussions, and unyielding field camaraderie.

References

Ackleson, S.G. 2003. Light in shallow waters: A brief research
review. Limnology and Oceanography , 48(1, part 2), 323-328.
Aspinall, R.J., Marcus, W.A., and Boardman, J.W. 2002. Consid-
erations in collecting, processing, and analyzing high spatial
resolution hyperspectral data for environmental investiga-
tions. Journal of Geographical Systems , 4, 15-29.
Chen, Z., Hu, C., and Muller-Karger, F. 2007. Monitoring
turbidity in Tampa Bay using MODIS/Aqua 250-m imagery.
Remote Sensing of Environment , 109, 207-220.
Chen, Z., Muller-Karger, F., and Hu, C. 2007. Remote sensing of
water clarity in Tampa Bay. Remote Sensing of Environment ,
109, 249-259.