This keynote paper addresses uncertainty in geo-engineering with particular emphasis on reducing uncertainty through updating. This is done in a systematic manner by first stating where uncertainty through lack of knowledge or randomness occurs, how it affects modeling and inference, and how observations and experiments in the field or laboratory or numerical experiments can be used to obtain new information. All this forms the basis for the main part of the paper, which shows how uncertainty can be possibly reduced through updating. This then logically leads to the well known observational method, which can be considered to be the practical consequence and application of updating. The entire process is illustrated with a coherent set of examples from the geo-engineering domains fractures and fracturing, landslides and tunneling.


The question of uncertainty in rock mechanics, geotechnical engineering in a wider sense and engineering geology is well recognized (Baecher and Christian, 2003;Vick, 2002; Einstein and Baecher, 1983). Hence, it is not surprising that the sources of uncertainty have been classified, e.g. based on Einstein and Baecher (1983): Inherent spatial and temporal variability Measurement errors (systematic, random) Model uncertainty Load uncertainty Omissions or alternatively e.g. Baecher and Christian, (2003); Nadim and Einstein, (2005): Aleatory uncertainty (randomness) Epistemic uncertainty (lack of knowledge) The two classifications overlap, e.g. inherent spatial and temporal variability has an aleatory and an epistemic component (see also Einstein, 2006). What one would like to do is to capture and represent uncertainty such that it can be considered in engineering decisions. It is important to note that engineering decisions might also lead to a reduction of uncertainties..


Given the extensive literature in all areas from which the following examples are taken, the examples represent only snapshots. In the first phase, and thus in this section, a distinction is made between "lacking or limited knowledge" on the one hand and "randomness" on the other. When looking at the examples more closely, it will become apparent that randomness is in many cases lack of knowledge on a lower (finer) scale. This corresponds to the well known concept discussed by Spetzler and Staël v. Holstein (1975) "decoding versus modeling".

2.1 Fracturing and faulting
2.1.1 limited or lacking knowledge

While it is well recognized that tension and shear mechanisms are the causes of fracturing and faulting, it is very often not known, which specific mechanisms or combination of mechanisms act. For instance, researchers such as Riedel (1929) and Cloos (1936, 1955) Lajtai (1969, 1974), Brace and coworkers (e.g. Brace and Bombolakis, 1963) as well as Horii and Nemat Nasser (1986) have shown that under applied external stress, both shear and tensile fractures can be produced. While these mechanisms can be reproduced in laboratory experiments (see also section 4) it is much more difficult to know what actually happens, particularly in the field where many mechanisms interact or act sequentially.

This content is only available via PDF.
You can access this article if you purchase or spend a download.