Managing risk and achieving reliable geotechnical designs using Eurocode 7 - Risk and Reliability in Geotechnical Engineering - page 403

Sucient quantity
( large data set )
Poor quality
( imprecise data )
Epistemic
uncertainty

Sucient quantity
( large data set )
High quality
( precise data )
Aleatory
variability

Insucient quantity
( small data set )
Poor quality
( imprecise data )
Epistemic
uncertainty

Insucient quantity
( small data set )
High quality
( precise data )
Epistemic
uncertainty

Precision (quality of information)

Figure 10.3 Uncertainty and variability as functions of the quality and quantity of information. (After Bedi, A.
and Harrison, J.P. 2013. Proc. International Society for Rock Mechanics Symposium—Rock Mechanics
for Resources, Energy, and Environment , Wroclaw, Poland, 23-26 September 2013. Taylor & Francis,
London.)

Since in geotechnical designs, only a few test results are usually available and because soil,
being a natural material, is significantly more variable than manufactured materials, there
are normally insufficient test results to identify the statistical distribution of the geotechnical
property with certainty, or determine the parameters such as the standard deviation to describe
it precisely, thereby taking into account the variability and selecting the characteristic value
objectively. Such a situation where there is insufficient information, or where the information is
not precise enough to determine the statistical distribution of a material property, is referred to
as epistemic uncertainty. If, in such a situation, the material is actually inherently aleatory, then
as more information is obtained the values of the material properties become more certain and
the material can be modeled as being aleatory. Many soils fall into this category.
The properties of rock masses are usually more difficult to determine than the properties
of soil. Some rock parameters cannot be determined precisely so that they are inherently
epistemic, that is, no quantity of data would enable a precise description to be determined.
Bedi and Harrison (2013) have presented a relationship between epistemic uncertainty and
aleatory variability as functions of the quantity of information and the precision of the data,
as shown in Figure 10.3. The epistemic nature of many rock mass properties poses difficul-
ties in applying the limit state design method in Eurocode 7 to designs involving rock as
outlined by Bedi and Orr (2014).

10.4.3.3 Selection of aleatory characteristic parameter values
Since, as shown in Section 10.4.3.1, it has been found that the current definition provided
in Eurocode 7 for the characteristic value of a geotechnical parameter can result in very
different values being selected for a particular design situation by different engineers from
a set of test results, the CEN sub-committee TC250/SC7, responsible for the development
of Eurocode 7, has established an evolution group to prepare more specific guidance on the
selection of characteristic parameter values for the next version of Eurocode 7. The chal-
lenge facing this evolution group is to provide appropriate guidance on using statistics for
the selection of characteristic values from a limited number of test results while at the same
time taking account of well-established and valuable experience so that geotechnical engi-
neers will select more consistent characteristic values.