English Abstract In the PROSA (PRObabilistic Safety Assessment) project,
a study is made of radioactive waste disposal in rock-salt formations using
probabilistic techniques. The purpose of PROSA is to determine the health
risks of humans and the safety relevant characteristics of a waste disposal
concept. The two generic local hydrogeological models considered in this
project are a shallow salt dome with an overburden of 230 m and a deep salt
dome with an overburden of 800 m. A prescribed hydraulic head distribution
at the top of the models results in two infiltration zones at the edges and
a seepage zone in the middle of the models, forcing the released
radionuclides from the top of the salt dome through the overburden into the
biosphere. Two different scenario's have been considered, a water intrusion
and a diapirism/subrosion scenario. The sensitivity and uncertainty
analysis of the groundwater compartment, performed with the UNCSAM package,
is presented in this report. The velocity values are derived from a large
number of groundwater velocity fields determined with the METROPOL-1 code.
The considered input parameters are the thickness, porosity and permeability
of the geohydrological units, the ratio of permeability between a fault
present in the middle of the models and the adjacent layer, and the absolute
value of the hydraulic gradient at the earth surface. Initial calculations
with the UNCSAM package, including all the input parameters, resulted in
unreliable values of the chosen sensitivity and uncertainty measures. To
improve the analysis, a total of four modifications were implemented in the
regression model including the use of the natural logarithm of the average
vertical particle velocity value as the dependent variable and the
elimination of the fault factor as a stochastic input parameter. The fault
factor is very dominant and since it is also correlated with the
permeability of the adjacent layers, it completely obscured the effects of
all other parameters in the UNCSAM analyses. Recognizing the fault factor
as the most influential parameter, it was left out in the UNCSAM analyses
discussed in this report. Of all other considered parameters, the
permeability of the clay layer (layer 1) is the most influential input
parameter. When the clay layer is removed from the models as a result of
diapirism/subrosion, the permeability of the layer with the cretaceous
deposits and the permeability of the Brussel's Sand/Ieper Clay layer (layer
1''') become the most influential parameters.