The reliability of the geotechnical enterprise

Uncertainty is a central part of geotechnical engineering. Although in the past the profession has dealt with uncertainty by such procedures as the observational method, probabilistic methods are increasingly widely employed. In using these approaches it is important to understand the differences between aleatory and epistemic uncertainty and to distinguish between the views of probability as defined by relative frequency versus degree-of-belief. In particular, degree-of-belief approaches lend themselves to Bayesian updating, which is in turn consistent with the observational approach in geotechnical engineering. The engineer also needs to be sensitive to the different ways that spatial and systematic errors contribute to the final error. In most practical cases estimation of statistical properties relies to some extent on eliciting expert opinion, a process that can be quite difficult. In general, groups of experts tend to be good at estimating mean trends, but they also tend to be overconfident in their estimates of ranges. Examples or recent practical application of probabilistic methods include the stability of a slope in Karst terrain, the decision on how to protect a tank farm against earthquake induced liquefaction, probabilistic estimates of costs and schedules for highway projects, and risk exposure for future hurricanes in southern Louisiana. (a) For the covering entry of this conference, please see ITRD abstract no. E216246.