See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/256925290

Estimation of oil spill behaviour parameters from readily-available oil

properties

Conference Paper · January 2007

CITATIONS READS

6 33

1 author:

Merv Fingas
Concordia University College of Alberta
388 PUBLICATIONS   7,929 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Working on new dispersion model View project

Oil Fates View project

All content following this page was uploaded by Merv Fingas on 09 March 2020.

The user has requested enhancement of the downloaded file.

 Estimation of Oil Spill Behaviour Parameters from Readily-Available Oil
Properties

Merv Fingas
Edmonton, Alberta
E mail: [email protected]

Abstract
Oils are very complex mixtures of hydrocarbons resulting in somewhat unique
oil properties for every oil. Some of this uniqueness may be bourne out in the simple
properties such as density and viscosity, others not. The solution to this uniqueness in
the past has been to model oils on typical characteristics such as light crude oil,
medium crude oil, etc. Another method has been to use very specific crude oil data.
When an oil or petroleum product is spilled there often is a need to be able to predict
oil behaviour until more comprehensive data can be found, with only simple data
such as density or viscosity.
Often oil behaviour parameters such as for evaporation, emulsification and
chemical dispersion, are needed and specific behavioural data are not available.
Despite the growth of oil property data bases, and initially at least, little data on some
spilled oils are available. Often only basic data such as density, and viscosity are
available. This paper presents a series of simple schemes to provide prediction for oil
evaporation, emulsification and chemical dispersion given readily-available data.
Schemes for prediction are given using only density and viscosity and then adding
SARA and other composition data. As the input data amount is increased, more
accuracy in the estimations are achieved.
The estimation schemes are compared to the measured data available and also
to the use of generic oil types sometimes used in models such as ‘light or heavy crude
oil’. It was found that the accuracy of these estimation schemes are more accurate
than the traditional predictions based on generic oil types.

1 Introduction
An important tool in oil spill planning and response is oil spill modeling. Oil
spill modeling relies on algorithms that require a number of oil property inputs.
Sometimes these inputs are not available, especially at the time of a spill. It is the
focus of this paper to report on the development of simple estimators to predict oil
evaporation, emulsification and chemical dispersion, using only density and viscosity.
Density and viscosity are often the only data available at the time of the spill. The
algorithm development will be described with the background of the significance of
spill processes and more detailed behaviour prediction algorithms, sometimes using
more sophisticated chemical data inputs.
The behaviour of oil in the environment is largely dictated by its chemical
composition. Oil composition data that are relatively available include the SARA or
Saturates, Aromatics, Resin and Asphaltene contents. As will be shown later,
sometimes these data are not sufficient to yield a truly accurate prediction. More
detailed compositional data are available for only a few oils (Oil Catalogue, 2007).
The most important behavioural aspect is felt to be evaporation. Evaporation
is responsible for the largest mass balance change in oils except for those rare times
when oils disperse naturally. Evaporation is generally necessary before
emulsification occurs and is a competing process to natural dispersion.
The chemistry of crude and refined oils defines their behaviour at sea. The
chemistry of crude and refined oils is in turn, dominated by the fact that they are
composed of many, in some cases, hundreds, of different compounds, each with their
own behaviour and properties. The combination of these compounds can be viewed
from a number of different perspectives; some as a combination of selected typical
compounds, others view oil as being represented by a single compound and still
others view each oil as a unique substance. All of these points of view have
disadvantages in modelling. If an oil is viewed as a combination of a few marker
compounds, some of the behavioural aspects are not accounted for - eg. asphaltenes
and their role in water-in-oil emulsification. Taking the oil as a single compound
removes many behavioural characteristics of an oil and thus is not useful for spill
modelling if behaviour is an important factor. Taking an oil as a unique combination
complicates spill modelling in that new oils are difficult to add. The importance of
understanding the composition and overall behaviour of a specific oil cannot be over-
emphasized.
A number of oil properties are important to the understanding of oils and their
behaviour. An important property is viscosity or resistance to flow. Technically
defined, viscosity is the shear resistance (force) divided by the shear rate (velocity).
The viscosity of an oil also defines its spreading rate and is related in some way to
most behavioural aspects. Density is the weight of specific volume of oil and in
terms of behaviour defines how an oil will float on water. Density is strongly
correlated to viscosity for most oils. American Petroleum Institute (API) gravity is a
different way of stating density (APIo = 141.5/d -131.5). Water content is an
important measurement for oils that have emulsified or formed water-in-oil
emulsions. This is an indication of the expansion of spill volume resulting from water
incorporation and to a lesser degree of the stability of the resulting emulsion. Surface
and interfacial tension are often cited as being important factors. The interfacial
tension is the energy difference between the oil layer and air or water. Interfacial
tension is not as important as some might stress since it cannot be used in isolation to
predict any behavioural aspects of the oil. Furthermore it varies little between oils.
Pour point is the temperature below which the oil ceases to pour as a fluid. Pour
point is a difficult measurement and subject to variance, furthermore with a multi-
component fluid such as crude oil and many of the distilled products, pour point is
not a meaningful measurement.

2 Evaporation
Evaporation is a very important process for most oil spills. In a few days,
typical crude oils can lose up to 40% of their volume (Fingas, 2004). Most oil spill
behaviour models include evaporation as a process and as an output of the model.
Despite the importance of the process, relatively little work had been conducted on
the basic physics and chemistry of oil spill evaporation until recently (Fingas, 1995).
The difficulty with oil evaporation is that oil is a mixture of hundreds of compounds
and this mixture varies from source to source and even over time. Much of the work
described in the literature focuses on ‘calibrating’ equations developed for water
evaporation (Fingas 1995).
Scientific and quantitative work on water evaporation is decades old (Jones,
1992; Brutsaert, 1982). The basis for the oil work in the literature is water
evaporation. There are several fundamental differences between the evaporation of a
pure liquid such as water and that of a multi-component system such as crude oil.
Most obviously, the evaporation rate for a single liquid such as water is a constant
with respect to time (Jones, 1992; Brutsaert, 1982). Evaporative loss, by total weight
or volume, is not linear with time for crude oils and other multi-component fuel
mixtures (Stiver and Mackay, 1984; Fingas, 1997).
Evaporation of a liquid can be considered as the movement of molecules from
the surface into the vapour phase above it. The layer of air above the evaporation
surface is known as the boundary layer (Monteith and Unsworth, 1990). The
characteristics of this air layer or boundary layer can influence evaporation. In the
case of water, the boundary layer regulates the evaporation rate. Air can hold a
variable amount of water, depending on temperature, as expressed by the relative
humidity. Under conditions where the boundary layer is not moving (no wind) or has
low turbulence, the air immediately above the water quickly becomes saturated and
evaporation slows or ceases. In practice, the actual evaporation of water proceeds at
a small fraction of the possible rate because of the saturation of the boundary layer.
The boundary layer physics is then said to regulate the evaporation of water. This
regulation manifests itself in the sensitivity of evaporation to wind or turbulence.
When turbulence is weak or absent, evaporation can slow down by orders-of-
magnitude. The molecular diffusion of water molecules is at least 103 times slower
than turbulent diffusion (Monteith and Unsworth, 1990).
If the evaporation of oil was like that of water and was boundary-layer
regulated one could write the mass transfer rate in semi-empirical form (also in
generic and unitless form) as:

where E is the evaporation rate in mass per unit area, K is the mass transfer rate of the
evaporating liquid, presumed constant for a given set of physical conditions,
sometimes denoted as kg (gas phase mass transfer coefficient, which may incorporate
some of the other parameters noted here), C is the concentration (mass) of the
evaporating fluid as a mass per volume, Tu is a factor characterizing the relative
intensity of turbulence, and S is a factor that relates to the saturation of the boundary
layer above the evaporating liquid. The saturation parameter, S, represents the effects
of local advection on saturation dynamics. If the air is already saturated with the
compound in question, the evaporation rate approaches zero. This also relates to the
scale length of an evaporating pool. If one views a large pool over which a wind is
blowing, there is a high probability that the air is saturated downwind and the
evaporation rate per unit area is lower than for a smaller pool. It should be noted that
there any many equivalent ways of expressing this fundamental evaporation equation.
Much of the pioneering work for evaporation work was performed by Sutton
(1934). Sutton proposed an equation based largely on empirical work:
where Cs is the concentration of the evaporating fluid (mass/volume), U is the wind
speed, d is the area of the pool, Sc is the Schmidt number and r is the empirical
exponent assigned values from 0 to 2/3. Other parameters are defined as above. The
terms in this equation are analogous to the very generic equation, (1), proposed
above. The turbulence is expressed by a combination of the wind speed, U, and the
Schmidt number, Sc. The Schmidt number is the ratio of kinematic viscosity of air (í)
to the molecular diffusivity (D) of the diffusing gas in air, i.e., a dimension-less
expression of the molecular diffusivity of the evaporating substance in air. The
coefficient of the wind power typifies the turbulence level. The value of 0.78 (7/9) as
chosen by Sutton, represents a turbulent wind whereas a coefficient of 0.5 would
represent a wind flow that was more laminar. The scale length is represented by d
and has been given an empirical exponent of -1/9. This represents, for water, a weak
dependence on size. The exponent of the Schmidt number, r, represents the effect of
the diffusivity of the particular chemical, and historically was assigned values
between 0 and 2/3 (Sutton, 1934).
This expression for water evaporation was subsequently used by those
working on oil spills to predict and describe oil and petroleum evaporation. Much of
the literature follows the work of Mackay (Mackay and Matsugu, 1973; Stiver and
Mackay, 1984). Mackay and co-workers adapted the equations for hydrocarbons
using the evaporation rate of cumene. Data on the evaporation of water and cumene
were used to correlate the gas phase mass transfer coefficient as a function of
wind-speed and pool size by the equation:

K m = 0.0292 U 0.78 X -0.11 Sc -0.67 (3)

where Km is the mass transfer coefficient in units of mass per unit time and X is the
pool diameter or the scale size of evaporating area. Stiver and Mackay (1984)
subsequently developed this further by adding a second equation:

N = kmAP/(RT) (4)

where N is the evaporative molar flux (mol/s), km is the mass transfer coefficient at
the prevailing wind (m/s), A is the area (m2), P is the vapour pressure of the bulk
liquid (Pascals), R is the gas constant [8.314 Joules/(mol-K)], and T is the
temperature (K).
Thus, boundary layer regulation was assumed to be the primary regulation
mechanism for oil and petroleum evaporation. This assumption was not tested by
experimentation, as revealed by the literature search (Fingas, 1995). The implications
of these assumptions are that evaporation rate for a given oil is increased by:
- increasing turbulence
- increasing wind speed, and
- increasing the surface area of a given mass of oil.
Experiments on evaporation were conducted from 1992 to 1998 (Fingas,
1997, 1998, 1999, 2004). The experimental series are summarized below:
1. Wind Experiments
Experiments on the evaporation of oil with and without wind were conducted
with ASMB (Alberta Sweet Mixed Blend), Gasoline, and with water. Water formed a
baseline data set since much is known about its evaporation behaviour. Regressions
on the data were performed and the equation parameters calculated. Curve
coefficients are the constants from the best fit equation [Evap = a ln(t)], t = time in
minutes, for logarithmic equations or Evap = a /t, for the square root equations. Oils
with few components evaporating at one time or components with similar evaporation
behaviour, have a tendency to fit square root curves (Fingas, 1997). A comparison of
oil and water evaporation clearly shows that water evaporation rate increased, as
expected, with increasing wind velocity. The oils, ASMB and gasoline, did not show
a significant increase with increasing wind speed. Any increase may only be a result
of the increase in evaporation in going from the 0-wind level to the other levels. In
any case, they do not show the U0.78 relationship that water shows. These data show
that oil is not boundary-layer regulated.
2. Evaporation Rate and Area
A light crude oil was also used to conduct a series of experiments with
varying evaporation area. The mass of the oil was kept constant so that the thickness
of the oil would also vary. However, the greater the area, the lesser the thickness and
both factors would increase oil evaporation if it were boundary-layer regulated.
Because of the lack of correlation between area and evaporation rate, one can
conclude that evaporation rate is not highly correlated with area and that the
evaporation of oil is not boundary-layer regulated.
3. Study of Mass and Evaporation Rate
Light crude oils were again used to conduct a series of experiments with
volume as the major variant. Alternatively, thickness and area were held constant to
ensure that the strict relationship between these two variables did not affect the final
regression results A strong correlation between oil mass (or volume) and evaporation
rate was shown. This suggests no or little boundary-layer regulation.
4. Study of the Evaporation of Pure Hydrocarbons - With and Without Wind
A study of the evaporation rate of pure hydrocarbons was conducted to test
the classic boundary-layer evaporation theory as applied to the hydrocarbon
constituents of oils. This experiment showed that the evaporation rates of the pure
hydrocarbons have a variable response to wind. Heptane (hydrocarbon number 7)
shows a large difference between evaporation rate in wind and no wind conditions,
indicating boundary-layer regulation. Decane (carbon number 10) shows a lesser
effect and hexadecane (carbon number 16) shows a negligible difference between the
two experimental conditions. This experiment shows the extent of boundary-
regulation and the reason for the small or negligible degree of boundary-regulation
shown by crude oils and petroleum products. Crude oil contains very little material
with carbon numbers less than decane, often less than 3% of its composition. Even
the more volatile petroleum products, gasoline and diesel fuel only have limited
amounts of compounds more volatile than decane, and thus are also not strongly
boundary-layer regulated.
Another consideration of evaporation regulation is that of saturation
concentration, the maximum concentration soluble in air. In fact saturation
concentration of water is less than that of many common oil components (Fingas,
2004). The saturation concentration of water is in fact, about two orders of
magnitude less than the saturation concentration of volatile oil components such as
pentane. This further explains why oil has no boundary layer limitation compared to
that of water.
Several studies showed that unique equations for each oil at each temperature
can be developed (Fingas, 2004).
Further examination of the temperature behaviour of oil evaporation was
conducted by determining the equations by which the evaporation rates, or equation
parameters, change with temperature. A series of correlations was performed,
between the evaporation rates, by both percentage and weight loss, using a linear
equation.
The empirically measured parameters at 15oC were correlated with both the
slopes and the intercepts of the temperature equations. Full details of this correlation
are given in the literature (Fingas, 1996). The resulting equations are:

Percentage evaporated = [B + 0.045(T-15)] (5)

where B is the equation parameter at 15oC and T is temperature in degrees Celsius.

The finding that unique equations may be needed for each oil is a significant
disadvantage to practical end use, and a way to accurately predict evaporation using
other readily-available data is necessary. One way to do so was to use distillation
data which are very common. New procedures to measure distillation data are very
simple, fast and repeatable (Jokuty et al., 1999).
Distillation data were directly correlated to the evaporation rates determined
by experimentation (Fingas, 2004). The percentage distilled at each temperature was
correlated with the equation parameter (sometimes referred to here as the evaporation
rate). Detailed correlation data are given in the literature (Fingas, 1999). The percent
mass distilled at 180 degrees was used to calculate the relationship between the
distillation values and the equation parameters. The equations used were derived from
correlations of the data.
The data from those oils that were better fitted with square root equations,
such as diesel, Bunker C light and FCC Heavy Cycle, were calculated separately.
The equations derived for the predication are:
For oils that follow a logarithmic equation:

Percentage evaporated = [.165(%D) + .045(T-15)]ln(t) (6)

For oils that follow a square root equation:

Percentage evaporated = [.0254(%D) + .01(T-15)]/ t (7)

where %D is the percentage (by weight) distilled at 180oC.

In addition, a large number of experiments were performed on oils to directly
measure their evaporation curves. The empirical equations that result are given in the
literature and some will be summarized here as well (Fingas, 2004).
The purpose of this paper is to present findings of another simple means to
calculate the evaporation rate or parameter B in equation (5). The evaporation
parameters (before temperature correction) are the simplest way to proceed. From the
work described above, these parameters are given as the B above (equation 95)) and -
0.675 for oils that follow a logarithmic curve and -0.195 for oils that follow a square
root curve (Fingas, 1999). Data on a number of oil evaporation rates, done
experimentally and basic properties of oil were collected, primarily from a previous
paper (Fingas, 2006), and also from literature sources (Jokuty et al., 1999; Oil
Catalogue, 2007). These data are shown in Table 1.
While it is known that density and viscosity are not necessarily good
indicators of other oil properties, these may be the only data available at the site of a
spill, at least until some time passes (Jokuty et al., 1995). Figures 1 and 2 show the
general trends of the relation between the evaporation equation parameter, from Table
1 for the oils that show logarithmic behaviour and the oil’s density and viscosity.
These figures show that the general trends are shown, with significant variance that is
only captured by specific oil experiments. The evaporation equation parameters were
measured in specific experiments, and often are the result of several repeat
experiments.
The data were correlated with the evaporation parameter with the objective of
deriving a simplified model using oil density and viscosity or SARA data. The curve-
fitting statistical programs, TableCurve 3D and Datafit, were used. Models were
judged on the basis of goodness-of-fit (regression coefficient or r2) for the simplest
equation. It should be noted that one can sometimes fit large-polynomial equations to
even unrelated data, therefore it is essential to use judgement and chose a realistic
equation out of the many that are available.
The curve fitting exercise with the logarithmic evaporation data and density
and viscosity data. resulted in a best-fit simplest equation:

Evap = 15.4 -14.5*density + 2.58/viscosity (8)

Where: Evap is the evaporation equation parameter and as in equation (5) is B

- 0.675 to compensate for the temperature parameter as explained above.
The results of the correlation and the prediction equation are shown in Table
2. The regression coefficient (r2) was 0.95 and the average standard deviation
between the input equation parameters and those predicted was 0.145.
Similarly the data for those oils that evaporate as a square root with time were
correlated with density and viscosity. The results of the correlation and predicted
parameters are shown in Table 3. The regression coefficient (r2) was 0.79 and the
average standard deviation between the input equation parameters and those predicted
was 0.37. It should be noted that the fit with the oils that evaporate as a square root
with time, is not as good as with the other oils. The reason for this is not apparent.
The equation for predicting those oils that evaporate as a square root of time
is:
Evap = 2.3 -1.47*density - 0.073*ln(viscosity) (9)

Where: Evap is the evaporation equation parameter.

Other data that may be available are SARA or Saturate Aromatic Resins and
Asphaltene data. As Figure 3 shows, the saturate content correlates somewhat with
the equation parameter. Statistical analysis shows that only the saturate content
correlates and this does not so well (r2 < 0.6). The addition of the other SARA data
only marginally assists in the correlation. However, an equation for the prediction of
the evaporation parameter using SARA data is:

Evap = 0.048*S +0.006*A -0.025*R -0.0003As + 0.27 (10)

Where: Evap is the evaporation equation parameter,

S is the saturate content in percent,
A is the aromatic content in percent,
R is the resin content in percent, and
As is the asphaltene content in percent.

It is important to understand how this data compares to other prediction

methods and empirical data. Table 4 lists the evaporation rate from empirical data,
from the prediction using equation (9), for a typical light crude, Alberta Sweet Mixed
Blend (ASMB) and equation (10) for diesel fuel. This is compared to empirical data
generated from several repeat measurements (Fingas, 2004), and to that of a generic
model. The generic model uses a boundary-regulated model and generic oil types in
the case of a light crude. These are illustrated as well in Figure 4 and Figure 5. Figure
4 shows the crude oil comparison. Figure 5 shows the comparison for the diesel fuel.
As can be seen in this figure, the difference between the generic model predication
and the empirical and estimated one, is large. This is caused by the fact that the
generic models do not consider the unique evaporative behaviour of diesel fuel and
the fact that a boundary-layer model vastly over-predicts the evaporation of diesel.

3 Emulsification
Emulsification is the process whereby water-in-oil emulsions are formed.
These emulsions are often called “chocolate mousse” or “mousse” by oil spill
workers. When emulsions form, the physical properties and characteristics of oil
spills change extensively. For example, stable emulsions contain from 60 to 80%
water, thus expanding the spilled material from 2 to 5 times the original volume. The
density of the resulting emulsion can be as great as 1.03 g/mL compared to a starting
density as low as 0.80 g/mL. Most significantly, the viscosity of the oil typically
changes from a few hundred mPa.s to about 100,000 mPa.s, a typical increase of
1,000. A liquid product is thereby changed to a heavy, semi-solid material.
Many researchers feel that emulsification is the second most important
behavioural characteristic of oil after evaporation. Emulsification has a significant
effect on the behaviour of oil spills at sea. As a result of emulsification, evaporation
of oil spills slows by orders-of-magnitude, spreading slows by similar rates, and the
oil rides lower in the water column, showing different drag with respect to the wind.
Emulsification also significantly affects other aspects of a spill, such as cleanup
response. Spill countermeasures are quite different for emulsions as they are hard to
recover mechanically, to treat, or to burn.
The basics of water-in-oil emulsification are now understood and well-
established (Fingas and Fieldhouse, 2006). The basic principle is that water-in-oil
emulsions are stabilized by the formation of asphaltene layers around water droplets
in oil. The layer or film is characterized by its high-strength visco-elastic properties.
Resins also form emulsions, to a certain extent, but resins do not form a stable
emulsions, and may aid in asphaltene emulsion stability by acting as asphaltene
solvents and by providing temporary stability during the time of the slow asphaltene
migration. The extent of knowledge on water-in-oil emulsion formation to this date is
strongly limited by the amount of knowledge of asphaltenes and resins themselves.
There is still very limited knowledge on the structure and chemistry of asphaltenes
(Fingas and Fieldhouse, 2006).
The availability of methodologies to study emulsions is very important. In the
past ten years, both dielectric methods and rheological methods have been exploited
to study formation mechanisms and stability of emulsions made from many different
types of oils (Fingas and Fieldhouse, 2006). Standard chemical techniques, including
Nuclear Magnetic Resonance (NMR), chemical analysis techniques, Near-infrared
spectroscopy, microscopy, interfacial pressure, and interfacial tension, are also being
applied to emulsions. These techniques have largely confirmed findings noted in the
dielectric and rheological mechanisms.
Several efforts have been undertaken to model the formation of water-in-oil
emulsions, especially from a practical standpoint. These efforts were not particularly
successful because they did not take into account the process by which emulsions are
formed (Fingas and Fieldhouse, 2006).
Four clearly defined states of water-in-oil have been characterized by a
number of measurements, the loss of water and by their visual appearance, both on
the day of formation and within one week later. Some emulsions were observed for
over a year, with the same results. The states are stable water-in-oil emulsions,
mesostable water-in-oil emulsions, entrained water, and unstable water-in-oil
emulsions. All four states were produced during the experiments in the laboratory and
in large test tanks.
Stable emulsions are brown solid materials with an average water content of
about 80% on the first day of formation and about the same one week later. Stable
emulsions remain stable for at least 4 weeks under laboratory conditions. The
properties of the starting oil required to form a stable emulsion are: density 0.85 to
0.97 g/mL; viscosity 15 to 10,000 mPa.s; resin content 5 to 30%; asphaltene content 3
to 20%; asphaltene-to-resin ratio 0.74; and average increase in viscosity of 1,100 at
day of formation and 1,500 one week later.
Mesostable water-in-oil emulsions are brown or black viscous liquids with an
average water content of 62% on the first day of formation and about 38% one week
later. Mesostable emulsions remain so less than 3 days under laboratory conditions.
The properties of the starting oil required to form a mesostable emulsion are: density
0.84 to 0.98 g/mL; viscosity 6 to 23,000 mPa.s; resin content 6 to 30%; asphaltene
content 3 to 17%; asphaltene-to-resin ratio 0.47; and average increase in viscosity 45
at day of formation and 3 one week later.
The greatest difference between the starting oils for stable and mesostable
emulsions is the asphaltene-to-resin ratio (stable - 0.74; mesostable - 0.47) and the
ratio of viscosity increase (stable 1,100, first day and 1,500 in one week; mesostable
45, first day and 30 in one week).
Entrained water-in-oil states are black liquids with an average water content
of 42% on the first day of formation and about 15% one week later. Entrained water-
in-oil states remain so less than 1 day under laboratory conditions. The average
properties of the starting oil required to form entrained water are: density 0.97 to 0.99
g/mL; viscosity 2,000 to 60,000 mPa.s; resin content 15 to 30%; asphaltene content 3
to 22%; asphaltene-to-resin ratio 0.62; and average increase in viscosity 45 at day of
formation and 30 one week later.
The greatest differences between the starting oils for entrained water-in-oil
compared to stable and mesostable emulsions are the narrow density range (entrained
= 0.97 to 99; stable = 0.85 to 0.99; mesostable = about the same as stable) and the
ratio of viscosity increase (entrained = 13, first day and 2 in one week; stable 1,100,
first day and 1,500 in one week; mesostable 45, first day and 30 in one week).
Furthermore, the viscosity of the starting oil is 2,000 to 60,000 mPa.s compared to 15
to 10,000 mPa.s for stable emulsions and 6 to 23,000 mPa.s for mesostable
emulsions.
Unstable water-in-oil emulsions are characterized by the fact that the oil does
not hold significant amounts of water and, when it does, the water remains for only a
short time, a matter of minutes. There is a much broader range of properties of the
starting oil than for the other three water-in-oil states. For example, viscosities are
very low or very high. Included in this group are light fuels such as diesel fuel and
very heavy, viscous oil products.
The stability of emulsions is due to the formation of asphaltene and resin films
at the oil and water interface. Asphaltenes form strong, elastic films which are largely
responsible for the stability of emulsions. While there is clear evidence of interaction
between resins and asphaltenes in forming emulsions, asphaltenes can form
emulsions without resins. The most stable emulsions are formed when the asphaltene-
to-resin ratio is less than about 1. The migration experiments show that asphaltenes
migrate to the interface very slowly and that this process can continue for longer than
one month. This leads to the possibility that the resins migrate very quickly and
temporarily stabilize water droplets before stronger asphaltene films form and
displace the weaker resin films.
Fingas and Fieldhouse (2006) proposed a model that takes into account
several compositional factors in the oil. The first step in this model is to obtain or
estimate the oil properties as they are at the weathering condition of concern. The
properties needed are the density, viscosity, and saturate, resin and asphaltene
contents. These values require transformation as summarized below:
Density: If the density is less than 0.97, then the density parameter is 0.97
less the density and if it is greater than 0.97, it becomes the density less 0.97. The
value used in the equation is this transformed value. (11)
Viscosity: Take the natural logarithm (ln) of the viscosity. If the natural log of
the viscosity is less than 8.7, then the viscosity parameter is 8.7 less the viscosity
natural log and if it is greater than 8.7, it becomes the natural log of viscosity less 8.7.
The value used in the equation is this transformed value. (12)
Resin Content: If the value of the resins is zero, then set this value to 20. If the
resin content is less than 5.4, then the resin content parameter is 5.4 less the resin
content and if it is greater than 5.4, it becomes the resin content less 5.4. The value
used in the equation is the transformed value. (13)
Asphaltene Content: If the value of the asphaltene content is zero, then set the
value to 30. If the asphaltene content is less than 12, then the asphaltene content
parameter is 12 less the asphaltene content and if it is greater than 12, it becomes the
 asphaltene content less 12. The value used in the equation is this transformed value.
(14)
The class of the resulting emulsion is then calculated as follows:

Class = 0.738 - 0.197*Dt - 0.0126*Vt -0.0007*Rt - 0.00358*At (15)

Where Class is the numerical index of classification

Dt is the transformed density as calculated in equation (11)
Vt is the transformed viscosity as calculated in equation (12)
Rt is the transformed resin content as calculated in equation
(13)
At is the transformed asphaltene content as calculated in
equation (14)

The class is then assigned according to the following scheme (Table 5):
Table 5 Emulsion Classification for Model
category class num ber
unstable <0.615
mesostable >0.615 <0.66
stable >0.66
entrained >.64 <.72
but density >0.96 or viscosity >10,000

Note that the entrained state overlaps the other states, but given that the
density of the oil is greater than 0.96 mg/mL or that the viscosity of the oil is greater
than 10,000 mPa.s, then the conditions for the entrained state apply.
The kinetics can be modelled as the time to formation after the oil is
weathered to the stated percentage:

Time to formation (min) = a + b/Wh1.5 (16)

Where: a is a constant and is 27.1 for a stable emulsion formation, 47 for

meso and 30.8 for an entrained water-in-oil class
b is a constant and is 7520 for a stable emulsion formation, 49100 for
meso and 18300 for an entrained water-in-oil class
Wh is the wave height in cm
It is difficult to model emulsification using only density and viscosity,
however, this is definitely better than using no estimator or an old water-uptake
model. To perform this correlation the same data set was used as above (Fingas and
Fieldhouse, 2006). The procedure was also the same, using only the values of density
and viscosity. Many iteration are required to achieve the greatest amount of
discrimination between states.
The class of the resulting emulsion is then calculated as follows:

Class = -21.4 + 28*D -0.00000073*V (17)

Where Class is the numerical index of classification

 D is the density
V is the viscosity
The class is then assigned according to the following scheme (Table 6):

The fit statistics on this estimation technique are not as good as on the full
model described above. The state mis-classifications are 58% for the simple model
compared to only about 20% for the full model. The regression coefficient for the
correlation was 0.81 for the simple model.

4 Dispersant Effectiveness
In an earlier work, a large set of data was correlated with over 15 oil
composition and properties characteristics to produce several prediction models
(Fingas et al., 2003). It was found that the parameters that correlate most highly with
the suite of parameters are those composition parameters that relate to smaller
compounds in the oil. These include n-C12, naphthalenes, and the sum of the C12 to
C16 components. Those that relate to the large compounds in the oil relate negatively
to the dispersibility, including C26 and resins. This indicates that dispersion primarily
affects only the smaller components of the oil.
The highest correlation was achieved with the n-C12 component. The next
highest correlation was for the naphthalene content. This also indicates a high
dispersibility for naphthalene. The third highest correlation is for n-C26 and this is an
inverse correlation. This indicates that the more n-C26, the less dispersion, especially
in the initial amount. This indicates that components of the size of C26 and greater
are poorly dispersed and in fact inhibit dispersion. The fourth highest correlation is
the PAH content and this correlates positively, namely that the higher the PAH
content, the higher the dispersion. This is somewhat surprising since the PAH
compounds, especially the larger PAHs such as phenanthrene and chrysene, were not
thought to be dispersible. This high correlation indicates that most of the PAHs are
dispersible, possibly because most of the PAHs consist of smaller-ring compounds.
The fifth highest correlation is that of the sum of the C12, C14, C16, and C18
components. This correlation is highly indicative that alkanes up to C18 are the
prime components dispersed along with the PAHs. The fact that C12 correlates the
highest of these n-alkanes and that this correlation rapidly drops off to C18 with no
useable correlation for C20 indicates that primarily hydrocarbons up to C18 disperse
and that past C20 compounds disperse poorly.
It should be noted that most of the dispersability data are available for Corexit
9500, both in the past and present study. Some data in the past study were available
for Corexit 9527, Enersperse 700, Dasic LTS and a few others.
In the past study 13 models were developed as summarized below. Detailed
statistics on the models are given in Fingas et al., 2003. It should be noted that all the
alkanes measured and used as predictor values are normal alkanes.
Model 1 uses the four of the higher correlating parameters of C12,
naphthalene, PAHs, C12 to C18, and the fraction that boils at less than 250oC. The
regression coefficient achieved was 0.98.
The model is:
Corexit 9500 dispersibility (%) = -11.1 -3.19(ln C12 content) +
0.00361(naphthalene content in ppm) - 7.62(PAH content squared) + 0.115(C12 to
C18 content squared) + 0.785(%fraction oil boiling below 250oC) (18)
The Prob(t) shows that all factors are relevant and are needed to form a reliable
prediction. Only boiling point has a low Prob(t) value. It should be noted that only 15
oils have the full data set to form this prediction set.
Model 2 uses the six highest correlating parameters of C12, naphthalene,
PAHs, C12 to C18, the C26 fraction (negative correlation), and the fraction that boils
at less than 250oC. The regression coefficient achieved was 0.98.
The model is:
Corexit 9500 dispersibility (%) = -10.7 -2.75(ln C12 content) +
0.00354(naphthalene content in ppm) + 0.113/C26 content - 7.48(PAH content
squared) + 0.0107(C12 to C18 content squared) + 0.761(% fraction oil boiling below
250oC) (19)
The Prob(t) shows that most factors are relevant and are needed to form a reliable
prediction. This prediction set is very similar to Model 1 and the predictions are
similar, but slightly more accurate.
Model 3 uses the 5 highest correlating parameters of C12, naphthalene, PAHs,
C12 to C18, the C26 fraction (negative correlation), and the viscosity of the oil rather
than the fraction that boils at less than 250oC. The regression coefficient achieved
was 0.94. The model is:
Corexit 9500 dispersibility (%) = -2.93 -1.29(ln C12 content) +
0.00368(naphthalene content in ppm) - 0.0185/C26 content - 8.65(PAH
content squared) + 0.0144(C12 to C18 content squared) + 100/viscosity (20)
The Prob(t) shows that there may be redundancy in adding viscosity. This prediction
set is very similar to Model 2 and the predictions are similar but less accurate as
viscosity is not as good a predictor as the values associated with the fraction boiling
below 250oC.
Model 4 is a simple 2-parameter predictor using only density and viscosity.
The regression coefficient is 0.71. The model is:
Corexit 9500 dispersibility (%) = -77.6 + 214e-density + 60/viscosity0.5 (21)
While this model produces a poorer prediction than most, it requires very little input
data and this data, the density and viscosity, are readily available. The overall
standard deviation is 4.6 as an average, but the maximum standard deviation is 32.
Model 5 is also a simple 2-parameter predictor using only density and the
fraction boiling below 250oC. The regression coefficient is 0.7. The model is:
Corexit 9500 dispersibility (%) = -68.8+67.4/density1.5+ 0.787BP1.5 (22)
This model produces a poorer prediction, similar to Model 4, but requires very little
input data and this data, the density and fraction that boils at less than 250oC, are
commonly available. The overall average standard deviation is 5 and the maximum
standard deviation is 30. Both the accuracy and other features of Model 5 are similar
to Model 4, although the maximum deviations with Model 5 are less. It should be
noted that as many as 295 data points were used to generate both Models 4 and 5.
Model 6 uses the SARA parameters of saturates, aromatics, resins,
asphaltenes, and the viscosity of the oil. The regression coefficient achieved was
0.68. The model is:
Corexit 9500 dispersibility (%) = -7.78 + 0.315(saturate content) +
3.44(square root of aromatic content in percent) - 4.32(ln resin content) - 1.81(ln
asphaltene content) + 58.9/viscosity (23)
The Prob(t) shows that there is little redundancy but none of the factors are highly
significant. As noted in Section 5, it was thought that the SARA analysis would yield a
good simple prediction system (Fingas, 2000), however this study shows that the
SARA fraction actually is a poor predictor of dispersibility. The reason for the poor
correlation achieved with SARA components, particularly the saturates and aromatics,
is that compounds grouped in these categories have variable dispersibility. For
example, the C12 to C18 group as described above are saturates and are highly
dispersible. On the other hand, the C20 fraction and above are not dispersible, but are
also saturates. The same situation exists for the aromatics group. Model 6 does not
show good predictability. Model 6 has the second poorest correlation coefficient of all
of the 13 models described in this study.
Model 7 uses the SARA parameters of saturates, aromatics, resins, asphaltenes,
and the sum of the C12 to C18 components. The regression coefficient achieved was
0.95. The model is:
Corexit 9500 dispersibility (%) = 296 - 1.86(saturate content) - 18.2(square
root of aromatic content in percent) - 33.6(ln resin content) - 9.03(ln asphaltene
content) + 0.0065(square of the C12 to C18 content in ppm) (24)
The Prob(t) shows that there is little redundancy in any input parameter but also that
none of the parameters are highly significant. This model is very much better in terms
of fit and accuracy than the very similar Model 6. This is because the C12 to C18
component provides the information to the model as to what is being dispersed. In
Model 6, this parameter was viscosity, which is a less powerful predictor.
Model 8 is similar and uses the SARA parameters of saturates, aromatics,
resins, asphaltenes, and the VOCs. The regression coefficient achieved was 0.71. The
model is:
Corexit 9500 dispersibility (%) = 73.4 - 0.0298(saturate content) - 2.24(square
root of aromatic content in percent) - 12.2(ln resin content) - 4.873(ln asphaltene
content) + 0.000681(VOC content in ppm) (25)
The Prob(t) shows that none of the parameters are more significant compared to the
other factors. Model 8 does not show good predictability as shown in Table 4 and
Table A2. The VOC content does not substitute for the high predictability of the C12
to C18 content as used in Model 7.
Model 9 uses only the SARA parameters of saturates, aromatics, resins, and
asphaltenes. The regression coefficient achieved was 0.68, the poorest of the 13
models described in this study. The model is:
Corexit 9500 dispersibility (%) = 62.7 - 0.103(saturate content) - 0.678(square
root of aromatic content in percent) - 13.3(ln resin content) - 4.38(ln asphaltene
content) (26)
The Prob(t) shows that there is redundancy in input parameters except somewhat for
the saturate component, which is a better predictor than the other input parameters.
Model 9 shows the SARA component does not provide good information upon which
to build a dispersibility model.
Model 10 is a larger model and uses all the composition components for which
data had been collected, including the SARA parameters of saturates, aromatics,
resins, asphaltenes, and the VOCs, the C12 to C18 component, and the C12, C14, C16,
C18, C26 naphthalene and PAH components. The regression coefficient achieved was
0.998. This is the second best model developed in the former study. The model is:
Corexit 9500 dispersibility (%) = 368 - 2.25(saturate content) - 15.4(square
root of aromatic content in percent) - 42.6(ln resin content) - 14(ln asphaltene content)
+ 0.000472(VOC content in ppm) + 0.074(C12 to C18 content squared) - 1.71(ln(C12
content) - 8.34(ln C14 content) - 17(C16 content) + 8.87(C18 content) + 0.821/C26
content + 0.00156(naphthalene content in ppm) - 1.36(PAH content squared). (27)
The Prob(t) shows that there is significance to all input parameters, especially the C12
and C14 parameters. Model 10 shows good predictability.
Model 11 is the largest model described in this study and uses many of the
composition components including the SARA parameters of saturates, aromatics,
resins, asphaltenes, and the VOCs, the C12 to C18 component, the C12, C14, C26
naphthalene, but physical components were substituted for those component
parameters that showed high redundancy in Model 10. The physical components
added were density, viscosity, and the fraction that boils at less than 250oC and less
than 200oC. The regression coefficient achieved was 0.998. This is the best model
developed in the former study, however, the fit is only marginally better than Model
10. The model is:
Corexit 9500 dispersibility (%) = 855(1/density) -250/viscosity - 7.09(saturate
content) - 72.6(square root of aromatic content in percent) - 69.7(ln resin content) -
11.6(ln asphaltene content) + 0.00045(VOC content in ppm) - 6.82(%fraction oil
boiling below 200oC) + 4.96(%fraction oil boiling below 250oC) - 0.0226(C12 to C18
content squared) + 11.4(ln(C12 content) + 2.8(ln C14 content) + 0.299(1/C26 content)
- 0.00414(naphthalene content in ppm) (28)
The Prob(t) shows that there is significance to all input parameters, especially the C12,
C14, and C26 parameters.
Model 12 is based on physical measurements. The physical components used
were density, viscosity, and the fraction that boils at less than 250oC and less than
200oC. The regression coefficient achieved was 0.71. The model is:
Corexit 9500 dispersibility (%) = - 95.6 + 90(1/density) + 22.9/viscosity -
0.443(%fraction oil boiling below 200oC) + 0.855(%fraction oil boiling below 250oC)
(29)
The Prob(t) shows that there is little significance to any input parameters compared to
other input parameters.
Model 13 is based on the same physical measurements as Model 12, but pour
point was added. The physical components used were density, pour point, viscosity,
and the fraction that boils at less than 250oC and less than 200oC. The regression
coefficient achieved was 0.69. The model is:
Corexit 9500 dispersibility (%) = - 124 + 121/density - 0.00071(pour point
squared)+ 15.3/viscosity - 0.488(%fraction oil boiling below 200oC) +
0.732(%fraction oil boiling below 250 oC) (30)
The Prob(t) again shows that there is all input parameters have low significance. The
model is poorer than Model 12 which includes the same parameters without pour
point. This shows that the addition of pour point actually decreases the accuracy of the
model. As discussed in the earlier paper, pour point is a very poor predictor and is not
a continuous variable (Fingas et al., 2003).
A new correlation was performed using only density and viscosity data. The
data set employed is given in Table 7. Correlation was performed directly using
TableCurve 3D software. One iteration was carried out to remove or correct bad data.
The model achieved is:
Dispersibility = 169 - 171*D + 62.4/V (31)

Where Dispersability is the Corexit 9500 dispersibility in percent

D is the density, and
V is the viscosity.

Table 7 gives the predicted values for the input data set and the standard
deviation. The average standard deviation between the predicted set and the input
values. The average standard deviation was 3.2 and the r2 for the regression model was
0.82. It should be noted that this model uses the same inputs as model 4 (equation
(21)) above, however a different equation resulted. Both equations yield roughly the
same variance in prediction capability. The model in equation (31) does not produce as
accurate a prediction as many of the models noted above, however is much better than
no model. In this case, the value of dispersibility predicted is with a 30% error of the
actual value. This compares to model 11, noted above which produces predications
within 5%, but requires detailed compositional data input. The large amount of scatter
in the relationship of density and chemical dispersibility is illustrated in Figure 6.

5 Conclusions
The use of only density and viscosity to predict evaporation, emulsification and
chemical dispersion was implemented and models developed. These were compared to
other model methods and to the original input data. While the models were generally
successful, the error in using these is higher than other approaches which require more
input data, especially for emulsification.
The model developed for evaporation where the oils evaporate as a logarithm
of time (most oils) is:
Evap = 15.4 -14.5*density + 2.58/viscosity (8)

Where: Evap is the evaporation equation parameter and as in equation (5) is B

- 0.675 to compensate for the temperature parameter.
The equation for predicting those oils that evaporate as a square root of time
(diesel fuels and crude oils having a high diesel component) is:
Evap = 2.3 -1.47*density - 0.073*ln(viscosity) (9)
In this case the evaporation coefficient is correct by subtracting 0.195 to
compensate for temperature.

The model for predicting emulsion formation first predicts the type or class of
the emulsion. The class of the resulting emulsion is calculated as follows:
 Class = -21.4 + 28*D -0.00000073*V (17)

Where Class is the numerical index of classification,

D is the density, and
V is the viscosity.
The class is then assigned according to the following scheme (Table 6):

The model developed to predict chemical dispersibility is:

Dispersibility = 169 - 171*D + 62.4/V (31)

Where Dispersability is the Corexit 9500 dispersibility in percent,

D is the density, and
V is the viscosity.

As noted in the development sections, these models will yield satisfactory

results until further detailed oil composition data are available. The addition of
detailed oil composition data improves the accuracy of the predictions.

6 References
Brutsaert, W. Evaporation into the Atmosphere, Reidel Publishing Company,
Dordrecht, Holland, 1982.

Fingas, M.F., “A Literature Review of the Physics and Predictive Modelling of Oil
Spill Evaporation”, Journal of Hazardous Materials, Vol. 42, pp.157-175, 1995.

Fingas, M.F., “The Evaporation of Oil Spills: Prediction of Equations Using

Distillation Data”, Spill Science and Technology Bulletin, Vol. 3, No. 4, pp.191-192,
1996.

Fingas, M.F., “Studies on the Evaporation of Crude Oil and Petroleum Products: I.
The Relationship between Evaporation Rate and Time”, Journal of Hazardous
Materials, Vol. 56, pp.227-236, 1997.

Fingas, M.F., “Studies on the Evaporation of Crude Oil and Petroleum Products: II.
Boundary Layer Regulation”, Journal of Hazardous Materials, Vol. 57, pp.41-58,
1998.

Fingas, M.F., “The Evaporation of Oil Spills: Development and Implementation of

New Prediction Methodology”, in Proceedings of The 1999 International Oil Spill
Conference, American Petroleum Institute, Washington, D.C., pp. 281-287, 1999.
Fingas, M.F., Z. Wang, B. Fieldhouse, and P. Smith, “The Correlation of Chemical
Characteristics of An Oil to Dispersant Effectiveness”, in Proceedings of the Twenty-
Sixth Arctic and Marine Oil Spill Program Technical Seminar, Environment Canada,
Ottawa, Ontario, pp. 679-730, 2003.

Fingas, M., “Modeling Evaporation Using Models That Are Not Boundary-layer
Regulated”, Journal of Hazardous Materials, Vol. 107, pp. 27-36, 2004.

Fingas, M.F. and B. Fieldhouse, “A Review of Knowledge on Water-in-oil

Emulsions”, in Proceedings of the Twenty-ninth Arctic and Marine Oil Spill Program
Technical Seminar, Environment Canada, Ottawa, Ontario, pp.1-56, 2006.

Jokuty, P., M. Fingas, S. Whiticar, and B. Fieldhouse, A Study of Viscosity and

Interfacial Tension of Oils and Emulsions, Environment Canada Manuscript Report
Number EE-153, Ottawa, Ontario, 43 p., 1995.

Jokuty, P.S., S. Whiticar, Z. Wang, M.F. Fingas, B. Fieldhouse, P. Lambert and J.

Mullin, Properties of Crude Oils and Oil Products, Environment Canada Manuscript
Report EE-165, Ottawa, Ontario,1999.

Jones, F.E., Evaporation of Water, Lewis Publishers, Chelsea, Michigan, 1992.

Mackay, D. and R.S. Matsugu, "Evaporation Rates of Liquid Hydrocarbon Spills on

Land and Water", Canadian Journal Chemical Engineering, Vol. 51, Canadian
Society For Chemical Engineering, Ottawa, Ontario, pp 434-439, 1973.

Monteith, J.L. and M.H.Unsworth, Principles of Environmental Physics, Hodder and

Stoughton, London, 1990.

Oil Catalogue, http://www.etc-cte.ec.gc.ca/databases/spills/oil_prop_e.html, 2007.

Stiver, W. and D. Mackay, "Evaporation Rate of Spills of Hydrocarbons and

Petroleum Mixtures", Environmental Science and Technology, Vol. 18, pp 834-840,
1984.

Sutton, O.G., “Wind structure and evaporation in a turbulent atmosphere”,

Proceedings of the Royal Society of London, A 146, pp. 701-722, 1934.
Figure 1 Graph of Oil Density versus Evaporation Parameter

Figure 2 Graph of Oil Viscosity versus Evaporation Parameter

Figure 3 Oil Saturates versus Evaporation Parameter

Figure 4 Comparison of the Evaporation of a Light Crude (ASMB) Using

Different Prediction Methods
Figure 5 Comparison of the Evaporation of Diesel Fuel Using Different Prediction
Methods

Figure 6 Corexit 9500 Dispersibility as a Function of Oil Density

View publication stats