16 Remote Sensing of Coral Reefs and Their Environments . . .
323
16.2.3 Benthic Communities
Beyond the geomorphological and depth attributes, habitats are also defined by the
type of benthic cover found on the seafloor. The high diversity of substrate types
(e.g. terrestrial and carbonate sand, rubble, boulders rocks, dead eroded corals, pavement) and benthic organisms (hard and soft corals, macroalgae, seagrass, sponges,
ascidians, microbial mats, etc) results in a virtually infinite number of mosaics of
communities. Most of the time, these communities are organized according to local
energy (due to wave and wind exposure) and depth gradients. Obviously, corals
are the focus of many coral reef remote sensing studies, but in the field, coral
dominated areas may be extremely narrow compared to vast expanses of sedimentary areas where other organisms, such as seagrass, dominate. The importance of
inter-connections and resulting fluxes between the different habitats are now fully
acknowledged in ecological studies and in management plans (e.g. see, for Zanzibar
Island, Dorenbosch et al. 2005).
The reflectance spectral properties of coral organisms from the Red Sea has
been studied by Minghelli-Roman et al. (2002) in the course of a hyperspectral
mapping project. The radiometric discriminant functions obtained from these local
studies were confirmed by measurements that have characterized globally the spectral signatures of a variety of coral reef end-members. Hochberg et al. (2003, 2004)
have measured the spectral differences between benthic end-members from all coral
reef regions worldwide, including samples from Mayotte Island in the Mozambique
Channel.
Measuring the spectral reflectance of individual organisms is not necessarily
immediately useful for mapping communities and habitats with images because
communities are made of the intricate spatial assemblages of organisms. These assemblages can also change with time, according to seasons and disturbances. Using
directly spectral signatures differences and scaling discriminant functions from organisms to communities and habitats to achieve a map is not a trivial task (Andréfouët
et al. 2004; Hedley et al. 2012). In the Red Sea, Purkis and Pasterkamp (2004) have
described the benefits of using in situ reflectance for habitat mapping, but spectral
measurements were achieved for meter-scale communities, which is already a step
ahead in terms of scaling and mixing compared to using organism signatures. As a
result, when mapping habitats, photo-interpretation, purely statistical classification
and object-based segmentation approaches often all appear more suitable especially
when using multispectral images limited in spectral resolution.
The number of satellite-based habitat mapping applications in the Red Sea and
Western Indian Ocean is quite high. We can cite the early Red Sea SPOT study
by Courboulès et al. (1987, 1988); Courboulès and Manière (1992) and Manière
and Jaubert (1985). Later, Landsat TM has been used by Purkis et al. (2002) and
Purkis and Pasterkamp (2004) also in the Red Sea, by Chapman and Turner (2004)
in Rodrigues Island and by Klaus et al. (2003) in Socotra Island. New generation
of very high resolution sensors such as IKONOS and Quickbird have been used in
Mayotte, Zanzibar and the Red Sea (Andréfouët et al. 2003; Knudby et al. 2010a,
Rowlands et al. 2012). Specifically with a focus on seagrass meadows, we can
323
16.2.3 Benthic Communities
Beyond the geomorphological and depth attributes, habitats are also defined by the
type of benthic cover found on the seafloor. The high diversity of substrate types
(e.g. terrestrial and carbonate sand, rubble, boulders rocks, dead eroded corals, pavement) and benthic organisms (hard and soft corals, macroalgae, seagrass, sponges,
ascidians, microbial mats, etc) results in a virtually infinite number of mosaics of
communities. Most of the time, these communities are organized according to local
energy (due to wave and wind exposure) and depth gradients. Obviously, corals
are the focus of many coral reef remote sensing studies, but in the field, coral
dominated areas may be extremely narrow compared to vast expanses of sedimentary areas where other organisms, such as seagrass, dominate. The importance of
inter-connections and resulting fluxes between the different habitats are now fully
acknowledged in ecological studies and in management plans (e.g. see, for Zanzibar
Island, Dorenbosch et al. 2005).
The reflectance spectral properties of coral organisms from the Red Sea has
been studied by Minghelli-Roman et al. (2002) in the course of a hyperspectral
mapping project. The radiometric discriminant functions obtained from these local
studies were confirmed by measurements that have characterized globally the spectral signatures of a variety of coral reef end-members. Hochberg et al. (2003, 2004)
have measured the spectral differences between benthic end-members from all coral
reef regions worldwide, including samples from Mayotte Island in the Mozambique
Channel.
Measuring the spectral reflectance of individual organisms is not necessarily
immediately useful for mapping communities and habitats with images because
communities are made of the intricate spatial assemblages of organisms. These assemblages can also change with time, according to seasons and disturbances. Using
directly spectral signatures differences and scaling discriminant functions from organisms to communities and habitats to achieve a map is not a trivial task (Andréfouët
et al. 2004; Hedley et al. 2012). In the Red Sea, Purkis and Pasterkamp (2004) have
described the benefits of using in situ reflectance for habitat mapping, but spectral
measurements were achieved for meter-scale communities, which is already a step
ahead in terms of scaling and mixing compared to using organism signatures. As a
result, when mapping habitats, photo-interpretation, purely statistical classification
and object-based segmentation approaches often all appear more suitable especially
when using multispectral images limited in spectral resolution.
The number of satellite-based habitat mapping applications in the Red Sea and
Western Indian Ocean is quite high. We can cite the early Red Sea SPOT study
by Courboulès et al. (1987, 1988); Courboulès and Manière (1992) and Manière
and Jaubert (1985). Later, Landsat TM has been used by Purkis et al. (2002) and
Purkis and Pasterkamp (2004) also in the Red Sea, by Chapman and Turner (2004)
in Rodrigues Island and by Klaus et al. (2003) in Socotra Island. New generation
of very high resolution sensors such as IKONOS and Quickbird have been used in
Mayotte, Zanzibar and the Red Sea (Andréfouët et al. 2003; Knudby et al. 2010a,
Rowlands et al. 2012). Specifically with a focus on seagrass meadows, we can