A Geostatistically Based Probabilistic Risk Assessment Approach - Evolving Application Domains of Data Warehousing and Mining - page 279

ProBABIlIStIc rISk
ASSeSSMent: clASSIcAl
APProAch

When the obtained risk is close to the accept-
ability limit, the estimation of a defined value of
the risk may not be exhaustive because it doesn't
take into account the uncertainty and variability
that have characterized the various phases of the
analysis (U.S. EPA, 2001).
Within a procedure of risk assessment there are
many sources of uncertainty (U.S. EPA, 2001):

The main probabilistic approach to risk assessment
is the Monte Carlo method (U.S. EPA 2001;APAT
2008). This method implies a random sampling of
each variable of the assumed probability distribu-
tion, considering frequently all the variables mutu-
ally independent and is developed in successive
phases. Each parameter is described throughout
a Probability Density Function (PDF). At each
iteration, by means of an algorithm of calculation,
a random value between zero and one is generated
that allows to extract a value for each variable
of the correspondent probability distribution. It
could be noticed that each number has the same
probability to be sampled. The obtained values are
inserted in the risk assessment model that generates
in turn a number representing one of the possible
risk values. At the next iteration the procedure is
repeated originating another result for the risk; in
this way a curve of the probability distribution of
the risk values is obtained, each characterized by
a determined probability of occurrence.
Once the risk probability function is deter-
mined, an accepted level of the risk has to be de-
fined. The simplest criterion is that of considering
acceptable the risk in the case in which the value
correspondent to the 95° percentile is lower than
the acceptable limit, both for the toxic hazard
index and the carcinogenic risk. If otherwise the
acceptable limits are comprised between the 50°
and the 95° percentile, it is necessary to carry
out a more in-depth study; in these cases a valid
help is given by a sensitivity analysis (US EPA,
2001) that allows to individuate the parameters
present in the risk assessment that mostly condi-
tion the final result that is to say the ones whose
light variation impacts in a significant way on the
curve of the distribution of risk. The available
monetary resources will be addressed to those

• uncertainty in the parameters (due to the

error measurement and the parameters'
spatial and temporal variation);

• uncertainty in the models (due to the im-

precision in the structures of the models
applied for contaminants propagation and
in the dose- response relationships);

• uncertainty in the scenery (due to scarce-

ness of data regarding the area and its level
of pollution).

In other words the uncertainty comes from
the lack of cognitive information on parameters,
phenomena or models and can be reduced, at least
partly, by means of the study and the acquisition
of new cognitive elements. Instead, the variability
characterizes each parameter within its domain of
existence and is linked to the complexity and het-
erogeneity of the physical and chemical processes
involved in the analyzed phenomena.
In those cases, the most appropriate approach
is a probabilistic analysis in which the aleatoriety
of the environmental variables can be described
and interpreted by means of probability distribu-
tions that provide both the interval of values and
the probability of exceeding a threshold value
for each contaminant. The choice of the most
adequate probability distribution for each of the
input parameters in risk assessment is based on
the evaluation of the available data and on the
specificity of the analyzed parameters, as well
as on the case studies reported in the specialized
literature.