Using Oil Spill Dispersants on the Sea (1989)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

2 Chemistry and Physics of Dispersants and Dispersed Oil Most oils spilled on water rapidly spread into a slick, with thick- ness from several millimeters down to one micrometer depending on the oil type and the area available for spreading. Wind-driven waves and other turbulence can break up the slick, producing more or less spherical droplets ranging in size from a few micrometers to a few millimeters. Sometimes, these droplets can be stabilized by natural surface-active agents (surfactants) present in the of] or contributed by the sea-surface microlayer in the region where the oil was spiked. These surfactants stabilize the droplets by orienting in the oil-water interface with the hydrophobic part of the surfactant molecule in the oil phase and the hydrophilic part in the water phase, thereby diminishing the interfacial tension. Applying chemical dispersants to an of} slick greatly increases the amount of surfactant available and can reduce oil-water interfacial tension to very low values it therefore tales only a small amount of mixing energy to increase the surface area and break the slick into droplets (Figure 2-1~. Dispersants also tend to prevent coalescence of of! droplets. The interface, stabilized by the surfactant, permits droplets to survive despite frequent collisions with adjacent droplets. The same stabi- lizing factors reduce adherence to hydrophilic solid particles, such as sediments, as well as other solid surfaces (discussed later in Chapter 4~. 28

CHEMISTRY AND PHYSICS OF OIL DISPERSANTS A 29 Application By_ B Hydrophilic Group Hydrophil Portion of Dispersant Prevents Droplet Coalescence Water - Lipophilic a~;g Surfactant-Stabilized c~°il Droplets FIGURE 2-1 Medh~sm of chemical dispersion. A. Surfactant locates at oil-water interface. B. Oil sliclc is readily dispersed into micelles or surfactant-stabilized droplets with mirumal energy. Source: Derived from C~evari (1969~. During the past 20 years, significant reviews and descriptions of dispersant chemistry include those by Poliakoff (1969), Dodd (1974), Canevari (1971, 1985), Wells (1984), Pastorak et al. (1985), Wells et al. (1985), API Task Force (1986), and Brochu et al. (1987~. COMPOSITION OF DISPERSANTS The key components of a chemical dispersant are one or more surface-active agents, or surfactants sometimes loosely caned "de- tergents." They contain molecules with both water-compatible (hy- drophilic) and oil-compatible (lipophilic or hydrophobic) portions. Most formulations also contain a solvent to reduce viscosity and facilitate dispersal. Chemistry of Surfactants The behavior of a surfactant is strongly affected by the balance betweeen the hydrophilic and lipophilic groups in the molecule. Grif- fin (1954) defined the hydrophile-lipophile balance. The useful range of this parameter is from 1 (most lipophilic) to 20 (most hydrophilic). Many organic compounds, like hexane, with no hydrophilic groups could have HLB as low as zero, and would not be surface active. In

30 USING OIL SPILL DISPERSANTS ON THE SEA the HLB range of 1 to 4, the surfactant does not mix in water; above 13, a clear solution in water Is obtained tHosen, 1978~. Bancroft's rule states that the dominant group of a surfactant tends to be oriented in the external phase (Bancroft, 1913~. Thus, a predominantly lipophilic surfactant (HLB, 3 to 6) would stabilize a water-in-oi} emulsion, and a predominantly hydrophilic surfactant (HLB, ~ to 18) would stabilize an oil-in-water emulsion. Surfact ant s used in of} spin dispersants tend to be of the latter type. Natural surfactants, which promote mousse (water-in-oil emulsion), tend to be predominately lipophilic. HLB is important in determining the effect of salinity on dis- persant performance, since hydrophobic portions of the surfactant molecule tend to be salted out. Laboratory measurements on weath- ered crude of} with a dispersant sensitive to salinity showed that, at a comparable treatment rate and mixing energy, the amount of oil dispersed is approximately 58 percent in seawater compared to 1 percent in fresh water. This formulation, balanced for effective performance in seawater, is too hydrophilic for freshwater service (Canevari, 1985~. Surfactants are also classed by charge type, as noted below (a list of formulations is given in Appendix A): Anionic. Examples include sulfosuccinate esters, such as sodium diocty! sulfosuccinate (e.g., Aerosol OT). Other examples are oxyalkylated Cue to Ci5 alcohols and their sulfonates. · Cationic. An example is the quaternary ammonium salt RN(CH3~3+CI-, but such compounds are often toxic to many or- ganisms and are not currently used in commercial dispersant formu- lations (I`ewis and Wee, 1983~. · Nonionic. These are the most common surfactants used in commercial dispersant formulations. Examples are sorbitan mono- oleate (HLB, 4.3), sold as Span 80, and ethoxylated sorbitan mono- oleate (HUB, 15), sold as Tween 80. In addition, polyethylene glyco! esters of unsaturated fatty acids and ethoxylated or propoxylated fatty alcohols are used. ~ Zwitterionic or amphoteric. These molecules contain both positively and negatively charged groups, which may balance each other to produce a net uncharged species. An example would be a molecule with both a quaternary ammonium group and a sulfonic acid group (refer to Appendix A), but such compounds are not found in current commercial formulations.

CHEMISTRY AND PHYSICS OF OIL DISPERSANTS 31 As surfactants become more concentrated, the interfacial tension between of] and water decreases until a critical miceDe concentration (CMC) is reached. Micelles are ordered aggregates of surfactant molecules, with the hydrophobic portions of the molecules together at the interior of the micelle and the hydrophilic portions facing the aqueous phase. Above the CMC level, there is little change in inter- facial tension, and additional surfactant molecules form new micelles. Below the CMC, additional surfactant molecules accumulate at the water-air or oil-water interfaces. The CMC can be estimated from the concentration at which a change in slope of a plot of interfacial tension (as measured by the drop weight technique) versus dispersant concentration occurs (Figure 2-2~. Some formulations can reduce interfacial tension to a few percent of the value without surfactant addition. For example, in a specially adapted Wilhelmy Plate instrument, the initial oil-water interfacial tension of 18 dye/cm was reduced to the minimum detectable value, approximately 0.05 dye/cm, within ~ min after dispersant was added (Ross anti Kvita, unpublished! data). Current Dispersant Formulations Early dispersant formulations were derived from engine room degreasers, and some were highly toxic (Chapter 1~. To reduce toxicity, nonaromatic hydrocarbons (or water-miscible solvents such as ethylene glyco! or glycol ethers), as well as less toxic surfactants, have been used in more recent formulations (Chapter 4 and Appendix A). Figure 2-3 illustrates how different surfactants become oriented at the oil-water interface. Compound A is sorbitan monooleate (HLB, 4.3), predominantly lipophilic. Compound B is Compound A that has been ethoxylated with 20 mol of ethylene oxide, rendering it more hydrophilic (HLB, 15~. A dispersant containing both A and B. with a larger amount of B. can stabilize an oil-in-water emulsion.* A blend of surfactants with different HLB, giving a resultant HLB of 12, wiD be more effective than a similar quantity of a sin- gle surfactant with HLB of 12. This is shown in Figure 2-3: the hydrophilic groups of B penetrate farther into the water phase, per- *As discussed earlier, the dominant group of a surfactant tends to be oriented in the external phase (Bancroft, 1913~.

32 USING OIL SPILL DISPERSANTS ON THE SEA 22 21 20 19 18 a, 16 to- 15 '.1~3 1 4 ,_ 13 J to 1 ~ O '' 11 10 9 8 7 6 ~ ~ ~ Dispersant C If\ \\ cmc = 5.3 + 2 ppm Dispersant F cmc = 38.2 + 6 ppm 1 ;\ Dispersant B cmc = 58.0 + 5 ppm I: 0 10 20 30 40 50 60 70 80 90 100 DISPERSANTIN SEAWATER(ppm) FIGURE 2-2 Interfacial tension as a function of Dispersant concentration showing the discontinuity in slope at the critical micelle concentration (cmc). Light Arabian crude with three dispersant formulations at 28°C aIld 38 percent salinity; interracial tension is measured by the dro~weight method. Source: Rewidlc et al., 1984. Reprinted, with permission, from the American Society for Testing and Matenals. ~ 1984 by ASTM. milting closer physical interaction between the lipophiles of both A and B. The overall result is to provide a stronger interfacial surfactant film and resistance to coalescence of dispersed of! droplets. A review of the patent literature (Appendix A), combined with discussions with several major suppliers of dispersants, indicates that a limited number of surfactant chemicals are used in the dispersant formulations most widely available today. The exact details of disper- sent formulations are proprietary, but the chemical characteristics of these formulations are broadly known (Canevari, 1986; Brochu et al., 1987; Wells et al., 1985; Fraser, private communication). Thus,

33 m ~ m ~ Z ~ to\ ~ ~ '. — _ ._ m Cal U ~ 0 ~ on

34 USING OIL SPILL DISPERSANTS ON THE SEA modern dispersant formulations containing one or more nonionic sur- factants (15 to 75 percent of the formulation) may also contain an anionic surfactant (5 to 25 percent of the formulation) and include one or more solvents. The surfactants used include the following: nonionic surfactants, such as sorbitan esters of oleic or lauric acid, ethoxylated sorbitan esters of oleic or lauric acid, polyethy- lene glyco} esters of oleic acid, ethoxylated and propoxylated fatty alcohols, and ethoxylated octy~phenol; and ~ anionic surfactants, such as sodium dioctyl sulfosuccinate and sodium ditridecanoyl sulfosuccinate. Dispersant formulations also contain a solvent to dissolve solid surfactant and reduce viscosity so that the dispersant can be sprayed uniformly. A solvent may be chosen to promote rapid solubility of the surfactant in the of} and to depress the freezing point of the dispersant mixture so that it can be used at lower temperatures. The three main cIasses of solvents are: (~) water, (2) water- miscible hydroxy compounds, and (3) hydrocarbons. Aqueous sol- vents permit surfactants to be applied by eduction into a water stream. Hydrocarbon solvents enhance mix~ng and penetration of surfactant into more viscous oils. Examples of hydroxy-compounc! solvents are ethylene glyco} monobuty] ether, diethylene glyco] mono- methy} ether, and diethylene glyco! monobuty! ether. An example of a hydrocarbon solvent is a low-aromatic kerosene. High-boiling solvents containing branched saturated hydrocarbons are also used since they are less toxic than aromatics. Appendix A gives a list of chemical formulations for use on oiT discharges listed in 1987 by the U.S. Environmental Protection Agency in its National Contingency Plan Product Schedule (Flaherty and Riley, 1987~. Composition of some of these formulations have been disclosed in patents, but all are proprietary. Only U.S. EPA listed formulations may be used to treat of} discharges in U.S. waters (Chapter 6~. Matching Dispersant Formulations With Oil Type for Increased Effectiveness Because oils vary widely in composition, it is reasonable to hy- pothesize that a particular dispersant formulation could be more effective with one of} than another; indeed that a dispersant could

CHEMISTRY AND PHYSICS OF OIL DISPERSANTS 35 be matched to a particular of! for increased, possibly optimal, ef- fectiveness. This is not possible at present for a number of reasons. Only a limited number of different dispersants have been used or are available. Such a narrow subset of all possible formulations has resulted from a convergent evolution in the industry. Furthermore, limited use has been made of chemical dispersants during accidental spins in the United States, and only a few research spills have been conducted using different dispersants with the same oil. Because it is not possible to conduct field tests of all dispersants with all oils, a more practical approach has been taken: formulations that perform best in laboratory studies are used on oil spills. If a dispersant works, it continues to be used. It should be noted, however, that the effect of dispersant appli- cation may be delayed, as was observed in Lichtenthaler and Daling's (1985) Norwegian offshore research studies. Effectiveness also may be reduced by of] resurfacing later (Bocard et al., 1987~. Temperature can affect dispersant performance in ways other than by changing the viscosity of the oil. The solubility of ethoxylated surfactants in water increases at lower temperature. For example, at similar dispersant- oil ratios, water salinity, and wave energy, dispersant OSD-1 was found to be 100 percent effective (aD of! was removed from the water surface) in laboratory tests at 15°C, but only 56 percent effective at 5°C. In contrast, dispersant EXP-A was 100 percent effective at both temperatures. This difference was explained by changes in water solubility and HLB of the dispersant (Becker and Lindblom, 1983~. Fate of Surfactants and Solvents in the Aqueous Environment Surfactants are used for many purposes other than treating of} spills, and their degradation in the aqueous environment has been a concern since the 1950s, when synthetic alky~benzene sulfonate (ABS) detergents were found to be resistant to biodegradation and produced persistent foam on waters receiving domestic sewage efflu- ent. This problem was solved by replacing the ABS detergents with more readily biodegradable surfactants. Manufactured surfactants in 1983 totalled about 24 m~lion metric tons worIdwide, most of which were employed as household detergents and industrial cleaners (Lay- man, 1984:49~. It is recognized that surfactants used in detergents may contribute to stream poDutant discharges that are continuous and often affect a large area over many years. In contrast, disper- sants and dispersed of] inputs to the sea are usually rare or infrequent

36 USING OIL SPILL DISPERSANTS ON THE SEA events that could cause temporary effects in the open sea, but cause large, short-term disruptions in restricted areas. Therefore, a direct comparison of continuous surfactant discharge from industrial and household use and dispersant surfactant loading is tenuous, at best, in regard to environmental effects of the two sources. Some discharged surfactants are biodegraded in sewage treat- ment plants, but many are not because much of the sewage is not treated. Some of the linear alky~benzenes (LAB), used in the man- ufacture of linear alky} sulfonates (LAS) remain as an impurity in LAS, and are found in suspended particles and sediments surround- ing municipal waste discharges. Eganhouse et al. (1983) used LAB as tracers and stated that they appear to be preserved in sediments for 10 to 20 years. Some work has been done on the rate of breakdown and environ- mental concentrations of surfactants in the water column (Kozarac et al., 1983; I'acaze, 1973, 1974; Penrose et al., 1976; Una and Gar- cia, 1983~. Surfactants are also transferred from water to air via sea spray, and increase the production of marine aerosol. They may thus encourage the transfer of of] slick components into the atmo- sphere (Fontana, 1976~. Adsorption of surfactn~nts onto sediments is discussed later in this chapter (see also Inoue et al., 1978~. Surfactants can also become bioconcentrated and metabolized in the tissue of fishes and invertebrates (Comotto et al., 1979; Kimerie et al., 1981; Payne, 1982; Schmidt and KimerIe, 1981~. Metabolic breakdown of the surfactants is rapid (85 percent in 4 days). The fate of solvents used in dispersant formulations might also be of concern. Hydrocarbon solvents are similar to portions of the of! being dispersed and tend to suffer a similar fate. Glyco] ether solvents are likely to be more readily biodegradable than the of] being dispersed, but nothing appears in the literature about the toxicity of their degradation products. However, the concentrations are usually small and decrease rapidly owing to dilution and mixing. FATE OF Off SPILLED ON OPEN WATER Slick Thickness Slick thickness is an important parameter in predicting optimum dispersant dosage (Chapter 5), but the thickness of an of} slick at sea cannot be readily determined. Reliable measurements of thickness over the whole area of a-slick have rarely been made (HoDinger and

CHEMISTRY AND PHYSICS OF OIL DISPERSANTS 37 Menella, 1973; Lehr et al., 1984; 0'Neill et al., 1983~. Infrared remote sensing provides an image of thick slicks, 10 to 50 I'm or greater (O'NeiD et al., 1983), and ultraviolet sensing can measure slicks down to the submicron range. Combined, infrared and ultraviolet remote sensors can be used to calculate areas and ratios of thin to thick slicks. The limits to infrared detection are unknown, however, and certainly vary with environmental conditions and of} type. Varying intensity levels in the infrared have been processed to yield additional contours, but assignment of thickness to such contours, although attempted, has been only relative (Ross, 1982~. Actual slicks at sea are nonuniform in thickness and distribution on the surface. The thickest portion of a slick can be as great as several millimeters and the "sheen" (m~crolayer) only 1 to 10 ~m. In one experiment, infrared thermography showed that thin areas of the slick were 10 to 20 Am and thick areas were 150 to 200 ~m. The thicker portion contained 28 of the 42 bb! of spired oil (Bocard et al., 1984), but covered only about one-fifth of the stick's surface area. Most estimates of slick thickness are averages based on the visual appearance of the of} or calculated by dividing the total volume of of} by its observed area (International Tanker-Owners Pollution Federation [ITOPF], 1982~. Some investigators believe that using average thickness, although formally consistent, is misleading and obscures one of the most important aspects of an oil slick from the cleanup team's point of view its nonuniformity. Also, from a biological point of view, exposures under a nonuniform stick are likely to be patchy. Nevertheless, many spills of widely varying size tend to reach a similar average thickness of about 0.1 mm rather quickly and this rule of thumb is widely employed by dispersant application specialists. Over several days, as the slick spreads, average thickness may decrease to 0.01 mm (API Task Force, 1986; McAuliffe, 1986~. The following is some evidence for such a generalization: · The Chevron Main Pass Block 41 C blowout, released of! at 1,680 to 6,650 bbl/day. The slick size varied, but was about 1-km wide and 10-km long, with thicker of] near the platform. The average thickness was 0.02 to 0.09 mm (McAuliffe et al., 1975; Murray, 1975~. · The Hasbah 6 of] well blowout in the Arabian Gulf released a viscous oil slick that extended for many miles. After 2 weeks, 9,930 bb} of of! were skimmed from an area of 11.3 km2. The slick was thus estimated to be at least 0.13-mm thick (Cuddeback, 1981~.

38 USING OIL SPILL DISPERSANTS ON THE SEA · In the API-EPA research spills, 20 bb} were released over 5 to 10 min. corresponding to rates of 3,000 to 6,000 bbl/day. After 15 to 30 min. before dispersant spraying began, the slicks covered 20,000 to 30,000 m2. Average thickness was therefore 0.1 to 0.2 mm (Johnson et al., 1978; McAuliffe et al., 1980, 1981~. · In a Norwegian test spill of 700 bb} over 2 hr (corresponding to 8,400 bbl/day), the thick part of the slick, containing 90 percent of the oil, covered 1 km2 after ~ hr and remained at 1.5 to 2 km2 for the next 5 days. The average spin thickness decreased from 0.06 to 0.013 mm (Audunson et al., 1984~. · Data from the 1983 Halifax trials showed average thicknesses of untreated slicks was of the order of 47 Em after ~ hr. decreasing to 40 ,um after 2 hr. The thin portion of the slicks was estimated to be of the order of 1-pm thick (Canadian Offshore Aerial Applications Task Force [COAATF], 1986~. A large release over a short time, such as would occur from a tanker accident, initially produces much thicker slicks near the release point. For example, release of 200,000 bb} into an area 100 m in diameter could create an average thickness of 20 mm, but if the slick spreads, the average thickness decreases. In cold climates and waters, however, the higher viscosity of of! can cause a stable slick to be thicker than 0.! mm. In addition, slick thickness could increase as the of! becomes emulsified to form mousse, which occurred during the Amoco Cadiz disaster and many other incidents. Slick Spreading Oil slicks are usually nonuniform in thickness because of the in- teraction of interfacial tension, gravity, and viscosity in spreading processes, the accumulation of oil at downwehing convergence zones procluced by water movement, and the formation of high-viscosity water-in-oi} emulsions (mousse). Furthermore, oils have different spreading tendencies, particularly on cold water (Tramier et al., 1981~. The work of Fay (1971) provided a mathematical prediction for spreading under the influence of interracial tension and gravity. His model predicts an area increasing to a maximum value propor- tional to the 3/4 power of the volume of oil spilled. Although the model included water viscosity, it did not include the effects of of! viscosity, emulsification, and evaporation, and considered only a calm water surface. Nevertheless, the model was successful in predicting

CHEMISTRY AND PHYSICS OF OIL DISPERSANTS 39 results of laboratory experiments and was subsequently used in more elaborate models (Huang and Monastero, 1982~. However, field tests revealed the limitations of Fay's approach. Jeffery (1973) reported the formation of an elongated slick with the major dimension increasing linearly with time over 4 days. The minor dimension increased rapidly during the first few hours, then remained constant. In the experiments reported by Cormack et al. (1978), the minor dimension increased at a rate in accordance with the Fay equations, but the major dimension increased at 10 times the expected rate. Attempts to deal with factors not addressed by the Fay approach include constraining the slick to be elliptical (Lehr et al., 1984) or imposing artificial nonhomogeneity on the slick by dividing it into "thick" and "thin" portions (Mackay et al., 1980a). Recent studies have shown that the dominant mechanism of oft spreading is the interaction of the of} droplets with diffusive and current shear processes in near-surface currents (Elliot, 1986~. Work by oceanographers, such as Bowden (1965), Fisher et al. (1979), and Okubo (1967), have shown that a patch wig elongate in the direction of flow. The length scales of the patch in the along- and cross-flow directions are approximated by equations that suggest that initally the spreading of of} droplets wiD be Fickian. that is, it is dependent on the spatial concentration gradients—and that spreading will grow linearly with time. These expressions also show that, with time, velocity shear is more influential. Subsurface release from a well blowout produces a thinner slick than a release directly onto the water surface because of the entrain- ment of water (and oil) in rising gas bubbles (Fannelop and Sjoen, 1980~. At the surface, water and of! flow away from the center of the plume, and the of} slick spreads faster than of! released on a quiet surface. A model of the bloc ~ blowout developed on these principles estimated that, at 1 km from a 30,000 bbl/day subsurface discharge, slick thickness would be 0.06 mm (Fannelop and Sjoen, 1980~. This may be compared with an average thickness of 0.07 mm calculated by Jernelov and Linden (1981) and McAuliffe (1986~. For a surface discharge, the calculation of Fannelop and Sjoen would indicate a thickness of 1.3 mm under the above conditions. *rick's first law states that the flux, or rate of diffusion, of a material (e.g., oil particles, chemicals) is proportional to the concentration gradient; this relationship as- sumes that flow is laminar.

40 USING OIL SPILL DISPERSANTS ON THE SEA In addition to the spread (or sometimes contraction) of the slick under control of surface tension and viscosity, a variety of of} types in test and accidental spins, including the bloc ~ spill, have been observed to spread into a complicated texture of thick "pancakes" and thin sheen. Formation of mousse is common. Windrows produced by surface convergences are further complications. Thicker patches have been observed to move downwind at a faster rate, leaving a thinner trailing sheen behind. Published observations of actual thickness variation downdrift from ~xtoc fare rare and none are quantitative. Several months after the well blew out, observers on research vessels, small boats, and helicopters, characterized the of} in the plume as occupying five zones (Atwood, 1980~: 1. A continuous light-brown water-oi] emulsion on the surface occurred in the immediate vicinity of the flames at the weDhead and extended no more than a few hundred meters down the plume. 2. The sea surface was 30 to 50 percent covered by a light-brown emulsion in disoriented streaks. This zone started a few hundred meters down-plume from the burn and extended several kilometers, depending on wind stress. At times it was virtually absent. 3. The sea surface was 20 to 50 percent covered by light-brown emulsion oriented in streaks parallel to the wind direction, apparently in the convergence zones of the Langmuir surface circulation.* The streaks were surrounded by a sheen, their width varied from a few centimeters to a few meters, and their length varied from one to tens of meters. The dimensions depended on wind stress. This zone extended from a few hundred meters from the flames to several kilometers down the plume. 4. The light-brown emulsion darkened until the streaks were black, apparently from photoox~dation, since the degree of darkening seemed to depend on light intensity. I`angmnir streaks were black- ened in the center and light brown at the edges. At times they coalesced into lines of blackened of} several kilometers long. At the brown edges were small balls of viscous emulsion (mousse). The softer balls would coalesce on contact; in some instances grapefruit- sized bass formed a raft 50 m in diameter. Most of these phenomena *The combination of wind forces and wave forces produce currents that rotate cylindrically. The convergences between these currents, or windrows, are straight run- ning in the direction of the wind, and their distance apart ranges from 20 to 50 m depending on wind speed (Falter, 1978~.

CHEMISTRY AND PHYSICS OF OIL DISPERSANTS 41 occurred 10 to 40 km from the burn. A light to heavy sheen of surface of! was always present in this zone. 5. An extensive sheen (1 to 10 ~m), visible because it changed the reflectivity of the water surface, covered more than 50 percent of the surface, usually in the form of Langmnir rows, and extended to the farthest extremity of the plume. Other observers noted that there were typically three to six "stringers," slicks much longer than they are wide, ~ to 3 mm thick (Fraser and Reed, 1982~. Observers from Petroleanos Mex~canos also noted the complicated nonuniform structure of the bloc ~ slick (Petroleanos Mex~canos, 1980, translated from the Spanish): The oil that did not disperse moved along the surface of the sea following the resultant marine surface currents and the direction of the winds, forming bands and strips of variable lengths and widths and other capricious shapes. During the first few days after the spill, the slick formed a maximum length of 15 km and a maximum width of 2 km. [As a result of] the combination of the containing barrier, recovery of spilled oil, and the reduction in flow of oil from the well, the size of the slick was reduced even though it advanced toward the coast along a corridor well defined by the currents. It can be shown that the slick did not advance as a compact and continuous mass, but as a series of strips that from aerial observation seemed lilac a web. Physical Processes of Dispersion Related to Water Motion In addition to spreading of an of} slick and dispersal of of] into droplets, spilled of! is distributed and transported by motion of the water mass in which it resides. Such motion includes drift of the slick caused by wind, tides, and other forces, motion of dispersed oil with the water mass, and redistribution of of} with respect to the water as a result of turbulent diffusion and vertical shear. Conventional concepts of physical oceanography are frequently used to describe these processes. The change with time of oil con- centration at a specific location is the sum of three terms: 1. advective changes, which depend on flow of the water mass; 2. turbulent diffusion changes, which depend on concentration gradients of the of] within the water mass and on wind- and current- induced turbulence; and 3. source or sink changes, which depend on physical, chemical, and biological interactions. Solid particles on which oil has adsorbed can settle out as sediment, dispersed oil droplets can be removed by

42 LL o CC ~ 30 o 111 CO By a: IIJ o USING OIL SPILL DISPERSANTS ON THE SEA 50 _ o -it TEST 5 NO DISPERSANT 300 RPM O 2.5 MIN · 5 MIN O 10 MIN 20 DROPLET SIZE (,um) FIGURE 2-4 Droplet size distribution for test of Kuwait crude dispersal. Source: Jasper et al., 1978. Reprinted, with permission, from the American Society for Testing and Materials. ~ 1978 by ASTM. recoalescence with the surface stick, or photochemical and biochemi- cal degradation can occur. Mechanical dispersion occurs primarily when waves break, which requires wind speeds greater than 10 kn. When wave action pro- vides sufficient energy to overcome interfacial tension and create new oil-water interfacial area, an of} slick breaks into small droplets, usually less than 10 Am in diameter depending on slick thickness, that become suspended in the water column. Larger drops form in smaller numbers, and their distribution (Figures 2-4 and 2-5) fans off more steeply than exponentially (Bonwmeester and Wallace, 1986a,b; Franklin and Lloyd, 1986; Jasper et al., 1978; Lewis et al., 1985; Norton et al., 1978; Shaw and Reidy, 1979~. Advection Bulk motion of water includes daily tidal currents; wind-induced currents (including the Langmnir surface circulation) that occur with daily sea breezes and longer-term time-scales (days to weeks); seasonal coastal currents; frontal eddies in boundary currents; and mesoscale eddies, which are common in the open ocean. Currents

CHEMISTRY AND PHYSICS OF OIL DISPERSANTS 43 have been charted for centuries in waters frequented by ships. This accumulated knowledge is used, for example, by the U.S. Coast Guard, Navy Supervisor of Salvage, and the Canadian Coast Guard to predict drift so that boats and other objects lost at sea might be found. Many currents are unpredictable, however, and advective flow is frequently complex and difficult to predict. For example, recent sateDite-based studies of Gulf Strewn eddies have shown the poten- tiaI complexity of mesoscale circulation (Backus et al., 1981~. This is also reflected in the subsurface distribution of dispersed of} in the 104 jo3 CO o C] 1 o2 o LLJ m 10 10°- 10 10-1 DROPLET DIAMETER (,um) . ~- ^\- \ \ ~ \\' \ ·] . I\ . 41 _ · tA 1` 1 1 1 1 1 1111 1 1 `.1 1 1111 10° FIGURE 2-5 Droplet size distributions of No. 2 diesel oil in wind-ware tank with no disperse. Initial (upstream) oil thickness = 0.15 mm. Wave height is condition II. Fetch: X = 10.6 m. Source: Boumeester and Wallace, 1986a.

44 25 30 35 USING OIL SPILL DISPERSANTS ON THE SEA DISTANCE FROM BLOWOUT (km) SW 10 0 10 20 . ~ too, ~ 10,000 ~1 ,. 5,00~00 30 40 50 70 80 90 100 NE ' ' ' 1 \s\ 2: ! / \ · / ~ ~ ~- . . . . Not Detectable . Concentration = 100 p9/1 FIGURE 2-6 Oil concentrations (pa/liter) in the water column following the Ixtoc I blowout. Note the horizontal extent of the oil, which suggests the role of midsection, while the downward penetration is indicative of the role of vertical turbulent diffusion. Source: Boehm and Fiest, 1982. Reprinted, with permission, from Environmental Science and Technology, Vol. 16, No. 2. ~ 1982 by American Chemical Society. bloc [plume (Figure 2-6). Observations of a section moving from the southwest to northeast suggested that of} was transported at least 40 km in subsurface water. Figure 2-6 shows substantial concentra- tions southwest of the blowout site ("upstream"~. The upstream oil most likely was the result of advection prior to the date on which these measurements were made, when the current was in a different direction. A second example of advective processes working at different time scales is shown in Figure 2-7. These data were derived from an experiment designed to test a mode} emphasizing physical processes, such as drift and spread, and weathering and vertical distribution (Johansen, 1984~. The surface distribution of the oil stick followed the track of buoys quite well. In Johansen's mode} (discussed in more detail later) advection was computed as a "vectorial sum of the drift induced by the local wind, tidal currents, and an assumed stationary background current," which adequately summarizes the advective components. Physical dispersion in many areas is dominated by unique local

CHEMISTRY AND PHYSICS OF OIL DISPERSANTS 1 Okm .. JUNE 25. 1 2GMT - JUNE 26. 1 8GMT lOkm 1 Okm JUNE 25. 20GMT 10km JUNE 27. 9GMT - JUNE29. 17GMT 1 Okm JUNE 29. 1 6GMT 45 1 Okm J I ~3 JUNE 26. 7GMT 1 Okm JUNE27. 19GMT 1 Okm JUNE 30. 1 2GMT FIGURE 2-7 Time variation of an oil slick observed by remote sensing during the Halten Bank experiment. With time, the slick extended in line of advection and took on shapes determined by tidal currents, inert oscillations, and wind events. Source: Johansen, 1984.

46 USING OIL SPILL DISPERSANTS ON THE SEA TABLE 2-1 Typical Turbulent (Eddy) Diffusion Coefficients ~ . . Location K 2 (cm /see) References Horizontal Diffusion New York Bight 5,500 Bering Sea 2,800 Harrison Bay 780 Bering Sea Beaufort Sea Okubo, 1971 Coachman and Charnell, 1979 Wilson et al., 1981 Vertical Diffusion 185 Cline et al., 1982 25 Liu and Leendertee, private communication SOURCE: Pelto et al., 1983. currents. For example, currents near a coast are primarily deter- mined by tidal flow, wind direction, and bathymetry, all of which are locally unique. Convergence zones created by internal waves can be a major factor in the onshore transport of slicks or tar lumps. Thus, in addition to models and theoretical studies, there is a need for knowledge of important physical processes specific to areas likely to be affected by of] spills. Turbulent Diffusion Turbulent diffusion coefficients are scale dependent since turbu- lent diffusion does not obey Fick's diffusion laws rigorously (some typical values are given in Table 2-1~. However, the order of mag- nitude of values is instructive: horizontal turbulent diffusion tends to be more than 10 times faster than vertical. For of] dispersion, this means that surface spills tend to stay near the surface and to be advected horizontally, rather than to diffuse downward and enter a less observable regime. Prediction of vertical movement may be highly simplified by using a "diffusion floor" (Mackay et al., 1982~. Dispersant-treated of] tends to be under greater infuence from vertical diffusion than untreated of} (Chapman, 1985), and this type of physical transport can be addressed using a depth-dependent diffusion coefficent. The classic study of turbulent ocean diffusion was made by Okubo (1971), and recent theoretical and observational studies that

CHEMISTRY AND PHYSICS OF OIL DISPERSANTS 47 give new information on the topic are Cline et al. (1982), Wilson et al. (1981), and Coachman and CharneD (1979~. Oil Concentration Under Slicks Oil dispersed into the water column can come into contact with marine organisms that would not otherwise be affected by an of! spill. It is important to know the concentration of of} and its distribution in the water column in order to assess the environmental impacts of using a dispersant to treat an of} spill. Vertical distribution of of! components in the water column de- pends on many physical factors, chiefly sea state, which breaks a stick into droplets and during storms can mix the upper layers of water to a depth of 10 m or more. (This vertical distribution has been studied in a number of field experiments described in Chapter 4.) Under untreated slicks, oil concentrations typically are a few parts per minion to less than 0.1 ppm, diminishing with depth and with increasing time. For example, in one experiment water sampled beneath two untreated control slicks 2 to 6 hr after the spill gave oil concentrations ranging between 3 to 5 ppm at 1 m and 0.03 and 0.63 ppm at 2 m (LichtenthaTer and Dating, 1985~. In Edition to redistribution of physically dispersed oil, the more soluble components of oil, such as benzene, toluene, xylene, and napEthalene, which are also more acutely toxic than other compo- nents, dissolve in the water and do not resurface if the droplets coalesce. Methods for analysis of oft in water are not consistent among researchers. For example, some researchers measure volatile hydrocarbons (C, to Clot. Others extract the water with a solvent and measure high molecular weight hydrocarbon by fluorescence, gas chromatography, or by weight. The toxicity of a hydrocarbon component to marine organisms depends on the aqueous solubility of the compound, how concen- trated it is in the portion of the water where the organism lives, the type of organism, the length of time it is exposed, and many other specific factors (see Chapter 3~. Evaporative Loss of Volatile Hydrocarbons During the first 24 to 48 hr of an oil spill, evaporation is the single most important weathering process affecting mass transfer and removal of toxic lower molecular weight components from the

48 USING OIL SPILL DISPERSANTS ON THE SEA slick. Evaporative loss is controlled by the composition, surface area, and physical properties of the of} and by wind velocity, air and sea temperatures, sea state, and solar radiation. The thinness of a slick and the small diameter of dispersed of! droplets allows the volatile hydrocarbons ACE to Coo) to evaporate quickly or go into solution. Evaporation greatly predominates over solution. For example, Harrison et al. (1975) demonstrated that evaporation is 100 times faster than solution for aromatics, and 10,000 times faster for alkalies. Biological toxicity of the remaining surface of! or droplets dispersed into the water column should thereby be greatly reduced (Anderson et al., 1974; McAuliffe, 1971, 1974; WeDs and Sprague, 1976~. From a physicochemical model, estimated evaporative loss of volatile hydrocarbons for different crude oils ranges from about 20 to 50 percent in 12 hr (Mackay et al., 198Oa; Nadeau and Mackay, 1978; van Oudenhoven et al., 1983~. The lower percentage occurs with more viscous oils. Field measurements have demonstrated rapid loss and low con- centrations of Cat to Coo hydrocarbons in of! samples from surface slicks, and in water under untreated and chemically dispersed crude of! sticks. McAuTiffe (1977) and Johnson et al. (1978) showed com- plete loss from a crude of} slick of low molecular weight aromatic hydrocarbons in ~ hr. for example, benzene and toluene within 1 hr. dimethy~benzenes by 5 hr. and trimethy~benzenes by ~ hr. Evaporative losses of specific compounds are difficult to pre- dict theoretically and thus limit theoretical modeling of weathering. Henry's law constants are required for hundreds of individual compo- nents whose rate of loss is usually assumed to be controlled only by individual mole fraction (which cannot be determined) in the crude or distillate product and pure component vapor pressure. This as- sumption can lead to some simplification, since pseudocomponents (or distillate cuts) of various boiling point ranges have been used in modeling evaporative behavior, and good agreement between ob- served and predicted behavior has been obtained (Page et al., 1983, 1984; Payne and McNabb, 1984~. Dissolved hydrocarbons, like the slick, are most concentrated near the of} spill release point, particularly if it is beneath the water surface. The regions in which the seawater is likely to be most toxic are therefore localized. As the slick spreads and evaporates, Tow molecular weight compounds are lost and the risk of subsurface exposure to these diminishes. Acute toxicity of of! on or near the

CHEMISTRY AND PHYSICS OF OIL DISPERSANTS 49 surface oil therefore tends to be diminished by evaporative weathering (Gordon et al., 1976; McAuliffe, 1986; WeDs and Sprague, 1976~. Photochem~cal Processes Oxygen-containing products, resulting from exposure of of} to air and sunlight, are likely to have some surface activity (Klein and Pilpel, 1974~. Photochemical oxidation of most of} on water occurs slowly, so the concentration of oxidation products and their effects on dispersibility are likely to be small in the first few days after an oil spill (Burwood and Spears, 1974; Hansen, 1975, 1977; Payne and McNabb, 1984; Payne and PhiBips, 1985; Wheeler, 1978~. However, for some oils (typically, waxy crudest, photolysis may have a significant effect on chemical dispersibility and on the formation and stability of water-in-oil emulsions after only a few hours of exposure (Daring, 1988; Daling and Bran~vik, 1988~. Although some oxygenated products have been isolated from samples taken at large of} spins, most predictions are based on smaD- scale laboratory experiments, and little or no fieldwork has been done on this process (Overtop et al., 1979, 1980; Payne and McNabb, 1984; Payne and Phillips, 1985~. Mousse Formation The formation of stable water-in-oi} emulsions appears to depend on the simultaneous presence of asphaltenes and paraffins (Bridle et al., 1980; Payne and Phillips, 1985~. Although their experiments were not aimed at establishing the limits of either component, Bridle et al. noted the following: · The original sample (Kuwait 200+ fraction, with 6.6 percent asphaltene and 9.S percent paraffin wax) readily formed mousse. · When waxes and asphaltenes were removed, no mousse could be formed. When either waxes or asphaltenes were added back to the basic of! singly, no mousse could be formed. · When both waxes and asphaltenes were added back to the basic oil, mousses were easy to form. . When only 10 percent of the original asphaltene content was added together with the waxes to the basic oil, mousses were easy to form.

50 USING OIL SPILL DISPERSANTS ON THE SEA Water-in-oi} emulsions can be destabilized by adding surfactants that displace the indigenous surfactants from the interface (Canevari, 1982~. One product performed well as an emulsion preventer at product-to-oi} ratios as low as i:5,000. The product has a high oil-to-water partition coefficient (10,000) compared to most disper- sants (10), and is a better dispersant than Corex~t 9527 at ratios between 1:400 ~.ncl 1:5,000 (Buist and Ross, 1986, 1987~. Similarly, dispersant formulation effectiveness may depend on interaction with the indigenous surfactants at the interface. Such an interaction ap- pears to affect dispersant performance even more than the physical properties of crude of! (Canevari, 1985~. BEHAVIOR OF OIL-DISPERSANT MIXTURES Criteria for Effective Dispersal The chemical nature of dispersants and the physical and chem- ical processes that affect untreated of} have been reviewed, and this chapter now turns to the physical and chemical criteria for effective chemical dispersal. (The various factors that influence success or failure of a dispersant response operation equipment design, dis- persant regulation, dosage, remote sensing, application strategy, and logistics are discussed in Chapter 5.) Some of the processes discussed earlier, particularly advection and turbulent diffusion, apply to chemically dispersed oil as wed as untreated oil. However, chemical dispersants can cause major changes in slick-spreading characteristics; droplet formation, stabi- lization, coalescence, and resurfacing; and adherence of of! to solid surfaces, suspended particulate matter, and sediments. Four basic criteria must be met for effective chemical dispersal of of} to occur: 1. Dispersant must be sprayed onto the slick. This obvious criterion can be a major problem in practice. Dispersants are usually sprayed from boats or aircraft onto a floating oil layer. 2. Dispersant must mix with of} or move to the oil-water inter- face. Dissolution of dispersant in bulk of] is not necessary as long as the surfactant molecules adsorb at the interface. The ideal dispersant application system produces drops small enough to just penetrate the slick; not so small that they remain at the oil-air interface or blow off-target with the wind, and not so large that they penetrate the slick Bind are lost in the water column (see Chapter 5; APT Task Force, 1986~.

CHEMISTRY AND PHYSICS OF OIL DISPERSANTS 51 3. The dispersant must attain the proper concentration at the interface; ideally, that is the concentration that causes a maximum reduction in interfacial tension (Rewick et al., 1981~. The ideal dosage (quantity per unit area) is an amount sufficient to cause dis- persion but not so high that dispersant is wasted. Although uniform dosage is the goal, actual dosage is often neither uniform nor pre- cisely known (Chapter 5; APT Task Force, 1986; Exxon Chemical Company, 1985~. If wind blows falling droplets outside the slick boundary, or if previous droplets cause herding the movement of the slick into narrow bands interspersed with clear water or dispersion, a falling droplet wiD be ineffective. In principle, thick areas require more dispersant and thin areas require less, but there is no practical way to vary the application rate to achieve a constant dispersant-to-oi] ratio. Spray equipment is set to provide a uniform quantity per unit area. Dispersants may be applied "neat" (undiluted) or diluted by water or a hydrocarbon solvent. In practice, average dosage required is determined from manufacturer's recommendations (e.g., Exxon Chemical Company, 1985) and experience. 4. The of! must disperse into droplets. Energy is required to increase the oil-water interfacial area. The lower the interfacial ten- sion, the less energy is required. Indeed, under optimum conditions (very low interfacial concentration for some dispersants), oil-water interfacial tension can be reduced to less than ~ dye/cm, and al- most any minor agitation wiD suffice. The energy can come from wind, waves, or mechanical stirring, although normal wave energy is usually adequate. Relation of Oil Composition to Dispersibility Earlier, the question whether dispersant formulation could be varied to match particular oil types was answered "no." Here, a related question is addressed: What of] composition factors influence the ease with which of} can be dispersed? ~ ~ ~ ~ ~ 1 A ~ 1 _ _ A ~ ~ 1 1 ~ 1 ~ ~ ~ t But o~ Usury contain some surIace-act~ve compounds, which are believed to contribute to mousse formation and can interact unpredictably with a dispersant's surface-active components (Payne and Phillips, 1985~. However, all oils are not equally dispersible (Canevari, 1987~. This is not surprising given the many variations in of] composition and physical properties. Available data about oil properties include volume yield during

52 USING OIL SPILL DISPERSANTS ON THE SEA TABLE 2-2 Labofina Effectiveness Results, Standard 1-Min Delay Before Sampling Dispersanta Crude Oil Emulsion Tendency A Kuwait Extremely strong 25 12 8 21 La Rosa Extremely strong 24 -- -- 20 North Slope Strong 26 -- -- 30 Guanipa Strong 18 -- -- 35 Loudon Strong 23 4 5 21 Murban Intermediate 17 10 10 21 Southern Relatively weak 20 11 4 15 Louisiana Ekofisk Weak 26 12 7 34 Saharan Blend Weak 19 -- -- 38 Goose Creek Weak 31 -- -- 29 C14 None 46 39 24 50 aIdentity of dispersant was not revealed in the publication. SOURCE: Canevari, 1987. distillation to temperatures that correspond usually to commercial product fractions, API gravity, viscosity, sulfur, nitrogen, nickel, vanadium, aromatics, naphthenes, smoke point, pour point, and aniline point (Bobra and Chung, 1986; Payne and McNabb, 1984~. Of these data, the only ones that have been correlated with dispersibility are viscosity and pour point. Information on the influence of oil composition on dispersibility is poorly understood and is insufficient to allow a rigorous prediction of the dispersibility of a particular oil. Testing is usually needed. Differences in performance of different dispersants with the same of} also occur, despite similarities in some dispersant compositions. An example from laboratory experiments with four dispersants on various crude oils is shown in Table 2-2 (Labofina test, discussed later in this chapter). Tetradecane (Cal) was used as a standard having no tendency to form water-in-oi! emulsions. The four dispersants performed on Ale in the order D > A > B > C, which was generally preserved for the 10 crude oils tested. However, differences varied greatly. For example: for Ekofisk crude oil, dispersant C's effective- ness was rated 7, and that for D was 34; for C:4, C was 24 and D was 50; for Guanipa crude, D was 35 and A was 18; for Kuwait crude, A was more effective at 25 than D at 21 (Canevari, 1987~.

CHEMISTRY AND PHYSICS OF OIL DISPERSANTS 53 Another example is the comparison of two crude oils adjusted to the same viscosity. Normally LaRosa crude (73 cSt viscosity at 16° C) is more difficult to disperse than Murban crude (6 cSt). When LaRosa crude was diluted with pure isoparaffin of} to a viscosity of 6 cSt, its dispersion efficiency (Mackay-Nadeau-Steelman tMNS] test, dispersant-oil ratio 0.003:1) was 50 percent compared to 78 percent for Murban crude of the same viscosity (Canevari, 1985~. Laboratory studies (NRC, 1985) have shown that metahopor- phyrins, which are naturally occurring components of crude oils, with some surface-active properties, favor formation of water-in- oil emulsions (mousse). Canevari (1985) predicted that these trace components would tend to inhibit dispersion (oil-in-water emulsion formation) rather than enhance it. The literature on enhanced oil recovery indicates that other oil components might also affect dis- persant effectiveness (Healy et al., 1976~. For example: · Aromatacity. Solubility in the oil phase of the oleophilic por- tion of surfactant molecules would be expected to increase with the concentration of aromatic constituents in the oil (Bancroft, 1913~. On the other hand, aromatics facilitate dissolution of dispersant in oil (Mackay, private communication). Experiments by Mackay et al. (1986) indicate that addition of toluene to Alberta crude oil in some cases increased effectiveness as measured by MNS and rotating flask tests. Naphthenic acids. These are carboxylic acids of cyclic hydro- carbons. They are not strongly surface active but could interact with added surfactants. U.S. West Coast crudes typically contain much higher concentrations of naphthenic acids than Gulf Coast crudes (acid numbers are 3.5 to 4.0 in California versus 0.4 to I.3 in the Gulf). · Paraffin. The pour point of an oil is strongly correlated with its paraffin (wax) content. As spired of! weathers and gradually loses its lower molecular weight components by evaporation, the higher molecular weight paraffins increase in relative amount and also become less soluble in the remaining oil, in some cases forming a separate solid phase. The result is an increase in pour point. If the pour point of the weathered oil becomes higher than the ambient water temperature, the oil will become solid or semisolid and be nondispersible. · Asphaltenes. Wax and asphaltenes are stabilizing agents for water-in-oil emulsions that, with their very high viscos~ty, are ex- tremely difficult to disperse. Hence asphaltenes can be important to

54 USING OIL SPILL DISPERSANTS ON THE SEA determining dispersibility. Unfortunately, they are not a factor com- monly determined in oil analysis. Lindblom (private communication) suggests that the 650+°F fraction might be taken as a surrogate in predicting whether mousse formation is likely. If that fraction is greater than 40 percent, the of} may tend to emulsify and wiD be difficult to disperse. Eject of Oil Viscosity, Time, and Other Parameters on Dispersion In this section, the influence of viscosity and other important physicochemical parameters will be considered in the context of opti- mizing dispersion. The effectiveness of of} dispersants as determined by laboratory tests is strongly dependent on of! viscosity. Viscosity has two effects: it retards dispersant migration to the oil-water in- terface, and it increases the energy required to shear off a drop from the slick. Forces applied to a viscous slick from water motion tend to be transmitted through the slick rather than being dissipated in the slick and causing dispersion (Mackay, private communication). Dispersants are most effective for of} viscosities less than about 2,000 cSt, and almost no dispersion occurs over 10,000 cSt (see Figure 2-8) (Cormack et al., 1986/87; Lee et al., 1981; Morris, 1981~. This viscosity limit was chosen by the United Kingdom as a reference point against which to consider the number of oils likely to be treatable by dispersants at sea (Cormack et al., 1986/87~. Other researchers use 2,000 or S,000 cSt as the limit (ITOPF, 1982~. Daling (1988) points out that the upper viscosity limit for chemi- cal treatment of oils and water-in-oi} emulsions is specific for different oils. It is therefore not possible to use a general viscosity limit, par- ticularly on water-in-oi} emulsions, where the dispersant has to break the emulsion into oil and free water at the surface before dispersion of the of} into the water column can take place. Criticality of Timely Response According to Cormack et al. (1986/87) most crude oils are treat- able when freshly spilled, but because viscosity increases rapidly with weathering, time is critical. This time relationship is shown in Figure 2-9, which indicates increased viscosity of different crude oils, floating on the sea and constrained in 4-m diameter floating rings (Cormack et al., 1986/87; Martinelli and Cormack, 1979~. In large accidental

CHEMISTRY AND PHYSICS OF OIL DISPERSANTS Dispersibility unlikely tlo4 Dispersibility diffioult 103 a Us o In > 102 10 _ SAFANly,4_ KUWAIT _ ~ . ... .. // ~ CLAYMORE _ I _ . , _ t' - I / /' . . .. . . . . . . . . . . ~ ~ ——FORTIES _ _ - - —N IN IAN ~ -TH I STLE ~ y/ BRE NT -a -1i 1 EKOFISK 1' I I , ~ , , , , , , , 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 TIME, h. AFTER SPILL 55 FIGURE 2-8 Increase of viscosity of several crude oils with weathering. Source: Comeback et al., 1986/87. spins, of! is often 24 hr old or older when dispersant is first applied, and may have formed mousse as well. The final viscosity values likely to be reached after weathering of of] at sea for 12 to 24 hr must be considered to determine whether and how long after a spill dispersants can be usefully applied (van Oudenhoven et al., 1983~. According to Cormack, all crude oils are initially amenable to dispersion except those crude oils with high initial viscosities, those that would be solid at seawater temperatures, and petroleum prod- ucts normally carried in heated cargo tanks. All light fuel oils are amenable. I.indblom (private communication) states that an oil is dis- persible if it wiD spread on water. If the of] will not spread (i.e.,

56 USING OIL SPILL DISPERSANTS ON THE SEA 90 80 70 - O 60 z 111 z llJ In 50 40 30 20 10 o l Kuwait it, Heavy Algerian ~~` ~ \\ at\ - o 1 00 1 ,000 1 0,000 1 00,000 VISCOSITY AT 10°C, Cf. 1oOS-1 SHEAR RATE FIGURE 2-9 Effect of viscosity on disperse efficiency using the Fina revolving flask test at the Warren Spring Laboratory. Source: Morris, 1981. if the water temperature is below the pour point of the oil), disper- sant drops will simply roll off of the of! layer. The dispersant formulation used also may influence the depen- dence of dispersibility on oil viscosity. For example, hydrocarbon- solvent based dispersants appear to work better on mousse and on high-viscosity oils than do dispersants with water, glycol, or glycol- ether solvents, probably because their dissolution in the of] keeps oil and dispersant in contact longer (COAATF, 1986; Fraser, private communication). Contact Time of Dispersant With Oil Contact time between the dispersant and of} is an important factor in effectiveness of dispersal. If the dispersant is water soluble it can be diluted in the water before the proper interfacial structure can be developed to stabilize droplets. The Labofina test as practiced at Warren Spring Laboratory uses a contact time of 1 to 1.5 min before the onset of agitation. For of} or mousse with a viscosity of 10,000 cSt, the standard WSL Labofina test gives almost zero effectiveness, but with a contact time of several minutes the effectiveness could

CHEMISTRY AND PHYSICS OF OIL DISPERSANTS 57 be as high as 40 percent (Cormack et al., 1987~. With shipboard applications systems, estimated contact time may be as short as 2 sec. Effect of Dispersant on S]ick Dynamics Because chemical dispersants lower interfacia] tension between oil and water, they greatly alter the dynamics of slick spreading and the development of slick structures described earlier. One of the most obvious and rapid effects, when surfactant is first applied to an oil slick, is herding (Chau et al., 1986; Delvigne, 1985; Mackay et al., 1986~. The surfactant lowers the surface tension of the water thereby causing the oil slick to contract in a few seconds. This herding soon subsides and is not important after a few minutes. On a longer time scale, of] slicks treated with dispersant appear to spread more rapidly than untreated slicks. Near St. John, New- foundland in 1981, investigators found that a treated slick initially spread slower than untreated ones, spread more rapidly after 2 hr 30 min. and after 3 hr 45 men was 33 percent larger (Goodman and MacNeill, 1984~. Three sets of slicks laid down during a test near Halifax in 1983 also exhibited this effect—the treated slick spread slower at first (COAATF, 1986~. The transition times and excess area varied, however both slicks were the same size after 4 hr 15 min for the Corex~t 9527 treated slick, 1 hr 45 min for Corex~t 9550, and 2 hr for BPMA 700. The Cored 9550 slick was 4 times larger than the untreated slick after 2 hr 30 min. Application of dispersant to a mousse patch from the bloc blowout was described in the cruise report of the ship Longhorn. On August IS, 1979, the ship encountered a large area of reddish to chocolate-colored viscous oil, containing much debris. A strong odor, likened to the smell of "an old gas station," permeated the area. The mousse was so viscous it made a "slurping" sound against the side of the ship. When the oil surface parted, schools of small silvery fish and some sharks could be seen swimming below. Shortly after the mousse was found, a DC-6 aircraft appeared and sprayed the patch, presumably with a dispersant; after the spraying had ceased, the Longhorn steamed back into the mousse and found the area of the patch appeared smaller, as if the oil had been "herded" by the chemical. Its texture was now more liquid, but the layer was thicker. (Hooper? 1981:186) These phenomena are rarely addressed by modelers, and not

58 USING OIL SPILL DISPERSANTS ON THE SEA much progress has been made in developing mathematical expres- sions for their effects (Zagorski and Mackay, 1981~. Payne and Mc- Nabb (1984) have written: It is not possible at this time to model spilled oil behavior beyond first order estunations of total area potentially covered for a defined range of slick thicimess. The variables affecting modeling of dispersed oil behavior are even more complex; these are addressed later in this chapter. Behavior of Droplets and Resurfacing Addition of a chemical dispersant reduces the energy required to break a slick into droplets. If the interfacial tension is sufficiently low, chemically aided dispersion can occur in the absence of breaking waves. In addition, under similar conditions the number of small drops tends to be larger for oil dispersed chemically than mechani- cally (Jasper et al., 1978~. Droplet size greatly affects the physical transport of oil. Because of} drops are generally less dense than water, larger droplets rise to rejoin the slick, smaller ones rise more slowly, and the smallest ones tend to remain suspended in the water column as the result of turbulent diffusion. The suspended droplets are then transported horizontally by subsurface currents and can diffuse deeper into the water column. Oil droplets can be removed from the water column by combining with sediment and other abiotic particles, or becoming bound to or ingested by biota, such as plankton. Size distribution of dispersed of} droplets in the water is an important measure of dispersant effectiveness (Mackay et al., 1986~. A spectrum of of! drop diameters can be characterized by a variety of terms: most-frequent droplet size or mean droplet size are used, as is volume mean diameter (vmd), the droplet size that contains average volume. The latter measure is quite different from mean droplet diameter or size since the volume of a sphere increases with the cube of the diameter. For example, a dispersion may have a mean particle diameter of 5 ~m, b,'t ~ ~rnl''m" m":;`n ~li~n~"tPr of An ``m {R`rf~rA at al., 1984~. Depending on the interracial tension produced by the dispersant and the energy with which the of] is dispersed, the vmd may range from 1 to 200 ~m. For example, the vmd of dispersed of} droplets (at dispersant-oi! ratios of 1:16 to 1:100) were 14 to 225 I'm in the MNS test, and 17 to 84 ,um in the 1`abofina test (Lewis et al., 1985~. Ideally, ~ —~^— -~^—~^ ~^ ~- ~ V —~1~ ~~,7 1~J$~ ~u

CHEMISTRY AND PHYSICS OF OIL DISPERSANTS 59 the dispersed of} droplets should be as small, if possible having a vmd as Tow as ~ to 5 am, in order to avoid resurfacing.* A laboratory study of the effect of temperature and energy on droplet size distribution (Byford et al., 1984) found that the vmd of of] droplets was most strongly influenced by dispersant formulation and energy input, but temperature (0° to 20°C) had only a minor effect. Dispersant-oi} ratio had a great effect on vmd for a low- performance dispersant, but not for a high-performance dispersant. Density had a greater influence on vmd than viscosity, at constant temperature. Delvigne (1987) showed that droplet size distributions became smaller with increasing energy input (d varied as E-° s) as well as duration. Oil type, weathering state, and temperature all affected viscosity, and droplet size was approximately proportional to the 0.34 power of viscosity. Droplet size distributions were similar whether they were drawn from submerged of} or a surface layer, and they were independent of salinity. To characterize the resurfacing of the larger droplets for a given observation time and turbulent regime, Mackay et al. (1986) used in their model the concept of a critical oil droplet diameter, above which most of] will resurface and below which most of! remains dispersed. The higher the mixing energy, the larger this critical droplet diameter will be. With time, the depth of dispersed of] particles increases (Mackay et al., 1986~. In practice, this depth depends on oceanographic conditions, especially turbulent diffusion and mesoscaJe currents. It is not necessary that all oil droplets remain in the water column indefinitely for the dispersant to work. It is apparent that for effective dispersal, oil droplets formed in the water should be as small as possible. The smaller they are, the longer they will stay in the water column without resurfacing (Gatellier et al., 1973; Mackay et al., 19SOb, 1986; Nichols and Parker, 1985~. Byford et al. (1984) summarized the situation by proposing that dispersed oil droplet size be regarded as a major factor in judging dispersant effectiveness. *The ideal vmd is an estimate based on observations; one review of this report suggested that a 20 Am vmd is acceptable.

60 USING OIL SPILL DISPERSANTS ON THE SEA Oil Concentration Under Dispersed Slicks Dispersal of an of} slick initially increases subsurface concentra- tion, which may be rapidly diminished by physical transport pro- cesses. The larger oil-water interfacial area of chemically dispersed oil may cause dissolution to be increased, and evaporation, which depends on oil-air surface, to diminish. Because the most water-soluble components of of} tend to con- tribute most to acute toxicity of dispersed oils (Abernathy et al., 1986; Anderson et al., 1974; Wells and Sprague, 1976), there is concern that the biological impact of chem~caDy dispersed of} on sub- surface organisms might be greater than for mechanically dispersed oil. This effect would be mitigated, in open water, by turbulent diffusion, which transports dispersed of] away from the surface and greatly dilutes its concentration. For a very large spin near shore, or for a smaller spin in a confined area, turbulent transport may not be adequate to disperse of} rapidly. Some field studies have been designed to quantify the effects of chemical dispersion on physical transport processes and to measure the exposure conditions that result. The concentrations of dispersed oil measured in these tests should be compared with toxic thresholds estimated for the organisms and crude oils reviewed in Chapter 3. For example, during tests off southern California, La Rosa crude was dispersed by aerial application of Corex~t 9527. An estimated 50 percent of the spired of} was dispersed. In samples collected 20 to 25 min after dispersant application, the highest concentrations were 2 to 3 ppm at ~ to 3 m and 0.5 ppm at 6 m (McAuliffe et a]., 1980~. Using the concept of ppm-hr,* the estimated exposure was 0.67 to 1.25 ppm-hr at 1 to 3 m depth and 0.17 to 0.21 ppm-hr at 6 m. After 100 min. the concentration at 1 to 3 m had not diminished, so that the exposure factor became 3.3 to 5 ppm-hr. Under a stick of Murban crude oil in the same test series esti- mated to be 90 to 95 percent dispersed 23 min after spraying, of! concentrations were 18 ppm at 1 m and 10 ppm at 3 m. After 50 to 57 min. concentrations had diminished to 3 to 4 ppm at 1 m. *The ppm-hr concept is an attempt to relate laboratory bioassay test data (in which oil concentration is usually held constant) to exposure conditions in nature, where the concentration of dispersed oil varies with time. The ppm-hr concept assumes that toxic effects are linear, both with concentration and time. Thus, both for variable concentration and for constant concentration, the integral of concentration times time is assumed to be comparable. There are situations, particularly involving extremes of concentration or time, where linearity should not be assumed.

ClIEMISTRY AND PHYSICS OF OIL DISPERSANTS 61 Exposure at 1 to 3 m ranged from 2.5 to 6.9 ppm-hr, and at 6 to 9 m it was 0.83 to 0.95 ppm-hr. For the best dispersed (approximately 80 percent) Pru~hoe Bay crude of! slick off southern California, McAuliffe et al. (1981) mea- sured higher total of} concentrations 15 min after dispersant appli- cation by aircraft: 30 to 50 ppm at 1 m, 10 ppm at 3 m. After 220 min concentrations at 1 to 9 m had decreased to 0.5 to 2.3 ppm. By graphical integration, assuming exponential decrease of concentra- tion with time, the exposure factors were 47 ppm-hr at ~ m depth and 12 ppm-hr at 9 m. Cormack and Nichols (1977) measured concentrations of Ekofisk crude oil chemically dispersed within the first 2 men after spraying from a boat: 16 to 48 ppm at 1 m. The oil concentration decreased to 5 to 18 ppm within 5 to 10 min after spraying, and to ~ to 2 ppm after 100 min. By graphical integration, the exposure factor in this case was 2 to 6 ppm-hr for the first 100 min. Aerially sprayed Statfjord crude of! slicks were sampled after 30 to 50 min. The highest of} concentrations were 25 to 40 ppm at 0.5 to 2 m depth; at 3 m only 4 ppm were found (Lichtenthaler and Daling, 1985~. Water samples collected 165 to 225 min after spraying contained up to 7 ppm at 0.5 to 1 m and 0.2 to 1.4 ppm of} at 3 m. Field experiments that measured of} concentration beneath nat- urally dispersed (untreated) slicks gave concentrations varying from less than 0.1 to 5 ppm, and exposures of ~ to 35 ppm-hr; the higher concentrations and greater exposure factors were found within the top meter of water. In the tests reviewed above, concentrations of chemically dispersed of} (up to 40 ppm) and exposure factors (up to 60 ppm-hr) are higher than those in undispersed cases. Evaporation and Weathering of Dispersed Oil Subsurface sampling of dispersed of} is seldom done because of logistical and measurement problems, and because concentrations decrease quickly to levels too low to be reliably analyzed. As a result, many questions about such interrelated processes as evaporation and weathering remain unanswered. For example, McAuliffe (1977) stated that "chemically dispersed of} appeared to weather in a manner similar to of] naturally dispersed under slicks." In contrast, Page et al. (1985) observed enhanced evaporation when of} was chemically dispersed.

62 USING OIL SPILL DISPERSANTS ON THE SEA Interaction of Dispersed Oil With Suspended Particulate Matter and Sediment Because a primary fate of spilled of} may be sedimentation, especially in estuarine and nearshore environments, it is important to know whether chemical dispersal increases or decreases transport of hydrocarbons to the sediment and whether sedimented oil is more available to benthic biota. In this section, physicochemical evidence for enhanced or reduced sediment transport is examined. Mackay and Hossain (1982) conducted a laboratory study of mechanically dispersed of] and suspended particulate matter (SPM) interactions and sedimentation of mechanically and chemically dis- persed crude of! (Alberta, Murban, and Lago Media crude and Corexit 9527 and BPllOOWD dispersants). They found that in the presence of sediment, dispersed of! was removed from the wa- ter column, and the settling velocity of oiled particulate matter was estimated to be as rapid as 1 m/hr. From 5 to 30 percent of the oil was incorporated in the settled sediment. Addition of dispersant decreased the fraction settled: for example, 30 percent with no dis- persant, 10 to 15 percent at a dispersant-oi} ratio of 1:10, and 6 percent at a dispersant-oi] ratio of 1:5. The most important factor appeared to be the degree of sorption of dissolved oil by organic matter in the sediment compared to the lower uptake of the mineral component (Karichoff et al., 1979~. As- sociation of of] and sediment particles changed the buoyancy of both; hence excess sediment tended to sink the oil, and excess oil (e.g., twice the volume of sediment) tended to keep fine sediment particles suspended. Smaller of} droplets (induced by chemical dispersion) were less susceptible to sedimentation than larger droplets. Even in the absence of sediment, of} droplets do not remain suspended indefinitely. Like other particles in the size range 1 to 10 ~m, they are scavenged by zooplankton, which absorb what nutrition they can and package the remainder into rapidly sedimenting fecal pellets (NRC, 1985; Sleeter and Butler, 1982~. Oil droplets can also be degraded by bacteria or fungi, which are comparable in size and capable of metabolizing hydrocarbons if they are given adequate nutrients. This group of processes is generally known as "biologically mediated transport." The agglomeration of both sediment particles and oil droplets is affected in a complex way by their interactions investigated by Harris and Wells (1979), Mackay and Hossain (1982), and Little et al. (1986 and earlier papers). Mackay and Hossain concluded that "prediction

CHEMISTRY AND PHYSICS OF OIL DISPERSANTS 63 of the environmental behavior requires a detailed knowledge of the prevailing sediment depositional regime." Nevertheless, their results suggested that chemical dispersion of of] leads to reduced interaction with suspended particulate matter or sedimentation. OIL FATE AND DISPERSION MODELS Model Types An ideal mode! simulates the following processes mathematically: · advection of slick and water masses; evaporation of of} components; dissolution of of} components in water; · dispersion of of} droplets in water; oxidation of of] components (particularly photooxidation); emulsification (mousse formation); · biodegradation of of} components; and sedimentation (including biologically mediated transport). All of these processes are time-dependent and must be described by dynamic models. State-of-the-art models include some, but not all, of these processes at varying degrees of sophistication; but field or laboratory experiments designed to calibrate or test models usually focus on only one process. Comprehensive models tend to be created in response to a need, such as the following (Mackay, 1986 and private communication): . real-time spin trajectory to facilitate countermeasures by on-scene commander (requires real-time environmental data, but can be used as a "game" for training); environmental impact assessment—to provide scenario for likely impact of oil-related developments such as offshore exploration and production or deep-water port; site-specific biological assessment- to provide an assessment of likely ecological effects; · ecosystem to provide long-term overall assessment of im- pacts and hazards on lakes, estuaries, bays, or open ocean. Chemical dispersion is only infrequently addressed in these models; typically the only dispersion modeled is produced by wind and surface turbulence. There is extensive literature on slick movement and oil fate models (Table 2-3~. The report of Huang and Monastero (1982) is

64 USING OIL SPILL DISPERSANTS ON THE SEA TABLE 2-3 Oil Spill Models Considering Dispersion Oil Spill Model Author DOOSIM Applied Science Associates Three-dimensional shear diffusion University of Toronto SINTEF University of Toronto A. D. Little SLIKFORCAST University of Toronto University of Rhode Island SEADOCK SLIKTRAKB USC/API Johansen, 1987 Spaulding, 1987 Elliott et al., 1986 Mackay, 1986 Johansen, 1985 Mackay et al., 1982 Ara~ramudan et al., 1981 Audunson et al., 1980 Mackay et al., 1980b Cornillon and Spaulding, 1978 Garner and Williams, 1978 Blaikley et al., 1977 Kolpack et al., 1977 SOURCES: Johansen, 1985a,b; Mackay et al., 1986. the most complete review to 1982; Spaulding (1986) updates it. Only Mackay et al. (1980b, 1982, 1986) and Johansen (1985a,b) consider chemical dispersion. Oil-weathering mathematical models developed by Payne et al. (1983, 1984) are based on measured physical properties and generate material balances for both specific and pseudocompounds (distiDa- tion cuts) in crude oil. They apply to open-ocean, estuary, lagoon, and land spills. Weathering processes included in the model are evaporation, dispersion into the water column, dissolution, water- in-oi} emulsification, and slick spreading. Good agreement has been obtained between predicted and observed weathering behavior. The material balance and weathered-oil composition predictions gener- ated as functions of time are useful for contingency planning, for as- sessing potential damage from spins, and in preparing environmental impact statements for outer continental shelf driding activities. Although of] is composed of hundreds of compounds, each with a distinct solubility (oil-water partition coefficient), volatility, reac- tivity, and diffusivity, most models treat "oil" as a single chemical species. This simplification is particularly inadequate when it is necessary to distinguish between components (such as benzene and toluene) that partition from oil into water and remain in the sub- surface water mass and insoluble hydrocarbons that comprise the

CHEMISTRY AND PHYSICS OF OIL DISPERSANTS 65 separate phase of} droplets that tend to rejoin the slick or attach to sedimenting particles. Nonuniform Slick Thickness Because slicks spread and have a nonuniform structure, some researchers (Mackay et al., 1986) have used a conceptual mode] that changes abruptly in moving from a small thick area to a sheen over the rest of the slick. This is important because uniform dispersant application based on such a mode} will result in overtreatment of the thin area by an estimated factor of more than 10 or more and undertreatment of the thick portions by a factor of 10 or more. Advection and Diffusion In modeling, it is frequently assumed that advection dominates the horizontal field, and turbulent diffusion dominates the vertical field and they are independent of each other. This greatly simplifies the differential equations and allows the two transport processes to be treated separately in numerical models (Mackay, 1984~. Resurfacing The most common supposition in modeling the natural disper- sion process is that breaking waves cause the of} layer to be propelled into the water column, thus forming a "shower" of of! droplets (Figure 2-10~. This depends on wind speed and on of} density and viscosity. Most of the of} particles rise again to the slick and coalesce there, but smaller droplets diffuse downward and are retained by sedimentation or biologically mediated transport (Mackay et al., 1980b; Sleeter and Butler, 1982~. Breaking Waves Dispersion rate is likely to be a function of slick thickness, oil- water interfacial tension, sea state, and fraction of the sea covered by breaking waves. It is wed known that of} causes turbulence damp- ing on a water surface; its presence thus presumably reduces the incidence of breaking waves and diminishes the fraction of sea they cover. It is also believed that dispersion occurs even in the absence of breaking waves, possibly as a result of "folding" of the of} when very

66 USING OIL SPILL DISPERSANTS ON THE SEA Nonbreaking Waves Breaking Waves Well Mixed Surface Layer e.g., 50cm 1- ~f Input `t Rising RN / C' Large Drops \ Small Drops / RS ~ / Small / Drops / Rising - Less Well- Mixed Diffusive Layer e.g., 10m Small Drops Diffusing Permanently to Lower Layer DIFFUSION FLOOR FIGURE 2-10 S<5hernatic diagram of dispersion processes. Source: Macilcay et al., 1980b. sharp, high-amplitude, short-wavelength waves pass through the oil layer and near-breaking conditions exist for a brief time. Buist (1979) has done the most detailed mathematical treatment of this process, along with experimental wind-wave tank measure- ments that resulted in equations with adjustable parameters. Integrated Approaches Bringing all the variables affecting oil dispersion together in a single mathematical model is a formidable challenge. Two groups have made substantial progress, however: Donald Mackay and co- workers at the University of Toronto, and Oistein Johansen of the Continental Shelf Institute (IKU), Trondheim, Norway. Mackay's Model Mackay's early models are primarily one-dimensional, emphasiz- ing mass balance, and do not attempt to predict a detailed three- dimensional distribution of dispersed oil. The equations developed by Mackay et al. (19SOb, 1986) predict an increasing dispersion rate as a slick becomes thinner. Thinner slicks damp turbulence less ef- fectively on the water surface, and fewer breaking waves are affected.

CHEMISTR Y AND PHYSICS OF OIL DISPERSANTS 67 The droplets formed from a thin slick are expected to be smaller and the rate of dispersion faster than for thicker slicks (Figure 2-10~. In a simple diffusion mode} to estimate the of! concentration beneath a dispersed spill, Mackay et al. (1982), using data from McAuliffe et al. (1980), predicted that concentration decreases expo- nentially with increasing depth and time at any time, it is directly proportional to oil volume. Thus a larger spiD would be expected (after dispersal) to yield higher concentrations of of} in water than would a small spill. A limitation on this calculation is the time al- lowed for spreading of of! on water; the calculations by Mackay et al. implicitly assume dispersant application and dispersion very shortly after the of} has been spiked, and before it has (for a large spill) spread very far. In the most recent model, Mackay et al. (1986) divide of! on the water surface into thick and thin slicks, the proportion of oil in each and the amount of dispersant sprayed on each are part of the input data. The influence of chemical dispersants is included in the form of an effectiveness factor, X; that is, the amount of of} dispersed is X times the amount of dispersant applied. A transition in dispersant effectiveness is recognized from a "performance-limited" regime at low-dispersant concentrations to an "access-lim~ted" regime of thin slicks in which herding and the relatively small amount of of! con- tacted by each dispersant drop make the amount of of} dispersed independent of the amount of dispersant added. Dispersal is assumed to be a quadratic function of wind speed; the coefficient decreases by a factor of 2 as the temperature decreases from 25 to 0°C. The size of dispersed droplets is assumed to be distributed according to the Weibull function; those larger than a critical diameter will resurface (following Stokes' law, modified for eddy diffusion effects) and expand the thin slick. In addition, eddy diffusion can transport the smaller drops to the surface, where they rejoin the slick, or to deeper water, where they are lost. Horizontal transport of of} dispersed in the water column produces a linear increase in plume diameter with time, as based on the dye patch data of Okubo (1971~. The mode} developed by Mackay et al. (19SOb) does not dis- criminate between dissolved and dispersed of} and does not include evaporation or any other process that might differ for components of the oil. It treats the sea as semi-infinite, without aDowing for shore- line, bottom topography, or stratification. Aspects of this mode} include:

68 USING OIL SPILL DISPERSANTS ON THE SEA . expression for of} drop diameter as a function of the effective- ness factor, X; · temperature and turbulence dependence of X; · validity of the resurfacing expressions; · validity of the horizontal diffusion expressions; · use of a "diffusion floor" to simplify modeling of vertical advection; and · calibration of the mode} with data on experimental spins. Johansen's Mode} Johansen's (1984) work illustrates the complexity of advection- diffusion processes and provides a new approach to including them in a mode} of dispersed oil. This mode} considers particles to be in one of three states: at the surface, entrained in the water column, or evaporated. The drift of a particle is determined by its state, and the transition from one state to another is determined by a random-number generator and a set of probability parameters: · probability for entrainment wind dependent; · probability for resurfacing dependent on density difference, droplet size distribution, time submerged, and wind force; · probability for evaporation—computed from established evaporation models. The advective movement of the spin is assumed to be dependent on the vectorial sum of the drift induced by the local wind, tidal currents, and an assumed stationary background current. These can be estimated by empirical correlations or from observations. This integrated mode] was extended to contrast the behavior of chemically treated with untreated of] (Johansen, 1985b). This is done by using two different sets of entrainment and resurfacing parameters, as well as two different droplet size distributions. The advantage of chemically dispersed oils, from the physical viewpoint, is that the of] is dispersed into the water column rather than remaining as a surface slick. Chemical dispersion thus tends to enhance the probability of entrainment and, since chemically dispersed of] droplets tend to be smaller and recombine less easily than physically dispersed untreated oil globules, the probability for surfacing is reduced. The most recent dispersion model, Dispersion of Oil on Sea Sim- ulation (DOOSIM), is a two-layer drift mode} using a random-walk algorithm for spreading calculations (Johansen, 1987~. A stochastic model is used for the mass budget, but an empirical mode] is used for

CHEMISTRY AND PHYSICS OF OIL DISPERSANTS 69 weathering. The mode! produces a color display with a generalized map, a detailed map, and a continual presentation of mass balance. Mode} Validation As more information is obtained in the laboratory and in field tests, models can be more thoroughly calibrated and their assump- tions tested. Good models help discipline thinking about complex processes and aid in designing and planning further work. The mod- els discussed above have been tested against data obtained in the field under realistic conditions. To make such a test meaningful re- quires carefully planned measurements on a controlled spill. Ideally, the field experiment should be designed to test a specific hypothesis. For example, Mackay et al. (1982) used a one-dimensional mode} to fit some data from the field studies by McAuliffe et al. (1980~. Summary Both untreated and chemically dispersed of! are transported by advective and diffusive processes. Oil left untreated on calm water tends to stay at or near the surface and thus tends to be controlled by wind-related surface drift. Untreated of} is more influenced by vertical turbulent diffusion as the seas become rougher, but it remains strongly influenced by surface currents. In contrast, dispersed of! enters the water column even in calm weather, is more influenced by vertical diffusion and vertical shear, and is less affected by horizontal advection. Resurfacing of of} in the water column is an important process that greatly complicates attempts to model, conceptually or numerically, the distribution of dispersed oil. These considerations emphasize the need to understand more precisely the role of vertical turbulent diffusion and the vertical distribution of dispersed oils. Areas of needed research include near- surface wave, current, and of! dynamics; oil-ice interaction; aigo- rithms for the time-dependent process of photooxidation; improved environmental data; and mode} validation with spin data. Trends include the development of comprehensive models, systems, develop- ment of portable stan(l-alone models, and the move toward interac- tive ("expert") system models (Spaulding, 1986~.

70 USING OIL SPILL DISPERSANTS ON THE SEA LABORATORY STUDIES OF EFFECTIVENESS Purpose of Laboratory Testing The general objectives of laboratory testing of dispersants in- clude the following: . testing a variety of dispersants to rank their relative effective- ness (e.g., Doe and Wells, 1978; Mackay and Szeto, 1981; Mackay et al., 1984; Martinelli, 1984; Rewick et al., 1981, 1984~; and · testing effectiveness of dispersants under carefully controlled conditions to assess the role of oil type, weathering state, dispersant- oil ratio, mixing energy, salinity, temperature, and application meth- ods (e.g., Byford et al., 1983; I,ebtinen and Vesala, 1984; Mackay et al., 1984; Payne et al., 1985; U.S. EPA, 1984~. Laboratory tests are also used to screen dispersant types prior to more expensive field testing (Meeks, 1981; Nichols and Parker, 1985~. They provide data for contingency planning; for stockpiling specific dispersants for particular environments, oil types, or deployment methods (Byford et al., 1983; U.S. EPA, 1984~; and ultimately for deciding whether or not to use a particular dispersant (Mackay and Wells, 1983~. Mathematical models for dispersal of oil can be partially vali- dated in the laboratory (Mackay, 1985~. Appropriate concentrations for toxicity testing can also be determined in laboratory tests (An- derson et al., 1985; Bocard et al., 1984; Mackay and Wells, 1983; Wells et al., 1984b). There are three generic types of laboratory tests in use as of 1987: 1. Tank tests with water volumes of 6 to 150 liters, including test vessels agitated using circulated seawater, and tests that employ breaking or nonbreaking waves to generate more realistic turbulent mixing energy. Examples are the EPA test (U.S. EPA, 1984), MNS test (Mackay et al., 1984), and the French Institute of Petroleum (IFB) test. 2. Shake-flask or rotating flask tests that are conducted on a 1-liter scale. Examples include the WSL Labofina test (Martinelli I984~. 3. Interfacial tension tests that measure properties of the treated oil instead of degree of dispersal in a system with given energy input (Rewick et al., 1984~. L7

CHEMISTRY AND PHYSICS OF OIL DISPERSANTS 71 The most common of these tests are compared later in this chapter. All tests reviewed establish an oil slick and then apply disper- sant in a defined dispersant-oi} ratio. Dispersant may be applied by spraying neat or mixed with seawater, by pouring into a ring on the water surface that contains the of} slick, by adding dispersant to seawater, or by premixing the dispersant with the oil. Mixing energy is applied by a high-speed propeller, by rotating a separatory funned containing the oil-dispersant mixture (Labofina), by a spray hose and circulation pump (EPA), or by a high-velocity air stream (MNS). Dispersant effectiveness is determined by one of the following four criteria: I. The amount of of! dispersed in the water. This can be mea- sured by visual observation, or by solvent extraction and spectropho- tometric analysis. The amount of dispersed oil may be determined under dynamic conditions (e.g., MNS and EPA tests) or after mixing has terminated (e.g., Labofina and MNS tests). It is also desirable to measure the amount of of! remaining in the surface slick to obtain a mass balance, but standard methods for doing so have not yet been developed (Nichols and Parker, 1985~. 2. Dispersed of} droplet size (Byford et al., 1984; Lewis et al., 1985~. This is another important criterion since larger droplets resur- face some time after dispersal in the water column. The volume mean diameter in the MNS test was 14 to 226 ,um depending on the dispersant-to-oi} ratio (DOR) and the dispersant used (Byford et al., 1984; Lewis et al., 1985~; in the Labofina test it was less than 154 am. 3. Dispersed droplet stability as a function of time and turbu- lence in both static and dynamic systems (Mackay et al., 1984; U.S. EPA, 1984). 4. Interfacial tension (Mackay and Hossain, 1982; Rewick et al., 1981, 1984) has been used to rank dispersant formulations, but the static character of the measurement makes correlation with dispersal under turbulent conditions unrealistic. Critical Factors In all tests, oil-water ratio, dispersant-oi] ratio, dispersant appli- cation method, mixing energy application, and methods of sampling and analysis were found to be critical factors in determining the pre- cision of results. Oil-water ratio is most important for the relatively

72 i: USING OIL SPILL DISPERSANTS ON THE SEA hydrophilic dispersant formulations (i.e., those with relatively high HLB) since greater dispersion occurs with higher concentration of surfactant; this will be the case if the volume of water is smaller for the same volume of oil and surfact ant. The nearly infinite capacity of the open ocean for diluting hydrophilic dispersant is not normally ac- counted for in laboratory tests. Typical oil-t~water ratios are 0.02:1 for the Labofina test, 0.0017:! for the MNS test, and 0.00077:1 for the EPA test. At high oil-water ratios, collisions (and possible coalescence) of dispersed droplets are more frequent; this too is an unrealistic simulation of the open ocean. Dispersant-oil ratio is especially important below 0.2:l, where a steep dependence of effectiveness on dispersant-oil ratio is observed (Rewick et al., 1981~. The method of applying the dispersant to the oil is a key factor In an effectiveness test. Types of application used include: dispersant premixed with water; dispersant premixed with oil; · neat (undiluted) dispersant, poured on slick; and · neat dispersant sprayed on slick (this is the only test that has direct field relevance). For dispersant sprayed on the slick, droplet diameters in the 200 to 700 Am range are desirable. These diameters are similar to those produced by dispersant spray systems used in practice, which is a fortunate match of circumstances (Chapter 5~. Dispersal tends to be more efficient with smaller drop sizes. Other factors affecting results include amount (volume) of dispersant in the slick surface, oil slick thickness, and drop-to-drop distance in the sprayed slick (Mackay et al., 1984~. These are areas that continue to be important in laboratory research. How mixing energy is applied in the laboratory is also a major factor. Recognizing that premixing dispersant with of} or water does not realistically represent field conditions, various methods have been employed to mix dispersant with oil: swirling flasks, water jetting, and air streams, for example. The mixing energy can be affected by the materials under study. In the MNS test, wave amplitude is reduced by No. 6 fuel oil and other viscous oils resulting in less dispersal (Mackay and Szeto, 1981; Mackay and Wells, 19S3; Mackay et al., 1984~. Oil-water ratio appears to be at least as important as viscosity and mixing energy.

CHEMISTRY AND PHYSICS OF OIL DISPERSANTS 73 Salinity and temperature are environmental factors that affect the results of all effectiveness tests. Salinity affects the hydrophilic- lipophilic balance, and salting-out effects diminish water solubility of ethoxylated surfactants. Lower temperatures tend to increase viscosity of both of} and dispersant as weD as changing solubility of various components. In some cases the effect of temperature can be so great that an of] dispersible at 15°C may not be dispersible at 5°C (Lehtinen and Vesala, 1984; Weds and Harris, 1979~. Sampling and analysis are the last factors to be considered. Be- cause of} drops resurface, the most reproducible results are obtained by sampling while mung is proceeding, or at predetermined times immediately after mixing is stopped (Welis et al., 1984a). Contami- nation of surfaces with of! is frequently a cause for major errors. Need for Standard Testing Oils A number of investigators (Canevari, 1985; Mackay et al., 1986; Fingas, private communication) have expressed the need for a sur- rogate or standard of} for dispersant testing, which would improve reproducibility of product testing and provide international intercal- ibration of methods and products. Canevari (1985) recommended that tetradecane be used. However, Mackay et al. (1986) have shown that long-chain paraffins are a primary inhibitor of effectiveness. Short-chain alkalies also reduce dispersant effectiveness, but not as dramatically, and aromatic compounds increase effectiveness. In view of these findings, a single compound as surrogate of] (e.g., te- tradecane) would not be representative of the dispersion properties of real oils. A multicomponent mixture containing aLkanes and aromatics might be more representative. In earlier work Mackay and Leinonen (1977) constructed such a synthetic oil, consisting of 10 components, to test evaporation rate modeling. Their mixture anticipated chemi- cal and natural dispersion. The surrogate oil consisted of the normal isomers of butane, hexane, octane, decane, dodecane, and hexade- cane, as well as benzene, toluene, naptthalene, phenathrene, and an "inert" component. From 1980 to 1982 the American Petroleum Institute in coop- eration with the U.S. Environmental Protection Agency set aside a number of oils for of] spill testing. They include Pru~hoe Bay and Arabian light crudes and No. 6 and No. 2 fuel oils. The U.S. EPA has analytical data on all of the stored oils. The actual samples 19

74 USING OIL SPILL DISPERSANTS ON THE SEA 55-gal (208-liter) drums of each of six oils are stored* at EPA's Oil and Hazardous Materials Simulated Environmental Test Tank (OHMSETT) facility at Leonardo, New Jersey (Kolde, private com- munication). At the same time, Canada's Environmental Protection Service set aside a reference oil (Environment Canada, 1984; Fingas, private communication). These oils have been available upon request to investigators in Canada, the United States, and other countries. Advantages and Disadvantages of Testing Methods To be useful, a laboratory test should be fairly easy and quick to perform. This is satisfied by shake-flask tests (Abbott, 1983), but not by the more sophisticated tests requiring specialized apparatus and trained operators. On the other hand, these more complex tests are designed to mimic actual field conditions more closely. A test should also be repeatable (within the same laboratory), reproducible (from one laboratory to another), and precise (variation coefficient of less than 20 percent). This is not always easy to achieve. Most significantly, the test should show a good correlation with real dispersant operations at sea. Although serious attempts have been made to mimic sea surface turbulence, so far no test satisfies this · ~ criterion. For example, the French Institute of Petroleum test, which uses a beating hoop just under the surface and changes the water contin- uously, was compared with the WSL Labofina test, using a variety of oil types and dispersants (Bocard et al., 1984; GiDot and Charlier, 1986~. No correlation on the basis of hydrophile-lipophile balance was observed, but rank order of effectiveness was often the same for both tests. Two primary differences are continuous washout of soluble materials in the IFP test, and greater mixing energy of the Labofina test (500 W/m3 versus 1.5 W/m3), although the longer run time of the IFP test (1 to 5 hr) brings total energy input into the same range. The IFP test process, by its continuous removal of water con- taining dispersed oil, tends to simulate at-sea flow conditions while other test processes use a fixed volume of water. This feature of the IFP process can be a disadvantage since the test system is more *Test oil stored at OHMSETT is kept under positive pressure and normal NO2 atmosphere, and stored at the ambient temperature of the facility. Oil is checked for indications of aging or deterioration before testing (Tennyson, private communication).

CHEMISTRYAND PHYSICS OF OIL DISPERSANTS 75 complex to set up and transport. The test has been used in France and Norway, but few researchers are familiar with the process. In addition, the control of turbulence generated by the IFP agitation device is significantly affected by the levier of the water surface. As a result, reproducibility of tests between laboratories is uncertain. Labofina Test The WSL Labofina test uses a standard separatory funnel in which the test fluids are mixed by a mechanical rotator (Martinelli, 1984~. It has one major advantage: it is simple and fast- 16 tests per day can be conducted. It is as reproducible as other tests (variation coefficient of 10 to 14 percent). The Labofina test shows a general bias toward lower effectiveness ratings for many oils owing to the relatively high oil-to-water ratio in this test (Daring, 1988~. However, this test tends to give a relatively high efficiency rating for high-density oils, which do not rise rapidly after agitation has stopped (Daring, 1988~. Byford and Green (1984) report good agreement between the MNS and Labofina tests, when used to identify optimum surfactant combinations. The parameter that caused the greatest effect on the Labofina test was the shape of the conical separatory funnel. Precision of timing was also important: the stopcock orifice diameter affected results since a narrower orifice lengthened the time to collect a 50-m} aliquot of dispersed of] for analysis. Oil adhering to the flask walls can also produce major errors. The major disadvantages of the Labofina test are that it uses an unreaTisticaBy high oil-water ratio, and mixing does not simulate the turbulence of an ocean surface. The samples are collected un- der static conditions, and the results depend on precisely when the samples are collected after mixing stops. Mackay-Nadeau-Steelman Test The MNS test uses a 20-liter closed glass, temperature-controlled tank, with a stream of air blowing tangentially on the water surface to generate reproducible waves and turbulence (Mackay and Szeto, 1981; Mackay et al., 197S, 1984; U.S. EPA, 1984; Wells and Harris, 1979~. It reproduces turbulence that closely simulates actual mix- ing conditions. In this dynamic measurement, effectiveness can be assessed as a function of time, and of} film thickness can be indepen- dently controlled. Airflow rate is critical, however, and satisfactory

76 USING OIL SPILL DISPERSANTS ON THE SEA waves cannot be generated unless the apparatus is precisely level (Byford and Green, 1984~. Reproducibility is good (variation coeffi- cient of 10 to 15 percent), and six tests can be completed per day by one operator. The disadvantages of the MNS test are that its mixing energy, while reproducible, is difficult to quantify, and wave dampening by the materials under study can affect the results. It also gives anoma- lous results at 0°C, because of viscous shearing (Fingas et al., 1987~. The airflow rate and direction must be critically adjusted, the appa- ratus is complex, and since it is hand-made tends to be expensive. The MNS and WSL tests were compared in a study involving 13 dispersants and two of} types (Daring, 1986~. Low-performance products were rated similarly by both tests, but the ranking of more effective products was generally different. The MNS test seems to be more reliable for low-viscosity oils, while the WSE test gives information for the higher-viscosity range. A recent comparison of effectiveness for 10 oils using the MNS, Labofina, and swirling flask methods demonstrated a poor correlation between results (Fingas et al., 1987~. U.S. Environmental Protection Agency Test The EPA test uses a 130-liter stainless steel tank,* that, with its larger water volume, allows lower oil-water ratios (U.S. EPA, 1984~. Use of a spray simulates application in the field from a boat. In this test also, effectiveness can be measured as a function of time, and samples can be obtained under dynamic conditions. Reproducibility is good for one operator using a consistent procedure (variation coefficient of less than 10 percent). Critical factors include rate of application of dispersant to the oil surface, height above the surface from which it was poured, exact height of the jet spray nozzle, and need for care to avoid splashing of} onto the tank wads. Disadvantages of the test are that mixing energy (from the water jet) is not only Circuit to quantify, it is difficult to reproduce. The test is sensitive to technique (Payne et al., 1985~; some operators produce more reproducible results than others. High-shear turbulence is generated by the circulation pump, *One reviewer recommends using a glass test container since it has hydrophilic properties; use of Lass would remove oil adherence problems that are common to stain- less steel containers.

CHEMISTRY AND PHYSICS OF OIL DISPERSANTS 77 and this affects results. The apparatus requires a skilled operator, is difficult to clean, and generates a large volume of contaminated water that must be disposed of. Only two tests per day are feasible. Flume or Wave-Tank Tests Flume or wave-tank tests do the best job of simulating turbu- lence from breaking or nonbreaking waves and currents even in ice-covered waters (Brown et al., 1987; Delvigne, 1985; To et al., 1987~. A flume, with its large volume, permits low oil-water ratios and greatly reduces wall effects. In a flume, the resurfacing of dis- persed of} droplets can be studied, and droplet size distributions in the water column can be measured in the course of the test. The disadvantages of a flume are obvious: it is expensive, com- plicated, and not portable. Large volumes of water are required, and the contaminated water must be heated a.nCI disposed of. Bocard et al. (1984) and Bocard and Castaing (1986), noting the dissolution processes in at-sea trials, suggested that laboratory tests should incorporate a flow-through seawater system because closed vessels cannot duplicate the dilution process that occurs in the field. It would be desirable to separate the rate of dispersion from the advective loss of dispersed droplets. The IFP dilution test was designed to do this. Four grams of of! are agitated on the surface of water in a 4-liter reactor; the water is replaced continuously at 0.5 liter/hr, and the percentage of of! washed out is measured with time (Bardot et al., 1984; Demarquest et al., 1985~. Deivigne (1985) concluded that one shortcoming of most labo- ratory tests is that evaporation, photooxidation, emulsification, and nonhomogeneity of of} layer thickness cannot be modeled. Therefore, in the Delft flume experiments, he evaluated the following parame- ters: evaporation (0, 12, and 30 percent) in a pan outside the flume; · photooxidation by ultraviolet exposure for 0.2 and 10 hr; · emulsification by premixing the of} and water in laboratory beakers to a total water content of 70 percent; and · layer thickness of 0.1, 0.5, and 2.5 mm. Visual observations during the flume experiments showed that dispersant droplets barely incorporating of! had moved into the water column after fading or slowly sinking through the of! layer. The oil layer contracted into small slicks with open areas (herding) immedi- ately after the dispersant droplets hit the of} and water surface.

78 USING OIL SPILL DISPERSANTS ON THE SEA Evaporation did not seem to affect dispersed of] concentrations with either naturally or chemically dispersed systems. It appeared, however, that of} droplet size may have decreased slightly with some evaporation. Photochem~cal oxidation increased naturally dispersed oil con- centrations, with no change in the chemically dispersed concentra- tion. In both cases, however, droplet size distribution shifted to smaller volumes, presumably due to the formation of surface-active compounds in the of} slick, which lowered the oil-water interfacial surface tension (Payne and Phillips, 1985~. Emulsification (i.e., premixing the of} and water) decreased dis- persed of} concentration in naturally dispersed experiments with Statfjord crude but had no effect in the chemically dispersed tests (DeIvigne, 1985~. Oil droplet size increased with emulsification in the naturally dispersed slick, but the of! droplet size did not increase despite emulsification in the case of chemical dispersion. Layer thickness effects on both dispersed of! concentrations and of} droplet sizes were minor. Deivigne concluded that his flume ex- periments showed that the lack of dispersant effectiveness in field tests cannot be explained completely by the various parameters ma- nipulated in his study. Summary Different laboratory tests give consistent results in discriminat- ing broadly between high- and low-performance dispersants. Each test has special advantages and disadvantages; all give reproducible results, although some mimic field conditions better than others. Each test appears to measure different physical and chemical phe- nomena in the sense that the weight assigned to, or simulation of, processes and effects, such as oil-dispersant mixing, turbulence, drop size, distribution, en cl resurfacing tendency, are quite different. It is thus not surprising that they rank the most effective dispersants differently, corresponding to different performance criteria. There is a consensus that it is impossible to simulate, in a labo- ratory system measured in tens of centimeters, turbulence character- istics that exist at the oceanic air-water interface, with its turbulent eddies Bunt breaking waves. It is clearly necessary to introduce some turbulence in laboratory tests to promote and maintain dispersion, but an entirely satisfactory method of accomplishing this has not

CHEMISTRY AND PHYSICS OF OIL DISPERSANTS 79 yet been devised. Even sophisticated, large-scale, experimental wa- ter tank systems cannot claim to simulate closely the ocean surface turbulence (Bonwmeester and Wallace, 1986a; Brown et al., 1987~. No attempt is usually made in laboratory tests to simulate photoox- idation. Only in the IFP tests is diffusive dilution simulated. A recent evaluation (Anderson et al., 1985) has compared a num- ber of effectiveness and toxicity tests, and evaluated them. There is no strong correlation between laboratory and field tests (see Chapter 4~. A simple strategy for screening dispersants is to apply a reliable test such as the rotating flask test or the MNS test (see CONCAWE, 1986~. NEED FOR RESEARCH There has been a regrettable lack of input into dispersant re- search on the basic interactions of dispersants with of} by profes- sional surfactant scientists, with the obvious exception of the notable contribution by those who have formulated the products. Unfortu- nately, little of the commercially funded research has appeared in print. Had there been a complimentary program of research in this area, the state of knowledge could have been greatly advanced. As a manifestation of the absence of research, no references have been found to recent papers in the peer-reviewed surfactant litera- ture, for example, the Journal of CoZioid and Interfacial Science or Journal of Dispersion Science and Technology. A considerable liter- ature exists in journals and texts, such as that by Eicke and Parfitt (1987), on the behavior of surfactants in hydrocarbon and water sys- tems. The thrust of most oil spill related studies has apparently been to determine if existing commercially available products work in the laboratory and at sea when applied by conventional methods. This would have been appropriate if it had been found that dispersants work in an efficient, predictable manner. In reality, the performance is often in doubt (see Chapter 4~. It can be argued that more ef- fort should have been devoted to determining why dispersants do, or do not, work on different oils under different application conditions. This research did not occur. As a result, there is an inadequate understanding of dispersant phenomena. A program of research is needed to elucidate the mechanisms by which droplets of dispersant contact an of! film, mix and pene- trate into it, how the surfactant forms various phases with the oil and migrates to the oil-water interface, and the microscopic process

80 USING OIL SPILL DISPERSANTS ON THE SEA by which emulsions actually form. This appears to be a complex transient process and will be difficult to observe, but a knowledge of these phenomena is fundamental to determining why dispersants work, and at times why they do not work. For example, it appears that oil composition (as distinct from physical properties) affects dispersibility, but the reason for this is not known. It is still not too late to start a program to Fit this lack of understanding.