Page 211 - Materials Chemistry, Second Edition

L1644_C05.fm Page 184 Monday, October 20, 2003 12:02 PM

distribution — will be shown as an example of uncertainty assessment. Applying
the uncertainty analysis to the risk assessment of the population living in its vicinity
will be demonstrated. Only the “direct risk” due to air emissions will be considered.
To account for variability and uncertainty, the Monte Carlo simulation will be applied
to estimate the set of risk estimates. In this example model and data uncertainty will
be considered. The sensitivity analysis will show how much each predictor variable
contributed to the uncertainty or variability of the predictions.

5.7.1 DETERMINATION OF DISTRIBUTION FUNCTIONS
The determination of which form of distribution function to assign to each parameter
depends on site-specific data and judgment based on statistical analysis. The distribution
employed in this example is assembled from site-specific data, data existing in the most
current literature, and professional judgment; they are considered to be the most up-to-
date description of the parameter (Katsumata, 1997). In the vegetation production in the
area of study (Tarragona), only the adult population was considered. In this example,
only exposition through the air, soil and 10% of the consumed vegetation will be
considered as a direct vies of exposition. Because the MSWI is located close to a city,
no impact to foods like meat, fish or dairy products will be considered.
Table 5.2 and Table 5.3 show a description of the Monte Carlo parameter
distribution for risk assessment evaluation due to direct exposure for people living
in the area surrounding the MSWI of Tarragona, Spain. After characterizing the
uncertainty and/or variability associated with each parameter, the uncertainty in the
risk can be estimated. For the risk assessment presented here, the commercially
available software package Crystal Ball (Version 4.0) was used. For analyzing the
results, the mean and median values and the percentiles 50 and 90% were extracted
and presented. In this example, ranges of exposure rather than single point estimates
are developed in order to account for the natural variability among members of a
population and for uncertainties in the input variables. Even if it were possible to
eliminate the uncertainty associated with the input variables, a probability density
function would still be required because of natural variability.
Figure 5.7 shows the distribution of the different variables of direct exposure
due to the incinerator emissions for the population living in the area surrounding
the plant. The distribution of total direct exposure is also depicted. Figure 5.8 shows
the sensitivity analysis for total direct exposure from the different exposure pathways
due to MSWI emissions. This figure shows that inhalation of air from the area
contributes to 50.4% of the variance and vegetation ingestion to 49.5%. The other
pathways — exposure, dermal absorption, soil ingestion, and inhalation of resus-
pended particles — are irrelevant.
Table 5.4 summarizes the exposure to PCDD/Fs by the population living in the
proximity of the MSWI of Tarragona, Spain, and Table 5.5 shows the PCDD/Fs
dose (ng I-TEQ/day/kg) for the population living around the incinerator. The toler-
able average intake levels of PCDD/Fs recently established by the WHO is between
1 and 4 pg I-TEQ/kg/day for lifetime exposure (Rolaf and Younes, 1998). Therefore,
–2
the current total exposures of 2.87 ¥ 10 pg I-TEQ/day/kg for the 50th percentile
–2
and 5.37 ¥ 10 for the 90th percentile (Table 5.5) are within this tolerable intake.
© 2004 CRC Press LLC