The presence of higher-MW, less soluble PAHs (i.e., chrysenes, benzo[e]pyrene) in excess of their computed solubilities is used to quantify the microdroplet phase. It is critical to have high-quality subcooled liquid solubility data for these compounds. Subcooled liquid solubilities estimated with SPARC at standard temperature and pressure compare favorably with available literature values (Fig. 3) 27–31, in particular with the values from Mackay et al. 26, a critically reviewed database. The available data span the range of compounds commonly found in crude oils from MAHs such as benzene and toluene to lower-MW PAHs such as naphthalene and phenanthrene to higher-MW, less soluble PAHs such as chrysene and benzo[a]pyrene.
Figure 3. Comparison of literature values and SPARC-calculated subcooled solubilities for commonly measured monocyclic aromatic (○), polycyclic aromatic (●), and saturated () hydrocarbons. Median and range of measurements are presented.
Download figure to PowerPoint
The physicochemical property estimates with SPARC compare favorably to measurements of subcooled solubilities for saturated aliphatic chemicals 32, 33. This is an important feature of the model because the presence of long chain aliphatic hydrocarbons in water is a strong indication of the presence of microdroplet oil as the solubility of these materials is very low (e.g., <10 ng/L). Reported subcooled solubilities 32, 33 for aliphatics range from approximately 900 µg/L for a C9 paraffin to approximately 0.010 µg/L for a C19 aliphatic. Of the 26 total entries for aliphatics, 10 are within a factor of 2, another 12 are within a factor of 5, and the remaining 4 are within a factor of 30. No detectable bias was apparent in the model-data comparisons along the range of C9 to C19 aliphatics, though the greatest deviations occur with the highest–carbon number chemicals. This suggests that additional work may be required to refine the solubility estimates for high–carbon number aliphatics whose behavior in aqueous environments is complex 34, 35. Because both the model and measured solubilities are very low and the model-data comparisons do not have any particular bias, the error in the solubility estimates of these materials is not likely to substantially impact the microdroplet estimates. Furthermore, microdroplet concentrations are determined by fitting the entire suite of hydrocarbons (see Methods, Eqns. 4–8), so the impact of the model error for any one constituent is moderated by the model performance for other constituents.
The SPARC-estimated subcooled liquid solubilities for the two least soluble PAHs in this model (benzo[a]pyrene and benzo[e]pyrene) are 1.93 and 1.94 µg/L and compare quite favorably with the values reported by Mackay et al. 26 of 2.11 and 2.14 µg/L, respectively. The SPARC estimates for other PAHs are within a factor of 2 of the values found in Mackay et al. 26 for a total of 19 compounds including 11 parent and three alkyl PAHs. A similar level of agreement was found for the five MAHs in the data set. Consequently, the SPARC estimates were used for all of the 46 compounds to maintain internal consistency with other SPARC-derived properties in this solubility model.
Highly soluble constituents found in WAFs are not affected by the presence of microdroplet oil since the relative concentrations of these constituents in droplet oil is small relative to their solubility in water. Therefore, due to their low solubility, only compounds with log(KOW) >5.3 were used to estimate microdroplet oil concentrations. An example of the computed dissolved fraction (Eqn. 10) for a WAF 21 is shown in Figure 4A. The estimated dissolved fraction for each PAH (Eqn. 10) is plotted against the log(KOW) for each PAH in a WAF. At a log(KOW) of approximately 5.3, the fraction dissolved is near 0.75 and decreases to 0.1 at a log(KOW) approximately equal to 7, indicating that concentrations of compounds with a log(KOW) >5.3 are more likely affected by entrained microdroplets of oil. The estimated microdroplet concentration for this sample was 68 µg/L.
Figure 4. Determination of log(KOW) cutoff for oil droplet concentration estimates. (A) The fraction of dissolved hydrocarbons (monoaromatic hydrocarbon [MAH] and polycyclic aromatic hydrocarbon [PAH]) in a water-accommodated fraction (WAF) 21 with 68 µg/L of microdroplets (CMD) versus log(KOW). (B) The ratio of measured to predicted total aqueous PAH in an oiled-gravel test 17 from the following selected sampling times: 13 d (●), 22 d (▿), 26 d (♦), and 38 d (□).
Download figure to PowerPoint
The data sets evaluated in the present study include standard WAFs and oiled-gravel tests. Due to the longer-term (>10 d) oil–water contact in the oiled-gravel experiments, which do not benefit from static renewals such as in WAF, biodegradation can be an important process. This process is not included in this model; as a result, the ratios of predicted total (droplet and dissolved phases) to measured total aqueous concentrations of individual constituents are quite small (<0.001–0.1) for compounds with log(KOW) <5.3, whereas for compounds with log(KOW) >5.3 the ratios are within a factor of 2 of the model prediction (Fig. 4B). Additionally, it appears that compounds with log(KOW) <5.3 are more prone to loss by biodegradation, as evident by the decreasing ratio with increasing duration. Therefore, the microdroplet concentration was estimated using only compounds with log(KOW) >5.3 based on the observation that the concentration of low-solubility constituents is most affected by the presence of microdroplets and that biodegradation has a more substantial effect on the more soluble constituents.
In the calculation of dissolved hydrocarbon concentrations (Eqn. 11) MWavg is important and substantially affected by MWResidual. The MWResidual value was determined by calibration to a data set of slow-stir, temperature-controlled WAFs performed for the present analysis.
The predicted versus observed total aqueous concentrations for all the measured components are presented in Figure 5. The top panels show the temperature series for naturally weathered oil calibrations and the lower panels are the neat oil calibration. The estimated MW of the residual fraction is higher in weathered North Slope Alaskan crude oil than neat oil, and the microdroplet concentrations in these test solutions varied. However, the higher viscosity of naturally weathered oil seems to result in a smaller concentration of oil droplets compared to neat oils (Table 2), though the processes that control the formation of oil droplets in these test systems are not well understood.
Figure 5. Calibration data set (predicted vs measured total polycyclic aromatic hydrocarbon [PAH] and monoaromatic hydrocarbon [MAH] concentrations) for naturally weathered (top panels) and neat (bottom panels) Exxon Valdez crude oil at three temperatures (5, 15, and 25°C). Data are from 21.
Download figure to PowerPoint
Table 2. Summary of calibration data set
| ||MWResidual (g/mol)||Microdroplet concentrations (µg/L)|
An example of the observed enrichment of measured high-log(KOW) compounds relative to the predicted dissolved concentrations is shown in Figure 6, which compares the predicted and observed total (i.e., sum of microdroplet and dissolved phases) concentrations of aqueous PAHs and MAHs and aliphatics for WAFs prepared with weathered crude oil in panels A and B 19 and a fresh oil in panels C and D 21. The data in panels C and D are the same measured data that are shown in Figure 1. Data-model comparisons shown with open symbols (upper panels A and C) do not consider the presence of microdroplets. The comparisons shown with filled symbols include the microdroplets. Measured values >10 µg/L are MAHs and are close to the predicted values; they show little impact of the concentration of microdroplets. Compounds with measured aqueous concentrations less than 0.5 to 0.002 µg/L are 3+ ring PAHs, including parent and alkyl (C1–C4 homologs) phenanthrene, chrysene, fluoranthene, and others. The lowest two points are benzo[b]fluoranthene and benzo[ghi]perylene, with measured aqueous concentrations near 0.002 µg/L. The ratio between the predicted dissolved and measured aqueous concentrations for the 3+ ring PAHs range from 0.1 to 0.001, which is similar to the predicted dissolved concentrations of the aliphatics (stars in panel A) that are, also, more than 1,000-fold lower than the measured concentrations.
Figure 6. Demonstration of solubility calculations with predicted dissolved-phase (open symbols) and predicted total, for example, microdroplet-adjusted, (filled symbols) aqueous concentrations of monocyclic aromatics (▵, ▴), dicyclic aromatics (◊, ♦), 3+ ring aromatics (○, ●), and aliphatics (), open symbols (A, C), and with adjustments for oil droplets, filled symbols (B, D). A and B contain data from a water-accommodated fraction (WAF) with weathered Exxon Valdez cargo crude oil 19. C and D contain the same measured data as Figure 1 from a WAF with fresh oil 21. Data are in Supplemental Data, Table S1.
Download figure to PowerPoint
The inclusion of a microdroplet phase (filled circles) dramatically improves the agreement of predicted and measured concentrations in both panels B and D. The estimated microdroplet oil concentration (CMD) was 374 µg/L of oil for the WAF with weathered oil and 430 µg/L for the WAF with fresh oil. The aqueous-phase concentration of each PAH is increased proportionally to the microdroplet concentration and the concentration of the individual oil-phase constituents. The composition of the higher-MW components in the WAF is similar to the composition of the parent oil, consistent with the presence of microdroplet oil in the WAF. The inclusion of microdroplet oil in the predicted total concentrations allowed us to explain the observed enrichment in both the 3+ ring PAHs and the sparingly soluble aliphatics.
As mentioned, the presence of aliphatic hydrocarbons in water is a strong indication of stable microdroplets. Also, the enrichment in the observed heavy PAHs relative to the model predictions is another strong indication. An alternate explanation might be the presence of dissolved organic carbon, which acts as a third phase that can show enrichment in the heavy PAHs. This hypothesis was considered insufficient to explain the observations for PAHs and SHCs. For example, the fraction of dissolved chrysene (log(KOW) 5.8) in the presence of typical dissolved organic carbon concentrations (2–10 mg/L) is expected to be around 90% (Eqn. 1 in 36), which is not sufficient to explain the observed enrichment in the model-data comparison that can be as high at 1,000-fold. Extrapolation of the dissolved organic carbon partitioning model to long chain aliphatics is uncertain due to a lack of validation data. Therefore, because the presence of microdroplets in oil–water preparations is considered likely, the modeling framework presented here provides a method for systematically estimating microdroplet concentrations and the fraction of dissolved material that is measured by chemical analysis.
Sensitivity analysis on microdroplet estimation method
Measurements of aqueous and oil-phase petroleum hydrocarbons, including saturates from an acute oil toxicity exposure to juvenile fathead minnows 19, were used to validate the microdroplet estimation method. For weathered and neat Exxon Valdez crude oil, WAFs were prepared and analyzed for concentrations of MAHs and PAHs in addition to a series of saturated C9 to C40 hydrocarbons. Saturated hydrocarbons are very insoluble in water, and their presence is a strong indication of the presence of colloidal oil in the exposure waters 37. Three different microdroplet concentration estimation methods were compared. The first used only PAHs with log(KOW) >5.3, the second approach used only SHCs (e.g., alkanes), and the third approach combined all three chemical classes. The three methods give similar microdroplet estimates (Table 3). The droplet estimates based on PAH-only data are approximately within a factor of 2 of the estimates using SHCs. These results support the use of measured aqueous PAHs if these are the only data available and demonstrate the usefulness of measuring long chain saturates in exposure waters.
Table 3. Summary of oil droplet concentration estimates using different approaches (µg/L)
The median microdroplet concentrations in the highest oil loadings among all studies range from 567 to 64 µg/L, with smaller to nonquantifiable concentrations at lower loadings (Table 4). In flow-through oiled-gravel tests, the effluent from the first few days of use had higher microdroplet concentrations that diminished quickly during the 10 to 20 d until microdroplet concentrations are no longer quantifiable. For example, data from each of the four oil-loading levels in Carls et al. 17 over the duration of the experiment are shown in Figure 7 as time-variable oil droplet concentrations in the aqueous effluents. The individual data series decrease generally with decreasing loading (see caption). Generally, higher microdroplet concentrations are observed in the higher oil-loading tests, which remain relatively constant during the first 21 d, corresponding to the LWO portion of the test, after which the concentrations begin to decline. This method estimated that there were no microdroplets in the effluents of the lower loadings by the end of the experiment. These patterns are consistent with the droplet estimates in the other oiled-gravel tests 18, 20 evaluated in the present study.
Table 4. Summary of microdroplet results and other calculated parameters
|Study ID||Median microdroplet concentration at highest loading (µg/L)a||Residual fractionb (%)||Weighted-averaged molecular weight (g/mol)||Source|
Figure 7. Time-variable oil droplet concentration (CMD) patterns in oiled gravel 17. Filled symbols are less-weathered oil (LWO) and open symbols are more-weathered oil (MWO) portions of study for high (), medium (▵, ▴), low (▿, ▾), and trace (○, ●) oil loadings. Microdroplet oil droplet concentration estimates at or below the method detection limit are also plotted (<).
Download figure to PowerPoint
Figure 8 summarizes the median and range of microdroplet concentrations of the highest oil loading for the various studies analyzed. Though the magnitude and occurrence of the microdroplets varied widely, they were estimated to be present in many different testing environments (i.e., standard WAF preparations to oiled-gravel exposures). This suggests that microdroplets were present in most crude oil exposures. The microdroplets appear to be stable, having remained in suspension even though there were several hours to days between sample collection and analysis.
Figure 8. Summary of microdroplet concentration estimates in highest loading or dilution stream from the studies in this analysis (Table 1). Symbols are plotted at median microdroplet concentration, whiskers represent the range, and the box is the inner two quartiles ( ±25%) of the concentration estimates over the duration of the experiment.
Download figure to PowerPoint