The role of absorption—Phytoplankton contain pigments that absorb light of different wavelengths. Chlorophyll a, the dominant pigment found in all oxygenic photosynthetic organisms, absorbs light broadly in the Soret, or blue (400 to 470 nm), region and narrowly in the red (660 nm) part of the electromagnetic spectrum (Fig. 1A). The green color of most plants, therefore, is due to reflected green light that is not absorbed by Chl a. However, other accessory pigments found in phytoplankton—such as chlorophylls b and c, phycobiliproteins, and carotenoids—allow organisms to harvest more of the incident blue and green light. Phytoplankton from different taxa generally have unique sets of accessory pigments that differentiate them from one another (Sathyendranath et al. 1987; Cullen et al. 1997). Even so, the overall absorption spectra from many different phytoplankton taxa are generally not distinguishable (Fig. 1B). Larger phytoplankton tend to have lower absorption per mass of Chl a than smaller phytoplankton because of packaging of the pigments within the cells (Bricaud et al. 1995; Ciotti et al. 2002). As noted in Fig. 1B, small cells (e.g., Prochlorophytes) tend to have higher specific absorption than the larger cells (e.g., Bacillariophyceae or diatoms). Furthermore, algal absorption properties for each taxon will be influenced by the light environment in which it is grown (see Kirk 1994). Dinoflagellates contain the carotenoid peridinin, which is considered to be reddish orange in color. The peridinin– chlorophyll–protein (PCP) complex broadens the blue absorption band from 470 to 555 nm (Fig. 1A). If peridinin were the cause of red tides, then the absorption spectrum for dinoflagellates containing peridinin should be significantly different from other phytoplankton taxa and should be shifted much further into the green region of the spectrum. However, absorption spectra from a variety of dinoflagellates are similar to other types of algae including diatoms, Bacillariophyceae (Fig. 1). Additionally, average absorption spectra from a variety of taxa show that most phytoplankton groups have accessory pigments that broaden the blue absorption band out to 550 nm (Stramski et al. 2001). Of the taxa considered, green algae (Chlorophyceae) absorb the least amount of green light, while red algae (Rhodophyceae) and Synechococcus absorb the most. When observed under magnification, individual phytoplankton cells most frequently appear green, yellowish green, or golden in color and are generally not the ruddy color of red tides. Although accessory pigment concentrations can alter the color of individual cells enough to make them distinctly red or even violet, this is not a necessary condition for formation of a red tide. The color of individual cells does not necessarily represent the color of the cells when suspended in a solution or when highly concentrated. For example, yellow food coloring can appear orange or red at high concentrations. As quantified here, red tides can be formed from cells that are not themselves red.
The absorption spectra in Fig. 1 represent average conditions. Spectral deviations in absorption will be observed in response to environmental conditions, especially irradiance. Absorption could potentially also deviate from the normal when the cells are growing rapidly, as is required for an intense algal bloom. However, significant changes in the shapes of absorption spectra have not been observed for most harmful algal species measured during stationary and exponential growth phases (Harding 1998; McLeroy-Etheridge and Roesler unpubl. data). No special absorption properties appear to be associated with phytoplankton known to form red tides, and many different algal taxa have sufficient accessory pigments to produce red coloration when concentrated.
The role of backscattering—The quantity and spectral quality of light reflected from the sea surface is not related just to the unabsorbed photons in the water column, but is also proportional to the amount of photons that are scattered in the backward direction out of the water column (i.e., backscattering). The reflectance of light at the sea surface (R) is governed by the ratio of backscattering (bb) to absorption (a) (Gordon and Morel 1983), such that
The amount and type of suspended materials (e.g., phytoplankton, sediment, minerals, etc.) will determine the magnitude and spectral shape of particulate backscattering (Stramski and Kiefer 1991). Backscattering properties of phytoplankton can vary with cell size, growth rate, and species composition. The phytoplankton taxa considered here have backscattering properties that vary spectrally (Stramski et al. 2001). As shown in Fig. 2A, small phytoplankton, like Synechococcus and Prochlorococcus, have enhanced backscattering of blue photons and the spectral shape is negatively sloping. Larger phytoplankton have backscattering spectra that are nearly constant with wavelength or positively sloping (higher in the red). Our simulations used chlorophyll-normalized particulate backscattering coefficients (bp), such that particulate backscattering increased linearly with increasing Chl a. This has been shown both theoretically (Morel 1988; Stramski and Kiefer 1991) and experimentally from a large database of in situ backscattering measurements with Chl a concentrations ranging over 100 mg m23 (Sullivan et al. 2005). In order for the sea surface to appear red, more red light must be backscattered out of the water than other colors of light. The definition of red light is generally considered to be light at wavelengths greater than 600 nm. Hyperspectral reflectance measurements made of the sea surface during red tides, however, typically show maximum spectral reflectance at wavelengths less than 600 nm (Fig. 2B; Carder and Steward 1985). Similar spectra are also derived from our modeling efforts (Fig. 2C,D). For many of the phytoplankton taxa considered, including the diatoms and dinoflagellates, our simulations showed a maximum spectral peak in both reflectance and water-leaving radiance at 570–580 nm, a region of the visible spectrum usually characterized as yellow light (Fig. 2C,D). The width and peak of the Rrs and Lw spectra, however, do vary by taxa. The spectral peak in reflectance for both Rhodophyceae and Synechococcus are shifted further into the red near 600 nm, while the spectrum for Chlorophyceae is less peaked and centered at 550 nm. A lesser peak in reflectance caused by solar-stimulated chlorophyll fluorescence is visible at 683 nm (Smith and Baker 1978), but this peak is generally lower in magnitude than the primary peak and not influential to the color perceived by the human eye (see following). Extremely dense algal blooms (.100 mg Chl a m23) can peak at or beyond 600 nm (Roesler and Boss 2003), but this is not required for the water to appear red. Color perception, however, involves more than just the peak spectral reflectance and requires consideration of the entire spectrum of light relative to receptors within the human eye. The role of human color vision—Understanding the perception of sea surface color requires a consideration of the physiological characteristics of the human eye. Human vision involves three different cone receptors and a fourth type of photoreceptor cell referred to as a rod. Rods are effective only in low light levels (referred to as night vision) and are not important for interpreting sea surface color. The three cone receptors in the eye, referred to as long-, middle-, and short-wavelength cones, are responsible for our color vision (Fig. 3). Each cone contains unique pigments that respond to a different range of visible wavelengths (Williamson and Cummins 1983). These cones are sometimes called red, green, and blue cones, although the red wavelength cone response is actually centered in the yellow region of the visible spectrum. The red cone pigment evolved 30–40 million years ago by a minor mutation in the green cone pigment that shifted the absorption peak about 30 nm to the red and is absent in individuals who are red/ green colorblind (Livingstone 2002). The color perceived by the human visual system depends on the total light or radiance incident upon each type of cone and the comparative response (i.e., contrast) between the three cone classes.
When phytoplankton become densely concentrated at the sea surface, blue/blue-green light is absorbed proportionally to the amount of Chl a (Fig. 4A), and the reflected light or water-leaving radiance incrementally shifts to longer wavelengths (Fig. 4B). Absorption by water rises steeply beyond 570 nm (Smith and Baker 1981; Pope and Fry 1997) and results in a maximum Lw peak between 570 and 600 nm. In addition, increasing phytoplankton concentrations typically result in increased Lw from 550 to 700 nm due to enhanced backscattering (Fig. 4B). When normalized to the peak radiance (Fig. 4C), changes in the shape of the radiance spectra are evident with increasing Chl a. The radiance distribution appears relatively similar in peak width as Chl a increases from 10 to 50 mg m23, but the entire spectrum swings like a bell toward the red as more blue light is absorbed and more red light is backscattered.
Perceived color was modeled from water-leaving radiance spectra using an empirically derived color coordinate system (CIE 1991). When the color is a red hue, the total light absorbed by the eye’s red cones is greater than that absorbed by the green and blue cones. Brown coloration can be considered a dull red and occurs when all three cones are stimulated, but with the greatest activation of the red cones. We modeled water-leaving radiance as a function of average absorption and backscattering properties for increasing concentrations of different phytoplankton taxa concentrated at the sea surface. As phytoplankton concentrations increased, more blue/green light was absorbed and more red light was backscattered out of the water. Perceived color progressed from blue, to green, and finally to red or red/brown for the majority of phytoplankton taxa (Fig. 5A). These results are also represented on a chromaticity diagram (Kelly 1943) for three representative taxa: Chlorophyceae, Bacillariophyceae, and Rhodophyceae (Fig. 5B). The x, y coordinates represent the fraction of the spectra intercepted by the red and green cones, respectively, and the residual from unity is the fraction intercepted by the blue cones (z). For example, light at 520 nm is represented by 10% red (x), 80% green (y), and by derivation 10% blue (z). Of the phytoplankton considered, only Prochlorococcus and certain species of Chlorophyceae (Cunaliella bioculata) did not produce red or brown coloration at high concentrations. At a modeled threshold concentration near 15 mg Chl a m23 for most phytoplankton taxa, the eye’s red cones (centered at 564 nm) are activated more than the green cones (centered at 534 nm) and the water appears reddish brown. This shift in color is not due to any special absorption or backscattering properties of the algae but is due to the human visual system. The peak in the waterleaving radiance spectra from dense algal blooms coincides with a critical visual hinge point where the green and red cones are nearly equally stimulated (570 nm). Subtle spectral changes in Lw about this critical hinge point produce large shifts in the observed color from green to red. Sensitivity analysis—To address uncertainties in our modeling assumptions, we conducted a sensitivity analysis varying the particulate backscattering and absorption properties of the water, as well as the incident light field. These simulations used the total upward radiance from the sea surface (Lu), which includes water-leaving radiance (Lw) plus the downwelling irradiance reflected off the sea surface. The simulations for scenario 1 used a fixed absorption coefficient (Dinophyceae) at 30 mg m23 (Fig. 6A) and variable particulate backscattering ratios ranging from 0.6% to 2% (Fig. 6B). The resulting Lu was characterized by a distinct peak at 570 nm regardless of the backscattering spectrum used in the model (Fig. 6C). This is because peak reflectance is largely determined by the combined absorption properties of pure water and phytoplankton. In addition, Lu spectra increase monotonically with increasing backscattering (Fig. 6C). When normalized to peak Lu, however, all cases had nearly identical spectral shapes resulting in a similar greenish-brown color (Fig. 6D). Changes in the magnitude of backscattering, without corresponding changes to the spectral shape of backscattering, will alter the perceived brightness or luminance of the water, but these changes have little impact on the color.
In scenario 2, three different backscattering spectra were used to illustrate how subtle variations in the radiance spectra about the 570 nm hinge point produce large changes in the perceived color of the sea surface (Fig. 6E–H). As shown in Fig. 6F, backscattering spectral shapes varied from (1) negatively sloped backscattering (greater in the blue than the red), (2) spectrally flat backscattering, and (3) positively sloped backscattering (greater in the red than the blue). The positively sloped simulation had the highest backscattering from 500 to 700 nm and resulted in the largest Lu peak (Fig. 6G). The three cases also produced significant differences in the spectral shape of Lu that resulted in widely differing colors. Even though the peak reflectance was the same for all cases, negatively sloped backscattering enhanced the blue reflectance and resulted in a greenish hue. In contrast, positively sloped backscattering enhanced red reflectance and resulted in a reddish-brown hue.
These results were not duplicated, however, in scenario 3 where green algae (Chlorophyceae) were the predominant phytoplankton (Fig. 6I–L). Lacking sufficient green-absorbing accessory pigments (Fig. 6I), these phytoplankton produced a broad and rounder radiance peak spanning the entire 500–600 nm region. Distinct spectral shapes were observed with the different backscattering shapes, but the resulting sea surface color was a similar shade of green for all simulations (Fig. 6L).
Scenario 4 explored variations in the incoming downwelling irradiance field on perceived water color (Fig. 6M– P). The simulations covered a range of solar zenith angles (SZA) from nearly overhead (10u) to nearly at the horizon (80u). The highest Lu occurred when the sun was nearly overhead (SZA 10u) because of the greater amount of light reaching the sea surface (Fig. 6O; note axis is four times higher than the other plots). When normalized to peak radiance, the case with 10u SZA showed elevated radiance across the whole spectrum (both the blue and red wavelengths). This is largely due to reflected skylight, which represents photons that are scattered by the sea surface without penetrating the water column. The resulting spectral whitening produced more luminance but did not significantly change the perceived color of the water (Fig. 6P).
Thus, the production of reddish water requires phytoplankton with sufficient accessory pigments to shift the radiance spectrum toward the critical visual hinge point (570 nm). Simply amplifying the amount of backscattering, without changing the spectral backscattering shape or simultaneously increasing absorption, does not change the spectral shape of Lu or the resulting color of the water. Phytoplankton with higher backscattering in the red produce brown/red–colored water at lower overall phytoplankton concentrations than those that backscatter more blue photons. A low sun angle can produce a more luminous spectrum because of reflected photons but is not responsible for brown/red–colored water. In general, algal absorption properties determine whether dense concentrations of a particular phytoplankton group can potentially cause red-colored water, but taxon-specific backscattering properties can influence the amount of phytoplankton necessary to produce red coloration.
Fluorescence and the red edge—Measured reflectance spectra from a variety of dinoflagellate red tides show a prominent secondary peak in the far-red region of the spectrum (Fig. 2A). At first glance, these peaks appear to be associated with chlorophyll fluorescence, an important pathway for the dissipation of light energy (Smith and Baker 1978). However, the peak wavelengths are not always centered at the chlorophyll fluorescence peak of 683 nm but can be shifted further into the near infrared (NIR) (.700 nm). For example, the reflectance spectrum from the dinoflagellate L. polyedra (Fig. 2A) has a peak at 710 nm. This is not from fluorescence alone, but appears to be caused by what the terrestrial remote sensing community calls the red edge.
Terrestrial vegetation and submerged macrophytes (sea grasses and seaweeds) exhibit strong reflectance in the NIR portion of the spectrum (700–1,600 nm, Hall 1994). This NIR reflectance is commonly attributed to scattering from cell and leaf structures (cell walls and membranes, organelles, air spaces, etc.) and forms the basis of the terrestrial biomass parameter normalized difference vegetation index (NDVI, Hall 1994). However, plant cell structures are capable of scattering visible, as well as NIR, radiation. In fact, reflectance from unpigmented white leaves lacks a red edge because light is reflected similarly across both the visible and NIR portions of the spectrum (Gitelson and Merzlyak 1994; Zimmerman unpubl. data). Thus, photosynthetic pigments produce the red edge signature by absorbing visible, but not NIR, radiation. NIR reflectance is generally ignored for dilute suspensions of microscopic phytoplankton because water strongly absorbs these photons. For most natural populations of phytoplankton, the peak reflectance observed in the far-red region can be assigned to fluorescence. However, pigmented microalgae also reflect NIR light with much higher efficiency than visible light (Fig. 7). For dense suspensions of algal cells at the sea surface (i.e., red tides), the infrared reflectance signal can be strong enough that it is not fully attenuated by the water, producing peaks in the reflectance spectra that are red-shifted relative to those produced by chlorophyll fluorescence (Vasilkov and Kopelevich 1982; Gitelson 1992).
While it is an important feature of red tide spectra, enhanced reflectance due to fluorescence and the red edge has little impact on the perceived color of the sea surface. The response of the human red cone, which is tuned to 564 nm, is very low at wavelengths greater than 650 nm (Fig. 3). Light at longer red wavelengths can be perceived when it originates from a concentrated source (i.e., a laser), but the low levels produced by chlorophyll fluorescence or red edge reflectance are simply not perceptible under most natural light conditions. Thus, the perceived sea surface color modeled for different phytoplankton (Fig. 5) did not change when reflectance greater than 650 nm was excluded from the analysis and is largely independent of chlorophyll fluorescence and the red edge. Enhanced radiance in the NIR is not important for color vision but has significant implications for understanding algal backscattering, for understanding atmospheric correction of remotely sensed imagery, and for developing new methods for remote sensing of red tides.
Red and black tides caused by CDOM and sediments— Our modeling indicates that optical properties of many different phytoplankton taxa can produce red coloration if cells are sufficiently concentrated at the sea surface. A red coloration can also result from nearly any constituent that is a blue/green absorber provided enough light is backscattered out of the water column. CDOM is a common component of coastal water that absorbs light in the blue region of the spectrum (Fig. 8A; Kirk 1976). The CDOM absorption spectrum decreases exponentially from blue to red. Suspended sediments also absorb light predominantly in the blue region of the spectrum (Fig. 8A), but the spectral shape depends on the suite of minerals present (Morel and Prieur 1977; Ahn 1990; Babin and Stramski 2004). Unlike CDOM, minerals have a large refractive index and high backscattering coefficients (Fig. 8B). Minerals suspended in water can produce reflectance spectra that are several orders of magnitude greater than water with CDOM, or even phytoplankton, alone. As originally noted by Morel and Prieur (1977) in their early delineation of Case 2 waters, waters dominated by sediments and CDOM have flatter and broader reflectance spectra from 600 to 650 nm (Fig. 8C,D) compared to water with dense phytoplankton concentrations. This is because the absorption coefficient for these constituents monotonically decreases with increasing wavelength, unlike phytoplankton pigments, which absorb both blue and red light.
Here, we modeled sea surface color as a function of increasing CDOM and suspended sediments or minerals (Fig. 9). Because CDOM does not scatter light, waters that are highly concentrated in CDOM alone are so dark that they appear black in color (Acker and Kempler 2004; Fig. 9A). An example of a black river that has high CDOM concentrations can be found in the Rio Negro, Brazil (Fig. 9C). Black water can also occur during intense phytoplankton blooms that have low backscattering (Carder and Steward 1985; Hu et al. 2004). High concentrations of CDOM will produce a red/brown coloration of the sea surface, however, when suspended materials that can scatter light out of the water column are also present in the water. Optically shallow water systems with a highly reflective bottom can also appear red/brown in the presence of high CDOM concentrations (Fig. 9D). Such high CDOM concentrations are seldom observed in the open ocean but are most likely to be found in estuarine waters where the terrestrial sources of CDOM are greatest (Kirk 1976; Twardowski et al. 2004).
Waters with high mineral content (.10–15 g m23) also produced red or brown colored water (Fig. 9B). From the limited set of minerals considered here, brown earth and red clay have the greatest absorption coefficient per mass and created red-colored water. Yellow clay and calcareous sand absorbed less of the blue/green light and produced a brownish hue when concentrated. Mineral particles with significant amounts of iron absorb more of the blue-green light (Babin and Stramski 2004) and would create even redder coloration at lower concentrations.
Remote detection of red tides—The radiance spectra modeled for water dominated by surface layers of different phytoplankton, CDOM, and minerals have unique spectral signatures that may allow us to discriminate these constituents remotely from satellites or aircraft. As shown in Fig. 2C, high concentrations of cryptophytes can result in a double peaked reflectance spectrum. Dense concentrations of red algae produce an asymmetric spectral peak that is shifted toward the red region of the spectrum. Green algae are distinguished by a broad spectral maximum centered at 540 nm. Unfortunately, the two most common red tide–forming phytoplankton in the coastal ocean, dinoflagellates and diatoms (Sournia 1995), have similar absorption characteristics and will be the most difficult to differentiate based solely on optical signatures (Schofield et al. 1999; Roesler et al. 2003). Waters dominated by CDOM and sediments produce broader reflectance spectra compared to phytoplankton-dominated waters.
Current space-borne ocean color sensors can view dense algal blooms (Kahru et al. 2004) and sediment plumes but have limited spectral capabilities. These sensors provide no information in the region of the spectrum where red tides peak and where most of the taxon-specific differences occur in water-leaving radiance (570–610 nm). Several airborne hyperspectral sensors can provide the spectral information needed to detect taxon-specific differences from ocean color (Chang et al. 2004; Ryan et al. 2005), and their use will be critical for monitoring red tide formation and remotely detecting phytoplankton composition in many coastal regions. Enhanced reflectance in the NIR due to fluorescence and the red edge may also be critical for remotely detecting surface blooms of algae.
Understanding the nuances of the color produced by increasing concentrations of different groups of algae or other water column constituents across the whole visible and NIR spectrum can lead to better remote monitoring and forecasting of dense algal blooms that may have harmful consequences. By combining hyperspectral imagery with a variety of monitoring platforms and parameters, it may be possible to develop probabilistic models of toxic algal blooms, or ecological algorithms (Schofield et al. 1999; Stumpf et al. 2003; Etheridge and Roesler 2004; Babin et al. 2005; Kudela et al. 2005). Future research may allow us to predict the probability of harmful algal bloom development, monitor the extent and longevity of the bloom, and forecast the coastal region to be affected by the bloom. Our research shows that the visual identification of red water is strongly affected by human physiology, in addition to the absorbing and scattering components of natural waters. Moreover, the presence of red water does not necessarily indicate a dense or harmful algal bloom, but can result from various combinations of phytoplankton, CDOM, and minerals.
Answers to all your Biology Questions
