20 Ocean-Colour Radiometry
353
But the development of algorithms for Case-2 waters cannot yet be considered
as finished: unresolved issues include development of algorithms that work across
many geographic regions (IOCCG, 2000). In Case-2 waters, the water-leaving radiance is influenced by a number of optically-active substances varying independently
of each other. The nature of these substances and the range of their concentrations
vary across location. Differences in their inherent optical properties, and insufficient information on the variability in these properties, hinder further progress. It
may well be that a global algorithm that would perform equally well in all regions is
not feasible. If such be the case, then branching algorithms would be needed, each
optimised for various conditions, such that regional algorithms could be stitched
together in a seamless manner. In Case-2 waters, one or more sea-water constituents
may mask the signals from other substances to such a level that the retrieval of the
minor constituents may become impossible.
Such limits on retrieval algorithms have to be established clearly, along with the
precision of the retrieved variables under all realistic conditions. Such information
would facilitate further applications of Case-2 algorithms: users need to understand
exactly what the satellite products are revealing, and the level of confidence that can
be placed in the products. Case-2 waters are influenced by phytoplankton, yellow
substances and suspended particulate material other than phytoplankton, all varying
independently of each other, whereas Case-1 waters are defined as those waters in
which optical variability is determined primarily by phytoplankton and substances
co-varying with them. As such, Case-1 waters can be considered a sub-set of Case-2
waters. Therefore, in principle, Case-1 algorithms can be subsumed within Case-2
algorithms, provided Case-2 algorithms are tested as rigorously as has been the case
for Case-1 algorithms.
Though the initial drive to improve spectral resolution in ocean-colour satellites
came from the desire to improve applications to coastal waters, high spectral resolution has also allowed us to move beyond the detection of just the concentration
of phytoplankton as indexed by chlorophyll concentration, towards identification of
various phytoplankton types that have distinct optical signatures (Nair et al., 2008).
Algorithms for identifying a number of phytoplankton types, such as diatoms, coccolithophores and certain types of blue-green algae have emerged in the last few
years. Algorithms that discriminate between phytoplankton on the basis of their
size have also been proposed. Algorithms for discriminating between types of phytoplankton may be classified as those that are based on abundance and those that are
based on spectral signatures. The abundance-based methods relate certain ranges
in the concentrations in chlorophyll (trophic status) with a particular phytoplankton community. Methods based on spectral signatures use discriminatory traits
in the optical properties of certain phytoplankton to distinguish them from other
phytoplankton.
Both approaches have their limitations and their advantages. The spectralsignature methods have the potential to improve in accuracy and expand in range of
phytoplankton types that may be so-identified, as instruments with higher spectral
resolution become available. The non-linear nature of the algorithms, the plasticity
in the optical properties of phytoplankton functional types and the small signals on
353
But the development of algorithms for Case-2 waters cannot yet be considered
as finished: unresolved issues include development of algorithms that work across
many geographic regions (IOCCG, 2000). In Case-2 waters, the water-leaving radiance is influenced by a number of optically-active substances varying independently
of each other. The nature of these substances and the range of their concentrations
vary across location. Differences in their inherent optical properties, and insufficient information on the variability in these properties, hinder further progress. It
may well be that a global algorithm that would perform equally well in all regions is
not feasible. If such be the case, then branching algorithms would be needed, each
optimised for various conditions, such that regional algorithms could be stitched
together in a seamless manner. In Case-2 waters, one or more sea-water constituents
may mask the signals from other substances to such a level that the retrieval of the
minor constituents may become impossible.
Such limits on retrieval algorithms have to be established clearly, along with the
precision of the retrieved variables under all realistic conditions. Such information
would facilitate further applications of Case-2 algorithms: users need to understand
exactly what the satellite products are revealing, and the level of confidence that can
be placed in the products. Case-2 waters are influenced by phytoplankton, yellow
substances and suspended particulate material other than phytoplankton, all varying
independently of each other, whereas Case-1 waters are defined as those waters in
which optical variability is determined primarily by phytoplankton and substances
co-varying with them. As such, Case-1 waters can be considered a sub-set of Case-2
waters. Therefore, in principle, Case-1 algorithms can be subsumed within Case-2
algorithms, provided Case-2 algorithms are tested as rigorously as has been the case
for Case-1 algorithms.
Though the initial drive to improve spectral resolution in ocean-colour satellites
came from the desire to improve applications to coastal waters, high spectral resolution has also allowed us to move beyond the detection of just the concentration
of phytoplankton as indexed by chlorophyll concentration, towards identification of
various phytoplankton types that have distinct optical signatures (Nair et al., 2008).
Algorithms for identifying a number of phytoplankton types, such as diatoms, coccolithophores and certain types of blue-green algae have emerged in the last few
years. Algorithms that discriminate between phytoplankton on the basis of their
size have also been proposed. Algorithms for discriminating between types of phytoplankton may be classified as those that are based on abundance and those that are
based on spectral signatures. The abundance-based methods relate certain ranges
in the concentrations in chlorophyll (trophic status) with a particular phytoplankton community. Methods based on spectral signatures use discriminatory traits
in the optical properties of certain phytoplankton to distinguish them from other
phytoplankton.
Both approaches have their limitations and their advantages. The spectralsignature methods have the potential to improve in accuracy and expand in range of
phytoplankton types that may be so-identified, as instruments with higher spectral
resolution become available. The non-linear nature of the algorithms, the plasticity
in the optical properties of phytoplankton functional types and the small signals on