Note from the committee: The work described in this appendix was commissioned by the committee with the purpose of better evaluating droplet size models—a key tool in modeling the oil fate in a subsea release. As discussed at length in Chapter 2, droplet models have been compared to varying degrees to laboratory experiments and one field study, but major questions remain as to how well these models scale to field conditions. Shortly after the committee began its work, Gros et al. (2017) published a dataset derived from the Deepwater Horizon measurements, which, after close review, the committee felt represented a reasonable benchmark for testing droplet models. Gros et al. (2017) only looked at two scenarios (0 and 0.4% dispersant-to-oil ratio [DOR]) of subsea dispersant injection using the VDROP-J model. The committee wanted to extend their original work by running the model with a DOR of 1%; a value which recent work suggests is much more optimal (Brandvik et al. 2014b). It also ran cases using droplet distributions by Spaulding et al. (2015, DWH NRDA) as well as a distribution predicted by the C-Image Consortium. Results from the consultants’ report are discussed in Chapter 6.
Report on Deepwater Horizon Simulations with Different
Droplet Size Distributions
Scott A. Socolofsky
Zachry Department of Civil Engineering,
Texas A&M University, College Station, TX 77845, USA
Zachry Department of Civil Engineering,
Texas A&M University, College Station, TX 77845, USA,
presently at GEOMAR Helmholtz Centre for Ocean Research, 24148 Kiel, Germany
We have completed simulations for the Deepwater Horizon blowout using different choices of subsea dispersant application rate and bubble and droplet size prediction models, as prescribed by the Committee on the Evaluation of the Use of Chemical Dispersants in Oil Spill Response. Each of these runs makes predictions for June 8, 2010, using our simulation package, the Texas A&M Oilspill Calculator (TAMOC) (Gros et al., 2017) and our 279 pseudo-component model of the Deepwater Horizon reservoir fluid (Gros et al., 2016). Model output includes dissolved concentrations in the deepwater intrusion layer, mass flow rate of compounds to the air/water interface, and water column concentrations between the intrusion layer and sea surface. All model parameters except for bubble and droplet size and interfacial tension matched those used in our paper (Gros et al., 2017), published in the Proceedings of the National Academy of Sciences of the United States of America (PNAS).
Table E.1 presents a description of each case simulated in this study. We report here the model results for all cases in Table E.1; Case 3 was not defined and has been deleted. Each case in the table represents a different approach for predicting the droplet size distribution and/or a different assumed DOR. For some methods (Cases 1 and 5), only the median droplet size d50 is provided by
TABLE E.1 Description of the Cases and Size Distributions Used in the Simulations
|Case||d50 [mm]||[—]||DOR [%]||dmax Rule||Micro Droplets [%]||Description and Source|
|1||10||0.5||0||d = dmax if the predicted distribution has d > dmax||0.2||Untreated upper limit of droplet size distribution|
|2||vdrop-j||vdrop-j||0||N/A||0.5||Untreated case already simulated and reported in Gros et al. (2017)|
|4||vdrop-j||vdrop-j||0.4||N/A||1.1||Treated case (hindcast value) already simulated and reported in Gros et al. (2017)|
|5||0.17; 3.3||0.5||30% treated; 70% untreated||N/A||–16.7||Bimodal distribution assumed using partial dispersant mixing|
|6||c-image||N/A||0||N/A||–1.7||Untreated case provided by C-IMAGE. Number size distributions were provided, which we converted to volume size distributions|
|7||vdrop-j||vdrop-j||1.0||N/A||–25.2||Treated case with optimal DOR|
the method. In these cases, the size distribution is estimated from a log-normal distribution of log standard deviation σN provided in the table. For Case 1, this approach predicts droplet sizes larger than the maximum stable droplet size. For this case, we truncate the distribution in order to retain the desired d50.
This report is organized as follows. Sections 1 to 6 report the results of each simulation case using the data comparison and analysis formats already reported in Gros et al. (2017). These sections compare measured and predicted fractionation indices for several compounds in the deepwater intrusion layer and at the sea surface. These sections also report the petroleum fluids mass balance among the deepwater intrusion, mid-ocean water column, and the sea surface. Throughout this report, we consider the term petroleum fluids to mean the whole reservoir fluid, including the C1-C5 compounds. Section 7 explains our method to compute dissolved concentrations between the intrusion layer and the sea surface, where individual Lagrangian bubbles and droplets transit the water column. This section also shows sample results for benzene. We conclude this report with a short discussion in Section 8.
1. Case 1: Untreated Upper-Limit of Droplet Size Distribution
Figure E.1 shows the gas bubble sizes (top panel) and oil droplet sizes (lower panel) used in the simulations for Case 1, designed as the untreated upper-limit of the droplet size distribution. The gas bubble sizes were computed using the empirical equations in Wang et al. (2018) with the gas properties computed by our model without dispersant addition. The oil droplet sizes have a d50 = 10 mm, and the distribution is truncated at the maximum stable droplet size (Clift et al., 1978). The oil droplet size distribution was truncated because any other redistribution of oil mass would
change the prescribed d50-value of the distribution. Because the maximum stable bubble size is quite large, the bubble size distribution did not have to be truncated.
We show the model predictions compared to measured field data in Figure E.2, using the same figure format as model-data comparisons in Figure 3 of our PNAS paper (Gros et al., 2017). This figure and all similar figures to follow are organized as follows. Panels A and B report the model predictions for the combined dissolved and liquid petroleum and compare the model values to measured data from Conductivity, Temperature, and Depth casts within 10 km of the wellhead. See the Supporting Information for Gros et al. (2017) for the details of the sources and time periods of the field data used here for comparison; we considered all data available within the 10 km radius of the wellhead and representative of conditions on June 8, 2010. Panel A reports the fractionation index relative to methane for three constituents of the released petroleum in the subsurface intrusion layer (900 m to 1,300 m water depth). Panel B reports fractionation indices relative to benzene in the intrusion layer for several additional selected components of the simulated oil. In Panel B, the blue bars represent the raw model output and the orange bars represent the model output with a fixed, constant fraction of microdroplets added to the intrusion layer. The fraction of microdroplets is adjusted for each case separately to achieve the best possible agreement between the model and the measurements for the 16 sparingly-soluble compounds from phenanthrene to pristane. This fraction of microdroplets is reported for each case in Table E.1. In Panel C, we report results for the fraction of spilled petroleum arriving at the sea surface in comparison to measurements by Ryerson et al. (2012).
The results for Case 1 show that less of the released oil entered the intrusion as dissolved or liquid petroleum than was observed (under-prediction of model results in Panels A and B). Instead, more volatile organic compounds are predicted to reach the sea surface than observed (over-predictions of model results in Panel C). These results are also summarized in Figure E.3, which reports the model prediction of the petroleum mass budget throughout the water column for Case 1 without the addition of the fitted microdroplets (e.g., similar to the blue bars in Figure E.2). The large oil droplets of this case dissolve slowly (low surface area to volume ratio), rise quickly (have a short time to dissolve), and bring more of the light components to the sea surface than were observed by the atmospheric measurements. Hence, this droplet size over-predicts the best-fit droplet size for our model for June 8, 2010.
2. Case 2: Untreated Sensitivity Simulation
Figure E.4 shows the gas bubble sizes (top panel) and oil droplet sizes (lower panel) used in the Case 2 simulations. These sizes were predicted from the VDROP-J model using the gas and oil properties we predict if no dispersants were used. This case was already simulated and reported in our paper (Gros et al., 2017). Because dispersants were injected on this day, this case represents a hypothetical sensitivity study with respect to dispersant effectiveness.
We show the model predictions compared to measured field data in Figure E.5, using the same figure format as in Figure E.2. In this case, the simulation under-predicts the fraction of released petroleum entering the intrusion layer. By adding microdroplets, the simulation results can match observations for insoluble components of the oil, but the lighter compounds remain under-predicted in the intrusion layer, suggesting less dissolution is occurring in the model than was observed. This is consistent with an over-predicted droplet size, and is corroborated by the predictions at the sea surface, which show more oil reaching the surface in the model than the observations. The petroleum mass budget predicted by our model for this case is shown in Figure E.6 without the addition of the fitted microdroplets. Like Case 1, this droplet size over-predicts the best-fit droplet size for our model for June 8, 2010.
3. Case 4: Best-Case Hindcast Simulation
Figure E.7 shows the gas bubble sizes (top panel) and oil droplet sizes (lower panel) used in the Case 4 simulations. These sizes were predicted from the VDROP-J model using the gas and oil properties we predict for a DOR of 0.4%, our estimate of the actual DOR during this day assuming full mixing of dispersant with the quantity of oil exiting the broken Macondo wellhead. This case was already simulated and reported in our paper (Gros et al., 2017). Because dispersants were injected on this day, this case represents our best-case hindcast of the behavior of the Deepwater Horizon oil spill on this day using our model.
We show the model predictions compared to measured field data in Figure E.8, using the same figure format as in Figure E.2. In the intrusion layer (Panels A and B) the model predicts the fractionation of light, soluble compounds well, with some constituents over-predicted (e.g., toluene, naphthalene) and others under-predicted (o-xylene, cyclohexane). With a small quantity of fitted microdroplets (1.1%, see Table E.1), the results for insoluble compounds are close to the observations, and the predictions for soluble compounds are mostly unchanged. At the sea surface (Panel C), model predictions for the fraction of oil reaching the surface corresponds well with the observations, with most compounds slightly under-predicted by the model. The petroleum mass budget predicted by our model for this case is shown in Figure E.9 without the addition of the fitted microdroplets. Because these results were obtained using a DOR representative of June 8, 2010, and the model results largely show good agreement with the observations, this droplet size corresponds to the best-fit droplet size for our model for this day.
4. Case 5: Partial Mixing of Dispersant
Figure E.10 shows the gas bubble sizes (top panel) and oil droplet sizes (lower panel) used in the Case 5 simulations. These sizes were predicted using a procedure similar to that used by RPS ASA used for June 8, 2010, in the NRDA documents and that was also reported in Li et al.
(2016). Following this method, dispersant is assumed to only treat some of the oil, and for Case 5, we assumed that the dispersant treatment fraction was 30%. The size distribution is taken as the sum of the distributions for the treated fraction (with d50 of 0.17 mm, see Table E.1) and untreated fraction (with d50 of 3.3 mm, Table E.1). Each size distribution was assumed log-normal with a log standard deviation = 0.5. Gas bubble sizes are computed using the formula in Wang et al. (2018) with an interfacial tension reduction factor of 5.4 times.
We show the model predictions compared to measured field data in Figure E.11, using the same figure format as in Figure E.2. In this case, the simulation significantly over-predicts the fraction of released petroleum entering the intrusion layer, especially for the insoluble compounds. The slanted black lines for the fluorene predictions indicate that the actual model value plots above the present y-axis. By removing microdroplets (–16.7%, see Table E.1), the simulation results can match observations for insoluble components of the oil, but the lighter compounds remain over-predicted in the intrusion layer, suggesting more dissolved petroleum is entering the intrusion in the model than was observed. The results at the sea surface (Panel C) also show a slight over-prediction relative to the measured data of the mass flow rate to the surface of the lighter compounds, indicating somewhat less dissolution overall occurring in the model throughout the water column than observed. The petroleum mass budget predicted by our model for this case is shown in Figure E.12 without the addition of the fitted microdroplets.
We can draw several conclusions from these results. Because the model predictions in the intrusion layer over-predict the observations, the modeled fraction of small droplets was over-predicted. When we subtract microdroplets, the predictions for soluble compounds remain over-predicted;
hence, the mass fraction of oil in the small droplets was over-predicted, yielding more dissolution than observed. On the other hand, the model predicts too much of several volatile organic compounds (VOCs) reaching the surface, which is consistent with over-predicting the mass fraction of large droplets for the untreated fraction of the distribution. Together, these results suggest to us that the hypothesis of incomplete dispersant mixing is not supported by these simulations: there are too many small droplets and large droplets for this type of distribution to produce results that fit the observations using our model. We conclude instead that all of the oil was treated somewhat (which would give a smaller maximum droplet size and less VOCs reaching the surface), and this more dilute treatment reduced the dispersant effectiveness (which would give a larger minimum droplet size and less dissolved compounds sequestered in the intrusion layer). Such a situation agrees with the observed fractionation indices and is close to the case of complete dispersant mixing across the plume (e.g., Case 4, above).
5. Case 6: Untreated Size Distributions Prescribed by C-IMAGE
Figure E.13 shows the gas bubble sizes (top panel) and oil droplet sizes (lower panel) used in the Case 6 simulations. These sizes were predicted by researchers in the C-IMAGE consortium and provided to us by Steve Murawski as number size distributions. We converted these number distributions to equivalent volume size distributions, and these latter volume size distributions are shown in the figure herein. Because dispersants were injected on June 8, 2010, this untreated case represents a hypothetical sensitivity study with respect to the true hindcast, which would include the effect of dispersant injection.
We show the model predictions compared to measured field data in Figure E.14, using the same figure format as in Figure E.2. In this case, the simulation strongly over-predicts the fraction of released petroleum entering the intrusion layer. By removing microdroplets (–1.7%, see Table E.1), the simulation results can match observations for the most insoluble components of the
oil (right half of Panel B), but the lighter compounds remain over-predicted in the intrusion layer, suggesting more dissolution is occurring in the model than was observed. This is consistent with an under-predicted droplet size, and is corroborated by the predictions at the sea surface, which show much less oil reaching the surface in the model than the observations. The petroleum mass budget predicted by our model for this case is shown in Figure E.15 without the addition of the fitted microdroplets. Based on these data, this droplet size under-predicts the best-fit droplet size for our model for June 8, 2010.
6. Case 7: Hypothetical Optimal Dispersant Treatment Scenario
Figure E.16 shows the gas bubble sizes (top panel) and oil droplet sizes (lower panel) used in the Case 7 simulations. These sizes were predicted from the VDROP-J model using the gas and oil properties we predict for a DOR of 1.0% and assuming full mixing of dispersant with the quantity of oil exiting the broken Macondo wellhead. VDROP-J predicts droplet sizes in bins of 100 µm most of the bubble and droplet sizes were small for this case, which is why so few bins were used in the simulations. This DOR is higher than actually occurred on June 8, 2010, and represents an optimal DOR to achieve greater dispersant effectiveness at a low DOR. Hence, this case represents a hypothetical estimate of what might have resulted were a higher DOR used at the Deepwater Horizon oil spill on this day.
tion of released petroleum entering the intrusion layer. By removing microdroplets (–25.2%, see Table E.1), the simulation results can match observations for the most insoluble components of the oil (right half of Panel B). This suggests that up to 25% more of the released insoluble compounds could have been funneled to the deep intrusion if this higher DOR were used. Additionally, these small droplets dissolve faster, and the predictions with and without microdroplets over-predict the fraction of soluble petroleum entering the intrusion, also increasing the total amount of petroleum sequestered in the deep ocean. This is consistent with the predictions at the sea surface, which show almost none of the compounds of the oil plotted in Panel C reaching the surface in the model. The petroleum mass budget predicted by our model for this case is shown in Figure E.18 without the addition of the fitted microdroplets. Hence, this droplet size distribution causes a significantly greater fraction of the released petroleum to be sequestered in the ocean than occurred at the actual DOR of 0.4% (Case 4, above).
7. Concentration in the Water Column Above the Main Intrusion
The simulations for each of the cases reported above tracked the concentrations of dissolved hydrocarbons and the masses of gas and liquid petroleum in the initial nearfield plume and the deepwater intrusion layer and the masses of gas and liquid petroleum in the water column between the intrusion layer and the sea surface. For injury assessment, the concentrations of dissolved hydrocarbons throughout the ocean water column are needed. These are not immediately available because oil droplets and gas bubbles rise as individual Lagrangian particles between the deep intrusion layer and the surface, and there is no associated control volume of seawater to use to track the dissolved concentrations. To predict these concentrations, we developed a new model for the dissolved phase concentration associated with a stream of Lagrangian particles.
Our new model for dissolved-phase concentration in the mid-ocean water column is based on a solution to the advection diffusion equation in seawater with the source term of dissolved mass
coming from the on-going dissolution of the Lagrangian particles. TAMOC predicts the steady-state dissolution for each Lagrangian particle by the mass balance equation
where m! í is the steady-state mass flow rate per unit length of chemical component i in Lagrangian particles of a given size, A is the surface area of the particles, β is the mass transfer coefficient for chemical component i, n! is the number flux of bubbles in the present particle class, β is the slip velocity of the particles, Cs,i is the solubility of component i at the bubble-water interface, and Ca,i is the concentration of component i in the ambient water, far away from the bubble, taken as zero for petroleum compounds.
If we use this mass flow rate m! í as the source term in the advection diffusion equation, an analytical solution exists if we assume that m! í is constant with height (an assumption that is approximately valid locally at each water depth) and that the horizontal transport is in the advection-dominant regime. The analytical solution in this case for a uniform crossflow of velocity U is
where E is the turbulent eddy diffusivity, x is taken parallel with U, and the source is injected at the coordinate (0,0). The model in Equation (2) is valid in the advection-dominated regime, at a location x, a distance larger than L downstream of the bubble stream, where
and α is a parameter, greater than 10. Our observations of turbulent diffusivity in the deep Gulf of Mexico give values of Et = 5 × 10–4 m2/s (Wang et al., 2016); hence, L is a very short distance downstream at typical ocean currents in the Gulf of Mexico of between 2 and 30 cm/s.
We compute these concentrations for each chemical component in the gas bubbles and oil droplets above the intrusion layer using our full 279 pseudo-component model of the Deepwater Horizon reservoir fluid. We compute all concentrations at a distance 10 km downstream of the broken Macondo wellhead, and the total concentration of a given chemical component is the superposition of the contributions from each bubble and droplet size in each simulation. Equation (2) is valid if the particles in a stream of bubbles or droplets do not spread out. However, in reality, the particles are already distributed across the plume width σ0 before exiting the intrusion layer and continue to spread by horizontal turbulent diffusion between the intrusion layer and a height z where Equation (2) is evaluated. We account for this spreading by computing Equation (2) for 1,000 particles Gaussian-distributed over the predicted cloud width, given by
where t is the travel time for a given particle from the intrusion layer to z. Because most of the spreading occurs in the plume (i.e., σ ≈ σ0 ), our results are weakly dependent on the value of Et. The total concentration is the superposition of the contributions for each of the 1,000 simulated bubbles or droplets.
Figure E.19 shows a sample result from the simulation matching the parameters of Case 4, our best-case hindcast simulation, for benzene. In order to capture the input from each of the modeled bubbles and droplets, we make the calculation at a distance x = 10 km downstream of the Deepwater Horizon wellhead. The colormap in Figure E.19 shows the computed concentrations on the yz-plane normal to the currents at this location.
The currents change direction with height, and at each depth, we assume that all bubbles are aligned on a single x-axis, parallel with the currents. Figure E.20 plots the currents we used for all our simulations, which were measured near the Deepwater Horizon wellhead by a near-surface, down-looking acoustic Doppler current profiler (ADCP) (Gros et al., 2017). Our assumption that the particles are always aligned along the currents yields the largest prediction for the dissolved concentrations. Comparing to Equation (2), we also see that concentrations are maximum where the dissolution m! í is large and/or where the currents U or the plume width σ are minimum. Because
this is a steady-state solution, concentrations increase as U → 0 , and because σ ≈ σ0 , the plume width is fairly constant. Based on these behaviors and Figure E.20, we expect the highest concentrations near the intrusion, where the currents are minimum and the oil and gas is the freshest (least weathered). The distribution of concentration in the map agrees with these expectations.
In Figure E.21, we present the profiles benzene for Case 4 at three different depths at a distance of x = 10 km downstream of the Deepwater Horizon wellhead, hence, profiles extracted from Figure E.19. Peak concentrations occur low in the water column, close to the intrusion layer, and are on the order of 10–7 kg/m3 (10–9 mol/l) over a slice about 400 m wide. The width of the concentration cloud is also largest at 850 m depth and decreases slightly with height. The narrower concentration plume at shallower depth is due to the currents: Oil droplets advect farther downstream at shallower depths; hence, the dissolved plume experiences less lateral diffusion between the location where the droplets are dissolving and the plane at 10 km downstream where the concentrations are evaluated as the depth reduces. The peak benzene concentrations also decrease with decreasing depth, dropping to 10–8 kg/m3 (10–10 mol/l) over a slice about 300 m wide at 400 m depth.
Similar behavior is observed for each of the dissolving components, with the values of the concentration depending on the solubility of each component in the Lagrangian bubbles and droplets as their compositions evolve with height. The properties of the 279 pseudo-component oil model as well as results of each of these concentration calculations in the form of ascii text files are included in the digital appendix to this report.
Herein, we have reported the results of our TAMOC simulations for the Deepwater Horizon accident on June 8, 2010, for six different prescribed subsea dispersant application rates and bubble and droplet size prediction models. We had previously reported the results of Cases 2 (no subsea dispersant injection) and 4 (reported subsea dispersant injection) using VDROP-J to predict bubble and droplet size distributions in our PNAS paper (Gros et al., 2017). Here, we also simulated a new case, Case 7, in which VDROP-J is used to predict bubble and droplet sizes for a theoretical optimal subsea dispersant injection rate of 1% DOR.
Figure E.22 compares the results of these three cases in the format of Figure E.2, together with the observed data. This figure shows that, using our model, Case 4 gives a prediction that is closest to the observations, and Case 7 predicts that a significantly greater proportion of the released petroleum would have been sequestered in the ocean within a radius of 10 km of the Deepwater Horizon wellhead had a higher DOR been applied subsea. From Figure E.18 for Case 7, 52% of the released petroleum would either be dissolved in the intrusion layer or the water column or suspended as small droplets in the subsurface intrusion. Figures E.9 and E.6 show that the actual DOR used during Deepwater Horizon is predicted to have sequestered 27% of the released petroleum (Case 4), and that if no dispersants had been used, 22% of the released petroleum would have been sequestered (Case 2).
These different dispersant injection rates also change the concentrations of light petroleum hydrocarbons in the mid-ocean water column. In Figure E.23, we plot the maximum concentration of benzene between the intrusion layer and sea surface at 10 km downstream of the broken Macondo wellhead for these cases using our new method in Section 7. The maximum concentrations are computed at y = 0 in Equation (2) and plotted as a function of depth z. The differences between the baseline case of 0.4% dispersant injection (Case 4) and no-dispersant (Case 2) are about an order of magnitude throughout the water column, with Case 4 having higher concentrations due to the larger amount of benzene dissolution with dispersant injection. The 1,000-fold reduction in atmospheric benzene emissions between these two cases results from the integral of this order-of-magnitude difference in benzene concentration over the full water depth. Similarly,
Case 7, with 1% dispersant injection, results in even greater benzene dissolution through the water column and nearly two orders-of-magnitude higher benzene concentration in the mid-ocean water relative to the baseline, Case 4. All of these benzene concentrations remain small, however, with maximum values less than 4 × 104 kg/m3 (5.1 µmol/l, or 0.4 parts per million).
The results for Cases 2 and 4 show that the actual amount of petroleum reaching the sea surface is fairly similar with and without subsea dispersant injection for the DOR used on June 8, 2010 (73% with subsea dispersant injection and 78% without). However, as we show in Gros et al. (2017), the composition of the surfacing petroleum is quite different due to the different dissolution occurring for the smaller droplets with dispersant injection. We previously reported that the mass flow rates of C1-C9 VOCs to the atmosphere reduced by 28% between Cases 2 and 4, and that the mass flow rate of the compound benzene reduced by 2,000 times with subsea dispersant injection (Case 2 compared to Case 4).
Similar results for other cases simulated here can also be extracted from the model results. Comparing the optimal DOR of 1% (Case 7) to the case of no dispersant injection (Case 2), we predict a reduction of C1-C9 VOCs of 84%, and that no benzene reaches the sea surface (e.g., infinite reduction of benzene mass flow rate to the atmosphere). Likewise, if we compare Case 1 (maximum possible droplet size) to Case 2 (VDROP-J no dispersant case), the flow rate of C1-C9 VOCs differs by only 10% (Case 1 being higher), and the benzene mass flow rates differ by 2.6 times (Case 1 being higher). If we compare Case 5 (droplet sizes following partial dispersant mixing) to Case 2, C1-C9 VOC emissions are reduced by 30% in Case 5 and benzene emissions reduce by two times. Hence, for the partial dispersant mixing model of subsea dispersant injection, one would not conclude that dispersants significantly affected atmospheric concentrations of VOCs. Finally, comparing Cases 6 (C-IMAGE size distribution) to Case 2, C1-C9 VOC emissions are reduced by 66% and benzene mass flow rate to the atmosphere is nearly suppressed (3 × 106 times reduction). This is similar to performance of the optimal subsea dispersant injection predictions using VDROP-J at 1% DOR.
Because the three Cases 2, 4, and 7 are all based on the same model assumptions and because Case 4 gives the best match between the model predictions and the observations, these cases may be considered as reliable predictors for the effect of subsea dispersant injection on the Macondo oil during the Deepwater Horizon accident. The DORs used during the accident were lower than optimal, but resulted in some liquid oil not reaching the sea surface and significantly improved air quality by suppressing atmospheric emissions of VOCs by 28%, including a 2,000 times reduction of benzene emission. Had a higher DOR of 1% been used, our model predicts that significantly more liquid oil would have remained subsea within the 10 km radius we have studied surrounding the Deepwater Horizon wellhead, atmospheric emissions of VOCs would have reduced by 84% relative to the no-dispersant case, and benzene emissions could have been entirely suppressed.
Clift, R., J. R. Grace, and M. E. Weber. 1978. Bubbles, Drops, and Particles. Mineola, NY: Dover Publications, Inc.
Gros, J., C. M. Reddy, R. K. Nelson, S. A. Socolofsky, and J. S. Arey. 2016. Simulating gas–liquid−water partitioning and fluid properties of petroleum under pressure: Implications for deep-sea blowouts. Environmental Science & Technology 50(14):7397-7408. DOI: 10.1021/acs.est.5b04617.
Gros, J., S. A. Socolofsky, A. L. Dissanayake, I. Jun, L. Zhao, M. C. Boufadel, C. M. Reddy, and J. S. Arey. 2017. Petroleum dynamics in the sea and influence of subsea dispersant injection during Deepwater Horizon. Proceedings of the National Academy of Sciences of the United States of America 114(38):10065-10070. DOI: 10.1073/pnas.1612518114.
Li, Z., M. Spaulding, D. French McCay, D. Crowley, and J. R. Payne. 2017. Development of a unified oil droplet size distribution model with application to surface breaking waves and subsea blowout releases considering dispersant effects. Marine Pollution Bulletin 114(1):247-257. DOI: 10.1016/j.marpolbul.2016.09.008.
Ryerson, T. B., R. Camilli, J. D. Kessler, E. B. Kujawinski, C. M. Reddy, D. L. Valentine, E. Atlas, D. R. Blake, J. de Gouw, S. Meinardi, D. D. Parrish, J. Peischl, J. S. Seewald, and C. Warneke. 2012. Chemical data quantify Deepwater Horizon hydrocarbon flow rate and environmental distribution. Proceedings of the National Academy of Sciences of the United States of America 109(50):20246-20253. DOI: 10.1073/pnas.1110564109.
Wang, B., S. A. Socolofsky, J. A. Breier, and J. S. Seewald. 2016. Observations of bubbles in natural seep flares at MC 118 and GC 600 using in situ quantitative imaging. Journal of Geophysical Research: Oceans 121(4):2203-2230. DOI: 10.1002/2015JC011452.
Wang, B., S. A. Socolofsky, C. C. Lai, E. E. Adams, and M. C. Boufadel. 2018. Behavior and dynamics of bubble breakup in gas pipeline leaks and accidental subsea oil well blowouts. Marine Pollution Bulletin 131(Part A):72-86. DOI: 10.1016/j.marpolbul.2018.03.053.