Introduction to Contaminated Site Management - Dealing with Contaminated Sites: From Theory Towards Practical Application - page 38

The procedure for the determination of probability density functions depends
on the purpose of the Risk Assessment. For 'generic Risk Assessment purposes',
for example, the derivation of Soil Quality Standards, the most effective way is to
incorporate both uncertainty and variability in the probability density functions. For
site-specific applications, however, most of the variability could be eliminated by
measurements, so that the probability density functions mainly cover uncertainty.
The most popular way of performing a probabilistic Risk Assessment is based on
Monte Carlo techniques (e.g., Seuntjes ( 2004 ), who assessed the risk of the leach-
ing of cadmium from soil, originating from the former presence of non-ferrous
industries, into the groundwater in Lommel, Belgium). Burmaster and Anderson
( 1994 ) described 14 principles of good practice to assist people in performing
and reviewing probabilistic or Monte Carlo Human Health and Ecological Risk
Assessment.
The result of a probabilistic Risk Assessment is a probability density function
of an important measure for risks, for example, human exposure or percentage of
soil organism affected, or of a Risk Index. The huge advantage of such a probabilis-
tic procedure is that the impact of uncertainties and variability is made transparent
in the resulting risk appraisal. However, a choice needs to be made for the level
of acceptability, in terms of a specific percentile of the probability density func-
tion as output of the Risk Assessment. Although this offers a more sophisticated
way of dealing with acceptable risks, there are no objective criteria to underpin this
choice.
Since Risk Assessment is a relatively unreliable process, it is of the utmost impor-
tance to describe each and every step taken, from the field survey on up to Risk
Management solutions. This should be done in such a detailed way that the Risk
Assessment is reproducible for third-party risk assessors. The report must explain
which political boundary conditions are incorporated in the Risk Assessment.
Furthermore, it should refer to all the Risk Assessment tools (including all input
parameter values) that were used, along with associated references.
Because of the characteristically limited reliability involved with Risk
Assessment, it is recommended to organise, at least for crucial reports, peer reviews
and/or second opinions. Peer reviewers cannot eliminate the uncertainties, but they
can judge whether risk assessors have made these uncertainties transparent and
also, very importantly, whether the uncertainties rectify the conclusions. Several
countries include peer review or second opinion procedures in their acts and laws.
Alternatively, these procedures are often included in national guidance documents.
1.5.4.3 Validation
The lack of reliability of Risk Assessment results is supported by numerous val-
idation, comparison and round-robin studies. It must be realised, however, that
validated models hardly (if at all) exist (Leijnse and Hassanizadeh 1994 ). In fact,
only model applications can be validated. The reason for this is that for each specific
model application, different equations and input parameters are the most relevant.
Therefore, in each specific model application a different part of the model is tested.