The role of mineral precipitation events in creating large patches of bright green water in the Salton Sea was investigated by comparing in situ inherent optical properties (IOPs) and constituent concentrations within and outside a green water region. While absorption was similar in both regions, scatter and backscatter were ∼2 and 3 times higher in green water, respectively. Ratios of scatter to absorption and backscatter to absorption had nearly identical spectral shapes but much higher magnitudes within green water. CIE chromaticity values were similar between stations, but luminance was 2.4 times greater in green water. Therefore, differences in observed water color were mostly due to increased brightness within green water. Further analyses of IOPs indicated that particles were small at both stations (average diameter ∼0.3 μm), but a larger proportion of particles present in green water were inorganic. Scanning electron microscopy analysis revealed the presence of small (up to 5 μm) particles consistent with gypsum. Because precipitated minerals only increase backscatter and do not by themselves affect water color, simple reflectance ratios will not always detect these events. Therefore, the magnitude of reflectance must be incorporated into analyses of precipitation events.
 Large-scale precipitation events result in high concentrations of suspended mineral particles that are visible as bright colorations in satellite imagery. Many investigations of these phenomena have utilized satellite-derived measurements to track the timing and location of events [e.g., Strong and Eadie, 1978; Weeks et al., 2002]. In several studies, precipitation events were characterized by decreased Secchi disk depth and light penetration, and increased turbidity, radiance, and reflectance [Effler et al., 1987; Ohde et al., 2007; Takeda et al., 1991; Weidemann et al., 1985]. Of the few studies that measured or estimated inherent optical properties (IOPs), increased reflectance and increased turbidity resulted from increased scatter or backscatter associated with particles [Effler et al., 1987; Ohde et al., 2007; Takeda et al., 1991]. However, none of these studies used IOPs to determine particle characteristics (size and index of refraction).
 For the last 50 years, spectacularly large patches of bright green water have been documented in the Salton Sea, a eutrophic, moderately hypersaline (48–50 g L−1) lake located in the southern California desert (Figure 1). Past studies have attributed the green water to gypsum (CaSO4 · 2H2O) particles, thought to form during the oxidation of hydrogen sulfide mixed into the upper water column during windstorms [Tiffany et al., 2007; Watts et al., 2001]. In this study, we compared in situ bulk IOPs at high spectral resolution of water located within and outside an extensive green water patch. Analysis of IOPs showed differences in brightness but not water color, and differences in particle composition (organic vs. inorganic) but not particle size. The size and composition of particles inferred from optical analyses matched those observed using scanning electron microscopy.
2. Materials and Methods
 Field collections were conducted on 20 August 2006. S-G was located within a green water patch (33° 25.0′N; 115° 55.0′W), and S-B was located just outside the green water (33° 20.58′N; 115° 55.99′W (Figure 1)). Vertical profiles of specific conductivity, temperature, and dissolved oxygen were measured using a YSI model UPG6000 Sonde. Discrete water samples (20 L) for particle analyses and optical measurements were collected at 0.25 m, 1.5 m, and 3.5 m.
 Duplicate water samples for chlorophyll a (chl a) measurements were filtered (Whatman GF/F; 0.7 μm) and placed in 90% buffered acetone (v/v) at −20°C for 24 h. Chl a concentrations were calculated using standard methods and a Turner TD-700 fluorometer [American Public Health Association (APHA), 1998a]. To determine the concentration of total suspended solids (TSS), whole water samples were filtered through pre-weighed filters (Whatman 934-AH; 1.5 μm), rinsed to remove salts, dried at 100°C for 2 h, and re-weighed [APHA, 1998b]. Environmental scanning electron microscopy (ESEM) was used to qualitatively explore particle composition. Whole-water samples were concentrated onto white polycarbonate membrane filters (Millipore; 0.2 μm) using gentle (<20 kPa) vacuum filtration. Filters were sputter-coated with carbon, mounted onto aluminum stubs, and imaged using an FEI Co. XL30 ESEM with a field emission gun at an accelerating voltage of 15 KeV. To determine elemental composition of particles, X-ray microanalysis was conducted with a Princeton Gamma Tech energy dispersive spectrometer (EDS) and a PRISM IG intrinsic germanium detector mounted on the ESEM.
 CDOM fluorescence was measured on discrete water samples using a WETStar CDOM fluorometer (WET Labs, Inc.) by introducing the sample into the flow tube and reading the output using a digital voltmeter. Raw fluorescence was converted to equivalent concentration of quinine sulfate dihydrate (QSD) using methods recommended by the manufacturer. Spectra of attenuation [c(λ)] and absorption [a(λ)] were measured by pumping discrete water samples through an ac-s spectral absorption and attenuation meter (25-cm pathlength, uncertainty ±0.001 m−1; WET Labs, Inc.). For each sample, spectra of whole water were measured. The water was then filtered (0.2 μm), and spectra of the filtered water (CDOM) were measured. Data files from S-B 3.5 m were corrupted and were not used in the analyses. To calculate spectra of the particulate fraction [cp(λ) and ap(λ)], filtered water spectra were subtracted from whole water spectra. Particulate scatter [bp(λ)] was calculated by subtracting scatter-corrected absorption from attenuation. Raw data files were processed in Matlab using ACS Graphic User Interface (GUI) v. 1.4.1 (WET Labs, Inc). The GUI subtracts pure water spectra from the measured spectra and performs temperature and salinity corrections according to Sullivan et al. . Unfiltered absorption spectra were corrected for scatter using the proportional subtraction method [Zaneveld et al., 1994]:
where acorr(λ) is corrected absorption, and “meas” indicates measured values.
 To simulate spectra that can be measured using ocean color satellite imagery, absorption [Morel and Prieur, 1977] and scatter [Morel, 1974] values of pure water were added to corrected a(λ) and b(λ) to calculate atot(λ) and btot(λ). For each sample, slopes of the cp(λ) spectra (γ) and of the particle size distributions (PSDs; ξ) were calculated following the relationships from Boss et al. . The spectral slope of CDOM absorption (S) was determined for each sample following Twardowski et al. . S and γ were estimated using a nonlinear least squares fitting approach (the fminsearch routine in Matlab) which better approximates the spectral slopes [Twardowski et al., 2004].
 Backscatter [bb(λ)] measurements were obtained from discrete samples using an ECO BB9 (WET Labs, Inc.). The BB9 measures the volume scattering function at 117° [β(117°, λ)] at nine wavelengths. Only six wavelengths are presented due to a calibration problem in the 406.5 nm channel, problems with salinity corrections in the 720.5 nm channel, and interference by chl a fluorescence in the 676 nm channel. Raw counts were converted to β(117°,λ) values using a scale factor supplied by the manufacturer and corrected for absorption along the pathlength with corresponding a(λ) measurements from the ac-s using a protocol described by the manufacturer. β(117°,λ) values for pure water [Morel, 1974], were subtracted from absorption-corrected β(117°,λ) values, and particulate backscatter coefficients [bbp(λ)] were determined using equations from Boss and Pegau . For use in the ratio of backscatter to absorption, total backscatter [bbtot(λ)] was calculated by adding backscatter values of pure water from Morel .
 Water-leaving radiance [Lw(λ)] just above the surface was estimated at each station from atot(λ), ctot(λ), and bbtot(λ) measurements using the two-component IOP model in Hydrolight (v. 4.2, Sequoia Scientific, Inc.). CIE chromaticity coordinates (x, y) and luminance were calculated from Lw(λ) using Matlab code by Dierssen et al. . A Km value of 683 lm W−1 was used to output luminance in units of lm m−2 sr−1.
 The precipitation event was first detected from satellite imagery and boat observation after a windstorm in late July 2006, concurrent with a dieoff of ∼3 million fish. At the time of sampling, a large green patch (∼390 km2) was present throughout the center of the lake (Figure 1). The water column was isothermal and anoxic (<0.5 mg L−1) at both stations, which is typical during these events (Table 1) [Watts et al., 2001]. Most indicators of dissolved (CDOM absorption and fluorescence, S) and particulate (TSS, chl a, ξ) components were similar at the two stations (Table 1).
Table 1. Measurements of Dissolved and Particulate Materials From Three Depths Collected at S-G and S-Ba
S, spectral slope of the CDOM absorption spectrum; ξ, slope of the particle size distribution.
Uncertainties based on instrument/method precision.
R(λ) values estimated using bbtot(λ)/[atot(λ) + bbtot(λ)].
 A comparison of bulk IOPs, chromaticity, and luminance revealed that differences in the visual appearance of water within and outside the green patch were mostly due to differences in brightness. While ap(λ) and aCDOM(λ) were similar for all samples, values of bp(λ) were 2 times higher and bbp(λ) was approximately 3 times greater at S-G (Figure 2). Luminance was 2.4 times greater at S-G (S-B = 290.97; S-G = 698.80 lm m−2 sr−1) indicating that the perceived brightness of green water was higher than ambient water. The magnitudes of btot(λ)/atot(λ), bbtot(λ)/[atot(λ)+bbtot(λ)], and Lw(λ) were higher at S-G, reflecting increased brightness at that station (Figures 3b and 3d). A narrow peak in btot(λ)/atot(λ) spectra centered at ∼576 nm was apparent in all samples. This same peak was observed in bbtot(λ)/[atot(λ)+bbtot(λ)] spectra using linearly interpolated bbtot(λ) values. The blue to green reflectance ratio [R(443)/R(555)], estimated using bbtot(λ)/[atot(λ)+bbtot(λ)], was nearly identical for all samples and indicated the presence of green water at both stations (Table 1). CIE chromaticity coordinates (S-B 0.3311, 0.4339; S-G 0.3388, 0.4457) and dominant wavelengths [Mobley, 1994] (S-B = 547 nm; S-G = 550 nm) were similar at both stations, falling within the yellowish-green portion of the chromaticity spectrum.
 Analyses of particulate IOP spectra showed particles were small at both stations, but a higher proportion of inorganic particles was present within the green water. Estimated PSD slopes (ξ) were nearly identical in all samples (Table 1). Using equations from Boss et al.  that assume a power-law PSD, estimated values of ξ, setting the minimum particle diameter to 0.2 μm (the pore size of the filter used), and the maximum diameter to 20 μm, the average particle size was ∼0.3 μm. The power-law PSD, however, tends to overestimate the contribution of small particles [Mobley, 1994] and the average particle size may have been larger. The backscatter ratios fell at or above a typical suspension of inorganic particles (0.02 [Roesler and Boss, 2008]) but indicated that the particle assemblage in the green water was more inorganic in nature (Figure 3a). The bulk index of refraction could not be estimated because ξ fell well outside the recommended range for this analysis [Twardowski et al., 2001]. The shape of the single scattering albedo [ωop(λ) = bp(λ)/cp(λ)] for all samples was similar to that of a typical algal cell [Babin et al., 2003], but the higher proportion of inorganic particles at S-G was apparent in its overall increase and a flattening of its shape at S-G (Figure 3c). Additionally, lower ratios of chl a to cp(660) indicated relatively low contributions by phytoplankton and large contributions by inorganic particles to attenuation at S-G (Table 1) [Boss et al., 2004].
 Based on the IOPs, we might expect a decrease in the concentration of organic particles (chl a), an increase in the mass of particles (TSS), and/or a change in particle size (ξ) at S-G. The difference in pore sizes of filters used to measure particulate IOPs and TSS means that particles ∼0.2–1.5 μm were included in measurements of bp(λ) but not TSS. Backscatter measurements (whole water) included even smaller particles. The estimated average particle size (∼0.3 μm) falls between the two pore sizes; therefore, a large proportion of particles included in particulate IOP measurements likely passed through the filter used for TSS. An increase in this small size fraction, which may not have been detected in TSS measurements, could account for the increased scatter and backscatter at S-G, especially if it consisted mostly of mineral particles.
 Particles with a geometric shape and EDS spectrum consistent with gypsum were observed under ESEM. Most gypsum particles were a few μm in size, but larger particles (20–50 μm) were also observed. With a refractive index of ∼1.13–1.15 relative to water [Nesse, 2004], these small gypsum particles could alter IOPs by increasing scattering to produce the bright green water observed in the Salton Sea. Minerals that have been shown to precipitate in other systems (e.g., calcium carbonate, elemental sulfur) were not detected.
4. Discussion and Conclusions
 This study is the first to measure IOPs in situ during a precipitation event in the Salton Sea, and one of the few studies of in situ IOPs during precipitation events in any aquatic system. Our measurements indicated the presence of green water at both stations, which is not surprising considering the high absorption at blue wavelengths by phytoplankton and CDOM and by water at red wavelengths. The observed water color difference resulted from increased brightness within the green water, which had to be due to increased scatter by particles. Further analysis of particulate IOPs revealed that a higher proportion of particles within the green water was of high index of refraction, which is inconsistent with organic particles. This study is the first to observe small (up to ∼5 μm) gypsum particles within green water at the Salton Sea. Tiffany et al.  used light microscopy without the use of an oil immersion lens and/or wavelength filters to identify particles. Resolving particles less than ∼2 μm and identifying visual features characteristic of gypsum is difficult using this method. Green water has been hypothesized to occur because of the presence of gypsum particles [Tiffany et al., 2007; Watts et al., 2001]. Our results are consistent with this hypothesis, but our methods cannot distinguish small differences in the bulk index of refraction (e.g., different mineral types) [Boss et al., 2004], and our sampling methodology may not have sufficiently preserved all sulfur species present. We therefore cannot rule out the presence of other small (0.2–2 μm) inorganic particles reported to play a dominant role in similar precipitation events [Takeda et al., 1991; Weeks et al., 2002]. Further studies are needed to fully characterize the size and type of particles present during these precipitation events.
 Differences in observed water color in the Salton Sea were consistently due to increased scatter and backscatter by particles within the green water. In other systems, the presence of inorganic particles (iron-containing dust, precipitated polysulfides) was detected using blue to green waveband ratios [Claustre et al., 2002] or a combination of waveband ratios and increased reflectance [Ohde et al., 2007]. Increased scatter due to precipitated minerals which do not contain iron will only affect the magnitudes of the reflectance spectra. Therefore, any differences in their shapes would depend on whether changes in the composition and concentration of other optically-active constituents (e.g., phytoplankton, CDOM) co-occur. Since no other reflectance spectra have been measured during Salton Sea precipitation events, it is unknown whether our results are widely applicable or are unique to the event sampled. Based on the data presented here, at least some precipitation events are not detectable using waveband ratios. Therefore, future studies of precipitation events should incorporate the magnitudes of reflectance spectra, especially if measurements of scatter or backscatter are unavailable.
 We thank B. Brinegar of Environmental Recovery Solutions for providing boat logistics, I. Cetinić for assisting with analysis of optical data, E. Boss for use of his BB9 and providing Matlab code, and K. Randolph for running Hydrolight models. We thank D.L. Valentine for providing partial financial support (NSF grant MCB-0604191). This project was funded in part by a CMIS California NASA Space Grant to B.K.S. Additional support was provided by a NASA ESS fellowship awarded to K.M.R. and a Philip and Aida Siff Graduate Fellowship awarded to B.K.S. Comments and suggestions by T. J. Swift, E. Boss, and an anonymous reviewer greatly improved this manuscript.