AN ASSESSMENT OF LUNG CANCER
FROM THE 1984 TIRE FIRE IN
EVERETT, WASHINGTON
On September 24, 1984 a fire broke out in a scrap tire dump east of Everett which
contained more than a million tires. The tires burned for more than two months. The
Puget Sound Air Pollution Control Agency requested that EPA perform a risk analysis
using data gleaned from selected ambient particulate matter samples from stations in
Everett and North Seattle operated by that agency. These samples were analyzed for
selected products of incomplete combustion (including benzo(a)pyrene). These data
coupled with emission parameter estimates were then used to estimate risk to the exposed
population. This document presents the results of the risk analysis.
October, 1986
U.S. Environmental Protection Agency
Region 10
1200 Sixth Avenue
Seattle, Washington 98101
-------
U.S. ENVIRONMENTAL PROTECTION AGENCY
REGION' 10
:200 £.:/T>- ^VLVJE'
SEATTLE. '.VASHINGTON 5cn.'< '
p«o^ GCT 8 1906
RF"IY
TI N OF
.,.<.
"/»>
Mr. Arthur Dammkoehler
Air Pollution Control Officer
Puget Sound Air Pollution Control Agency
P.O. Box 9863
Seattle, WA 98109
Dear Mr. Dammkoehler:
As you requested in your letter of November 20, 1985, we have evaluated
the long term health risk from the Everett tire fire. We have confined our
health assessment to estimating the increased cancer risk from inhalation
exposure to airborne particulates emitted during the fire. The specific
contaminants evaluated were benzo(a)pyrene (B(a)P) and a class of compounds
called Products of Incomplete Combustion (PIC). Attached is a report
summarizing the results of that assessment.
Using EPA risk assessment methods, the analysis indicated that those
individuals (4 of them) calculated to have the highest exposure to the tire
fire emissions may have an increased lung cancer risk from 2 in one million to
2 in ten thousand. Since exposure for other residents living in the vicinity
of the fire was less, their estimated lung cancer risks are also less
(substantially less for most).
Because of the methodologies used these estimates represent a likely upper
bound of lung cancer risk - the actual risk is somewhere between zero and
these numbers. It should also be stressed that there are many uncertainties
and assumptions involved in deriving these lung cancer estimates. These are
summarized in the report.
The U.S. EPA has not defined a cancer risk level which is considered to be
significant. However, excess cancer risk levels above 1 in a million to 1 in
100,000 (10~6 to TO'5) generally give some cause for concern and suggest
that exposures should be reduced. To put these numbers in perspective,
however, a lung cancer risk of 10~6 to 10~5 is equivalent to smoking about
5 to 50 cigarettes in a lifetime.
We appreciate the support your staff provided in performing this
assessment. Should you have any questions about the results or methodologies,
please contact Dana Davoli at 442-1757.
Sincerely,
Gary O'Neal, Director
Air and Toxics Division
Attachment
cc: Dave Peterson, Snohomish Health District
-------
On September 2£, 198^, a tire ^ire began burrvng e: trie old City of
Everett Landfill in Everett, Kashi ngtors. During t-ie first few weeks of the
fire large quantities of smoke were released. This fire continued to
smolder for about two months emitting smoke at gradually decreasing levels.
The City of Everett, which is west of the fire, was at times heavily
impacted by the plume. Smoke was also dispersed to the east and to the
south toward Seattle.
Limited air monitoring for organic vapors was done around the tire fire
by the U.S. Environmental Protection Agency's (U.S. EPA) Technical
Assistance Team on September 28, 1984. ("Tire Fire Investigation, TAT
Activities Report", U.S. EPA, Region 10, October, 1984). The results from
this one day of monitoring showed that high levels of some compounds (e.g.,
benzene) were present in or very near the fire but dropped off rapidly in
concentration within a half mile. At the request of the Snohomish Health
District, the Centers for Disease Control (CDC) reviewed these data. CDC
concluded that "concentrations of some chemicals in the immediate vicinity
of the fire were high enough to pose a potential immediate health threat to
individuals within 200 feet of the fire." Personnel at the scene of the
fire (e.g., firemen) were of most concern. Persons living beyond the
immediate area of the fire may have received transient exposure, according
to CDC, resulting in a "temporary increased risk of acute short term health
effects" (e.g., respiratory impairment; eye, throat and respiratory
irritation). CDC also stated that "we do not anticipate any significant
increased risk of long term health effects nor can we conclude that
chemicals in the smoke reached the public in sufficiently high concentration
to significantly affect health."
The Puget Sound Air Pollution Control Agency (PSAPCA) has monitors
located in Seattle and Everett to measure levels of particulates in the
air. A sample collected from the Everett monitor on September 28, 1984
(about 1.2 miles from the fire) was analyzed for selected particulate
organics by a lab at the University of Washington in January,1985. The
results showed levels of polycyclic aromatic hydrocarbons (PAHs) that were
above background. Several members of this class of compounds are known or
suspected of causing cancer. At the request of PSAPCA, the U.S. EPA's
Region 10 laboratory analyzed 24 additional samples collected by PSAPCA at
the Everett station and at two stations in Seattle (approximately 20 and 25
miles south of the tire fire) during the first few months after the tire
fire began. These results also showed elevated levels of PAHs in Everett
and at" both Seattle locations during the fire which lasted about 2 months.
Because of these results PSAPCA requested that.EPA quantitatively "assess
the long-term health risk from the tire fire emissions."
In response to PSAPCA's request, EPA has estimated the lung cancer risk
that may result from.emissions of benzo(a)pyrene
-------
(1) Although increased levels of participates were detected on
PSAPCA's samplers more than a mile from the tire fire, the limited
sampling data collected by EPA for gaseous organics suggest that levels
of these gaseous compounds dropped off quickly within a short distance
of the fire. CDC concluded from these data that these gaseous organics
did not reach the public in sufficiently high concentrations to cause a
long term health risk.
(2) Although other particulate organics were detected on PSAPCA's
samples, toxicity and potency (unit risk numbers) data are available
only for B(a)P and the generic class of incomplete combustion products,
PICs. (see below).
(3) Data from' the literature are available to make rough estimates of
the amounts of B(a)P emitted from the fire. These data were used in a
computer dispersion model to estimate ambient air exposure for
residents in the Everett-Seattle area. These modeled ambient air
results could also be compared to those levels measured by PSAPCA.
The scientific data now available make it extremely difficult, 'if not
impossible, to identify a level of exposure to cancer-causing agents that is
safe. Therefore, EPA and other federal agencies have taken the position
that cancer may occur at any level of exposure no matter how low. EPA has
also assumed that the risk of cancer increases as.exposure increases and
that this relationship is linear (e.g., when exposure doubles so does
risk). Thus, although a "safe" exposure can't be defined, estimates can be
made of the risk of getting cancer if exposure to a cancer-^causing substance
is known. To estimate the risk from the tire fire, EPA has combined two
different types of data: data on the B(a)P exposure for the populations
living within about 30 miles of the fire and data on the cancer potency of
B(a)P and PIC.
As a first step in calculating exposure levels, emission levels of
B(a)P from the fire were estimated from data on the number of tires consumed
during the fire and from literature data on the amounts of B(-a)P released
per pound of burning tire. A mathematical (dispersion) model used these
emission data as well as data on weather and geographic conditions to
estimate the concentrations of B(a)P at about 250. points around the fire
Within a 30 mile radius. This information was then combined with Bureau of
Census population figures to provide an estimate of the number of people
exposed to a given level-of B(a)P (see the Attachment for a more detailed"
explanation of this methodology).
The other type of data needed to estimate the public lung cancer risk
from the fire is that on the potency of B(a)P-this potency is expressed as a
unit risk number. The unit risk number is defined as the lifetime cancer
risk that would occur in a population which is exposed throughout their
lifetime (70 years) to one microgram per cubic meter of B(a)P in the air
they breathe. The unit risk number for B(a)P, which was derived using
experimental data on animals, is about 3 X 10~3 per ug/tn3 (micrograms of
B(a)P per cubic meter of air).
-------
-3- . •
The unit risk number for B(a)P and the estimated B(a)P exposure for
people living around the fire site were multiplied to give tne estimated
lung cancer risk if exposure to the fire had occurred for 70 years (over an
entire lifetime). This risk number was then divided by 420, "the number of
two month periods in 70 years, to adjust it for the fact that exposure
occurred only for the tire fire duration (i.e., about 2 months). For the
persons with the highest exposure to B(a)P emitted by the tire fire, their
increased lifetime cancer risk is approximately 2 X 10~5. That is, their
estimated risk of getting cancer as a result of the fire is 2 in 1 million
(see Table 3 in the Attachment). This risk decreases significantly as
distance from the fire increases.
This level of risk could be compared to the average expectation of
dying of all types of cancer which is about 1 in 5 and the lifetime risk of
dying from lung cancer for cigarette smokers (pack a day) which is about one
In 10. Another way of stating this is that a risk of 2 X 10~6 Is
equivalent to smoking about 10 cigarettes over a lifetime.
Another way of estimating risk from exposure to B(a)P utilizes a
different unit risk number, that for Products of Incomplete Combustion or
PIC. During combustion of organic material many compounds in addition to
C02 and water can be released because the combustion is not 100X efficient
and because of impurities in the materials being burned. These may include
the original organic material or other more or less complex compounds formed
during combustion. As previously mentioned, this complex mixture of
compounds is loosely defined as PIC and includes PAH (polynuclear aromatic
hydrocarbons) and possibly other organic compounds.
A unit risk number for PIC has been derived using B(a)P as a
surrogate. As an example, workers exposed to products of incomplete
combustion (e.g., roofers, gas workers) have higher lung cancer rates than
non-exposed workers. Although PICs are a complex mixture of compounds, most
of the worker exposure data are expressed in B(a)P concentrations since
B(a)P is a suspected carcinogen and is fairly easy to measure. Therefore,
in these studies, B(a)P serves as a surrogate or indicator of the PICs, and
cancer risk is expressed in excess cancers per unit measure (e.g., ug/m^)
of B(a)P. Results from many studies such as these (occupational and
non-occupational) were combined to estimate the PIC unit risk number. The
B(a)P exposure levels estimated to be produced as a result of the tire fire
can be used with the PIC unit risk number (4 X 10-1 per ug/m3) to
estimate lung cancer risk resulting from exposure to PICs emitted by the
tire fire. This results in an estimated lifetime cancer risk for thos-e -
persons with the highest exposure of 2 X 10~4 or about 2 in 10,000. This
is significantly higher than the estimate using the B(a)P unit risk number,
2 in one million.
It should be kept in mind that the risk estimates given here for both
B(a)P and PICs are for those few people with the highest exposure to
emissions from the tire fire. According to the model used by EPA, only 4
persons are living in this area of highest exposure (0.26 ug/m3 of
B(a)P). Exposures for other people living around the fire were less as can
be' seen in Figures 1 and 2 (i.sopleth maps) of the Attachment. In the area
-------
west of the fire and within about 0.6 miles, trie model we used predicts a
population of roughly 1200 residents. Their estimated exposure ranges from
about 0.023 ug/m3 B(a)P to 0.26 ug/m3 B(a)P. People living in
Snohomish, Monroe, and Lynnwood were exposed to B(a)P levels between 0.0002
ug/m3 and 0.002 ug/m3. Therefore their risk, is substantially less than
that of the highest exposure group.
It must also be stressed that there are many assumptions and
uncertainties involved in this type of risk estimate. For example, for
carcinogens, EPA assumes that a linear relationship exists between exposure
and cancer risk (e.g., a person who inhales one microgram of B(a)P per cubic
meter of air is one-tenth as likely to get cancer as a person who inhales 10
micrograms per cubic meter). A mathematical model based upon this
assumption is used to estimate the unit risk number; the model relies upon
laboratory data in animals (B(a)P) or studies of workers or community
exposures (PIC). Because this model is conservative, the risk numbers
generated represent upper bounds of risk rather than an actual expected
risk in terms of excess cancers is
value calculated here (2 in one million
Other assumptions and uncertainties are
level of risk. The actual level of
somewhere between zero and the risk
for B(a)P and 2 in 10,000 for PIC).
discussed below:
(1) Model ing - In Attachment A, Table 2, a comparison is made between the
ambient levels of B(a)P predicted from the dispersion model and those
measured (observed) by PSAPCA. The discrepancies between these two
numbers are likely a result of several factors in.cluding:
0 Emission estimates of B(a)P from the fire were based upon emission
rates obtained from the literature not measured data from the. fire
0 The dispersion model that estimates the ambient levels of B(a)P at
various points is limited in dealing with complex geographic and
meteorological conditions as well as non-constant emissions of
pollutants as was the case with the tire fire
0 The number of ambient air samples analyzed by EPA were too few in
number to consider them a very good representation of an average
concentration over the two-month period of the fire. Additionally,
these samples were analyzed more than 10 months after collection;
volatilization and decomposition of substances on the filter may have
occurred.
(2) Exposure
8 Much of the information available on carcinogens, including
development of potency numbers such as the unit risk numbers, are from
laboratory or occupational studies where exposure occurred over a long
time period. The use of such numbers for a two month exposure to B(a)P
as occurred with the tire fire may not be appropriate.
0 Exposure to B(a)P and PICs from inhalation of contaminated dust or
from ingestion of contaminated soils and dusts by children has not been
considered.
-------
c Tnere is a ever, more unce-tai nty with trie °IC unit risk nurcaer tnan
w;.tn orner unit risk numbers, in part oecause of tne way it was oe'ivec
anci because 6(a)P is used as a surrogate. For example, the PIC ur.it
risk numoer is derived from studies of workers and communities. Tne
rynes of chemicals present in tnese situation: may be very different
frotr. tnose emitted from the tire fire.
Cone It's ion
This assessment of the long term health effects from the tire fire
emissions was limited to estimating lifetime lung cancer risk for the
reasons already discussed. Using conservative assumptions, this increased
lifetime cancer risk may approach 2 in one million for B(a)P and 2 in 10,000
for PIC for a limited number of people (about 4) with the highest exposure
to emissions from the tire fire. It should be kept in mind that there are
many assumptions and uncertainties involved in this type of risk assessment
(e.g., assuming that a linear relationship between exposure and cancer risk
exists, estimating B(a)P emissions using literature values, using "lifetime"
unit risk numbers to estimate risk from a two month exposure to the tire
fire emissions). It should also be stressed that these risk estimates
represent upper bounds of lung cancer risk rather than an actual expected
level of risk; that is, the true risk is expected to be somewhere between
zero and the risk values calculated.
-------
ModeJinc Analysis to Assess T-j sk
fror, Everett Tire Fire LT.I scions
The purpose of this attachment is to briefly document a modeling
analysis of the air emissions from the Everett tire fire and the development
of risk estimates. The modeling approach employed the Industrial Source
Complex (ISC) Model to estimate concentrations, which were then input to
the Human Exposure Model (HEM) to estimate risk.
Emissions estimates for benzo-a-pyrene [B(a)P] employed in the risk
assessment analysis were developed based on the following information:
- 807,000 tires burned in. 60 days
- 20 pounds per tire
- 18 grams of B(a)P emitted per ton of tires burned *
This yielded an average B(a)P emission rate during the fire of 0.02802 grams
per second. The fire was simulated in the ISC model as a volume source
with horizontal dimensions of 100 meters by 100 meters, and a vertical
height of 20 meters. This accounts for initial dilution of the emissions
caused by the spreading out of the fire as it progressed. It also accounts
for the minimal rise of the plume during most of the two-ronth period. The
source was located at a latitude of 47° 57' 56" north, and a longitude of
122° 11' 30" west.
Meteorological data was supplied by the Puget Sound Air Pollution
Control Agency (PSAPCA). The data consisted of joint frequency distributions
of wind speed, wind direction, and stability class for the two-month period
from September 24 to November 23, 1984. Two distributions were developed
using the wind data from PSAPCA's monitoring stations at the Medical-Dental
Building on Colby Avenue in Everett, and at North 98th Street and Stone Way
in North Seattle. Stability classes were developed from concurrent cloud
cover and ceiling height observations at Seattle-Taccrre Airport. After
preliminary model runs were ccrpleted, it became evident that the Everett
wind data yielded higher modeled concentrations than the North Seattle
d£±a-^_Purthermore,_owing to the close proximity of the Everett station to
"fcheifire location, the Everett wind data are judged to be more representative
of the conditions which affected the fire than the North Seattle data. .
Thus, all subsequent modeling analyses utilized the Everett meteorological
data set.
Other meteorological data input to the ISC model are documented in
Table 1. The values are listed as.a function of stability class. The
temperatures and mixing heights are based on historical ciimatological data
for the Seattle area. The vertical potential temperature gradients and the
vertical wind profile exponents are normal default'values.
"Atmospheric Emissions from Open Burning;" Pdcharcl Gerstle and Eouglasl::-.
Kermitz, Journal of the Air Pollution Control Association, 'Wolume 7, •-- -
Number 5, May 1967, page 324.
-------
Concentrations were estirsterl using the rural none of trie Jone-term
version of the ISC model (EPA-450/4-79-030). ISC is e standard gaussian
model intended for use in flat or gently rolling terrain. Deviations in
plume trajectories resulting from v.inc flows which are altered by ccnplex
terrain are not simulated by ISC. Elevated stable plume irrpact on high
terrain is also not modeled by ISC. liowever, because cf the minimal plume
rise during most of the fire, stable plume impact should not be a significant
factor in this analysis. ISC does not, as all gaussian models do not,
simulate very low wind speed conditions well. Light and variable winds
were reported about 4% of the time during the fire. In spite of these
limitations, the ISC model is judged to be adequate for this analysis.
Receptors in the modeling analysis were set in a polar grid with the
fire at the center of the grid. Sixteen grid radials were spaced at azimuths
of 22.5° around a 360° circle. Concentric rings of receptors were located
at ranges from the fire in kilometers of 0.5, 1.0, 1.5, 2, 3, 4, 5, 7, 9,
12, 15, 20, 25, 30, 40, and 50. This grid layout .yields 256 receptors.
ISC was used to calculate average concentrations of B(a)P at all receptor
locations. The results are displayed in Figures 1 and 2. Figure 1 contains
the predicted spatial distribution of B(a)P within a few kilometers of the
fire location, while the predicted concentrations over the entire receptor
grid are shown in Figure 2. These figures show how concentration decreases
significantly with increasing distance from the fire. For example, the
highest concentration estimate (0.26 micrograms per cubic meter) was one-
half kilometer to the northwest of the fire, near the 1-5 freeway. Further
into Everett near the intersection of Hewitt and Broadway, the average
concentration was only about one-fifth of the maximum. And, near Oak Harbor
on Whidbey Island, the average concentration estimate dropped by about a factor
of 30C from the maximum. The concentration drops even more rapidly in other
directions from the fire.
Measured ambient concentration data collected by PSAPCA at their
Everett and. North Seattle monitoring stations were obtained. The measured
values were averaged over the duration of the fire for comparison with the_
model estimates. The model estimated concentrations were interpolated from—
the grid receptors to the monitoring locations so that a rough comparison- -. .-.-. •_
of predicted and observed concentrations could be made. The results of ' •""-" •
this comparison are listed in Table 2. •..--•
There are several factors which suggest that this comparison is crude
at best, and may in part explain some of the discrepancies between the
predicted and observed concentrations. While the emissions estimates are
based on the best information available, they must be considered very
approximate. The plume from the Everett tire fire was not sampled, so that
emission factors had to be obtained frcrr-. the literature. ISC assumes" that
the fire emissions are constant, while in reality the fire emissions were
obviously not constant "with' time. No background concentration was "subtracted '.'.
from the measurements, so that the measured values include contributions
from sources other than the tire fire. •• No. measurements' of sriierit: cdncentrsticns..
were available for November. The numbers of 24-hour averaged samples"for'""; "
B(a)F at the Everett and North Seattle monitors are really too small to
-------
consider them a very good representation of an average concentration over
the two-month period of the fire. Finally, the limited ability of the ISC
model to handle complex terrain and low wind speeds is also a factor in the
comparison.
It is not possible to draw any firm conclusions from the comparison.
Statistical significance of the differences (or similarities) between the
measurements and the model estimates can not be established. From this
limited evaluation it appears that the modeled concentration estimates are
probably good enough for order-of-magnitude risk estimates.
The output concentration estimates from the ISC model were input to
the HEM developed by Systems Applications, Inc. under contract to EPA. A
draft user's manual for HEM (October 1985) was obtained from the Pollutant
Assessment Branch at the EPA Office of Air Cuality Planning and Standards.
HEM uses the ISC concentration estimates and the population data from the
1980 census to estimate exposures. Risk estimates were developed with HEM
using two different unit risk factors: the B(a)P unit risk factor of
0.0033 per microgram per cubic meter, and the unit risk factor for products
of incomplete combustion (PIC) of 0.42 per microgram per cubic meter. The
HEM assumes a lifetime (70-year) exposure. Therefore, the risk estimates
were adjusted for the shorter two-month exposure resulting from the fire.
This amounted to dividing the lifetime values by 420, the number of two—month
periods in 70 years.
The results from the HEM are listed in Table 3. The maximum individual
risk estimates due to emissions from the fire are for people who live very
close to the fire location. These analyses extended to a radial distance
of 50 kilometers frcm the fire. The total number of people within this
area is approximately 1.23 million. The minimum risk is the lowest risk
value to which the entire population was exposed.
-------
Meteorciocicoi Leta Input to the IS:: r-v
Stability
Class
A
D
C
D
E
p
Temp.
(dec II)
286
285
284
282
280
279
Mixing
Height
(meters)
950
850
750
750
10000
10000
Foter.tiaJ
Temp. C-rec: -
(dec, K / m)
0.0
C.O
0.0
0.0
0.020
0.025
Wind Profile
Exponent
0.10
0.15
0.20
0.25
0.30
0.30
Table 2
Comoarisbn of Estimated and Measured B(a)P Concentrations
Monitoring Station
Everett
Forth Seattle
ISC Estimate
(rucrograms
0.0484
0.000276
Average Measured
Concentration
per cubic meter)
0.0142
0.0032
Number of
24-Hour
Samples
8
2
Table 3
Risk Estimates for B(a)P Emissions from the Everett Tire Fire
Calculated bv the Human Exposure I-'ocel
Unit Risk Factor Employed
C.0033
[for B(a)P]
0.42
[for PIC]
Maximum Risk to an Individual
as a "Result of Exposure
2.1 x ID"6
2.4 x 10-4
Number of People at
M? -v •» Tirtrn ^\T^ S"^"
Minimum Risk Level
3.5 x 10-10
4.4 x 10-8
-------
Figure 1- - 7his..jr.ap shows the spatial distribution of predicted concentrations -•
vithin a few kilometers of the fire location at~ ®~ The" concentrations* '". T"7"T '. ..~°
are in units of micrograms of B(a)P per cubic meter of air averaged over
the two—month period of the fire.
-------
Figure 2. This map shows the spatial distribution of predicted concentrations
-over^the -entire modeling grid,' within 50 -kllonie-l'-ers— of the fire. -location. -atir ®
^ n
The^conceTitraTions are in units of" mi cr-ogf ani s o 'f -B ( a )?"" ^p'e f~ cubl^. m'e't erJ ^f^SL"."
air averaged over the two— month period of the fire.
------- |