Oil irradiation experiments document changes in oil properties, molecular composition, and dispersant effectiveness associated with oil photooxidation

Abstract

While chemical dispersants are a powerful tool for treating spilled oil, their effectiveness can be limited by oil weathering processes such as evaporation and emulsification. It has been suggested that oil photooxidation could exacerbate these challenges. To address the role of oil photooxidation on dispersant effectiveness, outdoor mesocosm experiments with crude oil on seawater were performed. Changes in bulk oil properties and molecular composition were quantified to characterize oil photooxidation over 11 days. To test relative dispersant effectiveness, oil residues were evaluated using the Baffled Flask Test. The results show that oil irradiation led to oxygen incorporation, formation of oxygenated hydrocarbons, and higher oil viscosities. Oil irradiation was associated with decreased dispersant efficacy, with effectiveness falling from 80% to <50% in the Baffled Flask Test after more than three days of irradiation. Increasing photooxidation-induced viscosity seems to drive the decreasing dispersant effectiveness. Comparing the Baffled Flask Test results with field data from the Deepwater Horizon oil spill showed that laboratory dispersant tests underestimate the dispersion of photooxidized oil in the field. Overall, the results suggest that prompt dispersant application (within 2–4 days), as recommended by current oil spill response guidelines, is necessary for effective dispersion of spilled oil.

Graphical Abstract

Introduction

Chemical dispersants are an important tool for oil spill response.^1–3 They were designed⁴ for spray onto, and incorporation into, floating oil in order to lower the interfacial tension between the oil and underlying seawater. Dispersants are intended to promote the formation of smaller oil droplets so that surface oil can be entrained into the deeper water column by minimal wave action and be quickly diluted by ocean water to sub-toxic concentrations. Extreme weather, such as the Force 11 storm during the wreck of the Brea, is capable of dispersing surface oil without added dispersants,⁵ but dispersant application allows even mild weather to achieve this aim.¹ Nevertheless, the ‘window of opportunity’ for using currently available dispersants may be relatively short (approx. 2 to 4 days) due to weathering processes that change the composition and physical properties of the oil.⁶ The weathering processes that are typically considered are evaporation and the formation of a water-in-oil emulsion (i.e., emulsification of oil). Both processes increase the viscosity of the oil residue, thereby making it more difficult to break the oil into droplets.

A recent laboratory study found that in addition to evaporation and emulsification, photooxidation of oil decreased dispersant effectiveness.⁷ This study, which used a solar simulator for oil photooxidation and the Baffled Flask Test⁸ to test dispersant effectiveness, found a linear decrease of dispersant effectiveness with irradiation time. This decrease was attributed to a solvent incompatibility between the photooxidized oil residues and the co-solvents of the dispersants used to mix surfactants with the oil phase. The study investigators then directly used results from the Baffled Flask dispersant effectiveness test to conclude that photooxidation during the Deepwater Horizon oil spill might have decreased the effectiveness of aerial dispersants application to <45% for oil that was on the sea surface for more than 2 to 4 days (estimated distance of oil traveled within this time: 9 to 124 km; mean of 47 km).⁷ However, field data suggesting effective dispersion^9,10 were not taken into account.

To further investigate the role of photooxidation on dispersant effectiveness, we conducted a study to determine how photooxidation impacted the oil properties, chemical composition, and dispersant effectiveness of a Gulf of Mexico crude oil using outdoor mesocosms and natural sunlight that simulated field conditions. Specifically, our goals were (i) to characterize the effect of photooxidation on oil by determining changes in bulk oil properties and molecular composition by comparing irradiated oil samples to dark-control oil samples; (ii) to study the influence of photooxidation on the effectiveness of dispersants using a standardized laboratory-scale assay (Baffled Flask Test),¹¹ (iii) to evaluate changes in the chemical composition of weathered oil that may act as potential drivers affecting dispersant effectiveness, and (iv) to contrast the results of photooxidation and dispersant effectiveness tests to dispersant effectiveness observations and photooxidized samples collected during the Deepwater Horizon oil spill.^12,13

Materials and Methods

Mesocosm Experimental Set-Up.

The mesocosm setup is detailed in Fig. S1 in the Supporting Information (SI). The experiments were conducted at the Ohmsett oil spill test facility managed by the Bureau of Safety and Environmental Enforcement in Leonardo, NJ (40.425 °N, 74.068 °W). The mesocosms consisted of a series of six polyethylene (PE) tanks (117 cm long, 51 cm wide, 34 cm deep, grey 65-gal tote BC-4721, Bayhead Products Corp, NH). Each tank was divided into four sections (29 cm × 51 cm × 34 cm) with PE dividers that did not fully reach the bottom of the tanks to allow water to flow underneath replicate oil slicks. The tanks were plumbed together and connected to a pump and a 330-gal (1.25 m³) high-density PE intermediate bulk container as a reservoir. A water chiller and an air bubbler were immersed into the reservoir to control the temperature and prevent anoxic conditions. Approximately 1,500 L of seawater was continuously re-circulated through the system at a flow rate of approx. 2 L min⁻¹. The seawater was collected from Manasquan Inlet (40.103166°N, 74.035986°W; South of the Inlet; salinity: 30 ppt) during an incoming tide.

The light-exposure (“L”) tanks were placed on the deck of the wave tank at Ohmsett, where they were exposed to natural sunlight for 11 days. The tanks were covered with a non-transparent PE cover during rain events to avoid disturbance of the oil surface and limit the formation of emulsions. To simulate no-light conditions, identical experiments were conducted inside a nearby building. The dark-exposure (“D”) tanks were constantly covered with opaque PE covers to prevent exposure to ambient light.

A medium-light sweet Gulf of Mexico crude oil blend collected from the Hoover Offshore Oil Pipeline System (HOOPS) was chosen for our study (API gravity 35.2, sulfur content 1.15%, asphaltene content 0.3%).¹⁴ The experiment was started by adding 40 mL of fresh HOOPS oil to each of the four sections of the six mesocosm tanks to form an initial average oil film that was nominally 270 μm thick. This nominal thickness was chosen because it is thick enough to allow dispersant application while still thin enough for light to penetrate a significant depth into the oil film. A total of 960 mL crude oil (770g; 4 sections × 6 bins × 40 mL) was applied to a volume of approximately 1,500 L of seawater, resulting in an oil/water loading of 0.5 g L⁻¹.

Characterization of Solar Irradiation.

Natural sunlight irradiated these experiments. The solar irradiance was measured continuously with a silicon pyranometer (HOBO S-LIB-M003 sensor, spectral range 300–1100 nm; Onset, Bourne, MA). In addition, the spectral irradiance was measured with a miniature spectrometer (FLAME UV/VIS spectrometer 200–800 nm; Ocean Insight, Largo, FL). The spectral data were evaluated using the R software (www.r-project.org) and its Photobiology package.¹⁵

Sampling and Sample Extraction.

At eight time points during the 11-day experiment (Table S1), oil was removed from defined sections of the mesocosms by gently dragging two 10 × 40 cm² ethylene tetrafluoroethylene (ETFE) fabric sheets (Flourtex 9–250/36; General Oceanics, Miami, FL) per mesocosm section across the water surface. The ETFE sheets were immediately transferred into pre-cleaned wide-mouth glass jars and stored at 4°C until extraction and analysis. The ETFE sheets were extracted three times with dichloromethane (DCM), and the solvent was evaporated, as indicated by a constant weight of the samples. This process removed water from the samples, so viscosity and dispersant effectiveness measurements were not confounded by emulsified water.

Analytical Methods.

The apparent bulk saturated, aromatic, resin, and asphaltene fraction of the oil residues were quantified using thin layer chromatography - flame ionization detection (TLC-FID) using an Iatroscan MK-5 instrument (Iatron Laboratories, Tokyo, Japan) as detailed previously.^13,16 The oxygenated hydrocarbon fraction (“OxHC fraction) was calculated as the sum of the resin and asphaltene fraction.¹⁶ The bulk oxygen content was measured using an elemental analyzer (Elementar PyroCube) interfaced to an isotope ratio mass spectrometer (Isoprime VisION) at the Stable Isotope Facility of the University of California Davis. Select samples were analyzed for other elements (CHN on an elemental analyzer, S using the Schoniger combustion method¹⁷) at Midwest Microlab (Indianapolis, IN). The carbonyl index was measured using a previously-described Fourier-transform infrared spectroscopy (FT-IR) method.^13,18 We used an attenuated total reflectance crystal for measurement and integrated peak areas (carbonyl peak: 1,520 to 1,780 cm⁻¹: reference peak: 2,760 to 3,065 cm⁻¹) rather than peak heights, since this has been shown to improve the accuracy.¹⁹

The viscosity of the water-free oil residues was determined at 20°C using a microVISC viscometer (RheoSense Inc, San Ramon, CA) following published methods.⁷ Select samples were also measured using the Stabinger method for cross-comparison (see SI for details). The solubility of oil residues in dispersant co-solvents was measured gravimetrically (see SI for details).

PAHs and n-alkanes in the oil phase were measured using gas chromatography coupled to mass spectrometry (GC/MS; 7890B GC coupled to a 5977A MS; Agilent Technologies, Santa Clara, CA) following a published method.²⁰ In addition, select aliphatic photo-products (n-alkanoic acids and 2-alkanones) were measured using GC/MS according to methods described in the SI.

Representative oil samples were analyzed using Ion Mobility Spectrometry-Mass Spectrometry (IMS-MS). Each sample was diluted to a concentration of 1.5 mg mL^–1 with a 50:50 (v/v) mixture of HPLC-grade toluene and methanol. Sample analysis was performed using a 6560A Ion Mobility Q-TOF MS instrument (Agilent Technologies, Santa Clara, CA) using an atmospheric pressure photoionization source in positive ion mode (APPI⁺), facilitating the detection of aromatic compounds.²¹ IMS-MS raw files were processed for feature identification and molecular formula assignment (see SI for details) as detailed elsewhere.²²

Dispersant Effectiveness.

Dispersibility of fresh and experimental oil residues with and without Corexit EC9500A and Corexit EC9500B (Corexit Environmental Solutions LLC; Sugar Land, TX) were determined using the Baffled Flask Test.^8,23 Baffled Flasks (250 mL) contained 120 mL of 34 ppt synthetic seawater (Instant Ocean; St. Blacksburg, VA), 100 μL of test oil, and 5 μL of dispersant (dispersant-to-oil ratio 1:20). Two to four replicates were prepared per oil sample. The flasks were placed on an orbital shaker at 200 rpm for 10 minutes at room temperature, followed by a stationary, quiescent period of 10±0.25 minutes to allow undispersed and/or coalesced oil droplets to refloat to the surface. A 30-mL aliquot of the aqueous phase was collected from the bottom port of the baffled flasks and extracted with dichloromethane (DCM). The mass of dispersed oil was determined spectrophotometrically from the extracts. See SI for method details.

Results and Discussion

Irradiance and Visual Observations of Oil Slicks:

During the experiment, the average daily irradiance was 6.6 kWh m⁻² day⁻¹ (Fig. S2, Table S1). This irradiance was comparable to that in the Gulf of Mexico at the location of the Deepwater Horizon oil spill (Fig. S3 and S4). As a result of the irradiation, the oil on the irradiated tanks broke up into increasingly small patches that covered a smaller percentage of the water surface over time in each tank, while the oil on the dark-control mesocosm remained a coherent oil slick throughout the experiment (Fig. 1). We estimated that the average oil film thicknesses in the irradiated mesocosm increased from approx. 200 μm on day 1 (sample L1) to 350 μm on day 8 (L8; see SI for method details); a thicker film of 1,200 μm was measured on day 11 and was likely influenced by emulsification due to a rain event on the evening of day 9 when the mesocosms were not covered. The film thickness in the dark mesocosm was approximately 200 μm throughout the experiment.

Figure 1:

Photographs of the oil films during the experiment. L0 is the light-exposed sample at day 0, L1 the light-exposed sample at day 1, D5 the dark-control sample at day 5, etc. The scale bar corresponds to 5 cm. Indicated is the percent coverage of the oil films on the water surface.

Due to the light absorbance of light sweet crude oil, light with wavelengths relevant for oil photooxidation (approx. <500 nm)¹³ penetrates only tens of micrometers of an oil film, thereby limiting the photooxidation rate.^13,24 Similar limited oxygen diffusion into the slick has been hypothesized to serve as a kinetic constraint (i.e., the rate of photochemical O₂ consumption was faster than replenishment by diffusion).^13,24

Changes in Bulk Oil Properties:

Because oil photooxidation has been shown to produce oxygen-containing compounds^16,20 and modify the viscosity of oil residues,⁷ we investigated bulk and molecular changes in the oil. Oxygen content, carbonyl index, and OxHC increased by factors of 3.5 to 10 over eleven days in irradiated samples (Fig. 2, Table S2). The observed increases were linear with respect to the exposure duration and were higher in the irradiated samples than in the dark-control samples by at least a factor of four (Fig. 2B). The oxygen content increased from 1.3 to 6% throughout the experiment, and the OxHC fraction from 15 to 53%. At the same time, the aromatic fraction decreased from 38 to 10%, similar to irradiated oil in other studies.^25,26 In contrast, the changes in bulk properties in the dark-control mesocosm were smaller and were consistent with preferential enrichment due to evaporative loss of light hydrocarbons.

Figure 2:

(A) Bulk property changes for the light-exposed (L1, L2, etc.), dark-control (D1, D5, etc.), and HOOPS crude oil (HC) samples during the 11-day experiment. Given are TLC-FID-derived relative areas of saturated (Sat), aromatic (Aro), and polar fraction (OxHC), the carbonyl index derived from FT-IR, the oxygen content, and the measured viscosities (given as logarithm of the viscosity in cP at 20 °C; n.d. = not determined). (B) Changes in properties of panel A over time for the dark-exposed (red circles) and dark-control (blue triangles) samples. The numbers indicate the slopes of linear regressions.

We estimated the elemental composition of the photoproduct-containing OxHC fraction based on analysis of the oil residues (C, H, N, O, S; Table S2), following our previous approach.¹⁶ The O:C ratio for the total extract of irradiated samples was 0.04, and the OxHC fraction of irradiated samples had an average O:C ratio of 0.11 (assuming a constant molecular composition of C_xH_2.1x for the saturated fraction and C_xH_1.1xS_0.025x for the aromatic fraction). This ratio translates to one oxygen atom for approximately every 10 carbon atoms in the OxHC fraction. Previously, we determined higher O:C ratios (0.2) for highly weathered Deepwater Horizon oil-sand aggregates deposited on shores, where extensive biodegradation occurred in addition to photooxidation.¹⁶ Photooxidation can progressively oxygenate oil,²⁷ thereby increasing the O:C ratio. Furthermore, biodegradation has been hypothesized to have further increased the oxygen content in photooxidized Deepwater Horizon oil residues at advanced weathering stages.¹³ Biodegradation likely played a minor role in the oxygen content of the oil residues for the investigated time frame of this study (days of weathering).

The elemental analysis, carbonyl index, and OxHC results are consistent with those observed in Deepwater Horizon samples that were photooxidized on the sea surface.^13,16,28 Characteristic bulk property changes in these samples were similar to the changes observed in our irradiated samples (Fig. S6). This agreement provides evidence that findings from this study can be used to estimate the behavior of a similar oil in the field.

Changes in Viscosity:

The viscosity of oil slicks exposed to sunlight increased to a larger extent than those kept in the dark (Fig 1A, Table S1). While irradiated samples reached dynamic viscosities of up to 2.5×10⁶ cP after 11 days (sample L11), dark-control samples only reached 8.0×10² cP after 11 days (sample D11), showing that photooxidation had a greater impact on the increase in viscosity of oil residues than evaporation (Fig. 2B). Our results of the 1-day sample are comparable to those obtained using a solar simulator, where oil viscosity increased from 22 cP to 147 cP after 24 h of irradiation.⁷

Weathering Processes Affecting PAHs, n-Alkanes, and Oxygenated Compounds:

Evaporation and photooxidation affected the composition of n-alkanes, PAHs, and oxygenated compounds, as observed by GC/MS analysis (Fig. 3). Note that water-soluble compounds such as benzene, toluene, ethylbenzene, xylenes, and cyclohexane are also affected by dissolution.²⁹ However, we did not target these compounds in our analysis. GC/MS chromatograms of alkane traces show progressive loss of volatile compounds with increasing experimental duration (Fig. 3A). We calculated the remaining fractions of compounds relative to the original HOOPS crude oil using C₃₀-normalized data (similar patterns but more variable concentrations were obtained using hopane normalization, which is relatively recalcitrant towards weathering^25,28,30). The remaining fraction of n-alkanes vs. their retention times on GC/MS followed a sigmoidal relationship (Fig. 3B). Given that the retention times of compounds on an apolar GC column are mainly driven by their vapor pressure,³¹ this sigmoidal relationship is consistent with evaporative loss of n-alkanes.^12,32 Irradiated samples demonstrated a greater loss when compared to corresponding dark-control samples, consistent with higher maximum water temperatures of 36 °C experienced by irradiated samples as compared to 28 °C in the dark-control samples (Fig S5; the daily average temperature was comparable in the irradiated and the dark-control treatment: 26.8±1.0 °C and 26.7±4.3 °C, respectively). Furthermore, a slight increase in oil temperature (estimated 2 to 5 °C)³³ in the oil film on the water surface due to absorption of sunlight of the oil would lead to additional differences of evaporation between dark-control and irradiated treatments.

Figure 3.

Molecular changes in light-exposed and dark-control samples after 1, 5, and 11 days of exposure (samples L1, L5, L11, D1, D5, D11). (A) GC/MS chromatogram of the alkane traces showing weathering of the light-exposed and dark-control samples. The retention times are given as n-alkane carbon numbers. Progressive evaporation up to n-C₁₄ and n-C₁₅ was observed in the dark-control and light-exposed samples, respectively, after 11 days of exposure. (B) Remaining fraction of hydrocarbons relative to the original HOOPS crude oil in the same samples. Shown are n-alkanes (blue circles) and PAHs (red-shaded symbols), including two-ring species (inverted triangles; naphthalene and C₁ to C₄ alkylated congeners), three-ring PAHs (diamonds; fluorene, phenanthrene/anthracene, dibenzothiophene, and alkylated congeners), and four-ring PAHs (squares: pyrene/fluoranthene and C₁ to C₄ congeners; triangles: chrysene and C₁ to C₄ congeners). While n-alkanes mostly follow a sigmoid evaporation line, PAHs deviate from that line, especially for larger PAHs at later retention times. This behavior points to photooxidation. (C) Concentrations of C₁₀ to C₃₇ n-alkanes, total PAHs (tAlk and tPAH; left axis), and C₁₃ to C₃₀ n-alkanoic acids and 2-alkanones (tCA and tKet; right axis, note different scale). PAHs and n-alkanes were binned according to numbers of rings and carbon atoms, respectively (indicated by different degrees of shading). Evaporation mainly affected smaller compounds, whereas photooxidation also affected larger PAHs.

In contrast to n-alkanes, PAHs in irradiated samples did not follow a sigmoidal relationship, with many larger PAHs being more depleted than expected based on evaporation alone (Fig. 3B). This depletion of up to 80% is consistent with the effects of photooxidation, which is known to affect larger and more alkylated PAHs more than smaller compounds.³⁴ Furthermore, the absorption of light by oil compounds likely produced reactive species such as singlet oxygen (¹O₂), which specifically reacts with electron-rich structures such as aromatic rings (but not with saturated compounds).^24,35 In the absence of sunlight, the PAHs followed the same evaporative trend as n-alkanes (Fig 3B). The evaporative depletion pattern of n-alkanes and the photooxidation pattern of PAHs in our samples were similar to those in field-collected surface slick Deepwater Horizon oil spill samples (Fig. S6C and S7), showing that our experimental system effected molecular changes consistent with those occurring in the field on the sea surface.

Concurrent with the decrease in PAHs, we measured an increase in marker compounds for oil photooxidation in irradiated oil samples (Fig. 3C). Homologous series of 2-alkanones and n-alkanoic acids (carbon numbers C₁₀ to C₃₀) increased more than five-fold in irradiated samples over the course of the experiment but remained unchanged in dark samples (Fig. 3C, Table S4). These carboxylic acids and ketones are known products of photochemical degradation of n-alkanes,^36–39 and have been identified in irradiated oil residues.^20,40 The n-alkanes can be degraded by photochemically-produced radical species which can result from the degradation of photoreactive oil constituents such as aromatic compounds.³⁸ The magnitude of this radical-mediated n-alkane degradation seems to be relatively small, given the concentration of measured aliphatic photoproducts was only 2–5% of the corresponding n-alkanes (Fig. 3C).

Overall, our results show a relative persistence of n-alkanes during photooxidation, while aromatic compounds are more susceptible to transformation. This behavior is consistent with previous research that found that n-alkanes are less affected than aromatic compounds by oil irradiation.^25,34,41

Besides evaporation, we tested for oil biodegradation in the surface slick samples. We observed only a slight decrease in the heptadecane/pristane ratio (from 1.8±0.1 in the initial samples to 1.5±0.3 for all samples collected after 11 days; Table S3), indicating no significant biodegradation in the oil phase. Detectable biodegradation was likely limited by relatively high oil loading (approx. 500 ppm or 0.5 g oil per L of seawater) and the presence of an oil slick rather than dispersed oil. Previous work indicates that undispersed oil slicks with a nominal thickness of 350 μm only show minor signs of biodegradation during the first 28 days of incubation.⁴²

Untargeted Analysis of Photoproducts.

Given that initial photoproducts likely consist of a complex combination of compounds, we used an untargeted IMS-MS analytical method on representative samples. IMS-MS offers a rapid and high-resolution detection of spectral features, having unique mass-to-charge and drift times,⁴³ allowing for analysis of complex substances by mass and charge (m/z) with an additional spatial conformation known as collisional cross-section (CCS) based on the separation of ions based on their mobility. Using all high-abundance/quality features (n = 1,135), we observed clear changes in the overall chemical composition of the irradiated samples (Fig. 4A); Of these 1,135 features, a subgroup of n=59 (5.2% of total) was identified as oxygen-containing features (Table S5). Furthermore, hierarchical clustering (Spearman correlation and average linkage) led to the identification of two different clusters of features that have a higher abundance in samples that were exposed to light; therefore, they were characterized as “Light” induced features (n=175; Figure 4A). Amongst these 175 features, a subgroup of n=31 species (17.1%) was identified as oxygen-containing features (Table S5). These features significantly increased in light-exposed samples starting at four days of irradiation (Fig. 4B). The oxygen-containing putative oil transformation products span the C₁₂–C₂₇ range; C_15–20 and C_26–27-range features were detected only in samples exposed to light at day 11 (Fig. 4C). An increase in the C_21–25 range was observed in both dark- and light-exposed samples, yet their abundance was greater in light-exposed samples.

Figure 4.

Untargeted analysis using ion mobility spectrometry-mass spectrometry (IMS-MS). (A) A heatmap illustrating the relative abundance of IMS-MS features detected with APPI(+) (n=1,135). Hierarchical clustering (complete linkage) of samples (columns) and features (rows) was performed using data that was z-scaled for each sample with lower abundance features indicated by light blue and higher abundance features indicated by dark blue colors. Clusters of features that are most discriminating for “light” (n=175) conditions are outlined by red rectangles. (B) Relative abundance (compared to total feature abundance in each sample) of oxygen-containing features in all and “light” discriminating clusters of features. Features are shown using box-and-whiskers plots (vertical line is median, box is IQR, and whiskers are min-max with all features shown). The y-axis indicates experimental conditions, and light-exposed samples are colored in yellow. Each individual feature’s abundance is an average of three replicates in each sample. Asterisks (*) indicate a statistically significant (one-way ANOVA, p<0.05) difference from the HOOPS oil sample. (C) Line plots showing the relative abundance of oxygen-containing featured by carbon numbers. Shown are HOOPS Oil (black triangles) and oil slick samples from day 11 in dark (gray squares) and light (yellow circles) conditions.

All IMS-MS detected oxygen-containing hydrocarbons contained one oxygen per molecule (Table S5) and were, therefore, most likely ketones and/or alcohols. The increase in these compounds is consistent with the increase in bulk oxygen content (Fig. 2) and ketone and carboxyl photo-products identified using GC/MS (Fig. 3C). Our IMS-MS results are also in line with previous oil weathering studies using other ultra-high-resolution MS methods.^40,44–47 For example, Fourier-Transform Ion Cyclotron Resonance MS (FT-ICR-MS) measurements showed an increase of oxygen-containing molecules upon irradiation, including ketone/aldehyde O₁ compounds.⁴⁰ Diverse ketones and aldehydes were also identified in field-weathered Deepwater Horizon oil using liquid chromatography coupled to mass spectrometry.⁴⁸ Another FT-ICR-MS study, using electrospray ionization, found that the most prominent species were O₂ species, likely carboxylic acids.⁴⁴ We did not detect such higher oxygen classes since we used APPI(+) ionization, which preferentially ionizes aromatic compounds.⁴⁹ Overall, IMS-MS data confirm photochemical reactions across carbon number and compound classes, and complement the GC/MS and bulk analyses.

Oil Photooxidation Affected Dispersant Effectiveness:

We quantified the dispersibility of photooxidized oil using the Baffled Flask Test.¹¹ This test was intended to compare the relative effectiveness of various dispersants or oils for regulatory evaluation under standardized laboratory conditions.⁸ However, as summarized in a recent review, the Baffled Flask Test was not designed—and is not able—to predict absolute dispersant effectiveness in the field.⁵¹ In fact, the Baffled Flask Test generally underestimates the dispersant effectiveness in larger tank-based tests by a factor of two to three.^8,52 Therefore, we used this test to compare our data to previous research,⁷ but did not use these data to directly predict dispersion in the field (see discussion of field implication below). We used Corexit EC9500A and EC9500B with a dispersant-to-oil ratio of 1:20.

We found no correlation of dispersant effectiveness with experimental duration for the dark-control oil, and the D11 sample had similar dispersant effectiveness as the fresh crude oil (83±6 vs. 80±10% for crude oil and D11, respectively; Table S6). In contrast, the light-exposed samples showed a significant decrease in dispersant effectiveness after three days of irradiation (Fig. 5A). We found a correlation between oxygen content and dispersant effectiveness for the irradiated samples (Fig. 5B). Beyond three days of irradiation, samples had an oxygen content of >3% and low dispersant effectiveness values between 3 and 30%. Dispersant effectiveness also correlated to the OxHC fraction and carbonyl content (Figure S8B). Furthermore, the relationship between carbonyl index and dispersant effectiveness agreed with previously reported data from a laboratory irradiation experiment (Fig. S8A).⁷ It is noteworthy that different ways of generating photooxidized oil (solar simulator and a dry oil film on a glass plate in the cited study; oil film on water and natural sunlight in our case) led to consistent changes in Baffled Flask Test-dispersant effectiveness and carbonyl index.

Figure 5:

Baffled Flask Test dispersant effectiveness (DE) for dark-control (blue) and light-exposed (red) samples produced during the experiment. Shown are results for Corexit EC9500A (light-shaded colors) and EC9500B (solid colors). (B) Dispersant effectiveness of irradiated (red circles) and dark-control samples (blue triangles) vs. oxygen content, showing dispersibility below 40% for irradiated samples with an oxygen content larger than 3%. (C) Dispersant effectiveness plotted against dynamic viscosities of the oil samples for irradiated (red circles) and dark-control samples (blue triangles). For comparison, literature values of dispersant effectiveness and viscosities for a set of 25 different crude oils are given (asterisks; data from Holder et al., 2015).⁸

To isolate the effect of photooxidation from other effects (such as emulsification), we performed our experiment under calm conditions without wave or energy input. Furthermore, we intentionally broke up any possible emulsions by de-watering the oil residues at the end of the experiment by solvent extraction. This approach also allows comparison with a previous study.⁷ Note that emulsion formation would further increase the viscosity of the oil residues, and lower dispersant effectiveness values would be expected with the existence of emulsification.

Drivers of Observed Dispersant Effectiveness Changes:

To explore the factors potentially affecting the decrease in dispersant effectiveness upon irradiation, we investigated the paraffin content of the oil residues. It has been suggested that high paraffin contents in oil residues can be inversely correlated with dispersant effectiveness.⁵³ Previous studies found this correlation to be applicable for low-viscosity oils and not for oils with viscosities > 1,000 cP.⁸ In line with these studies, we did not find an inverse correlation between the paraffin content of the oil residues (total mass of n-alkanes per mass unit of total oil; Table S3) and measured dispersant effectiveness values (Fig. S9).

A second factor that might affect dispersant effectiveness is the solvent compatibility between the co-solvent system of the dispersant and the oil residues.⁷ We tested solvent compatibility of our oil residues with two types of co-solvent that are known to be present in Corexit 9500: two different hydrotreated light petroleum distillates and dipropylene glycol n-butyl ether (DPnB).^53–56 DPnB has been used as a marker compound for Corexit during the Deepwater Horizon oil spill.^57–61 We found that oil residues irradiated for 4 to 11 days had a reduced solubility in the petroleum distillates (average solubility: 68±13%); in contrast, DPnB, as well as 1:1 mixtures of DPnB and petroleum distillates, were able to completely dissolve irradiated oil residues (Fig. S10). To test whether the decreased solubility of irradiated oil residues in petroleum distillates could explain the measured reduction in dispersant effectiveness, we pre-mixed Corexit 9500A with photooxidized residue L8 prior to the Baffled Flask Test. However, pre-mixing did not improve dispersant effectiveness (12±2% without pre-mixing vs. 16±5 % with pre-mixing; Table S6), indicating that although we observed a decreased solubility of photooxidized oil residues in petroleum distillates, solvent incompatibility or mass transfer limitation between Corexit co-solvents and photooxidized oil residues were not major contributors to the reduced dispersant effectiveness of photooxidized oil.

Finally, we examined the role of viscosity in reducing the observed dispersant effectiveness. It is known from bench-scale, tank-based, and field tests that viscosity can affect the dispersibility of fresh and weathered crude oil.^8,53,62 In line with these studies, we also observed a decrease in dispersant effectiveness with increasing viscosity of irradiated oil residues (Fig. 5C). Our data showed a similar dispersant effectiveness-viscosity relation as a study that also used the Baffled Flask Test to measure the dispersant effectiveness of 23 different (non-photooxidized) crude oils (Fig. 5C).⁸ That study reported < 50% dispersant effectiveness for oils with viscosities above approx. 2,000 cP (measured at 15 °C). This agrees with our data: photooxidized oils with viscosities above 2,000 cP (at 20°C) all had dispersant effectiveness < 50% in the Baffled Flask Test. The consistency of the dispersant effectiveness-vs-viscosity relationship between our photooxidized oil residues and a set of 23 crude oils that were not photooxidized suggests that viscosity increases were likely the driver of our observed decreases in dispersant effectiveness. Photooxidation accelerated the rate and magnitude of viscosity increase, as seen by comparing our light and dark samples. The viscosity increase is likely driven by the formation of oxygenated compounds, which have a higher viscosity than corresponding hydrocarbons (Table S7). The ability of a compound to induce hydrogen bonding and dipole/dipole interactions has been shown to increase its viscosity.^62,63 Photopolymerization, which might occur during oil photooxidation,⁶⁴ would likely also increase viscosities.

Implications for Field Application:

The decrease of dispersant effectiveness for photooxidized oil has implications for field application of dispersants in oil spill response. In agreement with a previous study,⁷ our data indicate that irradiation of oil reduces the dispersant effectiveness determined in the Baffled Flask Test beyond what can be explained by evaporation alone. However, while this previous study directly used Baffled Flask Test dispersant effectiveness data to estimate dispersant effectiveness during the Deepwater Horizon oil spill, we take a more comprehensive approach and take into account field photooxidation as well as observations of field dispersant effectiveness.

Field observations of oil photooxidation and dispersant effectiveness during the Deepwater Horizon oil spill suggest that the role of photooxidation in dispersant effectiveness is minimal (Fig. 6). The degree of photooxidation achieved in our study matches that of surface slick samples collected during the Deepwater Horizon oil spill, with similar changes in bulk properties and hydrocarbon depletion in our samples (Fig. 3B) and the field samples (Fig. S6 and S7). Not surprisingly, large spatial heterogeneity in oxygen content can be found in Deepwater Horizon slick samples (Fig 6A), caused by mixing of weathered oil slicks with fresh oil, heterogeneity in oil slick thickness from corralling of oil in Langmuir cells, and emulsification.

Figure 6.

Oil photooxidation and dispersant application during the Deepwater Horizon oil spill. 50-km and 100-km radii from the wellhead are shown as concentric purple circles. (A) Location⁶⁸ and oxygen content¹³ in Deepwater Horizon oil slicks collected during the oil spill. See Table S5 for details. Samples with oxygen contents between 0.7 and 6.0% were collected within 100 km of the wellhead. Highly oxygenated slicks (oxygen content >5%) had variable modeled residence times at the sea surface between 1 to 18 days.¹³ (B) Field verification of dispersant effectiveness (data from refs^9,10). Flight tracks of performed dispersant application during the oil spill are indicated as lines (data from NOAA),⁶⁹ Effective dispersion denoted by blue symbols was observed in 77% of the verification measurements, including at locations where highly oxygenated oil was measured.

Based on field dispersant observations made during the Deepwater Horizon oil spill, there was no reduced dispersant effectiveness throughout the application area of dispersants, which was approx. within 100 km of the wellhead (Figure 6). These field observations were performed within approximately 30 minutes or less of either airplane or vessel application of dispersants during the Deepwater Horizon oil spill according to the Special Monitoring of Applied Response Technology (SMART) protocol.^58,65,66 The majority (77% of 25 observations) of the field measurements showed effective dispersion, based on fluorometry-determined oil concentration in the water column (Tier II/III; Fig. 6B), and the success rate was even higher based on visual observations (89% of 45 Tier I observations).^9,10 Note that while visual and fluorometry-based method as used in SMART have limitations (Table S10), a comparison study between fluorometry and GC/MS based measurements found a “reasonable agreement” between the two methods (Table S11).⁵⁸ This study also concluded that the “use of dispersants was still warranted despite the weathered nature of the targeted oil slicks”.⁵⁸ An oil spill modeling approach estimated residence times between 7 and 17 days for a set of Deepwater Horizon oil slicks that traveled between 30 and 80 km.¹³ Unlike in the Baffled Flask Test, where dispersant effectiveness is significantly decreased for irradiation times longer than three days (ref⁷ and our study), oil residues apparently still dispersed in the field after longer irradiation times. This is not surprising, given that dispersant effectiveness values are generally lower in the Baffled Flask Test than in larger-scale tests. For example, a study found that Baffled Flask dispersant effectiveness values lower than 50% (27±12%) still had 72±12% effectiveness in larger-scale tank tests,^8,52 and dispersant effectiveness can be even higher at sea.⁶⁷ This produces further evidence that the Baffled Flask Test, conducted in a very small closed system for only 10 minutes, should not be used to directly predict field dispersant effectiveness, as has been done in a previous study.⁷

We compared viscosities of our oil residues to those of Deepwater Horizon oil slicks. In our experiment, irradiation of an approx. 200 to 350 μm oil film for more than six days resulted in viscosities > 10,000 cP (Fig. 2A). Similar photooxidation-induced viscosities were observed during the Deepwater Horizon oil spill, where two oil slicks (“Slick A” and “Slick B”) had viscosities of 6,300 and 92,000 cP, respectively (at 30 °C; likely measured as emulsions).⁷⁰ Slick B (sample “SO25” in Table S5) had a modeled sea surface residence time of 8 to 9 days.¹³ After de-watering to break up emulsions, we measured an oil viscosity of 37,000 cP for this sample (at 20°C). Based on this limited observation, it appears that viscosities obtained in our mesocosm experiment match those of field-irradiated samples during the Deepwater Horizon oil spill after a comparable irradiation time.

Another way to interpret our results is using Baffled Flask Test-dispersant effectiveness values as conservative estimates for dispersant application for time scales and oil film thickness that are relevant in the field. In the field, target slicks are typically brown or black, which correspond to a thickness >100 μm.⁷¹ Thinner films are not targeted because dispersant droplets delivered by aircraft tend to penetrate through them and immediately transfer surfactants to the water column rather than the oil. Even without taking photooxidation into account, most guidelines recommend a time window for dispersants application of 2–4 days, after which oil is more difficult to disperse due to the formation of water-in-oil-emulsions.⁶ In our test, sunlight exposure of 200 to 350 μm films of HOOPS oil (original viscosity 6 cP) for three days still had Baffled Flask Test-dispersant effectiveness values larger than 60%, compared to 87% for the fresh oil. Considering that bench-top tests produce a lower dispersant effectiveness than large-scale tank tests^8,52 or field conditions (Fig. 6), our results imply that even though photochemistry will change the chemical composition and physical properties of a low-viscosity oil slick that would be targeted by aerial dispersion application, the slick would still be amenable to chemical dispersion when applied within the typical “window of opportunity,” which is within 2–4 days of the onset of an oil spill.

Synopsis:

Oil photooxidation can reduce dispersant effectiveness, but typical oil slicks will still be dispersible during marine oil spills.