Evolution of the optical properties of seawater influenced by the Deepwater Horizon oil spill in the Gulf of Mexico

Water samples were collected from stations around the Deepwater Horizon oilrig in the northern Gulf of Mexico during two cruises in October 2010 and October 2011 (figure 1). Both cruises were accomplished onboard the R/V Cape Hatteras and they covered 24 and 27 stations during October 2010 and October 2011, respectively. Detailed sampling locations and selected hydrographic data are listed in table 1.

Zoom In Zoom Out Reset image size

Water samples from different depths at each station were collected with Niskin or Go-flo bottles mounted on a CTD rosette system, including surface waters at ∼2–5 m depth, deepwater samples between 1100 and 1400 m, and bottom water samples (table 1). Immediately after sample collection, the water samples were filtered through pre-combusted glass fiber filters (0.7 μm, Whatman). Filtered water samples for dissolved organic carbon (DOC) were collected in 30 ml HDPE bottles and stored frozen, while samples for optical measurements, including UV–vis absorbance and fluorescence EEMs, were collected with pre-combusted (550 °C) 125 ml amber bottles and stored in the dark at 4 °C.

Concentrations of DOC were measured on a Shimadzu TOC-V total organic carbon analyzer using the high temperature combustion method [44]. For DOC measurements, samples were acidified with concentrated HCl to pH < 2 before analysis. Three to five replicate measurements, each using 150 μl samples were made, with a coefficient of variance of <2%. Calibration curves were generated before sample analysis. Nanopure water, working standards and certified DOC standards (University of Miami) were measured as samples every eight seawater samples to check the performance of the instrument. Total DOC blank, including water and instrument blanks, was normally less than 2–6 μM [44].

UV–visible absorption spectra of the samples were measured using an Agilent 8453 spectrophotometer and a 1 cm path-length quartz cuvette over the 200–1100 nm wavelength ranges with 0.1 nm increments. The water blank was subtracted, and the refractive index effect was corrected by subtracting the averaged absorbance between 650 and 800 nm [45]. Specific UV absorbance at 254 nm (SUV A₂₅₄) values were calculated by dividing the absorption coefficient (m⁻¹) at 254 nm (a₂₅₄) by the DOC concentration (mg C l⁻¹). The non-linear spectral slopes between 290 and 400 nm were calculated to provide information on the overall molecular weight of DOM [30, 46, 47].

A Shimadzu RF-5301PC spectrofluorometer was used to measure fluorescence signatures of water samples in a 1 cm path-length quartz cuvette. Each sample was scanned from 240 to 680 nm with 1 nm interval under excitation wavelengths from 220 to 400 nm with a 2 nm step. Ninety-one separate fluorescence emission spectra were concatenated to generate an excitation–emission matrix that was able to provide DOM component information for the water sample qualitatively and quantitatively [48, 49]. PARAFAC modeling was used to derive DOM fluorescence components [33] and to examine the spatial and temporal changes in DOM components in the Gulf of Mexico.

A water blank was scanned daily before sample analysis and its EEM was subtracted from each sample's EEM. An emission correction spectrum was generated using Rhodamine B and barium sulfate with the correction package from Shimadzu and multiplied to the EEM spectra. Quinine sulfate standards were also scanned daily for fluorescence calibration and for checking the instrument performance. All fluorescence intensities were converted to ppb-QSE units [49]. Data in two triangular areas, corresponding to the Rayleigh and Raman scattering peaks, were eliminated in the PARAFAC analysis to acquire better mathematical results.

PARAFAC modeling was applied to all field seawater samples collected from 2010 to 2011, using MATLAB (MathWorks R2010b) and the DOMFluor Toolbox [33]. Sample matrices were calibrated and corrected before running the PARAFAC analysis. A non-negativity outlier test was performed and no outlier samples were chosen for removal. Thus, no samples were removed before split-half analysis and model validation. The fluorescence intensities of each component in every sample were quantified as a result of the PARAFAC modeling.

The spatial distributions of salinity, DOC concentration, UV absorption coefficient at 254 nm (a₂₅₄) and specific UV absorbance at 254 nm (SUV A₂₅₄) during October 2010 are shown in figure 2. During October 2010, three months after the oil spill was capped, the DOC concentration and a₂₅₄ in surface waters did not show an obvious influence of oil and their abundance had dropped back to more naturally occurring levels, with the highest values found at stations close to the Mississippi River plume and a general decrease in DOC with increasing salinity (figure 3).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

These surface distribution patterns are distinctly different from those observed during the oil spill in the May and June cruises [14, 50], showing remarkable resilience of the surface waters. The distributions of DOC and chromophoric dissolved organic matter (CDOM) in surface waters during the early stages of the oil spill in May and June 2010 showed a profound influence of oil released from the Macondo Well in the northern Gulf of Mexico [50], with DOC concentrations as high as 6 mg C l⁻¹ found around the oilrig, which were considerably higher than the baseline values in the northern Gulf of Mexico [44, 51, 52]. Similarly, elevated DOC concentrations and absorbance values in deep waters between 1100 and 1400 m were also observed during May/June 2010 [50], consistent with the presence of oil plume observed in the deepwater [1, 2, 8].

Even though the surface water DOC and CDOM did not seem to have a significant oil signature by October 2010, as shown in figure 2, the relationship between DOC concentration and salinity in all water samples throughout the water column showed an abnormal deviation from a general conservative DOC–salinity relationship as observed before the oil spill in the Gulf of Mexico and in other oceanic environments [24, 44, 53–55]. Surprisingly, some of the DOC concentrations from those abnormal deepwater samples were significantly lower than those observed previously from deep waters in the Gulf of Mexico and North Atlantic Ocean (figure 3 and [44, 51, 52]). We hypothesized that the extremely low DOC concentrations measured for oil contaminated deep waters were the result of the scavenging or removal of DOC by oil droplets and subsequent sorption of oil on the glass fiber filters during water sample filtration.

Based on the correlations between salinity, DOC, a₂₅₄, and SUV A₂₅₄, two major types of DOM could be identified in the water column during October 2010, three months after the oil spill (figure 3). The first type of DOM, residing mostly in the upper water column, had natural DOM characteristics with a positive correlation between DOC concentration and SUV A₂₅₄ values. The second group of DOM, found exclusively in deep waters with a characteristic salinity of 34.96 ± 0.03, had an anomalously high optical yield and a negative correlation between SUV A₂₅₄ values and DOC concentrations, showing a strong influence of oil on deep waters in October 2010 (figure 3), although rapid recovery was observed in surface waters.

Similarly to the results observed during October 2010 (figures 2 and 3), surface water samples collected during October 2011 (15 months after the oil spill) had undetectable oil signals (figure 4), but deeper water samples again showed a strong presence of oil contaminated DOM (figure 5). While the oil signatures in surface waters identified from optical properties faded away quickly, the presence of oil in the deepwater column persisted even 15 months after the oil spill in the Gulf of Mexico was capped. Effective microbial and photochemical degradation in surface waters, water stratification in the deeper water column, as well as circulation in the Gulf of Mexico [56] are likely the major factors governing the distribution of oil and DOM and their fate and transport in the water column.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Fluorescence EEM spectra of crude oil and a time series of seawater samples taken from the same depth at ∼1050 m in the Gulf of Mexico in May 2010, October 2010 and October 2011 are shown in figure 6. The crude oil had its maximum fluorescence emission at 320–360 nm over excitation of 220–240 nm, centering on Ex/Em 226/340 nm. Another peak in the crude oil EEM was located at an emission wavelength of 322 nm under excitation between 260 and 280 nm, similar to that reported by Bugden et al [21].

Zoom In Zoom Out Reset image size

The fluorescence EEM signatures of the seawater samples taken in May 2010 strongly resemble those of the crude oil, indicating the presence of oil in the water column and its influence on the seawater samples [50]. However, the oil fluorescence signatures derived from the EEM spectra were weak in the samples collected during the October 2010 and October 2011 cruises (figure 6), indicating effective dilution, degradation and transformation of the oil in the water column. Since the data for DOC and other optical properties clearly demonstrated the presence of oil in deep waters even 15 months after capping the spill (see section 3.1), the weak oil fluorescence signatures observed after the oil spill likely also resulted from the application of a vast quantity of dispersants during the DWH oil spill and thus the interaction of oil with dispersants in the water column [57, 58]. Indeed, significant alterations in oil EEM spectra have been observed when dispersants are present with oil [21, 50, 59]. Other factors contributing to the weak fluorescence signatures included a possible sorption effect during the sample filtration processes since, in general, oil has a very low solubility in seawater, and can be readily removed on filters and sorb on the bottle wall during sample processing.

Despite low fluorescence oil signatures, oil components could be recognized from these seawater samples using PARAFAC modeling (figure 7). As shown in table 2, four DOM fluorescence components were identified using PARAFAC analysis of EEM data from seawater samples taken during and after the oil spill. A total of 228 fluorescence emission matrices collected at wavelengths from 240 to 680 nm over excitation wavelengths from 220 to 400 nm were decomposed into a four-factor PARAFAC model, including three oil-related components (C1, C2 and C3) and one humic-like DOM component (C4). The first component, C1, having an emission maximum at 224 nm under an excitation wavelength of 328 nm, was the most prominent oil component. The second and third components were identified at Ex/Em maximum wavelengths of 264/324 and 232/346 nm, respectively. The fourth component, with its maximum Ex/Em of 248/446 nm, was characterized as naturally occurring humic-like DOM (table 2, figure 7).

Zoom In Zoom Out Reset image size

As shown in figure 8, the fluorescence intensities of the oil component C3 had a broad correlation with a₂₅₄ in the Gulf of Mexico in October 2010, suggesting an oil component with similar quantum yields and optical activities. A more scattered relationship between a₂₅₄ and the fluorescence intensities was observed for oil components C2 and C1 (figure 8), suggesting that both C1 and C2 were more complex in composition and might have multiple production/degradation pathways.

Zoom In Zoom Out Reset image size

Since the oil-related fluorescence components identified by PARAFAC modeling were derived from seawater samples collected at different times during and after the oil spill, changes in the fluorescence intensities and component ratios between seawater samples would likely reflect the results of degradation and transformation of the oil in the water column. Additionally, the fluorescence component ratio is an intensive property, which is not related to the quantity or abundance of oil, and should be an ideal parameter or index to evaluate time series samples in the same water column. As shown in figure 9, even though the intensity/concentration of DOM fluorescence components decreased with time, the C2/C1 and C3/C1 ratios increased consistently in seawater samples from the middle of May to May/June to October 2010 and to October 2011, indicating that these oil component ratios are indeed correlated with oil degradation and can be used as an index for tracking the chemical evolution of the oil during its degradation and transformation in the water column. The increase in the C2/C1 and C3/C1 ratios during oil degradation in the water column suggested that the two oil components C2 and C3 had significantly lower degradation rates as compared to the C1 oil component, or that C2 and C3 were also degraded products of crude oil. Independent controlled laboratory experiments on the degradation of crude oil also provided similar variation trends of oil component ratios with increasing C2/C1 and C3/C1 ratios during oil degradation [50], further supporting the use of C2/C1 and C3/C1 ratios as an index to trace weathered and degraded oil in marine environments.

Zoom In Zoom Out Reset image size

Data from the October 2010 cruise also show a broad correlation between spectral slope values and oil component ratios such as C2/C1, C3/C1 and C2/C3 in the water column of the Gulf of Mexico although the correlations are somewhat scattered (figure 8). Given that the spectral slope values are inversely related with the aromaticity and average molecular weight of DOM [30, 46], these positive correlations between oil components and spectral slope values suggested that C2 and C3 were less aromatic, lower inferred molecular weight components compared with crude oil. Thus, the overall molecular weight of DOM in the water column is expected to decrease as the oil is degraded and as the C2/C1 and C3/C1 ratios increase (figure 9). A similar correlation between the C2/C3 ratio and the spectral slope (figure 8) further suggests a decreasing trend of average molecular weight in oil components from C1 to C3 and then to C2. However, detailed analyses of hydrocarbon composition are needed to confirm this conclusion derived from spectral slope measurements.

Interestingly, the C2/C1 ratios in surface water samples were, in general, slightly higher than those in deepwater samples regardless of sampling time, suggesting that either the C2 component was less sensitive to photochemical degradation, or its production rate from degradation was slightly higher than its degradation. In contrast, surface water C3/C1 ratios were in general lower than deepwater samples except for the May 2010 samples collected during the oil spill (figure 9), suggesting that the production of C3 in the surface waters could be lower than its degradation. We hypothesize that C2 is mostly derived from photochemical degradation and C3 is a degradation product from both microbial and photochemical degradation, while both of them are also subject to degradation in the water column. The inferred C2/C3 ratio was higher in surface water samples than in deep waters, and increased with time, which further confirmed the degradation preference of C2 and C3. Thus, changes in oil component ratios in the water column could be quantitatively linked to the fate, degradation and transformation pathways of crude oil in the water column.