Phytoplankton absorption and chlorophyll—Absorption coefficients for particulate material were determined using a Cary 50 dual-beam spectrophotometer. Water samples (25 mL) were filtered onto Whatman GF/F filters. Phytoplankton (aph) spectra were determined by subtraction of detrital absorption (Tassan and Ferrari 1995) from particulate absorption and correction for multiple scattering caused by the filter material (Cleveland and Weidemann 1993). Published absorption spectra were also used (Stramski et al. 2001). Phytoplankton absorption coefficients were normalized to the corresponding chlorophyll a (Chl a) concentration. Chl a was determined fluorometrically by filtration onto Whatman GF/F filters and extraction in acetone at 220uC for 24–48 h (Smith et al. 1981) using pure Chl a (Sigma) as a calibration reference. Hyperspectral reflectance measurements—Two different field methods were employed to measure remote sensing reflectance, Rrs, defined as water-leaving radiance, Lw, normalized to incident irradiance. In-water measurements were made with hyperspectral radiometers, Hyperspectral Tethered Spectroradiometer Buoy (HTSRB, Satantic, Inc.) and HyperPro II (Satlantic Inc.). The radiative transfer model, Hydrolight, was parameterized with coincident ac-9 measurements (WET Labs) and relevant ancillary data (e.g., latitude, longitude, day of the year, time) to develop site-specific spectral coefficients from the ratio of modeled water-leaving radiance and radiance modeled at the depth of the upwelling radiance sensors of the HTSRB and HyperPro radiometers (0.6 and 0.2 m, respectively). These coefficients were applied to the original hyperspectral upwelling radiance measurements to correct for absorption and scattering processes in the upper layer of water and extrapolate values across the air–sea interface. Abovesurface radiances were measured with a Field Spec Pro VNIR-NIR1 portable spectrometer system (Analytical Spectral Devices) along Monterey Bay and the California coast in conjunction with Center for Integrative Coastal Observation, Research, and Education (CICORE), California State University. Measurements were made with an 8u foreoptic focused at a 40–45u angle sequentially on a gray plaque, sea surface, and diffuse sky. Residual reflected sky radiance was removed from the calculated Rrs assuming that the mean reflectance from 750–800 nm was zero (Mobley 1999).
Reflectance of dense cultures of phytoplankton taxa was also measured in the laboratory using the Field Spec Pro VNIR-NIR1 portable spectrometer system (Analytical Spectral Devices) and integrating sphere attachment calibrated for baseline reflectance using a Spectralon plaque. Dense, log-phase cultures of the microalgae Thalassiosira sp., Coccolithophora sp., Porphyridium sp., and Amphidinium sp. were obtained from Wards Natural Science, Inc. Cells were captured onto Whatman GF/F filters using a gentle vacuum and immediately placed into the reflectance port of the integrating sphere. These measurements are relative reflectance spectra and have not been corrected for filter effects (Balch and Kilpatrick 1992).
Modeling water-leaving radiance—The radiative transfer model Hydrolight was employed to estimate the radiance leaving the water column composed of varying amounts of phytoplankton (Mobley 1994). Water-leaving radiance, Lw, was modeled assuming dense accumulations of algae at the sea surface (2–50 mg Chl a m23). Optical variability due to depth-dependent layers of algae was not considered here. Inputs to the model included: taxon-specific chlorophyll-normalized absorption (see Methods), chlorophyll-specific scattering coefficients (Stramski et al. 2001), and a semiempirical sky model with a solar zenith angle of 55u, wind speed of 5 m s21, and an infinitely deep water column that included Raman scattering and chlorophyll fluorescence (Mobley 1994). For the average case, a Fournier- Forand particle phase function (Fournier and Forand 1994) was used with a particulate backscattering ratio of 0.012 (Ulloa et al. 1994). This is within the range of backscattering ratios found in oceanic waters throughout the world (Twardowski et al. 2001; Sullivan et al. 2005) and was found to be an average value for achieving optical closure with Hydrolight in coastal waters sampled in conjunction with the CICORE program (0.7 to 50 mg Chl a m23). In the sensitivity analysis, the particulate backscattering ratio varied from 0.006 to 0.02. For the phytoplankton simulations, CDOM was set to be proportional to Chl a at 440 nm (Mobley 1994), although this is not necessarily the case in more optically complex waters. The absorption by CDOM in all simulations was much lower than either the absorption by phytoplankton or that needed to induce a red tide from CDOM alone (see following).
Hydrolight was also used to estimate water-leaving radiance due to increasing amounts of CDOM and minerals, respectively. A single exponential CDOM model was used with a spectral slope of 0.015 (Twardowski et al. 2004) and soluble absorption coefficients, ag (412), ranging from 0.1 to 15 m21. The model was also run with increasing amounts of four different types of minerals: brown earth, red clay, yellow clay, and calcareous sand. Mass-specific absorption and scattering coefficients were obtained from Ahn (1990), as supplied with the Hydrolight model (Mobley 1994). The absorption coefficients fall within the range of recent measurements from a variety of mineral particles (Babin and Stramski 2004). A Fournier- Forand particle phase function of 0.025 was used for the mineral simulations (Fournier and Forand 1994; Twardowski et al. 2001; Sullivan et al. 2005).
Color modeling—The Commission Internationale de l’Eˆ clairage (CIE) developed a universally recognized objective system of colorimetry whereby the spectral distribution of light can be used to derive Y, the luminance or brightness, and two chromaticity parameters, x and y, representing the hue and saturation (Williamson and Cummins 1983). This system is based on color matching functions (tristimulus functions) that have been derived for the average human subject and are considered to be reasonably accurate and reproducible. However, the apparent color is not only a function of the light that falls on the localized region of the retina, but can also depend on the surrounding field or background colors. Contrast phenomena that can alter the apparent color (e.g., bright sky vs. dark water) are not considered here. We assume the human observer is looking directly down at the sea surface with no sun glint or bright sky for contrast. The water-leaving radiance spectra were converted into color using the three tristimulus functions for a 10u field of view (x¯ ,y¯,z¯) (CIE 1991). The CIE color components were estimated as the products of the radiance spectrum and three tristimulus functions integrated over the visible spectrum (400–700 nm), such that
In calculation of the CIE chromaticity coordinates (x, y), the conversion factor Km cancels out of the equation and
However, the magnitude of Km is important for determining the perceived brightness or luminance of a spectrum and for converting from CIE coordinates into computer-based red–green–blue (RGB) coordinates. To determine whether the radiance spectrum is too low for human vision and the water appears dark or black, Km is set to the maximum luminous efficacy for photopic (daylight) vision of 638 lumen (lm) W21 (Mobley 1994). The water is determined to be black when the calculated luminance, Y, falls below 3 lm m22 per steradian (sr21), which is the luminance for a twilight sky (Mobley 1994). To determine the RGB color of a spectrum given its relative spectral shape, the conversion factor, Km, must be normalized to the luminance, Y, where
The empirical factor 0.4 produces colors in the midpoint of the brightness range for the RGB color system and adequately simulates seawater color. The CIE X, Y, and Z values are transformed into RGB primaries using a matrix transform based on the chromaticity coordinates and reference white of a standard computer monitor (ITUR 2002), such that
The resulting RGB coordinates range from 0 to 1 and represent the fractional amount of each primary needed to display a particular color on a computer monitor. Values calculated to be less than 0 were set to 0 and values greater than 1 were set to 1. The MATLAB code for determining the RGB color of a spectrum is posted on the web (Dierssen 2006).
Answers to all your Biology Questions
