Chapter 15 Remote Sensing of Seagrasses
357
determine water column depth and leaf area index of
Thalassia testudinum. Kutser et al. (2003) used Hydrolight RT-modeling to assess the airborne and
spaceborne hyperspectral data discrimination of substratum and substratum cover over coral reefs. They
could discriminate eight substratum types of coral,
algae, and cyanobacteria.
Analytical- and RT-based forward and inverse
models have several strong advantages over any of
the other methods applied, presuming that a remote
sensing image has already been corrected for atmospheric and air–water interface effects.
1. Repeatability—multitemporal images can be
compared quantitatively as the methodology is
objective and physics-rules based. Corrections
for changing water column depth (tides!) and
varying concentrations of water column constituents is possible.
2. Transferability—application of the models to
data from other sensors is straightforward and
only involves adaptation of the spectral bands.
3. Sensitivity and error analysis are exact and objectively determined. Once initialized, processing
of images is fast compared to human interpreter
based methods (seconds, minutes, or at the most
hours for a remote sensor image).
4. New knowledge can be added to the simulations,
and can be retrospectively applied to remote sensing images.
5. Archival remote sensing data (e.g. Landsat TM
data from 1984 onwards) can be processed according to the currently developed methodology.
The fact that no field measurements are available
from the past does not prevent analysis of the images; the only assumption is that the more recent
spectral information is correct in shape for each
of the components as illustrated in the Landsatbased seagrass change detection study by Anstee
et al. (2004).
VI. Conclusions, Recommendations,
and Outlook
Future developments (see also Malthus and Mumby,
2003) in the areas of airborne and spaceborne sensors, underwater optical instrumentation, spectral libraries, improved bathymetric datasets, simulation
and inversion methods, and the integration of remote
sensing with other methodologies will lead to accelerated development of methods for the remote sensing, mapping, and monitoring of seagrass meadows.
Remote sensing offers one of the most versatile and
accurate techniques for seagrass assessments at any
scale (down to ground resolutions of 0.5 m), where
field methods cannot be used to accomplish the required task in a reasonable time. There is a trend for
remotely sensed data to become more cost-effective,
either because of the real reductions in the cost of
raw data for multispectral imagery, or because of
the increase in the number of indicators that can be
retrieved using airborne hyperspectral data.
The sophisticated procedures applied to derive
benthic maps from digital multispectral or hyperspectral remote sensing require a combination of
mathematical, software, hardware, physics, and biogeochemistry skills that currently restrict management institutions from investing in this data acquisition capability. In contrast, seagrass maps can be
routinely produced by seagrass experts within an organization using aerial photographs. Despite high
spatial resolution, the poor spectral resolution of
aerial photography is insensitive to subtle spectral
variations and limits the successful discrimination
of submerged features (e.g. Holden and LeDrew,
1999). It would require only minimal training in
the use of remote sensing software for these same
staff to produce more accurate seagrass maps of
higher information content using the simpler empirical and semi-empirical methods for image analysis.
Though it may require higher investment to procure
the higher quality seagrass products based on analytical or RT modeling at present, the cost should be
balanced against the type, quantity, and accuracy of
information such techniques can provide. It is likely
that remote sensing from aircraft and satellites will
be the methods more often applied in the future. Indeed in the coral reef community, worldwide spectral library measurement programs (13,000 spectra
collected; see Hochberg et al. (2003)) have led to
a demand for remote sensing of coral reef ecosystems. The seagrass community should also carry out
a worldwide spectral library collection program (including the measurement of co-occurring benthic
micro-algae, macro-algae, sediment, and rock substratum), to mature the field of hyperspectral remote
sensing (by standardizing processing methods) for
use by seagrass biologists in their studies.
Merging the seagrass canopy structure geometry
analysis by Zimmerman in Chapter 13 into the underwater RT or analytical optical models required
357
determine water column depth and leaf area index of
Thalassia testudinum. Kutser et al. (2003) used Hydrolight RT-modeling to assess the airborne and
spaceborne hyperspectral data discrimination of substratum and substratum cover over coral reefs. They
could discriminate eight substratum types of coral,
algae, and cyanobacteria.
Analytical- and RT-based forward and inverse
models have several strong advantages over any of
the other methods applied, presuming that a remote
sensing image has already been corrected for atmospheric and air–water interface effects.
1. Repeatability—multitemporal images can be
compared quantitatively as the methodology is
objective and physics-rules based. Corrections
for changing water column depth (tides!) and
varying concentrations of water column constituents is possible.
2. Transferability—application of the models to
data from other sensors is straightforward and
only involves adaptation of the spectral bands.
3. Sensitivity and error analysis are exact and objectively determined. Once initialized, processing
of images is fast compared to human interpreter
based methods (seconds, minutes, or at the most
hours for a remote sensor image).
4. New knowledge can be added to the simulations,
and can be retrospectively applied to remote sensing images.
5. Archival remote sensing data (e.g. Landsat TM
data from 1984 onwards) can be processed according to the currently developed methodology.
The fact that no field measurements are available
from the past does not prevent analysis of the images; the only assumption is that the more recent
spectral information is correct in shape for each
of the components as illustrated in the Landsatbased seagrass change detection study by Anstee
et al. (2004).
VI. Conclusions, Recommendations,
and Outlook
Future developments (see also Malthus and Mumby,
2003) in the areas of airborne and spaceborne sensors, underwater optical instrumentation, spectral libraries, improved bathymetric datasets, simulation
and inversion methods, and the integration of remote
sensing with other methodologies will lead to accelerated development of methods for the remote sensing, mapping, and monitoring of seagrass meadows.
Remote sensing offers one of the most versatile and
accurate techniques for seagrass assessments at any
scale (down to ground resolutions of 0.5 m), where
field methods cannot be used to accomplish the required task in a reasonable time. There is a trend for
remotely sensed data to become more cost-effective,
either because of the real reductions in the cost of
raw data for multispectral imagery, or because of
the increase in the number of indicators that can be
retrieved using airborne hyperspectral data.
The sophisticated procedures applied to derive
benthic maps from digital multispectral or hyperspectral remote sensing require a combination of
mathematical, software, hardware, physics, and biogeochemistry skills that currently restrict management institutions from investing in this data acquisition capability. In contrast, seagrass maps can be
routinely produced by seagrass experts within an organization using aerial photographs. Despite high
spatial resolution, the poor spectral resolution of
aerial photography is insensitive to subtle spectral
variations and limits the successful discrimination
of submerged features (e.g. Holden and LeDrew,
1999). It would require only minimal training in
the use of remote sensing software for these same
staff to produce more accurate seagrass maps of
higher information content using the simpler empirical and semi-empirical methods for image analysis.
Though it may require higher investment to procure
the higher quality seagrass products based on analytical or RT modeling at present, the cost should be
balanced against the type, quantity, and accuracy of
information such techniques can provide. It is likely
that remote sensing from aircraft and satellites will
be the methods more often applied in the future. Indeed in the coral reef community, worldwide spectral library measurement programs (13,000 spectra
collected; see Hochberg et al. (2003)) have led to
a demand for remote sensing of coral reef ecosystems. The seagrass community should also carry out
a worldwide spectral library collection program (including the measurement of co-occurring benthic
micro-algae, macro-algae, sediment, and rock substratum), to mature the field of hyperspectral remote
sensing (by standardizing processing methods) for
use by seagrass biologists in their studies.
Merging the seagrass canopy structure geometry
analysis by Zimmerman in Chapter 13 into the underwater RT or analytical optical models required