How was the Variability Analysis Conducted?
EPA performed the variability analysis for the national-scale assessment in
the following 9 steps for each air toxic compound considered (steps
are numbered so the reader can refer back to previous steps):
-
Division of the geographic region. To account for variability
in the concentration of an air toxics compound throughout the
U.S., the country was divided into census tracts.
- Calculating variation in ambient air concentration. Sources
and their emission rates were determined, and the ambient air
concentrations were calculated throughout the U.S. using the ASPEN
dispersion model for that air toxic compound. These results then
were used to estimate the ambient air concentration at the centroid
of each census tract defined in the first step. EPA pooled the
results from all census tracts inside a geographic region (such
as county, state or nation) and the concentrations associated
with different percentiles were determined. For example, the combined
census tract results for a geographic region were used to estimate
the ambient air concentration below which 25% of the tracts in
that region fall; 50%; 75%; etc. This information depicts the
variation in ambient air toxics concentration between the different
census tracts in that geographic region (one estimate for each
census tract in that region and each air toxics compound).
-
Incorporating daily activity patterns. For each census
tract, the variation in daily activity patterns (the fraction
of time spent in each microenvironment and the fraction of time
spent commuting during a day) between individuals was determined
for each of the 10 receptor populations. The receptor populations
were defined by 5 age groups (a 0-5 year age group; a 6-11 year
age group; a 12-17 year age group; an 18-64 year age group;
and a 65+ age group) and 2 gender groups (male and female gender
groups).
-
Calculating annual exposures. For each census tract
and receptor population, the yearly activity pattern for 30
different random individuals in that receptor population was
simulated using Monte Carlo analysis. For the first individual,
365 daily activity patterns (one for each day of the year) were
chosen based on the variation in daily activity pattern for
that specific receptor population. This was repeated for 30
individuals in that receptor population, always using the same
ambient air toxic concentration specific to that census tract.
This produced 30 different estimates of the exposure over one
year of life for each of the 10 receptor populations in that
census tract (a total of 300 estimates in the census tract).
From these 30 samples, a median value was calculated and applied
to all individuals in that receptor population in that census
tract. This median value represents a typical
exposure for that receptor population and census tract. One
such median value was calculated for each receptor population
in each census tract.
-
Calculating lifetime exposure. For a given census tract,
the lifetime exposure was estimated using the results from Step
4. This was calculated by multiplying the exposure in an age
group (e.g. the 0-5 year age group) as determined by Step 4
by the fraction of a lifetime spent in that age group (e.g.
5/70 or 0.07 for the 0-5 age group); performing this calculation
for each of the 5 age groups; and summing these contributions
(one contribution from each age group). This produced a time-weighted
average lifetime exposure in each census tract for a typical
individual.
-
Calculating cancer risk. The variation in cancer risk
was obtained directly from the results of Step 5. For a given
census tract, the lifetime exposure was multiplied by the Unit
Risk Estimate to obtain the excess lifetime probability of cancer
for that air toxics compound. This process was possible because
the URE is not a function of age, gender or race (or at least
the function is unknown, and so is assumed to be a constant
for all individuals). The result was an estimate of the typical
lifetime excess probability of cancer in that census tract.
Note that it is not strictly the mean probability for the entire
population in that census tract, since it is based solely on
calculations of exposure at the centroid of the census tract.
This was repeated over all census tracts. From these data, the
fraction of census tracts characterized by a typical lifetime
excess probability of cancer below any particular value (e.g.
1 in 10,000) was determined. For example, the cancer risk below
which 25% of the census tracts fall was determined (as well
as other percentiles).
-
Calculating noncancer risk. The variation in noncancer
risk is obtained directly from the results of Step 5. For a
given census tract, the lifetime exposure was divided by the
Reference Concentration (RfC) to yield the Hazard Quotient (HQ)
for that air toxics compound. This process was possible because
the RfC is not a function of gender or race (or at least the
function is unknown, and so is assumed to be a constant for
all individuals). The results are two estimates of the HQ for
individuals in that census tract; one for typical young children
and one for typical adults. Note that these values are not strictly
the mean HQ for the entire population in that census tract,
since they are based solely on calculations of exposure at the
centroid of the census tract. This was repeated over all census
tracts. From these data, the fraction of census tracts characterized
by a typical HQ below any particular value (e.g. 0.01) was determined.
For example, the HQ value below which 25% of the census tracts
fall was determined (as well as other percentiles).
-
Calculating cumulative cancer risk. A cumulative carcinogenicity
risk assessment was performed by summing the lifetime excess
probability of cancer across air toxic compounds in a census
tract. From the resulting information (produced by performing
this calculation in each census tract and then pooling the results
from all census tracts), the fraction of census tracts characterized
by a typical cumulative lifetime excess probability of cancer
below any particular value (e.g. 1 in 10,000) was determined.
Separate analyses were performed for carcinogens for which human
data exist; carcinogens for which only animal data exist; and
all carcinogens combined. For example, the cumulative cancer
risk below which 25% of the census tracts fall was determined
(as well as other percentiles).
-
Calculating cumulative noncancer risk. Similarly, a
cumulative noncancer risk assessment was performed by summing
the HQs across all air toxic compounds in a census tract that
act on the same target organ or tissue to yield a Hazard Index
(HI). From the information (produced by performing this calculation
in each census tract and then pooling the results from all census
tracts), the fraction of census tracts characterized by a typical
HI below any particular value (e.g. 0.01) was determined. For
example, the HI value below which 25% of the census tracts fall
was determined (as well as other percentiles).
What are the components of variability?
Which components of variability did
the national-scale assessment include?
What are the results of the variability
analysis
How can these results be interpreted?
Return to Main Variability Page
|