Professional Documents
Culture Documents
To cite this article: Steven R. Hanna & Peter J. Drivas (1993) Modeling VOC Emissions and
Air Concentrations from the Exxon Valdez Oil Spill, Air & Waste, 43:3, 298-309, DOI:
10.1080/1073161X.1993.10467134
During the two-week period following the Exxon Valdez oil spill in March 1989 The primary objective of this study
in Prince William Sound, Alaska, toxic volatile organic compounds (VOCs) evapo- was to track the emission and atmos-
rated from the surface of the oil spill and were transported and dispersed throughout pheric transport and dispersion of VOCs
the region. To estimate the air concentrations of these VOCs, emissions and dis- in the atmosphere during the two week
persion modeling was conducted for each hour during the first two weeks of the period following the spill. Estimates of
spill. hourly VOC air concentrations were
A multicomponent evaporative emissions model was developed and applied to made for 15 compounds (benzene,
the oil spill; the model considered the evaporation of 15 specific compounds, in- ethylbenzene, toluene, o-xylene, m&p-
cluding benzene and toluene. Both mass transfer from the surface of the spill and xylenes, n-hexane, n-heptane, n-oc-
diffusion through the oil layer were considered in the emissions model. Maximum tane, n-nonane, n-decane, n-unde-
emissions of toluene were calculated to equal about 20000 kg/hr, or about 5 g/m2 cane, n-dodecane, n-tridecane, n-
hr, at a time of eight hours after the initial oil spill. Meteorological data were acquired tetradecane, and n-pentadecane).
from several sources and used to estimate hourly-averaged wind velocity over the
spill. Air concentrations of specific components were calculated using the ATDL
area source diffusion model and the Offshore and Coastal Dispersion (OCD) model. Data Collection Methods and Results
Maximum hourly-averaged concentrations were predicted not to exceed 10 ppmv In order to calculate VOC concen-
for any compound. trations in the air downwind of the oil
spill, several types of data are useful,
as outlined in the following two sec-
tions.
A ten million gallon oil spill oc- documented.1 There are many ongoing
curred during a five-hour period on research projects sponsored by govern-
March 24, 1989, when the T/V Exxon ment agencies and by the oil industries Emissions Data
Valdez ran aground on Bligh Reef in involving the environmental effects of Emissions data were required for the
Alaska's Prince William Sound. Most the liquid oil.2'3 However, the envi- 15 VOCs that were expected to evap-
of this oil was never recovered, and ronmental impacts of volatile organic orate relatively quickly from the oil
was spread by wind and ocean currents compounds (VOCs) that evaporated into spill. It would be useful to know the
over an area of hundreds and, ulti- the air during the first two weeks after liquid concentrations of the 15 com-
mately, thousands of square kilome- the spill are more subtle and have been pounds in the initial Exxon Valdez oil
ters. Figure 1 shows the Prince William less well-documented. It is generally spill (i.e., in the tanker) and, if pos-
Sound area, including the location of recognized that as much as 25 percent sible, at time increments during the two
the oil spill and the locations of several of the oil may evaporate during this weeks after the spill. Although thou-
meteorological stations. period,1 and all of the lighter VOCs, sands of samples were collected from
The direct impact of the liquid oil such as benzene, will evaporate within the Exxon Valdez oil spill, many of
on animals and plants has been well- a few hours after the spill. the samples suffered from QA/QC
problems, and these data were not being
distributed. We had to rely on gener-
alized historical data on the composi-
tion of Prudhoe Bay crude oil and the
Implications
weathering of crude oil.4-6 Some of the
The purpose of the research reported in this paper was data in the Payne et al. report6 refer to
to reconstruct the time and space variation of air con-
centrations of volatile organic compounds (VOCs) re- the Exxon Valdez oil spill, but the em-
sulting from the Exxon Valdez oil spill on March 24, phasis of that report is on the heavier
1989, in Prince William Sound, Alaska. The predicted hydrocarbons and insufficient infor-
maximum concentrations of VOCs did not exceed 10 mation is given on the 15 compounds
ppmv for any compounds. It is emphasized that this of interest.
exercise would have been unnecessary if networks of Maps of observations of the position
VOC monitors had been installed in the area of the ac-
cident within 12 hours after the initial spill. and size of the oil spill were obtained
Copyright 1993 — Air Waste Management Association
coefficient, with a Raoult's law liquid- To accurately calculate the total z = vertical distance in the
gas equilibrium at the liquid surface. number of moles, nT, Equation 2 would liquid layer (m)
This approach has been used in nu- have to be applied to each of the
To simplify the analysis, the rate of
merous oil spill and6 1liquid pool evap- hundreds of compounds in crude oil,
change of mass with vertical distance,
oration m o d e l s . 2 The primary and then the individual losses of each
dm/dz, in Equation 4 can be estimated
assumption is that the liquid layer is compound would have to be summed
by the approximate solution to the
well-mixed, and evaporation is thus to obtain the total number of moles
transient diffusion equation for a semi-
controlled by mass-transfer from the evaporated. To simplify the calcula-
infinite solid (Carslaw and Jaeger):15
liquid surface to the air. The basic tion of the total number of moles re-
evaporation equation is: maining, the assumption is made that m0 m0
the total molar evaporation rate will m, =
e 2.718
f decrease in an exponential manner.
These estimates are consistent with
Equation 2 and with the experimental when Az =
where: kg = mass transfer coefficient measurements of Payne et al.5 Thus,
the assumption is made that nTj/nT>0 =
Equation 4 can thus be approxi-
(gmoles/m2-atm-hr) mated by:
A = spill area (m2) e"kt, where nT>j and nT>0 are the total of
Pi = pure component vapor moles at hour j and hour 0, respec-
pressure (atm) tively. Because the total number of ~± = -0.632^) (5)
x . = liquid mole fraction of
moles usually changes very slowly, the
compound i exact time dependence assumed for to-
tal evaporation loss does not affect the For any given hour of the spill, the
nj = gmoles of compound i ratio (A/V) will be approximately con-
remaining results significantly.
stant. Equation 5 can thus be easily
nT = total gmoles of oil re- In summary, to calculate the mass- integrated over each one-hour time pe-
maining transfer-limited evaporation loss, hourly
values of wind speed (u), temperature
riod to yield the mass loss with time
t = time (hr) for a diffusion-limited evaporation sit-
(T), and spill area (A) are used to cal-
The emission rate in terms of mass is culate the mass remaining in Equation uation:
simply: 2 for each hour. The evaporation cal-
culations are then done sequentially, j - , exp (-1.264(-j VDlfVtJ -
_dnii with the calculated value of mass re-
where niy is the mass of compound i
(1) maining of compound i at the end of
\ remaining in hour j . Equation 6 is the
hour j , nijj, being used as the initial
mass for the following hour. The av- diffusion-limited evaporation rate
where: mj = mass of compound i re- expression analogous to the mass-
maining (g) erage emission rate of compound i for
each hour, E{>j, is then calculated using transfer-limited expression in Equation
Because wind and temperature data 2. The average hourly emission rate,
are typically recorded hourly and the Equation 3. '
Eifj, is again simply expressed by
spill area and total number of moles do Equation 3.
not change significantly over a one- The liquid molecular diffusion coef-
hour time period, the term (kgApi/nT) Diffusion-Limited Evaporation ficient, Dj, can be calculated by the
remains approximately constant during For high-volatility compounds such following correlation for organic/or-
each hour. Therefore, Equation 1 can as benzene and toluene, diffusion in ganic liquid diffusivity (Perry)16:
be integrated hour-by-hour to obtain the the liquid layer may limit the evapo-
solution: ration rate. For these compounds, the 6,
evaporation rate is so rapid that the
surface concentration becomes de-
pleted, and the limit to the evaporation
(2) rate is how fast the compound can dif-
fuse vertically through the thickness of where: D{ = liquid diffusion coeffi-
the liquid layer to the liquid surface. cient of compound i
where j is the indicator for the number
Payne et al.5 discuss the importance of (m2/hr)
of hours since the spill. The mass- (x = viscosity of liquid (cen-
transfer coefficient, kg, in the above diffusion for the evaporation of lighter
hydrocarbons, but do not include any tipoise)
equations is usually based on the work M{ = molecular weight of
of Mackay and Matsugu.13 diffusion-limited effects in their oil spill
model. compound i (g/gmole)
For each hour, the quantities u, A, Pi = liquid density3 of com-
and T are known, and the total number To calculate the effects of diffusion pound i (g/cm )
of moles, nT, does not vary signi- through the oil, consider the standard
cantly. The average evaporative emis- one-dimensional 14 mass-flux equation in Similar to the mass-transfer-limited
sion rate for each hour, in terms of g/ the liquid phase: case, the diffusion-limited evaporation
m 2 -hr, can be calculated by using calculations using Equations 6 and 7
Equation 2 and time-averaging over a are done sequentially, with the calcu-
^ = /M\!Ei (A) lated value of mass remaining of com-
one-hour period. The integrated solu- dt \ V / dz {V
tion is imply: pound i at the end of hour j , m,Jf being
where: D; = molecular diffusivity of used as the initial mass for the'follow-
m;: i
liquid compound i (m2/ ing hour. The average emission rate of
hr) compound i for each hour, Ey, is then
V — spill volume (m3) calculated using Equation 3.
Table II. Overview of Synoptic Situation in Prince William Sound during the Period Because of the limitations of the first
March 24,1989 through April 6,1989 based on NWS Surface Weather Maps and Valdez order NWS stations, it was desirable
Weather Observations. to find wind data from sites closer to
Day Valdez Weather Synoptic Situation the oil spill, in more representative lo-
cations. The four special temporary
March 24 Cloudy Low off Aleutians, Easterly flow over region stations mentioned above were set up
March 25 Mostly Cloudy Low off Aleutians, Front in western Gulf of Alaska, by NOAA for the purpose of providing
Dashed front 100 mi inland, E flow supporting data after the oil spill, but
March 26 Clear in pm, 40° Low off Western Alaska, another in SE Gulf of they were not installed until the last
Alaska, weak front 50 mi inland, E flow two or three days of the two week pe-
March 27 Clear, 25° Low South of Juneau, Cold front over Valdez, riod of interest. Consequently, during
Low forming over Middleton Island the first 11 or 12 days out of the 14-
March 28 Part Cloudy, 27° Low just East of Middleton Island, North winds day period, only the first-order NWS
over area station data were available. In order to
March 29 Part Cloudy, 32° Low still over Middleton Island, North winds, derive better relations among the NWS
front near Middleton Island station data and the special station data,
March 30 Part Cloudy, 30° Low SE of Middleton Island filling, N-NE winds the concurrent wind data during the
over region period April 4 to 15, 1989, were ana-
lyzed (i.e., days 12 through 23 after
March 31 Part Cloudy, Low 200 miles south of Middleton Island, E flow-
the oil spill). As a result of this analy-
34°, Light Snow troughs, weak fronts in area
sis, it was decided that the winds prior
April 1 Clear, 40° Low approaching south of Aleutians, E flow, to April 4 could be interpolated using
possible front south of area the observed winds at Potato Point and
April 2 Clear, 42° Low over Gulf of Alaska, possible inverted trough Middleton Island, and the observed
just west of region movement of the oil slick. The infor-
April 3 Part Cloudy, 40° Big low S-SE of region, E flow mation on the oil slick is available
April 4 Cloudy, 36°, Big low S-SE coming closer, inverted trough to through a series of maps produced by
Light rain W of region ADEC and by NOAA. The NOAA
April 5 Part Cloudy, 40° Big low just to SE, E flow, weak front nearing maps are based on a combination of
northwestern edge of region observations of the oil slick and pre-
April 6 Clear, 45° Low to SE & E, weak front passes through from dictions by their OSSM model. Time
NWtoSE series of figures containing the esti-
mated noontime position of the oil slick
were analyzed. During the entire two-
week period, the oil slick moved in the
same general direction, from NE to SW.
to local mountain ranges. For exam- Island, some brief wind direction shifts The fastest movement occurred on about
ple, the wind flow between Knight and were detected at that station. March 27 and 28, when a cold front
Montague Islands in Prince William passed over the spill area, bringing
Sound is usually from the NNE be- Wind Fields Interpolated for Modeling strong N to NE winds.
cause that is the direction along which As discussed previously, surface
Hourly wind vectors from Potato
the 1000 m mountain ranges on the wind data were acquired for two first-
Point and Middleton Island were drawn
Islands are oriented. The steep terrain order NWS stations (Valdez and Mid-
on a map of Prince William Sound,
also generates katabatic winds, which dleton Island), four routinely operated
and wind vectors over the location of
are gravity flows caused by the cooling stations whose data are archived by the
the spill during that hour were inter-
of air over glacier-covered coastal NWS (Cordova, Seward, Whittier, and
polated by eye. Time series of result-
mountains. These katabatic winds al- Potato Point), and four temporary sta-
ing hourly wind speeds and directions
ways flow away from the coastal tions (Johnstone Point, Dutch Group,
are given in Figures 5 and 6 for the
mountains and are channeled by steep Danger Island, and Sawmill Bay). Fig-
spill area, for Middleton Island, and
and narrow fiords. The effects of these ure 1 shows the locations of these sta-
for Potato Point. The persistent NE flow
local winds usually die off within about tions. Of the routine stations, Seward
over the spill can be seen in Figure 6.
10 km from the shorelines. and Whittier were not included in the
The weather maps and the hourly final analysis, since the winds at those
records of weather conditions at Val- sites reflect local drainage flows and Dispersion Models
dez were used to construct Table II, channeling. The Valdez and Cordova Two dispersion models were ap-
which provides an overview of the data have similar local influences, but plied to assess the air quality impacts
synoptic situation during the period. are retained because these stations are from VOC emissions associated with
This information leads to an under- not far from the location of the spill. the Exxon Valdez oil spill. The At-
standing of why the strong winds over The Potato Point station is the closest mospheric Turbulence and Diffusion
the oil spill area in the northern part to the oil spill, but that station nearly Laboratory (ATDL) area source dif-
of Prince William Sound on March 29 always shows a NE wind direction due fusion model21 was used to assess the
did not reach Middleton Island. The to the orientation of the fiord. The air quality directly over the spill. The
cold front stalled after it passed Valdez Middleton Island station is in the Gulf Offshore and Coastal Dispersion (OCD)
and was dissipated in a low pressure of Alaska and is the least subject to model22 was used for receptors at the
system that formed over Middleton Is- local influences of any of the stations, edge of the spill and over land. Ap-
land. As the low pressure system wan- but is unfortunately about 100 km south plication of a more complex mesoscale
dered back and forth over Middleton of the oil spill. wind field and dispersion model was
Table IV. Maximum hourly averaged concentration (ppmv) calculated over the center of the oil spill, at selected time periods, for each
compound. A dash indicates a concentration less than E-4 ppmv.
Time
March 26 Mar. 27 Mar. 28 Mar. 29 Mar. 31
Compound 01 06 12 18 12 12 12 12 Apr. 2
Benzene 4.9 .92 1.5E-2 _ __ _
Ethylbenzene .70 .33 2.3E-2 — — — — _ _
Toluene 8.2 1.6 3.2E-2 — _ — _ — —
o-xylene .45 .29 3.6E-2 _ _ — _ — —
m&p-xylenes 1.5 .83 7.7E-2 — — — _ — —
Hexane 1.3 .24 6.4E-3 — _ — _ _ _
Heptane 4.4 .82 2.5E-3 — — _ — — —
Octane 4.0 1.2 .05 _ _ — — — —
Nonane 1.4 1.0 .24 .01 — — _ _ _
Decane .42 .36 .25 .10 1.6E-3 — _ _ _
Undecane 9.2E-2 8.3E-2 8.8E-2 6.7E-2 1.7E-2 _ _ _ _
Dodecane 2.9E-2 2.7E-2 3.24E-2 3.0E-2 1.7E-2 7.4E-3 _ — _
Tridecane 1.9E-2 1.8E-2 2.2E-2 2.1E-2 1.4E-2 9.3E-3 2.2E-2 _ _
Tetradecane 3.6E-3 3.3E-3 4.2E-3 4.3E-3 3.4E-3 3.6E-3 2.6E-3 1.3E-3 3.8E-4
Pentadecane 1.0E-3 9.6E-4 1.2E-3 1.3E-3 1.0E-3 1.2E-3 1.0E-3 9.0E-4 6.4E-4
time intervals during the first two These calculations should be 2. D.D. Kelso, M. Kendziorek,
weeks of the spill. compared against similar calcu- "Alaska's response to the Exxon
• The accuracy of the wind field lations by other groups (e.g., oil Valdez oil spill," Environ. ScL
estimates would have been im- companies) once they become Technol 25: 16 (1991).
proved if the mesoscale wind available. 3. A.W. Maki, "The Exxon Valdez
network had been installed in As improved input data are pub- oil spill: Initial environmental im-
Prince William Sound immedi- lished (e.g., concentrations of in- pact assessment," Environ. ScL
ately after the spill, since the dividual compounds in the Technol. 25s 24 (1991).
maximum air concentrations of tanker), the models could be eas- 4. J.R. Payne, G.D. McNabb, Jr.,
VOCs are expected to occur dur- ily rerun to produce revised es- "Weathering of petroleum in the
ing the first few hours. Fortu- timates. marine environment," Marine
nately, during the March 24,1989 Technology Society J. 18: 24
through April 6, 1989 period the (1984).
wind directions affecting the spill Acknowledgments 5. J.R. Payne et al., "Mutlivariate
were nearly continuously from the This research has been sponsored by analysis of petroleum weathering
NE direction and the wind speeds the Department of Environmental in the marine environment-sub
could be easily interpolated from Conservation of the State of Alaska arctic," Volume 21 of Environ-
available NWS data. (ADEC) under Contract Number 18- mental Assessment of the Alaskan
The accuracy of atmospheric sta- 4018-90. The authors appreciate the Continental Shelf, Final Report of
bility and vertical mixing esti- guidance of the project monitor, Mr. Prinicipal Investigators, Outer
mates near the spill would have Gerald Guay. Some of the work de- Continental Shelf Environmental
been improved if instruments were scribed in this paper was carried out Assessment Program, U.S. De-
immediately installed that pro- by Laura Daly of Gradient Corporation partment of Commerce, 1984.
vide measurements of air-water and Donald DiCristofaro and Robert 6. J.R. Payne et al., "Oil-ice sedi-
temperature difference and mix- Mentzer of Sigma Research Corpora- ment interaction during freeze-up
ing depth. The current analysis tion. We also acknowledge the assist- and break-up." U.S. Department
used meager ship reports and dis- ance of dozens of scientists within of Commerce, NOAA, OCSEAP,
tant NWS radiosonde stations to ADEC and other government agencies Final report 64, 1989.
obtain these data. who have shared their data, their re- 7. P.J. Drivas, "Calculation of
To better understand wind flows ports, and their advice with us. In par- evaporative emissions from mul-
in the region, prognostic mesos- ticular, Dr. Gerald Gait and his staff ticomponent liquid spills," En-
cale wind field models should be at NOAA's Hazardous Materials Re- viron. ScL Technol. 16: 726
applied to Prince William Sound. sponse Branch have provided us with (1982).
The effects of the drainage winds much information in hard copy and on 8. W.C. Yang, H. Wang, "Model-
from the surrounding mountains floppy disks regarding the oil spill po- ing of oil evaporation in aqueous
and of channeling by the islands sition and the wind fields. environment," Water Research 11:
could be studied using these 879 (1977).
models. 9. G.M. Torgrimson, "The on-scene
The sensitivity of the estimates spill model," NOAA Technical
of emissions and air concentra- References Memorandum NOS OMA 12,
tions to variations in input data 1. B. Hodgson, "Alaska's big spill: Rockville, MD, 1984.
and model parameters should be Can the wilderness heal?" Na- 10. W. Stiver, D. Mackay, "Evapo-
investigated. tional Geographic 177: 5 (1990). ration of spills of hydrocarbons and