4.1 Introduction
4.1.1 Relevance to Coral Reef Management
Coral reefs are typified by colorful assemblages of organisms, and beyond the red,
green and blue multispectral capability of our eyes there is potential rich discriminating power in the spectral details of the color of reef benthos. This possibility is suggested by the reef dwelling mantis-shrimp, the eyes of which contain
more than ten different wavelength sensitivities over the 400–700 nm range
(Cronin and Marshall 1989). Clearly, evolutionary pressures have directed the
mantis shrimp to make use of the spectral details of the reef environment, and by
exploiting hyperspectral remotely sensed data we may do so too.
Viewed from above, the structure of a reef includes a number of components of
optical significance, the air–water interface, the water column and a bottom
reflectance that is dependent on the benthic types present. From a benthic mapping
perspective the optical contribution of the water column and surface reflectance is
just ‘noise’ that we effectively want to discard, leaving only the benthic reflectance
for analysis. Conversely, from the perspective of those only interested in water
constituents a variable bottom reflectance is the ‘noise’ and a confounding factor
for analysis. The benthic types that we wish to distinguish have spectral profiles
dependent on the pigments they contain. In some cases these pigments are the
same across different types, for example chlorophyll occurs in corals and algae;
but others are distinct, for example the colors of red, green and brown macroalgae
differ because they contain some distinct pigments (Hedley and Mumby 2002).
The premise of hyperspectral analysis is that the optical contributions of a coral
reef system’s components and differing benthic types have, to some extent,
spectrally distinct shapes with respect to wavelength. Multispectral sensors with
broadband responses have insufficient resolution to capture detailed peaks and
troughs in spectral absorption and scattering, but hyperspectral, or narrow-band,
sensors can reveal this detail (Fig. 4.1). The spectral absorption of light by colored
dissolved organic matter (CDOM) in the water has a highly characteristic curved
shape, as does absorption by chlorophyll in phytoplankton and the pure water itself
(Fig. 4.2). Similarly, spectral reflectances of benthic types (Fig. 4.3), while highly
variable, are to some extent distinct between types of interest (Hochberg et al.
2003a; Holden and LeDrew 1999). Hyperspectral data offers the promise of separating the contributions of these different components by means of their characteristic spectral shapes, hence separating the ‘noise’ from the ‘signal’ for a given
application. Nevertheless, the inherent variability of the optical properties of
individual components, and in particular the complex benthic spatial structure of
coral reefs, is a limiting factor to what can be achieved. It is important to
appreciate that components that are spectrally separable may not be in alignment
with the components that are desired to be separable from a scientific or management point of view. Corals and macroalgae are significantly different in an
ecological sense, but can have very similar spectral reflectances. Both contain
80
J. D. Hedley
4.1.1 Relevance to Coral Reef Management
Coral reefs are typified by colorful assemblages of organisms, and beyond the red,
green and blue multispectral capability of our eyes there is potential rich discriminating power in the spectral details of the color of reef benthos. This possibility is suggested by the reef dwelling mantis-shrimp, the eyes of which contain
more than ten different wavelength sensitivities over the 400–700 nm range
(Cronin and Marshall 1989). Clearly, evolutionary pressures have directed the
mantis shrimp to make use of the spectral details of the reef environment, and by
exploiting hyperspectral remotely sensed data we may do so too.
Viewed from above, the structure of a reef includes a number of components of
optical significance, the air–water interface, the water column and a bottom
reflectance that is dependent on the benthic types present. From a benthic mapping
perspective the optical contribution of the water column and surface reflectance is
just ‘noise’ that we effectively want to discard, leaving only the benthic reflectance
for analysis. Conversely, from the perspective of those only interested in water
constituents a variable bottom reflectance is the ‘noise’ and a confounding factor
for analysis. The benthic types that we wish to distinguish have spectral profiles
dependent on the pigments they contain. In some cases these pigments are the
same across different types, for example chlorophyll occurs in corals and algae;
but others are distinct, for example the colors of red, green and brown macroalgae
differ because they contain some distinct pigments (Hedley and Mumby 2002).
The premise of hyperspectral analysis is that the optical contributions of a coral
reef system’s components and differing benthic types have, to some extent,
spectrally distinct shapes with respect to wavelength. Multispectral sensors with
broadband responses have insufficient resolution to capture detailed peaks and
troughs in spectral absorption and scattering, but hyperspectral, or narrow-band,
sensors can reveal this detail (Fig. 4.1). The spectral absorption of light by colored
dissolved organic matter (CDOM) in the water has a highly characteristic curved
shape, as does absorption by chlorophyll in phytoplankton and the pure water itself
(Fig. 4.2). Similarly, spectral reflectances of benthic types (Fig. 4.3), while highly
variable, are to some extent distinct between types of interest (Hochberg et al.
2003a; Holden and LeDrew 1999). Hyperspectral data offers the promise of separating the contributions of these different components by means of their characteristic spectral shapes, hence separating the ‘noise’ from the ‘signal’ for a given
application. Nevertheless, the inherent variability of the optical properties of
individual components, and in particular the complex benthic spatial structure of
coral reefs, is a limiting factor to what can be achieved. It is important to
appreciate that components that are spectrally separable may not be in alignment
with the components that are desired to be separable from a scientific or management point of view. Corals and macroalgae are significantly different in an
ecological sense, but can have very similar spectral reflectances. Both contain
80
J. D. Hedley