Phytoplankton increased absorption

The best approach to the remote sensing of phytoplankton is undoubtedly that which makes use of the increased absorption, and therefore reduced radiance reflectance, at the blue end of the spectrum caused by the photosynthetic pigments. Clarke et al. (1970) measured from a low-flying aircraft, the spectral distribution of the emergent flux in different regions of the Northwest Atlantic Ocean with chlorophyll a concentrations varying from <0.1 to 3.0mgm-3. The general tendency (Fig. 7.8b) was that as the phytoplankton concentration increased, reflectance decreased in the blue (400-515 nm) and increased in the green (515600 nm). The increased reflectance in the green may be attributed to the

10 r

10 r

400 450 500 550 e00 e50 Wavelength (nm)

400 450 500 550 600 650 Wavelength (nm)

Fig. 7.8 Spectral distribution of reflectance above the Atlantic ocean off New England, USA (after Clarke et al., 1970). Measurements were made at zenith angle of 53°, using a polarizer to eliminate reflected skylight. (a) Dependence of reflectance on altitude: measurements were made at five altitudes over the same location (1° latitude east of Cape Cod). The large, and increasing, contribution of atmospheric scattering to measured reflectance at altitudes above 305 m is readily apparent. (b) Dependence of reflectance on phytoplankton content. Measurements were made at low altitude (305m). The phytoplankton chlorophyll level (mgchl am-3) is indicated above each curve. The three curves are for water on the northern edge of the Gulf Stream (0.3mgchl am-3), on the edge of the continental shelf (0.6mgchl am-3) and over the productive Georges Shoals (3.0mgchl am-3).

400 450 500 550 e00 e50 Wavelength (nm)

400 450 500 550 600 650 Wavelength (nm)

Fig. 7.8 Spectral distribution of reflectance above the Atlantic ocean off New England, USA (after Clarke et al., 1970). Measurements were made at zenith angle of 53°, using a polarizer to eliminate reflected skylight. (a) Dependence of reflectance on altitude: measurements were made at five altitudes over the same location (1° latitude east of Cape Cod). The large, and increasing, contribution of atmospheric scattering to measured reflectance at altitudes above 305 m is readily apparent. (b) Dependence of reflectance on phytoplankton content. Measurements were made at low altitude (305m). The phytoplankton chlorophyll level (mgchl am-3) is indicated above each curve. The three curves are for water on the northern edge of the Gulf Stream (0.3mgchl am-3), on the edge of the continental shelf (0.6mgchl am-3) and over the productive Georges Shoals (3.0mgchl am-3).

fact that phytoplankton, being refractive particles, increase scattering at all wavelengths, but in this spectral region absorb only weakly. Empirical methods of remotely estimating phytoplankton concentration generally take advantage of both the decreased reflectance in the blue and the increased reflectance in the green, by working in terms of the ratios or differences between reflectances in these two wavebands.

Morel and Prieur (1977) on the basis of their measurements of the spectral distribution of the upwelling and downwelling light flux in various oceanic waters found that the ratio of underwater irradiance reflectance (related in the manner described previously to the reflectance estimated by remote sensing above the water) at 440 nm to that at 560 nm decreases as the phytoplankton pigment content increases. If data for waters with high scattering coefficients due to suspended sediments are omitted then an approximately linear relation is obtained between the logarithm of R440/R560 and the logarithm of the chlorophyll plus phaeo-pigment (chlorophyll breakdown products) concentration over the range 0.02 to 20 mg total pigment m-3.941

Gordon et al. (1983) used the effect of phytoplankton on the spectral distribution of the upwelling flux to develop an algorithm by means of which Lw values obtained from the CZCS in the Nimbus-7 satellite could be used to calculate phytoplankton pigment concentration. They measured the upward nadir radiance within the water (which as we saw earlier is directly related to the emergent flux) at 440, 520 and 550 nm in marine waters with chlorophyll a (plus phaeopigment) concentrations ranging from 0.03 to 78 mgm-3. The data conformed reasonably well to relations of the type log10 C = logjQ A + B logjQ Rij (7.25)

where C is the chlorophyll a (plus phaeopigment) concentration in mg m-3, and Ri,j is the 440/550, 440/520 or 520/550 ratio of radiances. In the case of the 440/550 nm pair, for example, in Case 1 waters, log10 A is 0.053 and B is =1.71, i.e. as phytoplankton concentration increases so the ratio of upwelling radiance at 440 nm to that at 550 nm decreases. By substitution of a pair of values of upwelling radiance into the appropriate form of eqn 7.25 the chlorophyll (plus phaeopigment) concentration can be calculated. For chlorophyll concentrations below 0.6mgm-3 the radiances at 443 and 550 nm are most useful. At higher concentrations, however, Lw(443) becomes too small for accurate determination by the CZCS, and the radiances at 520 and 550 nm are used instead. Approximate agreement was found between the variation of chlorophyll concentration

Fig. 7.9 Phytoplankton distribution in the California Bight on 6 March 1979, derived from Coastal Zone Colour Scanner data (by permission, from Smith and Baker (1982), Marine Biology, 66, 269-79). The relation between chlorophyll a concentration (mgm-3) and the density of the image is indicated on the scale at the top of the figure.

Fig. 7.9 Phytoplankton distribution in the California Bight on 6 March 1979, derived from Coastal Zone Colour Scanner data (by permission, from Smith and Baker (1982), Marine Biology, 66, 269-79). The relation between chlorophyll a concentration (mgm-3) and the density of the image is indicated on the scale at the top of the figure.

determined by surface sampling along a ship course of several hundred kilometres in the Gulf of Mexico with that calculated for the same course from CZCS data:488 similar results were obtained in the Atlantic Bight and Sargasso Sea.487

During its operational life from 1978 to 1986, the CZCS accumulated a large amount of data, mapping phytoplankton distribution across the world's oceans, and the results of many such studies can be found in the literature. An early example may be found in Fig. 7.9, which shows the distribution of phytoplankton in the California Bight on a day in March 1979, derived by Smith and Baker (1982) from CZCS data.

Fig. 7.10 Distribution of phytoplankton in the ocean around Tasmania (Southern Ocean, south of the Australian continent), 27 November 1981, derived from CZCS data. Phytoplankton chlorophyll a concentration is colour coded from red (high) to blue (low). Courtesy of the NASA GSFC Earth Sciences Data and Information Services Center (GES DISC). See colour plate.

Fig. 7.10 Distribution of phytoplankton in the ocean around Tasmania (Southern Ocean, south of the Australian continent), 27 November 1981, derived from CZCS data. Phytoplankton chlorophyll a concentration is colour coded from red (high) to blue (low). Courtesy of the NASA GSFC Earth Sciences Data and Information Services Center (GES DISC). See colour plate.

Figure 7.10 shows the very complex distribution of phytoplankton around Tasmania, in the Southern Ocean, south of the Australian continent, on one November day in 1981.

To arrive at a suitable algorithm for SeaWiFS, O'Reilly et al. (1998) analysed a data set consisting of remote-sensing reflectance values (Rrs[1] = Lw[1]/Ed[0+, 1]) measured in the SeaWiFS wavebands at sea level, together with chlorophyll concentrations (range 0.02-32.8mgm-3) at 919 oceanic stations. Most of the observations were from Case 1 nonpolar waters, but included ~20 from more turbid coastal stations. The performance of 17 potentially suitable algorithms, using various combinations of wavebands, was evaluated. Two were identified as performing particularly well: ocean chlorophyll 2 algorithm (OC2), which uses two wavebands (490 and 555 nm), and ocean chlorophyll 4 algorithm (OC4), which uses four bands (443, 490, 510 and 555 nm). Both are empirical polynomials representing C, the phytoplankton chlorophyll concentration (mgm-3), as a function of R, the logarithm of the ratio of reflectance values in two wavebands

255 new oceanic stations were subsequently added to the original data set and used to arrive at new versions of the algorithms - OC2v4 and OC4v4.976 OC2 version 4 is a modified cubic polynomial, which take the form

where f (R) = 0.319 - 2.336R + 0.879R2 + 0.135R3 (7.28)

OC4 version 4 is a fourth-order polynomial, of the form, C = 10f (R), where f (R) = 0.366 - 3.067R + 1.930R2 + 0.649R3 - 1.532R4 (7.29)

The difference between the two algorithms is that whereas OC2 uses only the 490 and 555 nm wavebands, OC4 uses whichever of the three ratios -Rrs(443)/Rrs(555), Rrs(490)/Rrs(555) or Rrs(510)/Rrs(555), is the greatest for any given pixel. Algorithms such as OC4, where the reflectance ratio is selectable, are sometimes referred to as 'switching' algorithms. Numerous examples of the use of these algorithms for mapping phytoplankton distribution in the world's oceans from SeaWiFS data can now be found in the remote sensing and oceanographic literature, and also on NASA's SeaWiFS website (oceancolor.gsfc.nasa.gov). Figure 7.11 shows the average distribution of phytoplankton chlorophyll in the global ocean as revealed by SeaWiFS observations for the period September 1997 to July 1998.

The algorithms described above are purely empirical. There are others, however, which seek to make use of the approximate proportionality between the apparent optical property, Rrs, and the ratio of inherent optical properties, bb/a. These are known as semianalytic algorithms because as well as the analytic function, Rrs / bb/a, they make use of empirically derived relationships between reflectance ratios and optical

SeaWiFS Global Biosphere

Fig. 7.11 Average distribution of phytoplankton chlorophyll in the global ocean (and also terrestrial vegetation) over an 11-month period in 1997-98, derived from SeaWiFS data. Provided by the SeaWiFS Project, NASA/ Goddard Space Flight Center and ORBIMAGE. See colour plate.

properties such as the spectral distribution of backscattering, and the wavelength dependence of phytoplankton absorption etc. An example is that developed by Carder et al. (1999) for the MODIS instrument on the Terra and Aqua satellites. This uses Rrs at a number of MODIS wavebands to derive the absorption coefficient of phytoplankton chlorophyll at 675 nm, and the absorption coefficient due to the combined effects of CDOM and detritus at 400 nm. Chl a concentration is then calculated assuming an appropriate value for the specific absorption coefficient -0.0193 m2mg_1 for subtropical waters studied by Carder et al. Three versions of the algorithm were developed, parameterized for different bio-optical domains: (1) high light tropical/subtropical regions with minimal pigment packaging (see §9.2 for an explanation of this term), (2) upwelling and non-summer high-latitude regions where extensive pigment packaging is to be expected, (3) a transitional or global-average type. The appropriate domain is to be identified from space.

Another semianalytic algorithm in current use is the Garver-Siegel-Maritorena model (GSM01),440,866 which from measurements of normalized water-leaving radiance such as those provided by SeaWiFS simultaneously retrieves estimates for chlorophyll concentration, the absorption coefficient at 443 nm for dissolved and detrital colour, and the particulate backscattering coefficient at 443 nm. In tests of the model against a large body of archived data for non-polar Case 1 ocean waters, GSM01 was found to estimate chlorophyll with an accuracy similar to that of the standard SeaWiFS algorithm (OC4v4).

Algorithms such as those embodied in eqns 7.25 to 7.29, with appropriate values of the various coefficients, can be applied to all Case 1 oceanic waters, i.e. waters whose optical properties are almost entirely determined by phytoplankton and its associated debris and breakdown products. In Case 2 waters, whose optical properties are substantially affected by river-borne turbidity and dissolved yellow colour, or by resuspended sediments, relationships of these types still apply,171,491,917 but the values of the coefficients are different from those for Case 1 waters, and will also differ from one Case 2 water to another. Thus there is no universal Case 2 algorithm, so for these waters algorithms must be sought which are locality specific.

An example is one that Darecki et al. (2005) have developed for the southern Baltic Sea, a marine region of shallow depth and with much higher levels of yellow substance than the open ocean. On the basis of 700 simultaneous measurements of (in situ) remote sensing reflectance and chlorophyll at a large number of stations they arrived at a two-band algorithm

Chl a (mg m-3) = 10(-0 141-2 8652R) where R = log10[Rrs(490)/Rrs(555)], and a (switching) four-band algorithm

where, as in OC4v4 above, R = Rrs(443)/Rrs(555), Rrs(490)/Rrs(555) or Rrs(510)/Rrs(555), whichever is the greatest for any given pixel. For the Baltic these algorithms were found to provide more accurate estimates of chlorophyll than OC2 or OC4.

In the Mediterranean Sea, although the waters are mainly Case 1, use of the standard NASA algorithms, OC2v4 and OC4v4, with SeaWiFS leads to a significant overestimation of chlorophyll. Using a large data set of in situ measurements of chlorophyll and Rrs(4), Volpe et al. (2007) arrived at a modified version of OC4v4, referred to as 'MedOC4', implemented in the same way, i.e. switching between the bands, but using a form of eqn 7.29 with a different set of coefficients f (R) = 0.4424 - 3.686R + 1.076R2 + 1.684R3 - 1.437R4

where R is the maximum band ratio. In this region MedOC4 performs better than OC2v4 or OC4v4. They suggest that the poorer performance of the earlier algorithms in deriving chlorophyll concentration from Sea-WiFS radiances is due, not to a problem with the atmospheric correction term, but to peculiarities in the optical properties of the water column in the Mediterranean. The presence of Saharan desert dust has been suggested as a possible explanation.239

In the Irish Sea, a shallow shelf sea lying between Ireland and Great Britain, McKee and Cunningham (2006) found that in the majority of stations the water could be classified into one or other of two optical types. Using the ratio of particle backscattering (bbp) to the (non-water) absorption coefficient at 676 nm (an), the data fell into two clusters: Group A having bbp/an > 0.5 and Group B having bbp/an < 0.5. Particle inherent optical properties (IOPs) were dominated by mineral suspended sediments in Group A waters and by phytoplankton in Group B waters. McKee et al. (2007) used in situ measurements of optical properties and Rrs(1) to assess the performance of standard SeaWiFS algorithms at 102 stations in the Irish and Celtic Seas. They found that OC4v4 performed poorly for this region, with a strong tendency to overestimate chlorophyll concentration. The mean percentage error was 140%, but estimates for individual stations were up to a factor of six too high. Group A stations could be distinguished from Group B by higher values of normalized water-leaving radiance (nLw[lj) in the green and red wavebands. A threshold value of nLw(665 nm) = 0.1 mWcm-2nm-1 sr-1 was found to be a suitable flag with which to separate the data by optical water type. To develop improved algorithms for these waters, McKee et al. retained the form of OC4v4, but used their data to fine tune the coefficients. The modified versions of eqn 7.29 for the new regional chlorophyll algorithms are, for water type A, ISA-chl, in which f (R) = -0.2223 - 3.4118R - 3.6683R2 - 2.6599R3 - 0.8431R4 and for water type B, ISB-chl, in which f (R) = 0.1948 - 2.4851R - 2.4062R2 - 2.7332R3 - 1.5733R4 R is the maximum band ratio, selected as in OC4v4.

In Case 2 waters with significant amounts of CDOM, it is impossible with CZCS data to distinguish the lowered reflectance in the blue caused by this soluble yellow colour from that due to phytoplankton. The succeeding generation of ocean-viewing spaceborne radiometers, such as SeaWiFS and MERIS, are designed to carry out measurements at the extreme short-wave end of the visible spectrum, at —410 nm, where phytoplankton absorption has somewhat diminished (relative to that at 440 nm), but CDOM absorption is even higher. It was anticipated that the availability of this short-wavelength band would increase the feasibility of distinguishing phytoplankton from dissolved yellow colour,203,1175 but this has not yet proved easy to achieve. Gohin et al. (2002), using a data set derived from the English Channel and the continental shelf of the Bay of Biscay have created a lookup table that relates the triplet - OC4 maximum band ratio, Lw(412nm), and L^(555nm) - to chlorophyll concentration. Application of this table they believe can provide realistic chlorophyll maps for regions such as these.

The Landsat Thematic Mapper has a blue, as well as a green and a red, waveband. The limited amount of data so far available suggests that while it is not likely to be suitable for measuring the low levels of phytoplankton in oligotrophic ocean waters, it might provide useful data on the higher levels occurring in coastal waters, by means of a blue-green algo-rithm345,684,690 along the lines of eqn 7.25.

Eutrophic inland lakes very commonly develop blooms of cyanobac-teria, which present particular problems to water management agencies because of their frequently toxic nature. Early detection, and monitoring, of such blooms is therefore clearly desirable. A major photosynthetic pigment in these algae is the biliprotein, phycocyanin, which has an absorption peak at —620 nm. Cyanobacterial blooms therefore can give rise to a diminution in reflectance in this waveband. Algorithms for the detection of these blooms, which make use of the phycocyanin absorption band, have been developed for the MERIS spaceborne radiometer,1229 which has a band at 620 nm, and for hyperspectral radiometers used either in situ1180,1105 or from aircraft.298,664 Ruiz-Verdu et al. (2008) have compared the relative performance of these algorithms using data from lakes and reservoirs in Spain and the Netherlands.

Was this article helpful?

0 0

Post a comment