Article pubs.acs.org/est
Evolution of the Macondo Well Blowout: Simulating the Effects of the Circulation and Synthetic Dispersants on the Subsea Oil Transport Claire B. Paris,*,† Matthieu Le Hénaff,†,‡ Zachary M. Aman,§ Ajit Subramaniam,∥ Judith Helgers,†,⊥ Dong-Ping Wang,# Vassiliki H. Kourafalou,† and Ashwanth Srinivasan†,⊥ †
University of Miami, Rosenstiel School of Marine and Atmospheric Science, 4600 Rickenbacker Causeway, Miami, Florida 33149-1098, United States ‡ University of Miami, Cooperative Institute for Marine and Atmospheric Studies (CIMAS), 4600 Rickenbacker Causeway, Miami, Florida 33149-1098, United States § Colorado School of Mines, 1500 Illinois Street, Golden, Colorado 80401, United States ∥ Lamont Doherty Earth Observatory at Columbia University, 61 Route 9W, Palisades, New York 10964, United States ⊥ University of Miami, Center for Computational Science (CCS), Coral Gables, Florida 33124, United States # State University of New York, Stony Brook, New York 11794, United States S Supporting Information *
ABSTRACT: During the Deepwater Horizon incident, crude oil flowed into the Gulf of Mexico from 1522 m underwater. In an effort to prevent the oil from rising to the surface, synthetic dispersants were applied at the wellhead. However, uncertainties in the formation of oil droplets and difficulties in measuring their size in the water column, complicated further assessment of the potential effect of the dispersant on the subsea-to-surface oil partition. We adapted a coupled hydrodynamic and stochastic buoyant particle-tracking model to the transport and fate of hydrocarbon fractions and simulated the far-field transport of the oil from the intrusion depth. The evaluated model represented a baseline for numerical experiments where we varied the distributions of particle sizes and thus oil mass. The experiments allowed to quantify the relative effects of chemical dispersion, vertical currents, and inertial buoyancy motion on oil rise velocities. We present a plausible model scenario, where some oil is trapped at depth through shear emulsification due to the particular conditions of the Macondo blowout. Assuming effective mixing of the synthetic dispersants at the wellhead, the model indicates that the submerged oil mass is shifted deeper, decreasing only marginally the amount of oil surfacing. In this scenario, the oil rises slowly to the surface or stays immersed. This suggests that other mechanisms may have contributed to the rapid surfacing of oil− gas mixture observed initially. The study also reveals local topographic and hydrodynamic processes that influence the oil transport in eddies and multiple layers. This numerical approach provides novel insights on oil transport mechanisms from deep blowouts and on gauging the subsea use of synthetic dispersant in mitigating coastal damage.
■
with hydrodynamics.9−11 Measurement and modeling of the crude oil and gas chemistry have been presented by Reddy et al.,8 while estimates of rising fluid composition and velocity were presented by Ryerson et al. 2 and Camilli et al.12 The lack of information on dispersant-to-oil ratio that contribute to the oil atomization in droplets13,14 and their settling behavior15−17 complicates the assessment of the distribution and magnitude of subsurface oil. Indeed, coming from great depth at high
INTRODUCTION On April 20, 2010, the explosion of the BP Deepwater Horizon oil drilling platform, located 90 km offshore in the Mississippi Canyon, caused a 87-day blowout of the Macondo well. During this period, crude oil and natural gas flowed into the Gulf of Mexico from a water depth of 1520 m. After the permanent closure of the Macondo relief well on September 18, nearly five months later, the question turned to the fate of up to 260 million gallons of crude oil1−4 blended at depth with 721 000 gallon of dispersant injected at the wellhead5 to presumably prevent the oil from rising to the surface. The oil transport in the water column is governed by both physical and chemical processes2,6−8 that depend on the interaction of the buoyant oil © 2012 American Chemical Society
Received: Revised: Accepted: Published: 13293
August 7, 2012 November 16, 2012 November 20, 2012 November 20, 2012 dx.doi.org/10.1021/es303197h | Environ. Sci. Technol. 2012, 46, 13293−13302
Environmental Science & Technology
Article
velocities ranging from 0.2 to 0.8 m/s,12 the experimentally based droplet size model used here predicts the largest mean particle diameter at approximately 100 μm.28 On the basis of previous laboratory data,13 we estimated a dispersion of 25% Conroe crude oil in 75% water at 400 rpm. In these conditions, the particle size diameter ranges from approximately 1 to 300 μm, arranged in a non-normal distribution with an estimated peak at ca. 50−70 μm.13,24 The application of dispersants decreases water−oil interfacial tension and results in smaller oil particles (i.e., shifting the shear/interfacial force balance), with a peak in the droplet diameter distribution at approximately 10−20 μm,10,30 which is consistent with the interfacial behavior of both the crude oil surfactants and synthetic surfactants.20,21,23,25,32−35 This shift in mean particle size assumes that dispersant was applied in excess and fully mixed with the hydrocarbon stream at the wellhead, such that the water− hydrocarbon interfacial tension was significantly reduced.26−39 For the scenario presented in Figure 1, decreasing water−oil interfacial tension to 3 mN/m17 decreases mean particle diameter prediction to 7−20 μm for 0.5−1 m/s ejection velocity. Further, reservoir biodegradation of the crude oil is likely to result in an increased concentration of carboxylic acids, causing the stabilization of small particles.23,25,29 Our approach is based on the assumption of a single mode in the DSD between 20 and 100 μm, which agrees with a direct experimental study by Li et al.30 in a wave tank over a wide range of turbulent conditions. When Corexit 9500 was applied for the same experimental conditions,30 the oil particle diameter distribution showed peaks below 20 μm. Dispersed particle diameters below 300 μm are also suggested by Socolofsky et al.18,31 We acknowledge that several questions about initial and evolutionary droplet sizes, including the complex chemical and physical processes that govern such quantities, are still an area of ongoing research. It is possible that smaller droplets quickly coalesce into larger droplets, that the initial particle size distribution is bimodal, and/or that the liquid hydrocarbon particles did not separate effectively from evolving light ends. Indeed, the rapid formation of the surface slick, observed within hours of the blowout, suggests multiple contributing phenomena that are not addressed in this modeling study. One possible explanation includes form drag of gas bubbles with surrounding crude oil and water, where the increased buoyancy decreases the hydrocarbon rise time. That is, the initial free gas collected at the top of the reservoir may have provided an alternative transport mechanism for some oil to the surface, independent from the atomization process described by Socolofsky et al.,18,31 which is modeled here. The oil particle size from these estimates may be improved through a comprehensive suite of surface and interfacial tension experiments over multiple samples of the Macondo crude oil. Further insight in determining how deepwater environments affect hydrocarbon dispersion over changing physiochemical conditions19,21,22 may improve the simulation results presented herein. While the status and fate of the oil is time dependent,40−45 the assessment of the relative role of the physical dispersion by the blowout to that of the chemical dispersion by the synthetic surfactant is so far missing.46 Assembling all the pieces of the puzzle can best be achieved by numerical modeling.47−49 Our application couples a circulation model to a multifraction oil model9,50,51 with biodegradation and weathering.15,48 We simulate the oil discharge in deep waters using an approximation of a buoyant plume model18,31 and include the
pressure, the crude oil is likely dispersed, mixed with natural gas, and sheared into small, less buoyant particles.18 The droplet size distribution (DSD) produced by the blowout is of critical importance, as it controls the oil rising speed (and consequently the proportion of oil submerged at any time) and depends on three factors: the chemistry of the crude oil,7 the shear rate, and the temperature at the ejection point. Little work has been done to study the dispersion properties of the Macondo oil; the present work utilizes comparative reference oil compositions from literature.19 In addition to temperature considerations, the shear rate at the ejection point emulsifies the oil and generates small particles.20,21 In turbulent flow with water and crude oil, mean particle size may be estimated through a balance of shear forces (acting to break particles) and restorative interfacial forces.22 In the present work, we define the term “dispersion” to be a mixture of two immiscible phases, where the dispersed phase exists as droplets. This may be referred to as a simple “emulsion” in some fields.6,13,14,19,22−26 Assuming thus parallel behavior between water-in-oil and oil-in-water, simple estimates of the mean particle size can be computed as a function of ejection fluid velocity22 as a first approximation of the mean particle sizes relevant to this study (Figure 1).
Figure 1. Mean particle size for water-in-oil dispersions with respect to the ejection fluid velocity for three different flow diameters (0.1, 2, and 6 m), estimated using eq S4 in the Supporting Information from Boxall et al.22 with a water−oil interfacial tension of 20 mN/m.25 We used the classical definition of diameter for monodispersed spherical particles.
The purpose of this study is to understand subsea oil transport over large spatial and temporal scales and explore the effect of synthetic dispersant for deep-water oil spills. We assume hydrocarbon particles are continuously generated from a single source at the wellhead, although a more specific solution may account for the observed geometric and geographic fluctuations.12 We have also assumed a large range of blowout diameters, which is overall realistic considering the geometry of the various leaks.4 In particular, the diameter has not been artificially underestimated to favor the prediction of small droplets. Given flow rate estimates of 48 000−67 100 bbl/d of oil exiting from the blowout4,27 and an effective jet exit diameter between 0.2 and 0.5 m 1, the ejection velocities are of 2.8−3.8 m/s for 0.2 m diameter and 0.45−0.62 m/s for 0.5 m diameter (Figure 1). With estimated near-field 13294
dx.doi.org/10.1021/es303197h | Environ. Sci. Technol. 2012, 46, 13293−13302
Environmental Science & Technology
Article
fractions C3 with ρoil = 940−960 kg·m−3.7,5 Particle density ρoil is selected at random between the minimum and maximum values within each fraction. Individual particle size is selected at random from distributions of dispersed particle diameter ranging between 1 and 300 μm (Figure 1). These values encapsulate extreme estimates of oil droplet sizes expelled from the measured exit diameter of the broken riser pipe, diameter of flow path, and flow rate1−4 using the equation derived by Boxall22 (equation S4, Supporting Information). This range is compared to laboratory experiments of droplet size diameter for high shear light crude oil-in-water dispersions13,30 and contains the full range of residence time in the water column (Figures S1 and S2, Supporting Information). The scenario with synthetic dispersants (WD) is simulated by estimating the peak in the droplet size distribution toward smaller size value of 10−20 μm.24,25,30,32−34 This value was estimated by decreasing water−oil interfacial tension from 20 to 3 mN/m, as a first approximation of surfactant effectiveness.24,37,59,60 We further represent the hypothetical scenario without dispersant (ND) by a non-normal distribution with a mode value of 50−70 μm,13,14 corresponding to atomization and natural emulsification of the oil during the Macondo blowout (Figure S3, Supporting Information). The parameters employed are consistent with the interfacial behavior of both crude oil surfactants and synthetic surfactants.20,21,25,32−34 These two distribution scenarios were used to simulate the effect of the dispersant on the partition of the oil mass in the water column through time. It is noteworthy that the balance of the phase behavior of the oil can be upset both by mixing and chemical dispersants, which can easily be simulated by changing the overall concentrations of the fractions in the model. Independently from oceanic transport, oil/gas particles have inertial motion with a vertical terminal velocity UT computed with a fluid dynamics integrated formulation,61 solving for a broad range of gas bubble and oil particle sizes.62 We calculate UT for the regime of spherical shape and small particle size range
