140
E. V. Djagoua et al.
The similarity of the Sassandra river mouth with other test cases, in which the
spectral characteristics of water constituents have been highlighted by comparable
studies (see e.g. Miller et al. 2005), suggested to select 443, 555 and 670 nm as the
wavelengths of choice for our water constituents classification. The basic reasoning
underlying this choice is that the chlorophyll-like pigments of phytoplankton reflect
less at the shorter wavelengths, i.e. 443 nm (and 490 nm), and more at 555 nm.
Moreover, at 443 nm (and 490 nm), the shape of the reflection curve is not constant,
as it depends on the CHL concentration (Gower et al. 1988). Absorption by detritus
is strong at shorter wavelengths (i.e. 412 nm) and weak at longer wavelengths (i.e.
670 nm; Froidefond and Doxaran 2004) or, in other words, the reflectance associated
to DET is low at the shorter wavelengths and high at the longer wavelengths.
An increase in CDOM concentration leads to a lower reflectance at the shorter
wavelengths, from UV to blue-green (Bricaud et al. 1981). CDOM absorbs visible light, especially below 500 nm, and its absorbance increases exponentially
with decreasing wavelength. Nonetheless, Strömbeck and Pierson (2001) report that
absorbance of red light can be significant at high CDOM concentrations. So, as suggested by the literature above, the wavelengths 443, 555 and 670 nm appear to be
those that best allow the discrimination of the constituents of a given water body,
and they were selected for the present study.
7.2.3 How did we Compute the Percent Contribution to the Total
Reflectance of each major Constituent from R rs (443, 555,
670 nm)?
The RS reflectances at each of the selected wavelengths (i.e. 443, 555 and 670 nm)
were averaged pixel by pixel (i.e. points of constant latitude and longitude) over the
5 years considered (i.e. for the whole 2003–2007 period, as well as for each month
in the same multi-annual period), according to the equation:
R rs MEAN (λ, i, j ) =
1
n
n
t=1
R rs (λ, i, j , t)
(7.1)
where:
• R rs MEAN (λ, i, j ) is the mean (annual or monthly) RS reflectance, at the wavelength
λ and at the pixel location i, j;
• R rs (λ, i, j , t) are the RS reflectance values, at the wavelength λ and at the pixel
location i, j , which are being summed over the time t = 1, 2, 3 . . . n (i.e. all
images in the 2003–2007 data set, or all images for each climatological month in
the 2003–2007 data set);
• n is the number of valid pixels used in the sum.
(see also Fig. 7.4, for a graphical description of these variables).
E. V. Djagoua et al.
The similarity of the Sassandra river mouth with other test cases, in which the
spectral characteristics of water constituents have been highlighted by comparable
studies (see e.g. Miller et al. 2005), suggested to select 443, 555 and 670 nm as the
wavelengths of choice for our water constituents classification. The basic reasoning
underlying this choice is that the chlorophyll-like pigments of phytoplankton reflect
less at the shorter wavelengths, i.e. 443 nm (and 490 nm), and more at 555 nm.
Moreover, at 443 nm (and 490 nm), the shape of the reflection curve is not constant,
as it depends on the CHL concentration (Gower et al. 1988). Absorption by detritus
is strong at shorter wavelengths (i.e. 412 nm) and weak at longer wavelengths (i.e.
670 nm; Froidefond and Doxaran 2004) or, in other words, the reflectance associated
to DET is low at the shorter wavelengths and high at the longer wavelengths.
An increase in CDOM concentration leads to a lower reflectance at the shorter
wavelengths, from UV to blue-green (Bricaud et al. 1981). CDOM absorbs visible light, especially below 500 nm, and its absorbance increases exponentially
with decreasing wavelength. Nonetheless, Strömbeck and Pierson (2001) report that
absorbance of red light can be significant at high CDOM concentrations. So, as suggested by the literature above, the wavelengths 443, 555 and 670 nm appear to be
those that best allow the discrimination of the constituents of a given water body,
and they were selected for the present study.
7.2.3 How did we Compute the Percent Contribution to the Total
Reflectance of each major Constituent from R rs (443, 555,
670 nm)?
The RS reflectances at each of the selected wavelengths (i.e. 443, 555 and 670 nm)
were averaged pixel by pixel (i.e. points of constant latitude and longitude) over the
5 years considered (i.e. for the whole 2003–2007 period, as well as for each month
in the same multi-annual period), according to the equation:
R rs MEAN (λ, i, j ) =
1
n
n
t=1
R rs (λ, i, j , t)
(7.1)
where:
• R rs MEAN (λ, i, j ) is the mean (annual or monthly) RS reflectance, at the wavelength
λ and at the pixel location i, j;
• R rs (λ, i, j , t) are the RS reflectance values, at the wavelength λ and at the pixel
location i, j , which are being summed over the time t = 1, 2, 3 . . . n (i.e. all
images in the 2003–2007 data set, or all images for each climatological month in
the 2003–2007 data set);
• n is the number of valid pixels used in the sum.
(see also Fig. 7.4, for a graphical description of these variables).