Environmental Toxicology and Chemistry, Vol. 23, No. 10, pp.

2441–2456, 2004
q 2004 SETAC
Printed in the USA
0730-7268/04 $12.00 1 .00

Special Issue Honoring Don Mackay

OIL SPILL IMPACT MODELING: DEVELOPMENT AND VALIDATION

DEBORAH P. FRENCH-MCCAY
Applied Science Associates, Narragansett, Rhode Island 02882, USA

( Received 10 July 2003; Accepted 23 December 2003)

Abstract—A coupled oil fate and effects model has been developed for the estimation of impacts to habitats, wildlife, and aquatic
organisms resulting from acute exposure to spilled oil. The physical fates model estimates the distribution of oil (as mass and
concentrations) on the water surface, on shorelines, in the water column, and in the sediments, accounting for spreading, evaporation,
transport, dispersion, emulsification, entrainment, dissolution, volatilization, partitioning, sedimentation, and degradation. The bi-
ological effects model estimates exposure of biota of various behavior types to floating oil and subsurface contamination, resulting
percent mortality, and sublethal effects on production (somatic growth). Impacts are summarized as areas or volumes affected,
percent of populations lost, and production foregone because of a spill’s effects. This paper summarizes existing information and
data used to develop the model, model algorithms and assumptions, validation studies, and research needs. Simulation of the Exxon
Valdez oil spill is presented as a case study and validation of the model.

Keywords—Oil spill Model Fates/effects Impact Exxon Valdez

INTRODUCTION efforts for impact assessment relied on calculating the inter-

A large number of oil spill models have been developed section of oil trajectories with biota, assuming an impact
over the past 30 years, and several comprehensive reviews of threshold [31–38]. In some cases, wildlife (seabird or marine
oil spill trajectory and fate modeling have been performed [1– mammal) population and migration models were used to sim-
5] to assess the state of the practice, summarize key devel- ulate the distribution, behavior, and recovery of the affected
opments, and project future research. The vast majority of oil species, in conjunction with their intersection with oil trajec-
spill models (the focus of these reviews) are used to predict tories [39–45]. In these modeling efforts, the impact threshold
or hindcast the trajectory [6–10], weathering [11–14], and fate for wildlife appropriately was based on an oil thickness or
[12,15–28] of oil spilled on or near the surface of water, with mass for lethal or sublethal effects, although quantitative in-
the purposes of informing spill response (real time or in drills), formation definitively indicating what dose would be lethal
performing contingency planning, and evaluating the behavior was not available.
and mass balance. Relatively few models have been designed However, for aquatic organisms, impact thresholds based
and used to quantitatively evaluate the impacts of oil on aquatic on surface oil are inadequate, particularly when evaluating the
organisms and habitats (as described below). This paper de- change in impact caused by use of chemical dispersants on
scribes such a model, which represents the state-of-the-art of spills. Mackay and Leinonen [16] and Mackay et al. [17,18]
coupled oil fate and effects model systems. The Spill Impact suggested that meaningful assessments of impacts on aquatic
Model Application Package (SIMAP) was derived from the biota can be accomplished only after quantitative understand-
physical fates and biological effects submodels in the Natural ing of the natural and chemical dispersion processes (entrain-
Resource Damage Assessment Model for Coastal and Marine ment), as the major toxic effect of oil is from dissolved hy-
Environments, which were developed for the U.S. Department drocarbons, and the main dissolution mechanism is via dis-
of the Interior as the basis of Comprehensive Environmental persed oil droplets rather than surface slicks. Based on theo-
Response, Compensation, and Liability Act of 1980, and the retical calculations and modeling results, Mackay and
Natural Resource Damage Assessment (NRDA) regulations for Leinonen [16] showed that: Dissolution from surface slicks is
Type A assessments [29,30]. The technical development in- insignificant and can be neglected safely and dissolution time
volved several in-depth peer reviews, as described in the Final from subsurface droplets less than 100 mm in diameter has a
Rule [29]. The Spill Impact Model Application Package has half-life of less than 13 min. The dissolution rate is very sen-
been developed further for use in impact and ecological risk, sitive to the droplet size (because it involves mass transfer
as well as natural resource damage assessment. The physical across the surface area of the droplet), and the amount of
fates model draws on research and model development in com- hydrocarbon mass dissolved is a function of the mass entrained
mon with the larger universe of trajectory and oil fate models. and droplet size distribution. In turn, these are a function of
The biological effects model includes innovative algorithms soluble hydrocarbon content of the oil, the amount of evap-
to quantify the impact of oil on biological resources: Habitats, oration of these components before entrainment, oil viscosity
aquatic organisms (fish, invertebrates, aquatic plants, plank- (which increases as the oil weathers and emulsifies), oil surface
ton), and wildlife (birds, mammals, reptiles). tension (which may be reduced by surfactant dispersants), and
The reader is referred to the above reviews for discussion the energy in the system (the higher the energy the smaller
of oil trajectory and fate model development. Earlier modeling the droplets). Large droplets (greater than a few hundred mi-
crons in diameter) resurface rapidly, and so dissolution from
* [email protected]. those also is inconsequential.
2441
2442 Environ. Toxicol. Chem. 23, 2004 D.P. French-McCay

Table 1. Definition of four distillation cuts and the eight pseudo-components in the model (monoaromatic hydrocarbons, MAHs; benzene 1
toluene 1 ethylbenzene 1 xylene, BTEX; polynuclear aromatic hydrocarbons, PAHs)

Volatile and Semivolatile Low volatility Residual

Characteristic highly soluble and soluble and slightly soluble (nonvolatile and insoluble)

Distillation cut 1 2 3 4
Boiling point (8C) ,180 180–265 265–380 .380
Molecular weight 50–125 125–168 152–215 .215
Log(KOW) 2.1–3.7 3.7–4.4 3.9–5.6 .5.6
Aliphatic pseudo-components: Volatile aliphatics: Semivolatile aliphatics: Low-volatility aliphatics: Nonvolatile aliphatics:
No. of carbons C4–C10 C10–C15 C15–C20 .C20
Aromatic pseudo-component MAHs: BTEX, 2-ring PAHs: C4-ben- 3-ring PAHs: C3-, C4- $4-ring aromatics: PAHs
name: Included compounds MAHs to C3-ben- zenes, 2-ring to C2- naphthalenes, 3- to 4- with log(KOW) . 5.6
zenes naphthalenes ring PAHs with (insoluble)
log(KOW) , 5.6

Thus, both subsurface oil droplets and dissolved hydro- available for most locations and times of interest. Typically,
carbons explicitly must be simulated (in addition to surface- the most reliable and representative current data are those gen-
floating oil and associated processes) in an oil fates model to erated by a calibrated hydrodynamic model, unless site- and
evaluate exposure of aquatic biota to oil hydrocarbons and event-specific current measurements are available.
biological effects. A prime example is the North Cape oil spill This paper summarizes existing information and the current
of January 1996, which occurred during a severe winter storm state-of-the-art in oil fates and effects modeling, model al-
[46]. The barge North Cape spilled 828,000 gallons (2,682 gorithms and assumptions used in SIMAP, validation studies,
metric tons) of home heating oil (No. 2 fuel oil) into the surf and research needs. Simulation of the Exxon Valdez oil spill
zone on the south coast of Rhode Island, USA. Most of the is presented as a case study and validation of the wildlife
oil was entrained into the water column by heavy surf, resulting impact model. Validation and sensitivity studies indicate the
in high concentrations of dissolved components in shallow main sources of uncertainty, data requirements for an accurate
water that killed millions of water column and benthic organ- simulation, and research needs to improve our understanding
isms [46]. Though models based on surface oiling may be and quantification of oil spill effects.
useful in predicting the trajectory of slicks and impacts on
wildlife and shorelines, they cannot evaluate the extent of PHYSICAL FATES MODEL
impacts to aquatic biota caused by subsurface oil. The three-dimensional physical fates model estimates dis-
Moreover, because the many hydrocarbons in oil have vary- tribution (as mass and concentrations) of whole oil and oil
ing physical–chemical properties (most significantly those re- pseudocomponents (i.e., fractions of the oil treated as a single
lated to solubility and volatility), the oil fates model must chemical with properties representative of the included chem-
separately track chemical classes or pseudocomponents of the icals) on the water surface, on shorelines, in the water column,
whole oil with characteristics typical of the chemical group to and in sediments. Oil fate processes included are spreading
simulate their separate fates [21,28,30,47–49]. Most oil fates (gravitational and by shearing), evaporation from slicks, trans-
models (including SIMAP) employ a Lagrangian particle ap- port, randomized dispersion, emulsification, entrainment (nat-
proach, which enables the modeler to track physical and chem- ural and facilitated by dispersant), dissolution, volatilization
ical property changes as oil weathers, which is particularly of dissolved hydrocarbons from the surface water, adherence
needed when oil is released over time under varying conditions of oil droplets to suspended sediments, adsorption of soluble
[4]. This solution methodology also is used in SIMAP to track and semisoluble aromatics to suspended sediments, sedimen-
organisms’ movements and exposure to the bioavailable com- tation, and degradation. Oil mass is tracked separately for low-
ponents, those that are soluble or semisoluble. er molecular weight aromatics (1–3-ring aromatics), which
Potential and documented impacts of oil in aquatic envi- cause toxicity [50–62], other volatiles, and nonvolatiles. The
ronments have been reviewed by the National Research Coun- lower molecular weight aromatics dissolve from the whole oil
cil [50,51] as well as others [30,52–54]. A biological effects and are partitioned in the water column and sediments ac-
model that considers all impacts would include evaluation of cording to equilibrium partitioning theory.
exposure considering movements and amounts of both oil and In the model, the oil is treated as eight pseudocomponents,
biota; acute effects algorithms for direct impacts (lethal and characterized by volatility (boiling point), hydrophobicity (oc-
sublethal) in the short-term; consideration of sublethal effects tanol–water partition coefficient, Kow), and being aromatic or
of chronic contamination; indirect effects via reduction in food not (i.e., aliphatic): Volatile aliphatics, semivolatile aliphatics,
supply or habitat, or other changes in the ecosystem; behav- aliphatics of low volatility, nonvolatile aliphatics (i.e., in the
ioral changes resulting in reduced growth, survival, or repro- residual oil), volatile aromatics (monoaromatic hydrocarbons,
ductive success; impacts of response activities; and population [MAHs]), semivolatile aromatics (2-ring polynuclear aromatic
level impacts caused by mortality and sublethal effects. Sup- hydrocarbons, [PAHs]), aromatics of low volatility (3-ring
porting research and information is available to quantify some PAHs), and nonvolatile aromatics (in the residual; Table 1).
but not all of these effects, as described below. Six of the components (all but the two nonvolatile residual
Finally, the oil fate model, as well as the biological effects components) evaporate at rates specific to the pseudocompo-
model for planktonic organisms, should be driven by an ac- nent. Solubility is correlated strongly with volatility, and the
curate description of the winds and currents at all significant solubility of aromatics is higher than aliphatics of the same
spatial and temporal scales in order to judge the performance volatility, with the MAHs the most soluble, the 2-ring PAHs
of the oil fates and effects models. Wind data generally are semisoluble, and the 3-ring PAHs slightly soluble [63–66].
Oil spill impact assessment modeling Environ. Toxicol. Chem. 23, 2004 2443

Table 2. Physical–chemical properties for monoaromatic hydrocarbons. Molecular weight (MW), boiling point (BP), solubility, vapor pressure,
and log(KOW) are from Mackay et al. [63]

Vapor
MW Distillation BP Solubility pressure
Compound(s) Rings Cs (g/mol) cut no. (8C) (ppm) (atm) Log(KOW)

Benzene 1 6 78 1 80 1780 0.12534 2.1

Toluene 1 7 92 1 111 515 0.03750 2.7
Ethylbenzene 1 8 106 1 136 152 0.01253 3.1
o-Xylene 1 8 106 1 144 220 0.01155 3.2
p-Xylene 1 8 106 1 138 215 0.01155 3.2
m-Xylene 1 8 106 1 139 160 0.01086 3.2
Xylenes (mixture) 1 8 106 1 140 198 0.01132 3.2
Styrene 1 8 104 1 145 300 0.00868 3.1
Methylstyrenes 1 9 118 1 170 100 0.00264 3.4
1,2,3-Trimethylbenzene 1 9 120 1 176 70 0.00197 3.6
1,2,4-Trimethylbenzene (pseudocumene) 1 9 120 1 169 57 0.00266 3.6
1,3,4-Trimethylbenzene 1 9 120 1 169 57 0.00266 3.6
1,3,5-Trimethylbenzene (mesitylene) 1 9 120 1 165 50 0.00321 3.6
Trimethylbenzenes 1 9 120 1 170 59 0.00262 3.6
n-Propylbenzene 1 9 120 1 159 52 0.00444 3.7
Isopropylbenzene 1 9 120 1 154 50 0.00602 3.6
Ethyl-methylbenzenes (cumene) 1 9 120 1 163 85 0.00367 3.6
Isopropyl-4-methylbenzene 1 10 134 2 177 34 0.00201 4.1
Butylbenzenes 1 10 134 2 174 17.7 0.00225 4.1
Tetramethylbenzenes 1 10 134 2 200 3.48 0.00057 4.0

Both the solubility and toxicity of the nonaromatic hydrocar- docomponents) rapidly separates into different phases or parts
bons are much less than for the aromatics and dissolution (and of the environment, i.e., surface slicks; emulsified oil (mousse)
water concentrations) of nonaromatics is safely ignored. Thus, and tar balls; oil droplets suspended in the water column;
dissolved concentrations are calculated only for each of the dissolved lower molecular weight components (MAHs and
three soluble aromatic pseudocomponents. Tables 2 and 3 list PAHs) in the water column; oil droplets adhered and hydro-
physical–chemical characteristics of individual aromatics used carbons adsorbed to suspended particulate matter in the water;
to derive a mean molecular weight, boiling point, solubility, hydrocarbons on and in the sediments; dissolved MAHs and
and vapor pressure for each of the three volatile and soluble PAHs in the sediment pore water; and hydrocarbons on and
or semisoluble aromatic pseudocomponents (Table 4). The va- in the shoreline sediments and surfaces.
por pressure used to characterize each of the volatile aliphatic
pseudocomponents is equivalent to that for the aromatic pseu- Transport
docomponent of the same boiling point range. This number of Separate Lagrangian elements (spillets) representing sub-
pseudocomponents provides sufficient accuracy for the evap- lots of surface-floating oil, subsurface droplets, and dissolved
oration and dissolution calculations, particularly given the time components are used to simulate the movements of oil com-
frame (minutes) over which dissolution occurs from small ponents in three dimensions over time. Transport is the sum
droplets and the rapid resurfacing of large droplets (see dis- of advective velocities of currents input to the model, surface
cussion above, this section). wind drift (either using a constant percent of wind speed and
The schematic in Figure 1 shows oil fate processes simu- deflection angle, typically observed as 3–4% and 0–108 to
lated in the model in open water. The oil (whole and as pseu- right of downwind in the Northern Hemisphere [1] or based
on the model by Youssef and Spaulding [67]), and randomized
turbulent diffusive velocities in two (floating oil) or three (sub-
surface oil) dimensions. The magnitudes of the directional
components of randomized diffusion are scaled by horizontal
and vertical diffusion coefficients [68] using a random walk
technique [69]. The vertical diffusion coefficient (Dv, m2/s) in
the wave-mixed layer is computed as a function of wind speed
(at 10 m above the sea surface, W10), based on Thorpe [70]:
Dv 5 0.0015W10 (1)

The horizontal and deeper water vertical diffusion

coefficients are model inputs
Oil droplets also undergo vertical movements according to
buoyancy using a modified Stoke’s law. The rise velocity for
each droplet size i, wi (m/s), is
wi 5 di2g(1 2 ro/rw)/18yw (2)

Fig. 1. Simulated oil fate processes in open water. Monoaromatic where di is the droplet diameter (m), g is the gravitational
hydrocarbon (MAH); polycyclic aromatic hydrocarbon (PAH). acceleration (m/s2), ro is the density of the oil (kg/m3), rw is
2444 Environ. Toxicol. Chem. 23, 2004 D.P. French-McCay

Table 3. Physical–chemical properties for 2- to 4-ring polynuclear aromatic hydrocarbons. Molecular weight (MW), boiling point (BP), solubility,
and vapor pressure are from Mackay et al. [64–66]. Estimates of log(KOW) are based on Mackay et al. [63,64] and Neff and Burns [126]. (A
dash indicates no data available)

Vapor
MW Distillation BP Solubility pressure
Compound(s) Rings Cs (g/mol) cut no. (8C) (ppm) (atm) Log(KOW)

Tetralin 2 10 132 2 208 15 0.00052 3.8

Diphenylmethane 2 13 168 2 264 16 8.73 3 10207 4.1
Biphenyl 2 12 154 2 261 5.53 1.28 3 10205 3.9
Naphthalene 2 10 128 2 218 31 0.00010 3.4
C1-naphthalenes 2 11 142 2 243 26.5 8.80 3 10205 3.9
C2-naphthalenes 2 12 156 2 254 6.4 1.98 3 10205 4.4
C3-naphthalenes 2 13 170 3 267 — — 5.0
C4-naphthalenes 2 14 185 3 — — — 5.6
Acenaphthylene 3 12 152 3 270 16.1 8.88 3 10206 4.1
Acenaphthene 3 12 154 3 278 3.8 2.96 3 10206 3.9
Dibenzofuran 3 12 1 O 168 3 287 4.75 2.96 3 10206 4.3
Fluorene 3 13 166 3 295 1.9 8.88 3 10207 4.2
C1-fluorenes 3 14 181 3 — 1.09 — 5.0
C2-fluorenes 3 15 196 3 — — — 5.2
C3-fluorenes 3 16 211 3 — — — 5.5
Anthracene 3 14 178 3 340 0.045 9.87 3 10209 4.5
Phenanthrene 3 14 178 3 339 1.1 1.97 3 10207 4.6
C1-phenanthrenes/anthracenes 3 15 192 3 — — — 5.1
C2-phenanthrenes/anthracenes 3 16 207 3 — — — 5.3
C3-phenanthrenes/anthracenes 3 17 222 4 — — — 6.0
C4-phenanthrenes/anthracenes 3 18 237 4 390 — — 6.5
Dibenzothiophene 3 12 1 S 184 3 333 — — 4.5
C1-dibenzothiophene 3 13 1 S 199 3 — — — 4.9
C2-dibenzothiophene 3 14 1 S 214 3 — — — 5.5
C3-dibenzothiophene 3 15 1 S 228 4 — — — 5.7
Fluoranthene 4 16 202 3 375 0.265 — 5.2
Pyrene 4 16 202 3 404 0.013 — 5.2
C1-fluoranthenes/pyrenes 4 17 217 4 407 — — 5.7
Chrysene 4 18 228 4 448 0.0018 — 5.9
C1-Chrysenes 4 19 242 4 — 6.4

the density of the water (kg/m3), and yw is the kinematic vis- shore type. The algorithm and data are in French et al. [30].
cosity of water (m2/s). This relation holds for low Reynolds Shore widths by shore type are data inputs to SIMAP.
numbers (Re , 20). The Reynolds number describes the degree
of turbulence in the flow over the particle and is defined as Spreading
Re 5 Lv/yw, where L is the length scale (di in this case). Spreading of floating oil is modeled in three ways. The
rapid thinning and broadening of surface slicks caused by grav-
Shoreline stranding itational forces [73] is modeled as an increase in the diameter
The fate of spilled oil that reaches the shoreline depends of each floating spillet according to the spreading algorithm
on characteristics of the oil, the type of shoreline, and the empirically derived by Mackay et al. [17–19], who modified
energy environment. The stranding algorithm is based on data Fay’s approach and described the oil as thin and thick slicks
and analysis of Gundlach [71] and Reed and Gundlach [72]. (as described in French et al. [30]). They assumed the thick
In SIMAP, deposition occurs when an oil spillet intersects slick feeds the thin slick and that 80 to 90% of the total slick
shore surface and ceases when the volume-holding capacity area is represented by the thin slick. The formulation is cor-
for the shore surface is reached. Because subsequent oil com- rected for number of surface spillets [24]. Spreading is stopped
ing ashore is not allowed to remain on the shore surface, it is when an oil-specific terminal thickness is reached, based on
refloated and carried to sea by outgoing tidal currents and wind data from McAuliffe [61], as described in Table 5.
drift. The shoreline oil is then removed exponentially with In addition, oil spreads by so-called shear spreading, the
time. Data for holding capacity and removal rate are taken entrainment and resurfacing of oil, whereby the oil is affected
from Gundlach [71] and are a function of oil viscosity and differentially by the wind drift and subsurface currents [74,75].

Table 4. Mean physical–chemical properties for each aromatic pseudo-component, based on data in Tables 2 and 3; see Table 3 for acronym
definitions

Vapor
MW Distillation BP Solubility pressure
Pseudo-component Rings Cs (g/mol) cut no. (8C) (ppm) (atm) Log(KOW)

1-Ring aromatics 1 8.4 111 1 149 242.4 0.01525 3.3

2-Ring aromatics 1.7 10.9 142 2 222 17.3 6.20 3 10204 4.0
3-Ring aromatics 3.1 14.4 187 3 324 3.2 2.65 3 10206 4.8
Oil spill impact assessment modeling Environ. Toxicol. Chem. 23, 2004 2445

Table 5. Minimum oil thickness for gravitational spreading based on square of wind speed and previous water incorporation. Vis-
data in McAuliffe [61] cosity increases as water content increases in the oil, and the
Minimum slick thickness increasing viscosity feeds back in the model to slow the en-
Oil viscosity (mPa/s) (mm) trainment rate. As information becomes more available, a new
water incorporation algorithm accounting for the oil’s resin,
,10 0.01 asphaltene, and wax content, as well as the stability of the
10–20 0.05
20–1,000 0.1 water-in-oil emulsion, should be developed.
.1,000 1
Entrainment
Entrainment by surface-breaking waves is modeled based
This later process was observed by Reed et al. [76] and in on Delvigne and Sweeney [86] who, using laboratory and
validation studies of SIMAP using test spills in the Caribbean flume experimental observations, developed a relationship for
Sea [27]. entrainment rate as a function of oil droplet size, which in turn
Finally, the use of multiple spillets to simulate releases over is related to turbulent energy level (as breaking wave energy
time and in more than one location (such as along a ship’s times fraction of the sea surface covered by breaking waves)
path) effectively spreads the floating oil. Both surface and and oil viscosity. The data and relationships in Delvigne and
subsurface releases are modeled. For a subsurface release, the Sweeney [86] and Delvigne and Hulsen [87] are used to calculate
oil is initiated as droplets of an assigned droplet size distri- mass and particle size distribution of droplets entrained:
bution, the larger of which rise and surface over an area de- Qd 5 C*Dd0.57SFd0.7Dd (3)
termined by the height of the water above the release point(s)
and the currents in the water column. where Qd is the entrainment rate (kg/m2-s) for droplet diameter
d (m), C* is an empirical entrainment constant that depends
Evaporation on oil type and weathering state, Dd is the dissipated breaking
The evaporation algorithm in SIMAP is based on accepted wave energy per unit surface area (J/m2), S is the fraction of
evaporation theory, which follows Raoult’s law that each com- sea surface covered by oil, F is the fraction of sea surface hit
ponent evaporates at a rate proportional to the saturation vapor by breaking waves, and Dd is the oil particle interval diameter
pressure and mole fraction present for that component. Each (m). Using the data reported in Delvigne and Hulsen [87], the
pseudocomponent evaporates according to its mean vapor entrainment constant, C*, was fit to the following:
pressure, solubility, and molecular weight (Table 4). The mass If (m/r) , 132 cSt, C* 5 exp[20.1023 ln(m/ro ) 1 7.572]
transfer coefficient is calculated using the methodology of
Mackay and Matsugu [77], as described in French et al. [30] (4)
and analogous to that described by Jones [49]. Evaporation
If (m/r) $ 132 cSt, C* 5 exp[21.8927 ln(m/ro ) 1 16.313]
from surface and shoreline oil increases with the oil surface
area, temperature, and wind speed. As lighter components (5)
evaporate off, the remaining weathered oil becomes more vis-
cous. where m is the viscosity (mPa/s) and ro is the density (g/cm3)
Jones [49] compared the results of the pseudocomponent of the oil. The mean droplet diameter of entrained oil, d50 (mm),
model to laboratory data in Fingas [78] and the model of Stiver was fit with a curve to data in Delvigne and Sweeney [86] to
and Mackay [79], who used an analytical approach to predict yield:
the volume fraction evaporated using distillation data to es- d50 5 1,818E20.5(m/ro)0.34 (6)
timate needed parameters. The pseudocomponent model
agreed with the Fingas data under nearly equivalent conditions where E is the wave energy dissipation rate per unit volume
(20 g of oil 1.5 mm thick and no wind). The pseudocomponent (J/m3-s) with E set at 103 (J/m3-s) for breaking waves.
model predicts slightly lower evaporation rates than the Stiver Delvigne and Sweeney [86] found that there is a linear
and Mackay model, with the two models varying systemati- relationship between log(Ni) and log(di), where Ni is the num-
cally (parallel as a function of temperature). The difference is ber of particles in the size interval 0.5di to di, and di is particle
accounted for by the approximation of vapor pressure from diameter. Thus, the relative distribution of numbers, and vol-
distillation data for the Stiver and Mackay model. ume (or mass) for spherical droplets, can be calculated as a
function of droplet size. It follows (Delvigne and Sweeney
Emulsification [86], Eqn. 4) that
The formation of water-in-oil emulsions, or mousse, de-
O V 5 1.45V
i21

pends on sea state and the resin, asphaltene, and wax content j i (7)
j51
of the oil [78,80–82]. Oils vary in their ability to form stable
emulsions. Emulsified oil can contain as much as 80% water where Vi is the volume in size class 0.5di to di, and Vj is the
in the form of micrometer-sized droplets dispersed within a volume in each of the size classes smaller than 0.5di to di, and
continuous phase of oil [78,83,84]. These authors argue that, the sum is from j 5 1 to (i 2 1). This equation indicates a
for accuracy, an empirical approach with curves fit for each rapid decrease in volume as droplet size decreases. The min-
oil is needed to account for oil-specific emulsification rates imum and maximum droplet diameters entrained in the water
and the degree of stability of the emulsion. However, for gen- column are assumed to be 0.1d50 and d50, respectively. The
eral applicability of SIMAP, emulsification (water-in-oil, i.e., minimum is set at 10% d50 because volumes below this size
mousse formation) is modeled using the scheme of Mackay are relatively small (about 2% of the volume in the mean size
and Zagorski [85], as described in French et al. [30]. Water class), and can be neglected. The maximum value is set to
content increases exponentially, with the rate related to the equal the mean because, in numerical experiments and model
2446 Environ. Toxicol. Chem. 23, 2004 D.P. French-McCay

testing, droplets larger than d50 were found to resurface in less the entrainment rate (at a give level of turbulence) and de-
than one time step and so are not quantified as separate from creases droplet size. The droplet size distribution of chemically
surface slicks. In the model, six size classes are used, evenly dispersed oil is based on observed sizes in Lunel [89]. Typi-
distributed by diameter from the minimum to the maximum cally, dispersants are effective on oils up to a viscosity of
size. The dissipated wave energy, Dd (J/m2), is about 10,000 mPa/s [13,14,83], a threshold above which en-
trainment is negated in the model.
Dd 5 0.0034rwgH2 (8)
where rw is the density of water (kg/m3), g is the acceleration Surf entrainment
due gravity (m/s2), and H is the root mean square value of Wave height in the surf zone may be input to the model to
breaking wave height (m). The fraction of the sea surface hit induce entrainment. The above algorithm based on Delvigne
by breaking waves per unit time, F, is parameterized for Uw and Sweeney [86] is used to determine the droplet size dis-
# Uth [11] as tribution of surf-entrained oil, which is assumed well-mixed
in the surf zone. Settling of particles does not occur in water
F 5 3 3 1026(U3.5
w /Tw) (9)
depths where waves reach the bottom (taken as 1.5 multiplied
and, for Uw . Uth [86] as by wave height).
F 5 0.032([Uw 2 Uth]/Tw) (10) Dissolution
where Uw is the wind speed at 10 m above the sea surface (m/ The algorithm developed by Mackay and Leinonen [16] is
s), Uth is the threshold wind speed for onset of breaking waves used in SIMAP for dissolution from a surface slick. The slick
(;6 m/s), and Tw is the significant wave period (s). The total (spillet) is treated as a flat plate, with a mass flux [90] related
mass entrained into the water column in a time step Dt (sec), to solubility and temperature. It assumes a well-mixed layer
ME (kg), is with most of the resistance to mass transfer lying in a hypo-
ME 5 ADt O (Q Dd)
d (11)
thetical stagnant region close to the oil. For subsurface oil,
dissolution is treated as a mass flux across the surface area of
where A is the area of surface slick (m2). The intrusion depth, a droplet (treated as a sphere) as in Mackay and Leinonen [16].
Zm (m), is Dissolution rate increases the higher the surface area of the
oil relative to its volume. Because the surface area–to–volume
Zm 5 (1.5 6 0.3)Hb (12)
ratio is higher for smaller spherical droplets, dissolution is
where Hb is the breaking wave height (m). Wave height is higher for smaller droplets. The droplet size distribution and
calculated from wind speed, duration, and fetch (distance up- amount of oil entrained, both functions of turbulence, are the
wind to land), using the algorithms from the Coastal Engi- critical determinants of dissolved concentrations. As discussed
neering Research Center [88]. The mixing depth (except where above, dissolution from entrained small droplets is much faster
constrained by the seafloor) for each droplet size class, than from surface slicks in the shape of flat plates (which is
Zi (m), is insignificant). The soluble components also are volatile, and
evaporation from surface slicks is faster than dissolution into
Zi 5 max(Dv/wi, Zm) (13) the underlying water. Thus, the processes of evaporation and
where Dv is the vertical dispersion coefficient (Eqn. 1) and wi dissolution are competitive, with evaporation the dominant
is the rise velocity for the droplet size (Eqn. 2). process for surface oil.
Entrainment rate is slower and droplet size is larger for
Volatilization
higher viscosity oils and as oil viscosity increases by emul-
sification and evaporation loss of lighter volatile components. Volatilization of dissolved components from the water to
The droplet size determines how fast and whether the oil re- the atmosphere occurs as they are mixed and diffuse to the
surfaces. Droplets greater than 70 mm in diameter are assumed sea surface boundary and enter the gas phase. Volatilization
to float if they reach the water surface by vertical diffusion; rate increases with increasing temperature. The procedure out-
smaller droplets reaching the surface are mixed downward if lined by Lyman et al. [91], based on Henry’s law and mass
their rise velocity is overcome by the vertical diffusion velocity transfer [90], is followed in the model [30]. The volatilization
and resurface otherwise. The 70 mm threshold is based on depth for dissolved substances is limited to the maximum of
observations by Lunel [89] of droplet sizes that seem to be one half the wave height.
dispersed permanently. Resurfaced oil typically forms sheens.
As surface oil is blown downwind faster than the underlying Dissolved-suspended particulate partitioning, adherence,
water, resurfacing droplets come up behind the leading edge and sedimentation
of the oil, effectively spreading the slicks in the downwind Adsorption of dissolved components to particulate matter
direction. in the water occurs because the soluble components are only
Waves break beginning at about 12 knots (;6 m/s) of wind sparingly so. These compounds preferentially adsorb to par-
speed and wave breaking increases as wind speed becomes ticulates when the latter are present. The ratio of adsorbed to
higher. Thus, entrainment becomes increasingly important dissolved concentrations is computed from standard equilib-
(higher rate of mass transfer to the water) the higher the wind rium partitioning theory [92]. Adsorption increases with
speed. Below 12 knots of wind speed, the percentage of mass log(Kow) and the concentration of suspended particulates, with
entrained is not significant. As wind and turbulence increase, the particulate phase accounting for a significant percentage
the oil droplet sizes become smaller, although there is a feed- of the mass above 100 mg/L suspended sediments.
back slowing of the entrainment process as oil weathers and/ The model formulation developed by Kirstein et al. [21] is
or emulsifies, becoming more viscous. Application of chemical used to calculate the volume of oil droplets adhered to par-
dispersant decreases surface tension of the oil, and so increases ticles. The Stoke’s law formulation (Eqn. 2) is used to adjust
Oil spill impact assessment modeling Environ. Toxicol. Chem. 23, 2004 2447

vertical position of oil–sediment particles, using the density BIOLOGICAL EFFECTS MODEL
of the combined particle. If turbulence subsides sufficiently, The biological exposure model estimates the area, volume,
the oil–sediment agglomerates will settle. or portion of a stock or population affected by surface oil,
Sedimentation occurs when oil–sediment agglomerates and concentrations of oil components in the water, or sediment
particles with adsorbed semisoluble components (MAHs and contamination. The biological effects model estimates losses
PAHs) settle to the bottom sediments. Adherence and sedi- resulting from acute exposure after a spill (i.e., losses at the
mentation can be an important pathway of oil in nearshore time of the spill and while acutely toxic concentrations remain
areas when waves are strong and subsequently subside. Gen- in the environment) in terms of direct mortality and lost pro-
erally, oil–sediment agglomerates transfer more PAH to the duction because of direct exposure or the loss of food resources
bottom than sediments with PAHs that were adsorbed from from the food web. Losses are estimated by species or species
the dissolved phase in the water column. Resuspension of group for fish, invertebrates (i.e., shellfish and nonfished spe-
settled oil–sediment particles may occur if current speeds ex- cies), and wildlife (birds, mammals, sea turtles). Lost produc-
ceed threshold values where adhesive forces can be overcome, tion of aquatic plants (microalgae and macrophytes) and lower
assumed 20 cm/s in the model. trophic levels of animals also are estimated.
Bioturbation The area potentially affected by the spill is represented by
a rectangular grid with each grid cell coded as to habitat type.
Bioturbation, the process where animals in the sediments The habitat grid also is used by the physical fates model to
mix the surface sediment layer while burrowing, feeding, or define the shoreline location and type, as well as habitat and
passing water over their gills, effectively mixes the surface sediment type. A habitat is an area of essentially uniform
sediment layer about 10 cm thick (in nonpolluted environ- physical and biological characteristics that is occupied by a
ments). In the model, sediment concentration is calculated as group of organisms that are distributed throughout that area.
mass loading per area divided by 10 cm. Contaminant con- A contiguous grouping of habitat grid cells with the same
centrations in sediment are distributed between adsorbed and habitat code represents an ecosystem in the biological model.
dissolved states by equilibrium partitioning, as in the water Prespill abundance of fish, invertebrates, and wildlife, and rates
column. For this calculation, the particulate-to-interstitial wa- of lower trophic level productivity, are assumed constant for
ter ratio is taken to be 0.45 [88]. the duration of the spill simulation and evenly distributed
across an ecosystem. Though biological distributions are
Degradation
known to be highly variable in time and space, data generally
Degradation includes biodegradation, photooxidation, and are not sufficient to characterize this patchiness. Oil also is
other chemical reactions. A first-order decay algorithm is used, patchy in distribution. The patchiness is assumed to be on the
with a specified (total) degradation rate for each of surface same scale so that the intersection of the oil and biota is equiv-
oil, water column oil, and sedimented oil. Degradation rates alent to overlays of spatial mean distributions.
for the aromatics are from Mackay et al. [63–66] and for whole Mobile fish, invertebrates, and wildlife are assumed to
oil are from French et al. [30]. move at random within each ecosystem during the simulation
period. This is a reasonable assumption for the period of the
Fates model output simulation (generally a few weeks). Benthic organisms also
The physical fates model creates output files recording the may remain stationary on or in the bottom. Planktonic stages,
distribution of a spilled substance in three-dimensional space such as pelagic fish eggs, larvae, and juveniles (i.e., young-
and time as area covered by oil and thickness on the water of-the-year during their pelagic stage[s]), move with the cur-
surface (swept area); volumes in the water column at various rents.
concentrations of dissolved aromatics; volumes in the water Habitats include open water, reef (coral or mollusk), wet-
column at various concentrations of total hydrocarbons in sus- land, sea grass, kelp bed, and shoreline environments. Habitat
pended droplets; total hydrocarbon concentrations and dis- types are defined by depth, proximity to shoreline(s), bottom/
solved aromatic concentrations in surface sediment; and shore type, dominant vegetation type, and the presence of in-
lengths and locations of shoreline impacted and volume of oil vertebrate reefs. With respect to proximity to shoreline(s), hab-
ashore in each segment. The dissolved aromatic hydrocarbon itats are designated as landward or seaward. Landward portions
concentration in the water column is calculated from the mass are the nearshore rivers, estuaries, and inlets. The seaward
in spillets representing dissolved components, as follows. Con- portion is the more oceanic or main part of the water body.
centration is contoured on a three-dimensional Lagrangian grid This designation allows different biological abundances to be
system (of 200 3 200 cells in the horizontal and five vertical simulated in landward and seaward zones of the same habitat
layers), which is scaled each time step to just cover the volume type (e.g., open water with sand bottom).
occupied by dissolved aromatic particles, including the dis-
persion around each spillet center. This maximizes the reso- Wildlife
lution of the concentrations at each time step and reduces error In the model, surface slicks (or other floating forms such
caused by averaging mass over large cell volumes. Distribution as tar balls) of oils and petroleum products impact wildlife
of mass around the spillet center is described as Gaussian in (birds, marine mammals, sea turtles, and other reptiles). For
three dimensions, with one standard deviation equal to twice each of a series of surface spillets, the physical fates model
the diffusive distance (2Dxt in the horizontal, 2Dzt in the ver- calculates the location and size (radius of circular spreading
tical, where Dx is the horizontal and Dz is the vertical diffusion spillet) as a function of time. The area swept by a surface
coefficient, and t is particle age). The plume grid edges are spillet in a given time step is calculated as the quadrilateral
set at one standard deviation out from the outermost particle. area defined by the path swept by the spillet diameter. This
These data are used by the biological effects model to evaluate area is summed over all time steps for the time period the
exposure, toxicity, and impacts. spillet is present on the water surface and separately for each
2448 Environ. Toxicol. Chem. 23, 2004 D.P. French-McCay

Table 6. Combined probability of encounter with the slick and mortality once oiled, if present in the area swept by a slick exceeding a threshold
thickness. Area swept is calculated for the habitats occupied

Wildlife group Probability (%) Habitats occupied

Dabbling waterfowl 99 Intertidal and landward subtidal

Nearshore aerial divers 35 Intertidal and landward subtidal
Surface seabirds 99 All intertidal and subtidal
Aerial seabirds 5 All intertidal and subtidal
Wetland wildlife (waders and shorebirds) 35 Wetlands, shorelines, seagrass beds
Cetaceans 0.1 Seaward subtidal
Furbearing marine mammals 75 All intertidal and subtidal
Pinnipeds, manatee, sea turtles 1 All intertidal and subtidal
Surface birds in seaward only 99 All seaward intertidal and subtidal
Surface diving birds in seaward only 35 All seaward intertidal and subtidal
Aerial divers in seaward only 5 All seaward intertidal and subtidal
Surface birds in landward only 99 All landward intertidal and subtidal
Surface diving birds in landward only 35 All landward intertidal and subtidal
Aerial divers in landward only 5 All landward intertidal and subtidal
Surface diving birds in water only 35 All subtidal
Aerial divers in water only 5 All subtidal

habitat type where the oil passes. Spillets sweeping the same chronic exposures, it is the aromatics, and specifically the
area of water surface at the same time are superimposed. The PAHs, that cause effects either directly or indirectly via bio-
total area swept over a threshold thickness by habitat type is accumulation and uptake via the food web [50].
multiplied by the probability that a species uses that habitat The model evaluates mortality and sublethal effects of dis-
(0 or 1, depending upon its behavior) and a combined prob- solved aromatic concentrations in the water or sediment. Mor-
ability of oiling and mortality. This calculation is made for tality is a function of duration of exposure: The longer the
each surface-floating spillet and each habitat for the duration duration of exposure, the lower the threshold for effects
of the model simulation. [30,60–62,95–101]. After a certain period of time, all indi-
The portion of the wildlife in the area swept by the slick viduals that will die at a given concentration have done so and
over a threshold thickness that are assumed to die is based on no further mortality is observed. The lethal threshold concen-
probability of encounter with the slick multiplied by the prob- tration, also termed the incipient lethal level, is the concen-
ability of mortality once oiled. The probability of encounter tration where mortality occurs after this sufficiently long ex-
with the slick is related to the percentage of the time an animal posure [95,102]. The incipient lethal concentration to 50% of
spends on the water or shoreline surface. The probability of exposed organisms, LC50`, is the asymptotic LC50 reached
mortality once oiled is nearly 100% for birds and fur-covered after infinite exposure time (or long enough that that level is
mammals (assuming they are not treated successfully) and approached). The standard mortality model is utilized, with a
much lower for other wildlife. The products of the two prob- lognormal relationship between percent mortality and concen-
abilities for various wildlife behavior groups are in Table 6. tration, and the LC50 the center of the distribution.
Estimates for the probabilities were derived from information In SIMAP, LC50` is input to the model for the mixture of
on behavior and field observations of mortality after spills [30]. dissolved MAHs and PAHs originating from the type of oil
The wildlife mortality model has been validated with more spilled (see below). For each of a series of aquatic biota be-
than 20 case histories, including the Exxon Valdez and other havior groups, the model evaluates exposure duration, and
large spills, verifying that these values are reasonable [46,93]. corrects the LC50 for time of exposure and temperature to
(However, see discussion below in the Validation section.) calculate mortality. Movements of biota, either active or by
Area swept is calculated for the habitats occupied by each current transport, are accounted for in determining time and
of the behavior groups of wildlife listed in Table 6. Species concentration of exposure. Lagrangian elements are used to
or species groups are assigned to behavior groups to evaluate represent schools or groups of animals. The elements move or
their loss. The threshold is 10 microns (;10 g/m2) thick oil, remain stationary according to the behavior type, and con-
based on data on minimum dose to impact a bird and calcu- centration and duration of exposure are recorded. Exposures
lations described in French et al. [30]. Wildlife mortality is are integrated over space and time by habitat type to calculate
directly proportional to abundance per unit area and the percent a total percentage killed.
mortalities in Table 6. Behavior groups are used to represent species or stages
within species covering the possible movement patterns (or
Fish, invertebrates, and aquatic plants lack thereof) for aquatic organisms, i.e., planktonic (moving
In the model, aquatic biota (fish, invertebrates, and plants with currents), demersal and stationary (on the bottom exposed
in the water column and on/in the sediments) are affected by to near bottom water), benthic (in the sediments and station-
dissolved aromatic concentrations in the water or sediment. ary), demersal fish and invertebrates (on the bottom exposed
This rationale is supported by the fact that soluble aromatics to near bottom water and moving slowly), small pelagic fish
are the most toxic constituents of oil [50–61,94]. Exposures and invertebrates (moving randomly and slowly in the water
in the water column are short in duration [51], therefore, effects column), and large pelagic fish and invertebrates (moving ran-
there are the result of acute toxicity. In the sediments, exposure domly and rapidly in the water column). Pelagic fish move at
may be both acute and chronic, as the concentrations may about 0.5 body length per second [103], which amounts to 9
remain elevated for longer periods of time. In either acute or km/d for a 20-cm small pelagic fish and 45 km/d for a 100-
Oil spill impact assessment modeling Environ. Toxicol. Chem. 23, 2004 2449

cm large pelagic fish (sizes from French et al. [30]). For de- pounds, such as the lower-molecular-weight aromatics in oil
mersal species, movements are much slower, assumed to be (MAHs and PAHs), is related to Kow [30,62,99,101,107–113].
0.5 km/d. Demersal organisms always remain in the bottom Chemicals that have a narcotic mode of action impact organ-
layer within 1 m of the bottom, whereas pelagic fish move isms by accumulating in lipids (such as in the cell membranes)
vertically within the water column. Lagrangian elements are and disrupting cellular and tissue function, such that the more
used to distinguish organisms by behavior group and in six hydrophobic the compound the more accumulation and the
habitat types: Seaward (offshore) open water, landward (es- more severe the impact. However, the more hydrophobic the
tuarine) open water, seaward (offshore) wetland and seagrass, compound, the less soluble it is in water, and so the less avail-
landward (estuarine) wetland and seagrass, seaward (offshore) able it is to aquatic organisms. Compounds of log(Kow) . 5.6
reef, and landward (estuarine) reef. These six habitat categories are considered insoluble, and so are not bioavailable and thus
account for the fact that fish and other aquatic biota tend to not acutely toxic to aquatic biota [62]. Thus, impact is the
prefer one or more of these types [30,104]. result of a balance between bioavailability (dissolved-com-
Mortality is calculated as percent loss by habitat and be- ponent exposure) and toxicity once exposed.
havior group. This is translated into the equivalent area of The acute toxic effects of narcotic chemicals are additive
100% loss. That area may be divided by the total area of habitat [30,62,105,106,113]. The Toxic Unit (TU) Model is used to
available in the region of interest to estimate a percentage of estimate the toxicity of a mixture of narcotic chemicals, with
a population affected. The percent mortality of the exposure TU defined as the exposure concentration divided by the LC50.
group may be multiplied by abundance at the time exposed For a mixture, the toxic units are additive: When STU 5 1,
and in the habitat type to calculate the species’ mortality as the mixture is lethal to 50% of exposed organisms.
numbers or biomass (kg). It has been shown [62] that the LC50 of the mixture
The mortality of each species is evaluated as a population (LC50mix) is related to the LC50 of each chemical i in the
loss using species-specific natural and fishing mortality rates mixture and the fractional concentration of chemical i (Fi) in
and standard fisheries models. Production foregone is calcu- the total mixture:
lated as the lost growth the killed individuals would have
undergone over their remaining life span. The population and
production foregone models are described in French-McCay
Fi 5 Cw,i @1O C 2 w,i (15)
[46].
Lost production of plants and animals at the base of the where Cw,i is the dissolved concentration of chemical i in the
food chain also is integrated in space and over time using the water.
effective concentration to reduce growth to 50% of normal, to
parameterize a lognormal function. For each time step and for
LC50mix 5 1 @ O (F / LC50 )
i i (16)
each of the concentration grid cells output by the physical The values of Fi may be measured in the field or, if field
fates model, lost primary, zooplankton, and benthic production samples are not available, Fi may be estimated from the source
(PL) are calculated as follows: oil composition. French-McCay [62] showed that for surface
waters, where turbulent entrainment of oil has occurred, the
PL 5 (1 2 Fk)V/Zd 3 t (14)
values of Fi nearly are proportional to the source oil aromatic
where Fk is the fraction of the uninhibited rate of production composition. The LC50mix is calculated including those aro-
that is realized at the contaminant concentration, V is volume matics that are measured in the oil and dissolved in the water
contaminated (m3), Zd is water depth, and t is the days con- (with log(Kow) # 5.6) for long enough times for exposure to
taminated. Total production loss is summed over time by hab- aquatic organisms to be significant. Typically (except for gas-
itat type and the integrated losses are summarized as m2 2 oline), only the PAHs are dissolved in sufficient quantity and
days of equivalent 100% loss of production. These may be remain in the water long enough for their TU values to be
multiplied by production rates (g dry wt m22 d21) to estimate significant.
production losses. Lost production of dependent larger animals The values of LC50i for MAHs and PAHs were estimated
in the food web, due to reduction of food supply, is estimated using a regression model relating LC50 to Kow [62], with the
using a simple food chain model [30]. Production losses of 95% confidence range of this regression providing LC50s for
lower trophic levels typically are very small because of their average (50th percentile), sensitive (2.5th percentile), and in-
short generation times and quick recovery after a spill. They sensitive (97.5th percentile) species:
have not been measured in the field because the impact is less
than natural variability. log10(LC50`) 5 log10(f) 1 g log10(Kow) (17)
Based on 278 bioassays on individual aromatics, the intercept
Oil toxicity
and slope of the regression are: log10(f) 5 4.8926 and g 5
The following summarizes the oil toxicity model used to 21.0878. This regression describes the mean response for all
determine appropriate values of LC50` to input to the SIMAP species (i.e., the response of the average species). The slope
exposure model. The full development of the oil toxicity model of this relationship is constant for all species [106]. The in-
and data upon which it is based are in French-McCay [62]. tercept varies by species, with 95% of species falling within
The oil toxicity model utilizes the accepted toxic units ap- the range log10(f) 5 3.9704 (sensitive species) and log10(f)
proach for organic compounds whose primary acute effect is 5 5.8147 (insensitive species) [62].
narcosis, which also is being used by the U.S. Environmental The exponential decrease of LC50 with increasing duration
Protection Agency in the development of PAH water and sed- of exposure is due to the accumulation of toxicant over time
iment quality criteria [105,106]. The oil toxicity model has up to a critical body residue (tissue concentration) that causes
been validated using laboratory oil bioassay data [62]. mortality. The accumulation is more rapid at higher temper-
It has been shown that toxicity of narcotic organic com- ature, such that LC50 at a given (short) exposure time de-
2450 Environ. Toxicol. Chem. 23, 2004 D.P. French-McCay

creases with increasing temperature. The following algorithm spills [46,93], as well as test spills designed to verify the model
was developed in French-McCay [62]: [27]. The validation studies show that the model is capable of
hind-casting the oil trajectory and shoreline oiling, given ac-
LC50` 5 LC50t (1 2 e2«t ) (18)
curate observed wind data following the spill and a reasonable
log 10 («) 5 «1 2 «2 log 10 (Kow ) (19) depiction of surface currents, both tidal and background. As
winds and currents are the primary forcing variables on oil
d« / dT 5 t T (20) fate, obtaining accurate data on these is very important to the
where t is time of exposure, LC50t is LC50 at time t, LC50` accuracy of any simulation.
is LC50 at infinite time of exposure, Kow is the octanol–water The biological effect model has been validated using sim-
partition coefficient, «1 5 1.47 and «2 5 0.414, T 5 temperature ulations for 28 spill events where data are available for com-
(C), and t 5 0.11. parison [46,93]. In most cases [93] only the wildlife impacts
The SIMAP exposure model uses a rearrangement of Equa- could be verified because of limitations of the available ob-
tion 18 to correct the LC50 for time of exposure and temper- servational data. However, in the North Cape spill simulations,
ature, as recorded by movements of Lagrangian elements rel- both wildlife and water column impacts (lobsters) could be
ative to toxic concentrations (i.e., greater than the concentra- verified [46].
tion lethal to 1% of exposed organisms, LC1, approximated In most cases, impact information for the spill primarily
as 1% of LC50`). Exposure time is the total time concentration consists of counts of rescued or dead wildlife. The model
exceeds LC1 and the concentration is the average over that results either agree well with field estimates, or underestimate
time. The percent mortality is then calculated using the log- the observed kill. The cases where wildlife kills are under-
normal function centered on LC50t. estimated are where an unusually high abundance of certain
The dissolved concentrations are estimated by the physical species occurred in the spill path, or where data on abundance
fates model for both the water column and sediments. Dis- of animals were not available. Many of these cases were in
solved concentrations in the water column result mainly from fact considered significant spills because of the presence of
dissolution of entrained oil droplets, as the soluble compounds unusually high aggregations of certain wildlife species [93].
evaporate faster from floating oil. In the sediments, exposure Modeling results show that the impact algorithms in the model
and mortality of benthic organisms are a function of the dis- are valid when input data on abundance are accurate. The
solved concentrations in pore water. This methodology has uncertainty in the model results is proportional to uncertainty
been validated by Swartz et al. [113] and used in sediment in abundance.
quality criteria for PAHs [106]. For the North Cape oil spill of January 1996 in Rhode
Island, field observations were available for both surface and
Impacts to habitats subsurface oil fates and impacts. The model’s prediction of
Habitat impacts result from smothering by greater than a the oil’s fate agreed with observations of surface oil move-
threshold thickness of oil on intertidal habitat or sediment areas ments and measurements of total hydrocarbons and aromatics
or lethal concentrations of dissolved aromatics in the water in the water. The model estimate of birds oiled was 2,200 to
column (e.g., for coral reef species forming the structural basis 4,400, depending on the prespill abundance data assumed. (Un-
of the reef). The model performs these evaluations for each certainty in the model results is proportional to that for the
cell in the habitat grid. The threshold for smothering by whole abundance estimates.) The midpoint of this range is 8.5 times
oil is 14 mm, based on literature reviewed in French et al. the number of birds collected on beaches, in agreement with
[30]. The lethal dissolved concentration in the water column estimates for other spills. (The negotiated settlement used a
is the LC50 input to the model. factor of 6 for this spill.) Impacts on water column organisms
Long-term impacts from habitat loss are integrated in the bi- were validated by comparison of the model estimate to the
ological effects model over the period of recovery. The recovery field observations of the lobster kill (8.3 million and 9 million,
curve is simulated as a sigmoid function described by respectively) [46]. The model estimates of impacts to other
marine species were used in the government’s natural resource
dPR
5 a r PR (1 2 PR ) (21) damage claim to the responsible party. This claim was settled
dt based on these estimates of injury.
where PR is portion recovered and ar is a constant. The value
of the constant ar is derived from solution of the equation Exxon Valdez
assuming PR at t 5 0 is 0.01 and PR at t 5 trec (the recovery The current version of the model, SIMAP, was used to
time) is 0.99, whereupon Equation 21 may be solved as PR 5 simulate in more detail (than previously [93]) oil fates and
1/(1 1 99 exp[2art]) with ar 5 9.19/trec. Recovery times by biological effects from acute exposure to the Exxon Valdez
habitat are input to the model. Typical recovery periods are oil spill (EVOS), as more information is available for this spill
three years for benthic and intertidal habitats, 10 years for than any other in U.S. waters. Descriptions of the oil obser-
seagrass beds, 15 years for salt marshes, and 30 years for vations and ultimate fate are available from several sources
mangroves [30]. Impacts to dependant species and life stages [114–118]. The analysis and simulation was for oil fate pro-
are assumed proportional to lost habitat function (PR), and cesses and impacts occurring in Prince William Sound ([PWS]
production foregone is integrated over time to estimate a total AK, USA) over the first 60 d after the release of 250,000 bbl
impact. (34.8 3 106 kg) of Alaskan North Slope crude oil on March
24, 1989, from 00:00 to 10:00 h (at Bligh Reef, 60.88N,
VALIDATION
146.888W, Fig. 2).
Summary of previous validation studies Water temperature was 28C and suspended sediment con-
The physical fates model has been validated with more than centration was assumed 1 mg/L, a typical value for PWS [30].
30 case histories, including the Exxon Valdez and other large Depth data for PWS were obtained from Hydrographic Survey
Oil spill impact assessment modeling Environ. Toxicol. Chem. 23, 2004 2451

Fig. 2. Simulated currents in Prince William Sound (AK, USA) as a Fig. 3. Cumulative map of areas where surface-floating oil passed by
seasonal mean for spring. The spill site is indicated by the circled March 27, 1989, at noon. The curved line on the map indicates the
cross at Bligh Reef (AK, USA). southwestward extent of observed oiling by this date.

Data supplied on CD-ROM by the U.S. Department of Com- ON, personal communication): Density, 0.8761 g/cm3; vis-
merce, National Oceanic and Atmospheric Administration cosity, 16 mPa/s; surface tension, 27 dyne/cm; and maximum
(NOAA), National Geophysical Data Center. Shore types, water content of mousse, 70%. The spilled oil was assumed
available in digital form from the Environmental Sensitivity to contain 30.66 g/kg MAHs; 3,750 mg/kg 2-ring PAHs; 6,622
Atlas Geographical Information System (provided by NOAA, mg/kg 3-ring PAHs; and 18.9%, 13.3% and 20.0% in aliphatic
Hazardous Materials Response and Assessment Division, Seat- distillation cuts 1, 2, and 3, respectively. Thus, using these
tle, WA, USA), were gridded with a cell size of 196 3 389 data, 4.1% of the oil potentially was toxic to aquatic biota and
m, and the diagonal of 276 m assumed the length of a shore up to 56% of the oil could volatilize, dissolve, or biodegrade.
segment. Shore widths were assumed: 3 m for rocky, 10 m Figures 3 through 5 show cumulative maps for the simu-
for gravel, 20 m for sand, and 300 m for wetlands [30]. lation of areas where surface-floating oil passed during the
A seasonal mean and tidal current field was simulated pre- first 60 d after the release. Figures 3 and 4 indicate the south-
viously using a hydrodynamic model applied to PWS (Fig. 2 westward extent of oiling observed after the spill, as well as
[23,119]). The boundary of the modeling domain was at the the model simulation. The model results are in agreement with
two entrances: Hinchinbrook Entrance and Montague Strait observations: The oil remained near and just west of the spill
(AK, USA; Fig. 2). Tidal forcing functions for the major har- site for the first 2 d (March 24 and 25). On March 26, during
monic constituent (M2) and seasonal mean flows at the en- the windstorm, the oil emulsified and was dispersed and trans-
trances were derived from the larger domain hydrodynamic ported rapidly to the southwest, reaching the north end of
model application for the Gulf of Alaska by Isaji and Spaulding Knight Island (AK, USA) by March 27. Floating oil was trans-
[119]. This current field is similar to that used by Galt et al.
[115] in their simulation of the spill. The flow generally was
towards the southwest and out of PWS at Montague Strait
(Fig. 2).
Wind data from several nearby stations in PWS were com-
piled [93], choosing the closest station to the floating oil on
each date. On March 24 and 25, winds were light, but on
March 26 there was a large windstorm involving strong north-
easterly winds that emulsified and dispersed the oil widely in
central PWS [115]. Wind drift was varied from 3 to 5% with
angles applied from 0 to 208 to the right of downwind, with
3.5% and 08 proving to provide the best fit and timing of oil
movements. The horizontal diffusion coefficient was varied
from 10 to 100 m2/s, which is a reasonable range for open
waters and storm conditions [120]. The value 50 m2/s provided
the best fit to the observations. The vertical diffusion (ran-
domized mixing) coefficient below the wave-mixed layer was
assumed a typical 0.0001 m2/s, but model results for this spill
were insensitive to this parameter as water column contami-
nation was very low.
Oil properties and pseudocomponent content were obtained Fig. 4. Cumulative map of areas where surface-floating oil passed by
from Environment Canada’s Emergencies Science Division oil March 30, 1989, at noon. The curved line on the map indicates the
property database (Z. Wang, Environment Canada, Ottowa, southwestward extent of observed oiling by this date.
2452 Environ. Toxicol. Chem. 23, 2004 D.P. French-McCay

of oiling is lower than the assumed 35% for these species, and
7 to 9% probability would fit the observations. However, the
uncertainty of the abundance data is on the order of a factor
of 2 or more as well. Thus, it could be error in abundance
estimates and/or probability of oiling that accounts for the
differences.
The model estimates were 3,555 otters and 26 seals oiled
with PWS. The government estimates for all areas were 3,500
to 5,500 sea otters and 200 seals [125]. Most of the otters
were impacted in PWS, but seals were impacted in other areas
as well as inside PWS. Thus, the model estimates of impacts
for these marine mammals are reasonable.

CONCLUSION

Success of a model simulation is dependent on both the

algorithms and the accuracy of the input data. The validation
exercises verified that the model algorithms provide reasonable
results. The most important input data in determining accuracy
Fig. 5. Cumulative map of areas where surface-floating oil passed of the physical fates model results are winds, currents, and
during the first 60 d after the release.
assumed randomized diffusion rates. These data can be ac-
curately estimated with appropriate measurements and hydro-
ported by prevailing southwesterly currents through Montague dynamic modeling.
Strait, first exiting PWS on March 30. Oil continued to move With respect to the biological effects model, the results are
generally in the southwesterly direction over the first two sensitive to the assumed toxicity (LC50s), probability of oiling
months, with wind-driven and localized dispersion causing (including avoidance behavior), and biological abundances of
shoreline oiling over a wide area (the modeled path in agree- the affected species. The range of toxicity values is well un-
ment with the summary map in Gundlach et al. [118]). derstood, based on the analysis in French-McCay [62]. How-
Table 7 summarizes the abundances of birds by species ever, few species have been measured in bioassays such that
group occurring in seaward habitats affected in PWS [30]. The their degree of sensitivity to oil hydrocarbons is known. Per-
assumed probability of oiling was based on Table 6, the species formance of bioassays on individual aromatics, which logis-
feeding and sleeping behavior, and observations of avoidance tically are easier to perform than with whole oil, would provide
of oil made by Day et al. [121]. Field estimates of the number the necessary data to determine species sensitivity.
of birds oiled in PWS (Table 7) were obtained using data in Estimation of biological abundance always will be uncer-
Piatt et al. [122] on the carcasses collected in PWS, divided tain due to the natural patchiness and variability of species
by the probability of a killed bird being found in PWS, 35% distributions. In many analyses, such as in ecological risk as-
(accounting for loss at sea, scavenging, search effort, and suc- sessments, evaluation of percent losses for populations of in-
cess), from Ford et al. [123]. Schempf et al. [124] estimated terest may be the appropriate measure of impact. This would
192 bald eagles oiled in PWS, a factor 6 times the number of avoid the difficulty of quantifying abundance at the time and
carcasses collected. location of the spill.
Model-estimated bird kills were in good agreement for spe- The probabilities of oiling, given oil passing through an
cies groups that were not observed to avoid oil. For those area where wildlife are present, are based on limited data avail-
where avoidance was observed, model estimates were a factor able in the literature. More detailed and quantitative infor-
4 to 5 higher than the field estimates. Likely the probability mation on bird and other wildlife behavior in the presence of

Table 7. Abundances of bird species groups in seaward habitats affected by the Exxon Valdez oil spill in Prince William Sound (AK, USA) [30],
assumed probability of oiling, avoidance behavior (based on Day et al. [121]), estimated number killed based on field data (based on Piatt et al.
[122], Schempf et al. [124], and Ford et al. [123]), and model estimated kill

No. No.
Probability Avoid killed killed
Species group No./km2 (%) oil (field) (model) Model/field

Loons 0.138 99 No 835 389 0.5

Grebes 0.337 35 (Some spp.) 1,132 563 0.5
Sea ducks 6.005 35 Yes 2,389 9,313 3.9
Procellariids 0.021 35 No 384 22 0.1
Cormorants 0.631 35 Yes 1,535 946 0.6
Gulls 4.013 5 Yes 173 687 4.0
Murres 4.672 35 Yes 1,458 5,282 3.6
Murrelets 2.767 35 Yes 1,113 5,369 4.8
Guillemots 0.182 99 No 451 514 1.1
Puffins 0.003 99 No — 9 —
Other alcids 0.048 99 No 77 137 1.8
Eagles 2.323 35 Yes 192 34 0.2
Total birds 21.141 — — 9,940 23,264 2.3
Oil spill impact assessment modeling Environ. Toxicol. Chem. 23, 2004 2453

oil would improve greatly the knowledge base and ability to Canada, June 14–16. Emergencies Science Division, Environ-
estimate impacts from spills. As the largest impacts of oil spills ment Canada, Ottawa, ON, pp 525–539.
13. Aamo OM, Reed M, Daling P. 1993. A laboratory-based weath-
typically are to birds and fur-bearing marine mammals, these ering model: PC version for coupling to transport models. Pro-
data needs are all the more pertinent. ceedings, 16th Arctic Marine Oil Spill Program Technical Sem-
The biological effects model in SIMAP evaluates exposure, inar, Calgary, AB, Canada, June 7–9. Emergencies Science Di-
considering movements and amounts of both oil and biota; vision, Environment Canada, Ottawa, ON, pp 617–626.
14. Daling PS, Aamo OM, Lewis A, Strom-Kritiansen T. 1997. SIN-
impacts of response activities, such as booming and dispersant TEF/IKU oil-weathering model: Predicting oil’s properties at
use; acute effects (lethal and sublethal) in the short-term; in- sea. Proceedings, 1997 International Oil Spill Conference, Fort
direct effects of acute exposure via reduction in food supply Lauderdale, FL, USA, April 6–9. API Publication 4651. Amer-
or habitat; and nondensity-dependent population level impacts ican Petroleum Institute, Washington, DC, pp 297–307.
caused by mortality and sublethal effects. What have yet to 15. Kolpack RL, Plutchak NB, Stearns RW. 1977. Fate of oil in a
water environment—Phase II, a dynamic model of the mass
be addressed in a quantitative manner are sublethal effects of balance for released oil. API Publication 4313. University of
chronic contamination; indirect effects via changes in ecosys- Southern California, American Petroleum Institute, Washington,
tem structure; and behavioral changes resulting in reduced DC.
growth, survival, or reproductive success. The latter are de- 16. Mackay D, Leinonen PJ. 1977. Mathematical model of the be-
havior of oil spills on water with natural and chemical dispersion.
pendent on understanding of population and/or ecosystem Economic and Technical Review Report EPS-3-EC-77-19. Fish-
structure and dynamics, areas of active study that are difficult eries and Environment Canada, Ottawa, ON.
to quantify but a challenge for future research. 17. Mackay D, Buist I, Mascarenhas R, Paterson S. 1980. Oil spill
processes and models. Report EE-8. Environmental Emergency
Branch, Department of Fisheries and Environment, Environment
Acknowledgement—The SIMAP Model has been developed over Canada, Ottawa, ON.
many years with the assistance and advice of many individuals at 18. Mackay D, Paterson S, Trudel K. 1980. A mathematical model
Applied Science Associates. The contributions of M. Reed, E. An- of oil spill behavior. Environmental Emergency Branch, De-
derson, D. Mendelsohn, M. Spaulding, V. Kolluru, T. Isaji, E. Howlett, partment of Fisheries and Environment, Environment Canada,
H. Rines, C. Galagan, N. Whittier, J. Rowe, and T. Giguere are rec- Ottawa, ON.
ognized and appreciated. 19. Mackay D, Shiu WY, Hossain K, Stiver W, McCurdy D, Peterson
S. 1982. Development and calibration of an oil spill behavior
REFERENCES model. Report CG-D-27-83. U.S. Coast Guard, Research and
1. Huang JC, Monastery FC. 1982. Review of the state of the art Development Center, Groton, CT.
of oil spill simulation models. Final Report. American Petroleum 20. Mackay D, McAuliffe CD. 1988. Fate of hydrocarbons dis-
Institute, Washington, DC. charged at sea. Oil Chem Pollut 5:1–20.
2. Huang JC. 1983. A review of the state of the art of oil spill fate/ 21. Kirstein BE, Clayton JR, Clary C, Payne JR, McNabb D Jr,
behavior models. Proceedings, 1983 International Oil Spill Con- Fauna G, Redding R. 1987. Integration of suspended particulate
ference, San Antonio, TX, USA, February 28–March 3, Amer- matter and oil transportation study. OCS Study MMS87-0083.
ican Petroleum Institute, Washington, DC, pp 313–322. Minerals Management Service, Anchorage, AK, USA.
3. Spaulding ML. 1988. A state-of-the-art review of oil spill tra- 22. Daling PS, Mackay D, Mackay N, Brandvik PJ. 1990. Droplet
jectory and fate modeling. Oil Chem Pollut 4:39–55. size distributions in chemical dispersion of oil spills: Towards
4. American Society of Civil Engineers Task Committee on Mod- a mathematical model. Oil Chem Pollut 7:173–198.
eling of Oil Spills. 1996. State-of-the-art review of modeling 23. Anderson E, Howlett E, Knauss W, French DP, Spaulding ML,
transport and fate of oil spills. J Hydraulic Eng-ASCE 122:594– Reed M, Puckett S, Isaji T, Mendelsohn D. 1990. The Alyeska
609. tactical oil spill model. Mar Technol Soc J 24:33–39.
5. Reed M, Johansen O, Brandi PJ, Doling P, Lewis A, Fiasco R, 24. Kolluru V, Spaulding ML, Anderson E. 1994. A three-dimen-
Mackay D, Prentiss R. 1999. Oil spill modeling towards the sional subsurface oil dispersion model using a particle-based
close of the 20th century: Overview of the state-of-the-art. Spill approach. Proceedings, 17th Arctic and Marine Oil Spill Pro-
Science and Technology Bulletin 5:3–16. gram Technical Seminar, Vancouver, BC, Canada, June 8–10.
6. Yapa PD, Shen HT, Wang DS, Angammana K. 1992. An inte- Emergencies Science Division, Environment Canada, Ottawa,
grated computer model for simulating oil spills in the St. Lawr- ON, pp 815–835.
ence River. J GT Lakes Res:34–51. 25. Spaulding ML, Howlett E, Anderson EL, Jayko K. 1992. OIL-
7. LaBelle RP, Johnson WR. 1993. Stochastic oil spill analysis for MAP: A global approach to spill modeling. Proceedings, 15th
Cook Inlet/Shelikof Strait. Proceedings, 16th Arctic Marine Oil Arctic Marine Oil Spill Program Technical Seminar, Edmonton,
Spill Program Technical Seminar, Calgary, AB, June 7–9. Emer- AB, Canada, June 9–11. Emergencies Science Division, Envi-
gencies Science Division, Environment Canada, Ottawa, ON, pp ronment Canada, Ottawa, ON, pp 15–21.
573–585. 26. Spaulding ML, Kolluru V, Anderson E, Howlett E. 1994. Ap-
8. Leech M, Walker M, Wiltshire M, Tyler A. 1993. OSIS: A plication of three-dimensional oil spill model (WOSM/OIL-
Windows 3 oil spill information system. Proceedings, 16th Arc- MAP) to hindcast the Braer spill. Spill Science and Technology
tic Marine Oil Spill Program Technical Seminar, Calgary, AB, Bulletin 1:23–35.
Canada, June 7–9. Emergencies Science Division, Environment 27. French DP, Rines H, Masciangioli P. 1997. Validation of an
Canada, Ottawa, ON, pp 549–572. Orimulsion spill fates model using observations from field test
9. Proctor R, Elliott AJ, Flather R. 1994. Forecast and hindcast spills. Proceedings, 20th Arctic and Marine Oil Spill Program
simulations of the Braer oil spill. Mar Pollut Bull 28:219–229. Technical Seminar, Vancouver, BC, Canada, June 10–13. Emer-
10. Lehr WJ. 1996. Progress in oil spread modeling. Proceedings, gencies Science Division, Environment Canada, Ottawa, ON,
19th Arctic and Marine Oil Spill Program Technical Seminar, pp 933–961.
Calgary, AB, Canada, June 12–14. Emergencies Science Divi- 28. Reed M, Daling PS, Brakstd OG, Singsaas I, Faksness LG,
sion, Environment Canada, Ottawa, ON, pp 889–894. Hetland B, Efrol N. 2000. OSCAR 2000: A multicomponent 3-
11. Lehr WJ, Overstreet R, Jones R, Watabayashi G. 1992. ADI- dimensional oil spill contingency and response model. Pro-
OS—Automatic data inquire for oil spill. Proceedings, 15th Arc- ceedings, 23rd Arctic Marine Oil Spill Program Technical Sem-
tic and Marine Oil Spill Program Technical Seminar, Edmonton, inar, Vancouver, BC, Canada, June 14–16, 1992. Emergencies
AB, Canada, June 9–11. Emergencies Science Division, Envi- Science Division, Environment Canada, Ottawa, ON, pp 663–
ronment Canada, Ottawa, ON, pp 31–45. 952.
12. Lehr WJ, Wesley D, Simecek-Beatty D, Jones R, Kachook G, 29. U.S. Department of the Interior. 1996. CERCLA Natural Re-
Lankford J. 2000. Algorithm and interface modifications of the source Damage Assessment, Type A, Final Rule. US Fed Reg
NOAA oil spill behavior model. Proceedings, 23rd Arctic and 61:20559–20614.
Marine Oil Spill Program Technical Seminar, Vancouver, BC, 30. French D, Reed M, Jayko K, Feng S, Rines H, Pavignano S,
2454 Environ. Toxicol. Chem. 23, 2004 D.P. French-McCay

Isaji T, Puckett S, Keller A, French FW III, Gifford D, McCue of Principal Investigators, Vols 21 and 22. U.S. Department of
J, Brown G, MacDonald E, Quirk J, Natzke S, Bishop R, Welsh Commerce, National Oceanic, and Atmospheric Administration,
M, Phillips M, Ingram BS. 1996. The CERCLA type A natural Ocean Assessment Division, Juneau, AK.
resource damage assessment model for coastal and marine en- 48. Payne JR, Kirstein BE, Clayton JR Jr, Clary C, Redding R,
vironments (NRDAM/CME), technical documentation, Vols 1– McNabb GD Jr, Farmer G. 1987. Integration of suspended par-
6. Contract 14-0001-91-C-11. Final Report. U.S. Department of ticulate matter and oil transportation study. Contract 14-12-
the Interior, Office of Environmental Policy and Compliance, 0001-30146. Final Report. U.S. Department of the Interior, Min-
Washington, DC. erals Management Service, Alaska OCS Region, Anchorage,
31. Reed M, Spaulding ML. 1984. Response of Georges Bank cod AK.
to periodic and nonperiodic oil spill events. Environ Manag 8: 49. Jones RK. 1997. A simplified pseudo-component oil evaporation
67–74. model. Proceedings, 20th Arctic and Marine Oil Spill Program
32. Reed M, French DP, Calambokidis J, Cubbage J. 1989. Simu- Technical Seminar, Vancouver, BC, Canada, June 10–13. Emer-
lation modelling of the effects of oil spills on population dy- gencies Science Division, Environment Canada, Ottawa, ON,
namics of northern fur seals. Ecol Model 49:49–71. pp 43–61.
33. Spaulding ML, Saila SB, Lorda E, Walker HA, Anderson EL, 50. National Research Council. 1985. Oil in the Sea: Inputs, Fates,
Swanson JC. 1983. Oil spill fishery interaction modelling: Ap- and Effects. National Academy, Washington, DC, USA.
plication to selected Georges Bank fish species. Estuar Coast 51. National Research Council. 2002. Oil in the Sea III: Inputs,
Shelf Sci 16:511–541. Fates and Effects. National Academy, Washington, DC, USA.
34. Spaulding ML, Reed M, Anderson E, Isaji T, Swanson JC, Saila 52. Neff JM, Anderson JW, Cox BA, Laughlin RB Jr, Rossi SS,
S, Lorda E, Walker H. 1985. Oil spill fishery impact assessment Tatem HE. 1976. Effects of petroleum on survival respiration
model: Sensitivity to spill location and timing. Estuar Coast and growth of marine animals. In Sources, Effects, and Sinks
Shelf Sci 20:41–53. of Hydrocarbons in the Aquatic Environment. American Insti-
35. Laevastu T, Fukuhara F, eds. 1984. Quantitative determination tute of Biological Sciences, Washington, DC, pp 515–539.
of the effects of oil development in the Bristol Bay region on 53. Neff JM, Anderson JW. 1981. Response of Marine Animals to
the commercial fisheries in the Bering Sea. NWAFC Processed Petroleum and Specific Petroleum Hydrocarbons. Applied Sci-
Report 84-06. U.S. Department of Commerce, National Marine ence, London and Halsted Press Division, John Wiley, New
Fisheries Service, Northwest and Alaska Fisheries Center, Seat- York, NY, USA.
tle, WA. 54. Capuzzo JM. 1987. Biological effects of petroleum hydrocar-
36. Trudel BK. 1984. A mathematical model for predicting the eco- bons: Assessments from experimental results. In Boesch DF,
logical impact of treated and untreated oil spills. In Allen TE, Rabulais NN, eds, Long-Term Environmental Effects of Off-
ed, Oil Spill Chemical Dispersants. STP 340. American Society shore Oil and Gas Development. Elsevier, New York, NY, USA,
for Testing and Materials, Philadelphia, PA, USA, pp. 390–413. pp 343–410.
37. Trudel BK, Belore RC, Jessiman BJ, Ross SL. 1989. A micro- 55. Rice SD, Short JW, Karinen JF. 1977. Comparative oil toxicity
computer-based spill impact assessment system for untreated and and comparative animal sensitivity. In Wolfe DA, ed, Fate and
chemically dispersed oil spills in the U.S. Gulf of Mexico. Pro- Effects of Petroleum Hydrocarbons in Marine Ecosystems and
ceedings, Oil Spill Conference, San Antonio, TX, USA, Feb- Organisms. Pergamon, New York, NY, USA, pp 78–94.
ruary 13–16, American Petroleum Institute, Washington, DC, 56. Rice SD, Moles A, Taylor TL, Karinen JR. 1979. Sensitivity of
pp 139–151. 39 Alaskan marine species to Cook Inlet crude and No. 2 fuel
38. Jayko K, Reed M, Bowles A. 1990. Simulation of interactions oil. Proceedings, 1979 Oil Spill Conference, Los Angeles, CA,
between migrating whales and potential oil spills. Environ Pollut USA, March 19–22. American Petroleum Institute, Washington,
63:97–127. DC, pp 549–554.
39. Samuels WB, Lanfear KJ. 1982. Simulations of seabird damage 57. Tatem HE, Cox BA, Anderson JW. 1978. The toxicity of oils
and recovery from oil spills in the northern Gulf of Alaska. J and petroleum hydrocarbons to estuarine crustaceans. Estuar
Environ Manag 15:169–182. Coast Mar Sci 6:365–373.
40. Ford RG. 1985. A risk analysis model for marine mammals and 58. Malins DC, Hodgins HO. 1981. Petroleum and marine fishes:
seabirds: A southern California Bight scenario. MMS 85-0104. A review of uptake, disposition, and effects. Envir Sci Technol
Final Report. U.S. Department of the Interior, Minerals Man- 15:1272–1280.
agement Service, Pacific OCS Region, Los Angeles, CA. 59. Anderson JW. 1985. Toxicity of dispersed and undispersed Prud-
41. Ford RG. 1987. Estimating mortality of seabirds from oil spills. hoe Bay crude oil fractions to shrimp, fish, and their larvae.
Proceedings, 1987 Oil Spill Conference, Baltimore, MD, USA, Publication 4441. American Petroleum Institute, Washington,
April 6–9. American Petroleum Institute, Washington, DC, pp DC.
547–551. 60. Anderson JW, Riley RG, Kiesser SL, Gurtisen J. 1987. Toxicity
42. Ford RG, Wiens JA, Heinemann D, Hunt GL. 1982. Modelling of dispersed and undispersed Prudhoe Bay crude oil fractions
the sensitivity of colonially breeding marine birds to oil spills: to shrimp and fish. Proceedings, 1987 Oil Spill Conference,
Guillemot and kittiwake populations on the Pribilof Islands, Be- Baltimore, MD, USA, April 6–9. American Petroleum Institute,
ring Sea. J Appl Ecol 19:1–31. Washington, DC, pp 235–240.
43. Brody A. 1988. A simulation model for assessing the risks of 61. McAuliffe CD. 1987. Organism exposure to volatile/soluble hy-
oil spills to the California sea otter population and an analysis drocarbons from crude oil spills—A field and laboratory com-
of the historical growth of the population. In Siniff DB, Ralls parison. Proceedings, 1987 Oil Spill Conference, Baltimore,
K, eds, Population Status of California Sea Otters. Contract MD, USA, April 6–9. American Petroleum Institute, Washing-
14-12-001-30033. U.S. Department of the Interior, Minerals ton, DC, pp 275–288.
Management Service, Pacific Outer Continental Shelf Region, 62. French-McCay DP. 2002. Development and application of an oil
Los Angeles, CA, pp 191–274. toxicity and exposure model, OilToxEx. Environ Toxicol Chem
44. French DP, Reed M, Calambokidis J, Cubbage J. 1989. A sim- 21:2080–2094.
ulation model of seasonal migration and daily movements of the 63. Mackay D, Shiu WY, Ma KC. 1992. Illustrated Handbook of
northern fur seal, Callorhinus ursinus. Ecol Model 48:193–219. Physical–chemical Properties and Environmental Fate for Or-
45. Seip KL, Sandersen E, Mehlum F, Ryssdel J. 1991. Damages to ganic Chemicals, Vol 1—Monoaromatic Hydrocarbons, Chlo-
seabirds from oil spills: Comparing simulation results and vul- robenzenes, and PCBs. Lewis, Boca Raton, FL, USA.
nerability indexes. Ecol Model 53:39–59. 64. Mackay D, Shiu WY, Ma KC. 1992b. Illustrated Handbook of
46. French-McCay DP. 2003. Development and application of dam- Physical–chemical Properties and Environmental Fate for Or-
age assessment modeling: Example assessment for the North ganic Chemicals, Vol 2—Polynuclear Aromatic Hydrocarbons,
Cape oil spill. Mar Pollut Bull 47:341–359. Polychlorinated Dioxins, and Dibenzofurans. Lewis, Boca Ra-
47. Payne JR, Kirstein BE, McNabb GD Jr, Lambach JL, Redding ton, FL, USA.
R, Jordan RE, Hom W, deOliveria C, Smith GS, Baxter DM, 65. Mackay D, Shiu WY, Ma KC. 1992c. Illustrated Handbook of
Gaegel R. 1984. Multivariate analysis of petroleum weathering Physical–chemical Properties and Environmental Fate for Or-
in the marine environment—Sub-Arctic, Environmental As- ganic Chemicals, Vol 3—Volatile Organic Chemicals. Lewis,
sessment of the Alaskan Continental Shelf, OCEAP. Final Report Boca Raton, FL, USA.
Oil spill impact assessment modeling Environ. Toxicol. Chem. 23, 2004 2455

66. Mackay D, Shiu WY, Ma KC. 1992d. Illustrated Handbook of ment of the Army, Waterways Experiment Station, U.S. Army
Physical–chemical Properties and Environmental Fate for Or- Corps of Engineers, Vicksburg, MS.
ganic Chemicals. Vol 4—Oxygen-, Nitrogen-, and Sulfur-Con- 89. Lunel T. 1993. Dispersion: Oil droplet size measurements at sea.
taining Compounds. Lewis, Boca Raton, FL, USA. Proceedings, 16th Arctic Marine Oil Spill Program Technical
67. Youssef M, Spaulding ML. 1993. Drift current under the action Seminar, Calgary, AB, June 7–9. Emergencies Science Division,
of wind waves. Proceedings, 16th Arctic Marine Oil Spill Pro- Environment Canada, Ottawa, ON, pp 1023–1056.
gram Technical Seminar, Calgary, AB, Canada, June 7–9. Emer- 90. Hines AL, Maddox RN. 1985. Mass Transfer Fundamentals
gencies Science Division, Environment Canada, Ottawa, ON, and Application. Prentice-Hall, Englewood Cliffs, NJ, USA.
pp 587–615. 91. Lyman CJ, Reehl WF, Rosenblatt DH. 1982. Handbook of Chem-
68. Okubo, A. 1971. Oceanic diffusion diagrams. Deep-Sea Res 8: ical Property Estimation Methods. McGraw-Hill Book, New
789–802. York, NY, USA.
69. Bear J, Verruijt A. 1987. Modeling Groundwater Flow and Pol- 92. Di Toro DM, Zarba CS, Hansen DJ, Berry WJ, Swartz RC,
lution with Computer Programs for Sample Cases. Kluwer Ac- Cowan CE, Pavlou SP, Allen HE, Thomas NA, Paquin PR. 1991.
ademic, London, UK. Annual review: Technical basis for establishing sediment quality
70. Thorpe SA. 1984. On the determination of Kv in the near surface criteria for nonionic organic chemicals using equilibrium par-
ocean from acoustic measurements of bubbles. J Phys Oceanogr titioning. Environ Toxicol Chem 10:1541–1583.
14:861–863. 93. French DP, Rines H. 1997. Validation and use of spill impact
71. Gundlach ER. 1987. Oil holding capacities and removal coef- modeling for impact assessment. Proceedings, 1997 Interna-
ficients for different shoreline types to computer-simulated spills tional Oil Spill Conference, Fort Lauderdale, FL, USA, April
in coastal waters. Proceedings, 1987 Oil Spill Conference, Bal- 7–10. API Publication 4651. American Petroleum Institute,
timore, MD, USA, April 6–9. American Petroleum Institute, Washington, DC, pp 829–834.
Washington, DC, pp 451–457. 94. Craddock DR. 1977. Acute toxic effects of petroleum on arctic
72. Reed M, Gundlach E. 1989. Hindcast of the Amoco Cadiz event and subarctic marine organisms. In Malins DC, ed, Effects of
with a coastal zone oil spill model. Oil Chem Pollut 5:451–476. Petroleum on Arctic and Subarctic Marine Environments and
73. Fay JA. 1971. Physical processes in the spread of oil on a water Organisms, Vol 2, Biological Effects. Academic, New York,
surface. In Proceedings, Joint Conference and Control of Oil NY, USA, pp 1–93.
Spills, June 15–17, Washington, DC. 95. Sprague JB. 1970. Measurement of pollutant toxicity to fish. II.
74. Johansen O. 1984. The Halten Bank experiment—Observations Utilizing and applying bioassay results. Water Res 4:3–32.
and model studies of drift and fate of oil in the marine envi- 96. Kooijman SALM. 1981. Parametric analysis of mortality rates
ronment. Proceedings, 11th Arctic Marine Oil Spill Program in bioassays. Water Res 15:107–119.
Technical Seminar, Vancouver, BC, Canada, June 7–9, 1992. 97. French DP, French FW III. 1989. The biological component of
Emergencies Science Division, Environment Canada, Ottawa, the CERCLA type—A Damage Assessment Model system. Oil
ON, pp 18–36. Chem Pollut 5:125–163.
75. Elliot AJ, Hurford N, Penn CJ. 1986. Shear diffusion and the 98. McCarty LS, Ozburn GW, Smith AD, Bharath A, Orr D, Dixon
spreading of oil slicks. Mar Pollut Bull 17:308–313. DG. 1989. Hypothesis formulation and testing in aquatic bio-
76. Reed M, Turner C, Odulo A. 1994. The role of wind and emul- assays: A deterministic model approach. Hydrobiologia 188/
sification in modeling oil spill and drifter trajectories. Spill Sci 189:533–542.
Technol Bull 1:143–157. 99. McCarty LS, Mackay D, Smith AD, Ozburn GW, Dixon DG.
77. Mackay D, Matsugu RS. 1973. Evaporation rates of liquid hy- 1992. Residue-based interpretation of toxicity and bioconcen-
drocarbon spills on land and water. Can J Chem Eng 51:434– tration QSARs from aquatic bioassays: Neutral narcotic organ-
439. ics. Environ Toxicol Chem 11:917–930.
78. Fingas M. 1995. Water-in-oil emulsion formation: A review of 100. McCarty LS, Mackay D, Smith AD, Ozburn GW, Dixon DG.
physics and mathematical modelling. Spill Science and Tech- 1992. Toxicokinetic modeling of mixtures of organic chemicals.
nology Bulletin 2:55–59. Environ Toxicol Chem 11:1037–1047.
79. Stiver W, Mackay D. 1984. Evaporation rate of oil spills of 101. Mackay D, Puig H, McCarty LS. 1992. An equation describing
hydrocarbons and petroleum mixtures. Environ Sci Technol 18: the time course and variability in uptake and toxicity of narcotic
834–840. chemicals to fish. Environ Toxicol Chem 11:941–951.
80. Payne JR, McNabb GD Jr. 1984. Weathering of petroleum in 102. Buikema AL Jr, Neiderlehner BR, Cairns J Jr. 1982. Biological
the marine environment. Marine Technology Society Journal monitoring. Part IV—Toxicity testing. Water Res 16:239–262.
18:24–42. 103. Durbin AG, Durbin EG, Verity PG, Smayda TJ. 1981. Voluntary
81. Payne JR, Phillips CR. 1985. Petroleum Spills. Lewis, Boca swimming speeds and respiration rates of a filter-feeding plank-
Raton, FL, USA. tivore, the Atlantic Menhaden, Brevoortia tyrannus (Pisces: Clu-
82. Payne JR, Phillips CR. 1985. Photochemistry of petroleum in peidae). Fish Bull 78:877–886.
water. Environ Sci Technol 19:569–579. 104. Odum EP. 1971. Fundamentals of Ecology. W.B. Saunders, Phil-
83. Daling PS, Brandvik PJ. 1988. A study of the formation and adelphia, PA, USA.
stability of water-in-oil emulsions. Proceedings, 11th Arctic Ma- 105. Di Toro DM, McGrath JA, Hansen DJ. 2000. Technical basis
rine Oil Spill Program Technical Seminar, Vancouver, BC, Can- for narcotic chemicals and polycyclic aromatic hydrocarbon cri-
ada, June 7–9, 1992. Emergencies Science Division, Environ- teria. I. Water and tissue. Environ Toxicol Chem 19:1951–1970.
ment Canada, Ottawa, ON, pp 153–170. 106. Di Toro DM, McGrath JA. 2000. Technical basis for narcotic
84. Fingas M, Fieldhouse B, Mullin JV. 1997. Studies of water-in- chemicals and polycyclic aromatic hydrocarbon criteria. II. Mix-
oil emulsions: Stability studies. Proceedings, 20th Arctic and tures and sediments. Environ Toxicol Chem 19:1971–1982.
Marine Oil Spill Program Technical Seminar, Vancouver, BC, 107. Nirmalakhandan N, Speece RE. 1988. Structure–activity rela-
Canada, June 10–13. Emergencies Science Division, Environ- tionships, quantitative techniques for predicting the behavior of
ment Canada, Ottawa, ON, pp 21–42. chemicals in the ecosystem. Environ Sci Technol 22:606–615.
85. Mackay D, Zagorski W. 1982. Water-in-oil emulsions. Environ- 108. Hodson PV, Dixon DG, Kaiser KLE. 1988. Estimating the acute
ment Canada Manuscript Report EE-34. Environment Canada, toxicity of waterborne chemicals in trout from measurements of
Ottawa, ON. median lethal dose and the octanol–water partition coefficient.
86. Delvigne GAL, Sweeney CE. 1988. Natural dispersion of oil. Environ Toxicol Chem 7:443–454.
Oil Chem Pollut 4:281–310. 109. Blum DJ, Speece RE. 1990. Determining chemical toxicity to
87. Delvigne GAL, Hulsen LJM. 1994. Simplified laboratory mea- aquatic species. Environ Sci Technol 24:284–293.
surements of oil dispersion coefficient—Application in com- 110. McCarty LS. 1986. The relationship between aquatic toxicity
putations of natural oil dispersion. Proceedings, 17th Arctic QSARs and bioconcentration for some organic chemicals. En-
Marine Oil Spill Program Technical Seminar, Vancouver, BC, viron Toxicol Chem 5:1071–1080.
Canada, June 8–10. Emergencies Science Division, Environment 111. McCarty LS, Mackay D. 1993. Enhancing ecotoxicological
Canada, Ottawa, ON, pp 173–187. modeling and assessment. Environ Sci Technol 27:1719–1728.
88. Coastal Engineering Research Center. 1984. Shore Protection 112. Varhaar HJM, VanLeeuwen CJ, Hermens JLM. 1992. Classi-
Manual, Vol 1—Coastal Engineering Research Center, Depart- fying environmental pollutants, 1: Structure–activity relation-
2456 Environ. Toxicol. Chem. 23, 2004 D.P. French-McCay

ships for prediction of aquatic toxicity. Chemosphere 25:471– 120. Okubo A, Ozmidov RV. 1970. Empirical dependence of the
491. coefficient of horizontal turbulent diffusion in the ocean on the
113. Swartz RC, Schults DW, Ozretich RJ, Lamberson JO, Cole FA, scale of the phenomenon in question. Atmospheric and Ocean
DeWitt TH, Redmond MS, Ferraro SP. 1995. SPAH: A model Physic 6:534–536.
to predict the toxicity of polynuclear aromatic hydrocarbon mix- 121. Day RH, Murphy SM, Wiens JA, Hayward GD, Harner EJ,
tures in field-collected sediments. Environ Toxicol Chem 14: Smith LN. 1995. Use of oil-affected habitats by birds after the
1977–1987. Exxon Valdez oil spill. In Wells PG, Butler JN, Hughes JS, eds,
114. Exxon Valdez Oil Spill Trustee Council. 1989. State/federal nat- Exxon Valdez Oil Spill: Fate and Effects in Alaskan Water.
ural resource damage assessment plan for the Exxon Valdez Oil STP 1219. American Society for Testing and Materials, Phila-
Spill, August 1989. Public Review Draft. Exxon Valdez Oil Spill delphia, PA, pp 726–761.
Trustee Council, Juneau, AK, USA. 122. Piatt JF, Lensink CJ, Butler W, Kendziorek M, Nysewander DR.
115. Galt JA, Lehr WJ, Payton DL. 1991. Fate and transport of the 1990. Immediate impact of the Exxon Valdez oil spill on marine
Exxon Valdez oil spill, Part 4 of a five-part series. Environ Sci birds. Auk 107:387–397.
Technol 25:202–209. 123. Ford RG, Bonnell ML, Varoujean DH, Page GW, Carter HR,
116. Payne JR, Clayton JR Jr, Daniel McNabb G Jr, Kirstein BE. Sharp BE, Heinemann D, Casey JL. 1996. Total direct mortality
1991. Exxon Valdez oil weathering fate and behavior: Model of seabirds from the Exxon Valdez oil spill. Am Fish Soc Symp
predictions and field observations. Proceedings, 1991 Interna- 18:684–711.
tional Oil Spill Conference, San Diego, CA, USA, March 4–7, 124. Schempf PF, Bowman TD, Bernatowicz J. 1992. Assessing the
American Petroleum Institute, Washington, DC, pp 641–654. effect of the Exxon Valdez oil spill on bald eagles. Bird Study
117. Wolfe DA, Hameedi MJ, Michel J, Galt JA, Watabayashi G, 4. Annual Progress Report—1991. November 22, 1991, revised
Payne JR, Short J, O’Claire C, Rice S, Braddock J, Hanna S, May 12. Exxon Valdez. Oil Spill Trustee Council, Juneau, AK,
Sale D. 1994. The fate of the oil spilled from the Exxon Valdez. USA.
Environ Sci Technol 28:561–568. 125. Exxon Valdez Oil Spill Trustee Council. 1991. Summary of the
118. Gundlach E, Pavia EA, Robinson C, Civeaut JC. 1991. Shoreline effects of the Exxon Valdez oil spill on natural resources and
surveys at the Exxon Valdez oil spill: The state of Alaska re- archaeological resources. Juneau, AK, USA, Filed by U.S De-
sponse. Proceedings, 1991 International Oil Spill Conference, partment of Justice, U.S. District Court, Alaska, Civil Action
San Diego, CA, USA, March 4–7, American Petroleum Institute, No. A91-082.
Washington, DC, pp 519–529. 126. Neff JM, Burns WA. 1996. Estimation of polycyclic aromatic
119. Isaji T, Spaulding ML. 1987. A numerical model of the M2 and hydrocarbon concentrations in the water column based on tissue
K2 tide in the northwestern Gulf of Alaska. J Phys Oceanogr residues in mussels and salmon: An equilibrium partitioning
17:698–704. approach. Environ Toxicol Chem 15:2240–2253.