A Geostatistically Based Probabilistic Risk Assessment Approach - Evolving Application Domains of Data Warehousing and Mining - page 301

verification are needed for the eventuality of
accepting the risk value corresponding to the
probability of overcoming of 90%, it could be
possible to ascertain that, outside the surface
area enclosed by this probability (Figure 14) not
relevant underestimation errors are detected, that
is to say values flatly lower than 10% in most
of the area reaching values of 40% just in some
restricted zones.

In the field of cleanup risk assessment it
is operatively impossible to know exactly the
boundaries of a “potentially contaminated” area
for the exceeding of the admissible concentration,
as the available data contain a certain degree of
uncertainty. In this context geostatistics proves
to be versatile in providing elements that take
into account the nature of the data themselves,
particularly their uncertainty.

conclusions: comparison
of three Approaches fInAl concluSIon

This study is aimed at offering alternatives to the
classical risk assessment that bases itself on the
“worst case” principle, providing the possibil-
ity to quantify and integrate the uncertainty and
variability of input parameters in the risk assess-
ment procedure by means of the application of
geostatistical techniques.
In this study the following elaborations have
been carried out:

Geostatistics has proved to be a useful decision
support tool in order to treat in an efficient way
huge quantity of data and to cross various order of
information. Through the application of geostatis-
tical techniques it is possible to catch and describe
in a synthetic way the relations and tendencies
that are intrinsic in the data set, characteristics
that are neglected by a traditionally used classi-
cal approach.
The aim of this paper is to present different case
studies that focus on the utility of the application
of geostatistical approaches within risk assess-
ment. Geostatistics intervenes in this procedure
first of all in the calculation of the probability
distributions characterizing the most significant
variables that condition the expression of the
risk. It follows that an accurate characterization
of these parameters is of priority before carrying
out a risk assessment. Geostatistics provides a
useful help in this task, in that it allows to model
the uncertainty associated to the estimation and its
application within the decisional process can help
to formulate hypothesis regarding remediation and
the eventuality to proceed with further sampling
in some areas. The geostatistical criterion makes
it possible not only to design a sampling scheme
making reference to optimization criteria, but also
to improve the cost- effectiveness of a sampling
campaign, allowing to insert new points where the
knowledge is more approximate and to remove
others where the information is redundant.

• determination of Source Representative

Concentration by means of classical and
geostatistic methods
delimitation of potentially contaminated ar- •

eas by means of Indicator Kriging in terms
of map of probability of exceeding the
Threshold Concentration of Contamination
and evaluation of overestimation and un-
derestimation error

Both the classical and the geostatistical ap-
proach lead to the determination of one single
value of Source Representative Concentration,
the former clearly higher than the latter, due to
the smoothing effect of Kriging in attenuating
extreme values. The geostatistical approach of
Indicator Kriging on the contrary, lead to determine
a range of Source Representative Concentration
values and subsequently to estimate and diagram
the risk in function of the probability of exceeding
the threshold value.