352
S. Sathyendranath and T. Platt
water-leaving radiances is often significantly lower than what could be expected
from the sensor specifications which determine the precision and accuracy at the top
of the atmosphere. It is therefore essential to continue to improve atmospheric correction procedures, if we are to derive full benefits from the advanced ocean-colour
sensors now available.
But have we already reached the peak in this area of research? Perhaps not.
Scientists are still finding new ways to improve atmospheric correction: multispectral approaches, use of longer wavebands in the infrared, Fraunhofer line filling
(Vountas et al., 2007), and the use of UV wavebands in addition to those in the
infrared. All these are promising routes, as are inversion techniques that treat the
ocean and atmosphere as a coupled system. But is it likely that when the limit of
atmospheric correction is reached, it would set the limit to in-water properties or
quantities that might be retrieved from ocean-colour data?
Not necessarily. The spectral characteristics of atmospheric constituents are quite
distinct from those of many oceanic constituents, such that errors resulting from
atmospheric correction are likely to have a spectral form different from those of
in-water constituents. There is, therefore, scope for development of algorithms that
exploit these differences: the need is for algorithms that would be insensitive to
potential systematic errors in the water-leaving radiances arising from atmospheric
correction. When atmospheric correction errors are relatively high, one anticipates
that algorithms for retrieval of concentrations of substances such as phytoplankton
pigments that have distinctive peaks and troughs in their inherent optical properties would work better than those designed to quantify substances such as yellow
substances or suspended material whose inherent optical properties are monotonic
functions of wavelength in the visible domain. The fluorescence line height might
be considered one such algorithm that exploits a distinctive peak associated with
chlorophyll fluorescence. In principle, the fluorescence line height would be insensitive to systematic offsets in the background signal, due, for example, to errors in
atmospheric correction, or to the presence of non-pigmented scattering particles in
the water. Thus, when the limits of atmospheric correction are known, it would be
essential to understand the nature of the residual errors and incorporate them into
the in-water algorithms.
The algorithms for Case-2 waters have been improving over the years (IOCCG,
2000), facilitated by the higher spectral resolution in the visible, improvements
in atmospheric-correction procedures, and application of novel mathematical and
statistical tools such as neural networks (Doerffer and Schiller, 1998). The need
to address Case-2 problems provided impetus for development of in-water algorithms that are based on inversion of theoretical models of ocean colour that rely
on our understanding of the inherent optical properties of oceanic constituents and
of radiative transfer in the ocean (Sathyendranath et al., 1989; IOCCG, 2006),
whereas, initially, models developed for Case-1 waters were mostly empirical in
nature. These theoretical developments helped extend the line of products from
ocean-colour data to include concentrations of suspended sediments and yellow substances. Improvements in Case-2 algorithms have also helped extend applications of
ocean-colour data to many issues related to coastal zone management.
S. Sathyendranath and T. Platt
water-leaving radiances is often significantly lower than what could be expected
from the sensor specifications which determine the precision and accuracy at the top
of the atmosphere. It is therefore essential to continue to improve atmospheric correction procedures, if we are to derive full benefits from the advanced ocean-colour
sensors now available.
But have we already reached the peak in this area of research? Perhaps not.
Scientists are still finding new ways to improve atmospheric correction: multispectral approaches, use of longer wavebands in the infrared, Fraunhofer line filling
(Vountas et al., 2007), and the use of UV wavebands in addition to those in the
infrared. All these are promising routes, as are inversion techniques that treat the
ocean and atmosphere as a coupled system. But is it likely that when the limit of
atmospheric correction is reached, it would set the limit to in-water properties or
quantities that might be retrieved from ocean-colour data?
Not necessarily. The spectral characteristics of atmospheric constituents are quite
distinct from those of many oceanic constituents, such that errors resulting from
atmospheric correction are likely to have a spectral form different from those of
in-water constituents. There is, therefore, scope for development of algorithms that
exploit these differences: the need is for algorithms that would be insensitive to
potential systematic errors in the water-leaving radiances arising from atmospheric
correction. When atmospheric correction errors are relatively high, one anticipates
that algorithms for retrieval of concentrations of substances such as phytoplankton
pigments that have distinctive peaks and troughs in their inherent optical properties would work better than those designed to quantify substances such as yellow
substances or suspended material whose inherent optical properties are monotonic
functions of wavelength in the visible domain. The fluorescence line height might
be considered one such algorithm that exploits a distinctive peak associated with
chlorophyll fluorescence. In principle, the fluorescence line height would be insensitive to systematic offsets in the background signal, due, for example, to errors in
atmospheric correction, or to the presence of non-pigmented scattering particles in
the water. Thus, when the limits of atmospheric correction are known, it would be
essential to understand the nature of the residual errors and incorporate them into
the in-water algorithms.
The algorithms for Case-2 waters have been improving over the years (IOCCG,
2000), facilitated by the higher spectral resolution in the visible, improvements
in atmospheric-correction procedures, and application of novel mathematical and
statistical tools such as neural networks (Doerffer and Schiller, 1998). The need
to address Case-2 problems provided impetus for development of in-water algorithms that are based on inversion of theoretical models of ocean colour that rely
on our understanding of the inherent optical properties of oceanic constituents and
of radiative transfer in the ocean (Sathyendranath et al., 1989; IOCCG, 2006),
whereas, initially, models developed for Case-1 waters were mostly empirical in
nature. These theoretical developments helped extend the line of products from
ocean-colour data to include concentrations of suspended sediments and yellow substances. Improvements in Case-2 algorithms have also helped extend applications of
ocean-colour data to many issues related to coastal zone management.