10 Ocean Colour Remote Sensing of Harmful Algal Blooms in the Benguela System
193
at stations around − 32.08 N and 18.26 E, with additional sampling where bloom
patches were evident. Detailed laboratory and processing methodology can be found
in Bernard et al. 2009.
Two algorithms are used in the study. The Maximum Peak Height (MPH) algorithm (Matthews et al. 2012) is an empirical top-of-atmosphere algorithm designed
for MERIS in high biomass waters, returning only Chl a concentrations. The Equivalent Algal Population (EAP) algorithm (variant of Bernard 2005) is a semi-analytical
reflectance algorithm, based upon inversion of IOPs from modelled algal populations,
which gives a more comprehensive suite of returns.
The EAP algorithm can be summarised as follows. Inputs are MERIS atmospherically corrected multi-spectral normalised water leaving reflectance, on a pixel-bypixel basis. The model uses five solvable unknowns: chlorophyll a concentration (Chl
a, mg m
−3 ), algal effective diameter (D eff , μm), the relative concentration of two representative algal groups (diatoms/dinoflagellates and nanoflagellates/chlorophytes),
combined gelbstoff and detrital absorption (e.g. a gd (400), m
−1 ), and small particle
backscattering (e.g. b bs (550), m
−1 ). Gelbstoff/detrital absorption and small particle
backscattering employ constant spectral shapes and variable magnitude (Bernard
2005).
The EAP algorithm uses an unconstrained non-linear minimisation (Nelder and
Mead 1965), with constant initial values except for chlorophyll, for which an initial
estimate is provided via an empirical, band-ratio switching algorithm on a per pixel
basis. This algorithm extends the maximum band ratio approach used in the standard MERIS algal 1 product, by adding an additional ratio between 665 and 709 nm
bands. The EAP algorithm and all subroutines are coded in Matlab R14. As the underlying model specifically does not account for sun-induced natural fluorescence,
the convergence weighting for the Nelder-Mead solution is set to negligible values
between 665 and 715 nm—the spectral region affected by natural algal fluorescence.
The algorithm thus does not seek to match spectral reflectance values at these fluorescence wavelengths, and in effect offers a means of discriminating fluorescence
effects, as have earlier models of a similar nature (Roesler and Boss 2003). This
allows the derivation of algal fluorescence quantum yield as an additional algorithm
product, utilising integrated fluorescence (calculated from fitting a Gaussian distribution to modelled R rs subtracted from measured R rs at fluorescence wavebands),
algal absorption as returned by the algorithm and incident scalar irradiance calculated
independently (Gregg and Carder 1990).
10.4 Case Studies: Application to Bloom Types
10.4.1 Transitions Between Dominant Species Detected Through
Changes in Effective Diameter Using the EAP Algorithm
The EAP algorithm approach allows for determination of a wide range of chlorophyll
concentrations and an “effective diameter” relating to the particle size distribution of
193
at stations around − 32.08 N and 18.26 E, with additional sampling where bloom
patches were evident. Detailed laboratory and processing methodology can be found
in Bernard et al. 2009.
Two algorithms are used in the study. The Maximum Peak Height (MPH) algorithm (Matthews et al. 2012) is an empirical top-of-atmosphere algorithm designed
for MERIS in high biomass waters, returning only Chl a concentrations. The Equivalent Algal Population (EAP) algorithm (variant of Bernard 2005) is a semi-analytical
reflectance algorithm, based upon inversion of IOPs from modelled algal populations,
which gives a more comprehensive suite of returns.
The EAP algorithm can be summarised as follows. Inputs are MERIS atmospherically corrected multi-spectral normalised water leaving reflectance, on a pixel-bypixel basis. The model uses five solvable unknowns: chlorophyll a concentration (Chl
a, mg m
−3 ), algal effective diameter (D eff , μm), the relative concentration of two representative algal groups (diatoms/dinoflagellates and nanoflagellates/chlorophytes),
combined gelbstoff and detrital absorption (e.g. a gd (400), m
−1 ), and small particle
backscattering (e.g. b bs (550), m
−1 ). Gelbstoff/detrital absorption and small particle
backscattering employ constant spectral shapes and variable magnitude (Bernard
2005).
The EAP algorithm uses an unconstrained non-linear minimisation (Nelder and
Mead 1965), with constant initial values except for chlorophyll, for which an initial
estimate is provided via an empirical, band-ratio switching algorithm on a per pixel
basis. This algorithm extends the maximum band ratio approach used in the standard MERIS algal 1 product, by adding an additional ratio between 665 and 709 nm
bands. The EAP algorithm and all subroutines are coded in Matlab R14. As the underlying model specifically does not account for sun-induced natural fluorescence,
the convergence weighting for the Nelder-Mead solution is set to negligible values
between 665 and 715 nm—the spectral region affected by natural algal fluorescence.
The algorithm thus does not seek to match spectral reflectance values at these fluorescence wavelengths, and in effect offers a means of discriminating fluorescence
effects, as have earlier models of a similar nature (Roesler and Boss 2003). This
allows the derivation of algal fluorescence quantum yield as an additional algorithm
product, utilising integrated fluorescence (calculated from fitting a Gaussian distribution to modelled R rs subtracted from measured R rs at fluorescence wavebands),
algal absorption as returned by the algorithm and incident scalar irradiance calculated
independently (Gregg and Carder 1990).
10.4 Case Studies: Application to Bloom Types
10.4.1 Transitions Between Dominant Species Detected Through
Changes in Effective Diameter Using the EAP Algorithm
The EAP algorithm approach allows for determination of a wide range of chlorophyll
concentrations and an “effective diameter” relating to the particle size distribution of