324
S. Andréfouët
cite the Landsat TM processing by Dahdouh-Guebas et al. (1999) in Kenya, the
Landsat TM/ETM + change detection work by Gullstroem et al. (2006) and Knudby
et al. (2010b) both around Zanzibar Island, Tanzania, and the IKONOS mapping by
Knudby and Nordlund (2011) also next to Zanzibar.
From the digital airborne hyperspectral realm, Despinoy and Naim (1997) have
mapped reefs of La Reunion Island and Borstad et al. (1997) looked at Mauritius
reefs and lagoons. Red Sea reefs have been investigated by Hamylton (2011a) and
Rowlands et al. (2012). Possibly the largest hyperspectral mapping exercise was
conducted in Amirantes, Seychelles. This work yielded a well-documented atlas
of the reefs and islands of this archipelago (Spencer et al. 2009). All the above
hyperspectral work relied on a CASI.
In addition to “classic” satellite and airborne missions, an ultra-light aircraft has
been used in La Reunion Island to acquire very high resolution vertical images
of coral bleaching events (Pennober and Borius 2010). Finally, hand-held digital
photographs taken by astronauts from the International Space Station served as a
background image for mapping reefs in Iles Eparses (Samani 2005).
Methodologically, multispectral image processing techniques for habitat mapping
have significantly evolved during the two last decades. A number of the aforementioned studies had a strong innovative methodological goal (Purkis and Pasterkamp
2004), whereas others were more thematically oriented and relied on standard processing methods (Andréfouët et al. 2003; Knudby et al. 2010b). Part of the habitat
mapping challenge is to maintain both high thematic richness (number of mapped
habitat classes) and high classification accuracy. Andréfouët (2008) discusses the
relative merits of producers and users approaches in addressing this challenge, and
devises the benefits of combining simple visual interpretation methods with traditional image processing methods, especially in a capacity building context. Examples
of habitat maps produced with this user-oriented approach are shown Fig. 16.4.
16.2.4 Processes and Functional Studies
Habitat maps, once created, should be used for some type of applications. A map is not
necessarily an end in itself. However, few studies go the extra-step, beyond the habitat
mapping exercise. Fundamental science and resource management benefited poorly
from habitat maps. Thus far, functional studies looking at processes controlling
coral reef communities remain scarce. They are limited to the characterization of
hydrodynamic regime, geological processes (see Purkis et al. 2010 as discussed
above), and fish community modelling. Finally, there is an increasing effort to map
coral reef resilience potential at local and regional scale.
The spatial distribution of living communities is partly constrained by the hydrodynamic regime. Water movements around reefs can often be inferred on visible and
near infra-red images, especially when significant sunglint can offer an image of the
sea surface. The patterns obviously only capture snapshots in time if only one image is
used. Nevertheless, surface features visible on SPOT images have been used to infer
the relative difference of hydrodynamic exposure among different reef sections in the
S. Andréfouët
cite the Landsat TM processing by Dahdouh-Guebas et al. (1999) in Kenya, the
Landsat TM/ETM + change detection work by Gullstroem et al. (2006) and Knudby
et al. (2010b) both around Zanzibar Island, Tanzania, and the IKONOS mapping by
Knudby and Nordlund (2011) also next to Zanzibar.
From the digital airborne hyperspectral realm, Despinoy and Naim (1997) have
mapped reefs of La Reunion Island and Borstad et al. (1997) looked at Mauritius
reefs and lagoons. Red Sea reefs have been investigated by Hamylton (2011a) and
Rowlands et al. (2012). Possibly the largest hyperspectral mapping exercise was
conducted in Amirantes, Seychelles. This work yielded a well-documented atlas
of the reefs and islands of this archipelago (Spencer et al. 2009). All the above
hyperspectral work relied on a CASI.
In addition to “classic” satellite and airborne missions, an ultra-light aircraft has
been used in La Reunion Island to acquire very high resolution vertical images
of coral bleaching events (Pennober and Borius 2010). Finally, hand-held digital
photographs taken by astronauts from the International Space Station served as a
background image for mapping reefs in Iles Eparses (Samani 2005).
Methodologically, multispectral image processing techniques for habitat mapping
have significantly evolved during the two last decades. A number of the aforementioned studies had a strong innovative methodological goal (Purkis and Pasterkamp
2004), whereas others were more thematically oriented and relied on standard processing methods (Andréfouët et al. 2003; Knudby et al. 2010b). Part of the habitat
mapping challenge is to maintain both high thematic richness (number of mapped
habitat classes) and high classification accuracy. Andréfouët (2008) discusses the
relative merits of producers and users approaches in addressing this challenge, and
devises the benefits of combining simple visual interpretation methods with traditional image processing methods, especially in a capacity building context. Examples
of habitat maps produced with this user-oriented approach are shown Fig. 16.4.
16.2.4 Processes and Functional Studies
Habitat maps, once created, should be used for some type of applications. A map is not
necessarily an end in itself. However, few studies go the extra-step, beyond the habitat
mapping exercise. Fundamental science and resource management benefited poorly
from habitat maps. Thus far, functional studies looking at processes controlling
coral reef communities remain scarce. They are limited to the characterization of
hydrodynamic regime, geological processes (see Purkis et al. 2010 as discussed
above), and fish community modelling. Finally, there is an increasing effort to map
coral reef resilience potential at local and regional scale.
The spatial distribution of living communities is partly constrained by the hydrodynamic regime. Water movements around reefs can often be inferred on visible and
near infra-red images, especially when significant sunglint can offer an image of the
sea surface. The patterns obviously only capture snapshots in time if only one image is
used. Nevertheless, surface features visible on SPOT images have been used to infer
the relative difference of hydrodynamic exposure among different reef sections in the