Crude Oil Spills Disappearance of Aromatic and Aliphatic Components from S.mall Sea-Surface Slicks Wyman Harrison*
Energy and Environmental Systems Division, Argonne National Laboratory, Argonne, 111. 60439 Mitchell A. Winnik and Paul T. Y . Kwong Lash Miller Laboratories, University of Toronto, Toronto, Ont., Canada M5S 1 A l
Donald Mackay
Institute for Environmental Studies, University of Toronto, Toronto, Ont., Canada M5S 1 A 4
Experimental data are presented for the weathering of five small (1.04 m3) ocean spills of South Louisiana crude oil. The oil was spiked with cumene and the concentrations of cumene and several alkanes were measured for up to 5 hr after the spill. Comparison of the rates of loss of cumene and n-nonane, which have similar volatilities but different solubilities, is made in an attempt to quantify the relative rates of evaporation and dissolution. An approximate model of the evaporation-dissolution process is derived which suggests that cumene is lost principally by evaporation. The effects of whitecapping and the existence of different weathering rates in the same spill are described.
Water-soluble aromatic and aliphatic hydrocarbons may have sublethal effects on marine organisms a t concentrations of 10-100 ppb, lethal toxicity a t 0.1-1.0 ppm for most larval stages, and lethal effects a t 1-100 ppm for most adult organisms ( I ) . I t is thus of interest to investigate the rates of disappearance of specific aromatic and aliphatic components of crude oil slicks under conditions of essentially constant water temperature and nearly constant air temperature. Previous investigations of the “weathering” of crude oil slicks on the sea have demonstrated that all of the lowerboiling components evaporate or dissolve within a few hours of slick initiation (2-5). Although little is known of the relative percentages of loss of these slick components due to evaporation and dissolution, it is assumed that they are mainly lost by evaporation, a t least under conditions of low sea-surface roughness. Detailed studies of the fate of slick components during the early stages of slick aging are crucial because the lower-boiling fractions contain almost all of the lethal components of the slick (6). Although the individual spills in this study are small (275 U.S. gal), they adequately represent chronic coastal waters oil spills typical of crude oil transfer accidents due to the coupling and uncoupling of hoses a t bunkers or pumpout of oil-tank waters. Some extrapolation can also be made to predict the behavior of larger oil spills. This study had thus the dual objective of developing the methodology for investigating the fate of aromatic and aliphatic hydrocarbons in a slick and of obtaining data on the rates of loss under typical environmental conditions. Experimental Two (kO.1) U S . gal of liquid were removed from each of the twenty-five 55-gal. drums of South Louisiana crude oil. This was replaced with an equal volume of Phillips
technical grade (est. >98%) cumene (isopropyl benzene). The resultant solutions were approximately 0.34 mol/l or 4.2% cumene by weight. Acting as an internal standard and tracer, cumene made it possible to monitor the disappearance of one of the many aromatic components of each spill. Cumene has the advantage that under oil-spill conditions it has a “half-life” of about 20 min, thus providing a reasonable time for sampling. It also has an appreciable solubility in water, estimated to be about 50 mg/l (7). (A slightly lower solubility is expected in seawater.) Cumene should thus be detectable in the water beneath a slick if there is appreciable dissolution. Lower aromatics, such as benzene or toluene, are experimentally inconvenient as tracers since they evaporate too fast, whereas higher aromatics or paraffins are inconvenient because they are too insoluble in water to be readily detectable in solution. Cumene is thus a convenient tracer to identify the rate of loss of toxic aromatics from a spill. The rate of loss of other aromatics can then be estimated by taking into account the relative vapor pressure and aqueous solubility. Each 275-gal (1.04 m3) spill of crude oil was made by lining up five 55-gal drums on the stern rail of the research ship and opening the bungs simultaneously. The ship’s rail was 1.3 meters above the sea surface and the barrels emptied in about 3 min. The ship maintained a very slight headway to prevent the spreading oil slick from engaging the hull or being stirred into the water column by action of the ship’s propeller. A 10-in. diameter glass funnel, fitted with a stopcock and handle, was used to scoop samples of each slick off the water surface. The seawater component of each sample was then immediately run out through the stopcock and the remaining oil samples were run into 6-oz brown, wide-mouth Boston round bottles. The bottles were covered immediately with tight-fitting caps lined with aluminum foil. At least 10 ml (but usually 100 ml) of oil were collected. We explicitly assume that scooping the oil samples did not perturb the composition of the oil. Samples were kept in the funnel less than 1 min before completing transfer to the bottle. The large sample sizes diminish the likelihood that evaporative losses were significant, relative to those of natural slick aging. Sampling of crude oil was confined to the thickest (oilpool) regions of each slick. All sample bottles were sealed with “Parafilm” and kept a t 0°C until ready for GLC analysis. Sodium azide (100 mg) was added to each sample upon return to the laboratory to prevent bacterial growth in the aqueous phase. Crude oil (5 ml) was diluted with 10 ml of reagent grade n-pentane and filtered to remove trace amounts of precipitated asphalts. Portions of this solution (2-3 111) were injected onto the column of a Varian Aerograph (Model Volume 9,Number 3 , March 1975
231
1400) gas-liquid chromatograph with a flame-ionization detector. The temperature was programmed manually to achieve a temperature rise of 4"C/min between 60" and 300°C. Peaks on the GLC trace for cumene, nonane, dodecane, and docosane were identified by adding minute quantities of pure material to a sample of crude oil prior to injection. Analysis of the GLC traces indicated that the area ratios of heptadecane (c17),octadecane (CIS), nonadecane (CIS), and eicosane ( C ~ O were ) constant. The C17 peak was the largest and most symmetrical; it was chosen as the base peak for subsequent GLC trace analysis. Peak heights gave results identical to those obtained from peak areas. Values of peak-height ratios, Cm/C17 were calculated from the n-alkanes of length m and for cumene for each sample. GLC traces were averaged from each of two samples, taken from the cumene-spiked crude oil before spillage, to obtain time-zero values. The ratio R = (C,/ C17)t/(Cm/C17)r = O represented the fraction of m remaining a t time t after the spill. Calibration of the GLC method by triplicate injections of several samples indicated a standard deviation of k0.15 R. This value is taken as the typical uncertainty in R for the GLC analysis of all other samples. One-litre samples of the water under slicks 3 and 4 (Figure 1) wdre taken with a Van Dorn sampler a t a depth of 0.5 meter. The dissolved and emulsified hydrocarbons were extracted into about 50 ml of carbon disulfide and stored in tightly sealed bottles. These solutions were later analyzed by GLC by injecting 5 rl samples. Cumene was identified by retention time. Only qualitative results are presented here for these analyses since the cumene levels were close to the detection limits. Further, it is believed that substantial errors may be introduced by contamination during sampling and by incomplete extraction. It has also been suggested by Mackay and Wolkoff (8) that analysis of aqueous solutions of lower hydrocarbons by extraction may introduce substantial errors through evaporation. Pertinent conditions for the waters off the south shore of Grand Bahama Island where the experiments were conducted in February 1973 were as follows: water temperature, essentially constant a t 23.6%; air temperature, 20.5-24.1"C; relative humidity, 60-79%; wind (measured a t an elevation of 3 meters with a Belfort wind measuring system), calm to 18 mph, with gusts to 22 mph (9.8 m/ sec); and sea-surface conditions ranging from calm with gentle swell to extensively whitecap covered. All slicks were sampled until it became impossible to find coherent oil pools. Of particular interest in interpreting the experimental results are the relative disappearance rates of cumene and nonane. These hydrocarbons have very similar vapor pressures [cumene 4.2 and nonane 3.9 mm Hg a t 23.6"C ( 9 ) ] ; thus, their evaporation characteristics will be similar. Cumene is, however, considerably more soluble in water [50 vs. 0.22 mg/l (IO)]; thus, it will dissolve in water at a much greater rate. A comparison of the relative rates of loss of these hydrocarbons may yield information on the contributions of evaporation and dissolution as mechanisms for the removal of specific hydrocarbons from oil spills.
Mathematical Model Although the complete mathematical modeling of the fate of hydrocarbon components from an oil spill is presently not possible, it is instructive to develop a simple model which quantifies the approximate relative importance of evaporation and dissolution. 232
Environmental Science & Technology
Consider a slick with a thickness corresponding to M mol/m2 and containing a component of mol fraction x t . This compound has a saturation vapor pressure Psi (atm), activity coefficient y,,and aqueous solubility St (mol/m3). The rate of evaporation can be expressed by a mass transfer coefficient K E (m/sec) using a corrected Raoult's Law ) R the gas constant is 82 X 10-6 as K E X , ~ , P , / ( R Twhere m3atm/mol K . The rate of dissolution can similarly be expressed by a liquid phase dissolution mass transfer coefficient Ko(m/sec), if it is assumed that the water immediately in contact with the oil is in solution equilibrium with the oil. The interfacial aqueous concentration of the compound will thus be approximately S l x i y l (m0l/m3). This assumption has been shown to be justified by Leinonen and Mackay (12). The rate of dissolution will thus be K , T J S J mol/m2 ~ ~ ~ sec. If we assume that the oil slick is well mixed, the differential equation describing the change in x t with time will thus be d ( M x , ) / d t = -K,x,y,Ey/(RT) - K,S,X,Y, Rearranging, and assuming that M is constant and integrating between limits of t = 0, x i = x l o and t = t x i = x i , give xi = xio exp (-Kt) where K = ( K Ey,PS,/RT
+ KuSly,)/Ms-l
The time a t which 63% of the component has been lost will thus be l/KS.
I--
MINUTES AFTER SPILL Figure 1. Percent of low-boiling components of South Louisiana crude oil remaining in slicks and total wind movement ( k m / h r X h r ) as a function of time. Moment of onset of widespread whitecapping shown by arrow for slicks from spills 2 , 4 , and 5
The assumption that M is constant is invalid since as the spill spreads, M decreases in proportion to the spill thickness. Also as the oil evaporates and dissolves, the amount of oil remaining decreases, and this causes a further decrease in M and causes a change in the oil composition. It is presently impossible to quantify these effects accurately since this would require information on the spill area as a function of time and on the oil composition. Whatever the variation of M is with time it will have a similar effect on both cumene and nonane; thus, comparison of the relative rates of loss of these hydrocarbons will yield information on the relative importance of dissolution and evaporation. A mean value of M can thus be used for this approximate analysis. Fortunately, bo!h cumene and nonane are fairly volatile, about 85% of the oil being less volatile than these hydrocarbons. As a result, the effect of their concentration being increased as the more volatile material is lost is relatively small. The use of these equations to interpret the data is thus justified although it is invalid to use them for prediction of absolute rates of loss. Implicit in this model is the assumption that diffusion or mixing in the oil slick is fast relative to the evaporation or dissolution rates. This is increasingly valid as the spill spreads and becomes thinner. Only if the oil is fairly viscous is it likely that there will be a buildup of appreciable concentration gradients in the slick. In the present experimental conditions, M corresponds to about 1 m3 spread over an average of l o 3 m2 and is thus approximately 5 mol/m2. Mackay and Matsugu, (12) have correlated KE against wind speed and pool size, typical values being about 10-2 m/sec. Assuming the activity coefficient to be unity and taking a vapor pressure of 5.3 x 10-3 atm (4.1 mm Hg) the group K E ~ , P , / ( R T Mwill ) have a value at 297 K of approximately 5 X 10-4sec-l; 63% of the component will thus be evaporated in 2000 sec, or 33 min. This calculation is very sensitive to the value of M selected and since M changes during the spreading process, such times can be predicted accurately only if M is known as a function of time. The above analysis suggests that 33 min will be a typical time for 63% of a component such as cumene or nonane to evaporate from a spill of thickness 1m m . Little is known about the value of KO,the dissolution mass transfer coefficient. Liss and Slater (13) have suggested that a typical liquid phase mass transfer coefficient a t a sea surface is 5.5 x m/sec. Probably a lower value will be encountered due to turbulence damping under an oil sliclk. When we use the above figure, the group K ~ l s , y , / Mwill have a value about 5 X 10-6 sec-1 for cumene. This is about two orders of magnitude lower than the evaporation rate. For nonane the dissolution rate will then be about four orders of magnitude slower than the evaporation rate. An alternative modeling approach is to assume unsteady state penetration diffusion of the hydrocarbon into the air and water phases. This is invalid for the air phase since there is continual replacement of the air above the spill. It is more valid for the water phase although there will be some rlelative motion between the oil and the water.
Results and Discussion Details of the ‘(disappearance’’ (evaporation plus dissolution) of the aromatic and aliphatic components are shown on Figurls 1 for the first few hours of slick aging. Cumene and all lower-boiling aromatics disappear within the first 90 min. In spills 1, 2, and 3, the cumene consistently disappears faster than the nonane. In spills 4 and 5,
the results are inconclusive. The time to achieve 63% loss in spill 1 is 13 min for cumene and 28 min for nonane; 25 and 35 min, respectively, in spill 2, and 20 and 27 min, respectively, for spill 3. Averaging these data gives K values of 0.52 min-l for cumene and 0.033 min-l for nonane. Since the dissolution rate of nonane is negligible, it can be assumed that the rate constant of 0.033 min-l reflects only evaporation. The faster rate of loss of cumene can be attributed to experimental error and to three factors, an 8% higher vapor pressure, a higher dissolution rate, and a higher activity coefficient in the oil phase. It is not possible a t present to distinguish between these effects but the dominant effect may be that of the activity coefficient. Since the dissolution rate is slow compared to the evaporation rate, it cannot be determined accurately by subtracting the evaporation rate from the total loss rate. Dissolution rates are thus best determined by measuring dissolved hydrocarbon concentrations in the water under a slick. The analytical technique used to determine aqueous cumene concentrations was insufficiently sensitive to give quantitative data below 1 mg/l. In two samples no hydrocarbons were detected and in two samples only cumene was detected, a t a concentration of about 0.3 mg/l. It is thus concluded that the dissolution rate is considerably slower than the evaporation rate and that only organisms in water that are or have been in close proximity to a spill for an extended time are likely to suffer toxic effects. The measurements of the composition of the oil slick indicate that the dissolution rate is less than half the evaporation rate. The model indicates that the dissolution rate may be as low as 1% of the evaporation rate, with the dissolved cumene concentrations tending to confirm this figure. More accurate estimates of dissolution rates can only be obtained by more sensitive analysis for dissolved hydrocarbons. It is suspected that most of the dissolved aromatics from small slicks are lost by evaporation from aqueous solution. The most profound biological effects will occur under large, thick slicks and under quiescent conditions in shallow water when substantial concentrations of aromatics may be achieved in the underlying water and maintained for a considerable time. There is evidence in spill five, from the apparent increase in cumene concentration, that different pools of oil in the slick weathered a t different rates. Variation in weathering in different parts of the same slick is an important problem. It is probably associated with uneven spill thickness. It can be inferred that naphthalene, which has the same boiling point as dodecane, should disappear in 3-8 hr, depending on wind conditions. Thus, the majority of the toxic fractions (boiling point below 220°C) are inferred to disappear from a typical slick some 3-8 hr after a spill, most having disappeared a t the earlier end of this time interval. The relatively rapid decrease in the concentrations of CIZ and C13 in slicks 2 and 4 is related to the sudden onset of capping on the sea surface during a brief gusty period. This has also been observed by Smith and MacIntyre ( 4 ) . These rates of disappearance can be compared to those of slick 5 , in which the wind speed increased very gradually until about 115 min after the spill when widespread whitecaps suddenly appeared. It would seem, therefore, that there is a significant discontinuity in the rate of disappearance of the aromatic and aliphatic. components between wind speeds below and wind speeds above those which cause the onset of extensive capping. Possibly, evaporation is suddenly enhanced by increased air turbulence, and the spill may undergo a significant area increase. It would thus seem desirable to develop Volume 9, Number 3, March 1975 233
loss-rate curves for the two different sea-state-roughness regimes, one in which there is little or no whitecapping and one which displays extensive capping. A further complication is that the greater turbulence during capping may result in the formation of oil-in-water and water-inoil emulsions. In conclusion, data have been presented for the change in specific hydrocarbon concentrations in small oil slicks undergoing weathering in typical open sea conditions. It is suggested that the relative rates of dissolution and evaporation can be estimated by spiking the oil with an aromatic such as cumene. Unequivocal determinations of dissolution rates can only be achieved by measuring dissolved hydrocarbon concentrations to levels as low as about 0.1 mg/l in the water under the slick. This was attempted with only partial success in the present study. It is suggested that future studies should employ the same technique but with more sensitive water sampling and analytical techniques. An approximate interpretive model has been developed to quantify evaporation and dissolution rates from a slick. The acceleration of weathering due to the onset of whitecapping and the variation in weathering in different parts of the same slick have also been observed experimentally. It is hoped that experimental studies such as this may yield observations and data which will be used in the development of more complex and realistic models of the physical, and ultimately biological, behavior and effects of oil spills on the oceans.
Acknowledgments Our thanks to C. L. Smith for reviewing an early draft of the manuscript and to W. Y. Shiu for assistance with analyses.
Literature Cited (1) Moore, S. F., Dwyer, R. L., Katz, A. M., Mass. Inst. Tech. Sea Grant ReDt. 73-6. 1973. (2) James, W. P., et al., Texas A&M Univ. Sea Grant Prof. Rep. 73-201, College Station, Tex., 1973. (3) Kinney, P. J., Button, D. K., Schell, D. M., Proc. Joint Conf. Prevention and Control of Oil Spills, Amer. Petrol. Inst., New York, N.Y., 15-17, 1969. (4) Smith, C. L., MacIntyre, W. G., ibid., Washington, D.C., 1971. (Bj-Skadier,H . O., Mikolaj, P . G., ibid., p 475, 1973. (6) Blumer, M., Enuiron. A f f a i r s , 1 ( l ) ,54 (1971). (7) McAuliffe, C . , J . Phys. Chern., 70, 1267-75 (1966). (8) Mackay, D., Wolkoff, A. W., Enuiron. S e i . Technol., 7, 611-14 (1973). (9) Zwolinski, B. J., Wilhoit, R. C., “Handbook of Vapor Pressures and Heats of Vaporization of Hydrocarbons and Related Compounds.” Amer. Petrol. Inst. Project 44, Washington, D.C., 1971. (10) McAuliffe, C., Science, 158,478-9 (1969). (11) Leinonen. P . J., Mackay, D., Can. J . Chern. Eng., Sl, 230(1973). (12) Mackay, D., Matsugu, R. S., ibid., pp 434-9. (13) Liss, P . S., Slater, P. G., N a t u r e , 247, 181-4 (1974).
Received f o r reuieu M a y 14, 1974. Accepted N o u e m b e r 4, 1974. Work supported by Erindale College iUnzniuersity of Toronto) and Seabulk International Inc.
Cumulative Chemical Light Meter G. D. Dixon* and D. H. Davies Westinghouse Research Laboratories, Pittsburgh, Pa. 15235 J. D. Voytko Westinghouse Environmental Systems, Monroeville, Pa. 15146
A cumulative light meter has been devised based on the light-induced degradation of aqueous polymer solutions. The viscosity of the solution can be related to the total amount of incident light absorbed. The system possesses a number of significant advantages over competitive actinometers. It is very cheap, needs no extra equipment or skilled personnel, and can be simply modified to respond to any desired spectral region.
Quantification of total incident light provides information about the health of an ecosystem by relating its typical primary productivity to the amount of available energy. Pitts and co-workers ( I ) have developed a chemical actinometer, suitable for such field studies, which consists of a thin film of polymethyl methacrylate in which is dissolved o-nitrobenzaldehyde. Radiation, between 280 and 410 nm, causes photoisomerization to produce o-nitrosobenzoic acid with a quantum efficiency of 0.5. Another system (2) is based upon the photodimerization of anthracene. Since the anthracene is soluble in benzene, but the dianthracene is not, spectrophotometry can be used to measure the amount of anthracene remaining. 234
Environmental Science & Technology
We have developed a light-monitoring device based upon the photochemical degradation of a polymer. The degradation causes a decrease in the molecular weight of the polymer which results in a decrease in the viscosity of solutions containing the polymer. If one now carries out the degradation of a polymer solution in a transparent tube, one can follow the progress of the degradation by measuring the viscosity of the solution using the GardnerHoldt method of time of rise of a bubble in the solution. The viscosity can then be related to the quantity of light which has fallen on the tube, which means that the tubes can be examined in the field if necessary, a stopwatch being the only apparatus needed. It is well established that the spectral sensitivity region needed for the photopolymerization of organic monomers can be altered by using reactive dyes. We have extended this principle to cause the degradation of polymers. By using dyes, one can then use radiation in the visible part of the spectrum to initiate these reactions. The polymer chosen was hydroxyethyl cellulose (HEC) which is susceptible to photodegradation. It is also water soluble and does not precipitate out of solution a t high temperatures. Further, the system does not suffer from an “inner filter” effect often found in actinometer systems. This is because the polymeric cellulose ether fragments have absorption characteristics similar t o the original polymer.