8

The role of epistemic uncertainties

Section structure not completely clear yet. Types of uncertainties: aleatory, epistemic but also other taxinomies could be covered in the intro text. Stress that main part is devoted to logic tree framework for epistemic uncertainties Sections could follow the structure

8. 1 The conceptual framework (Logic tree as ensemble discrete distribution)

8. 2 Use and misuse of logic trees. Planting the tree requires model selection and weighting

8. 3 Model selection and ranking
8. 3. 1 Pre-selection for GMMs  (Cotton et al., 2006) and Bommer (2010). Similar for SSC models. In general, preselection involves many decision under uncertainties. Subject to heuristics and biases (Kahnemann and Tversky) ->  
8. 3. 2 Cognitive biases 
8. 3. 3. Data driven model selection and ranking 
a) Hypothesis testing 
b) Information theoretic approaches 
-> The meaning of information: matHSHADocu2/tutorial/InformationIllustration
Example: matHSHADocu2/tutorial/GMMSelectionUsingIntensities

8.4 The role of branch weights and their elicitation. To date not yet possible to let the data speak exclusively. Therefore, quantification of experts knowledge still crucial. How to.

8.5 Processing of hazard curves:  Which hazard curve to use in case of logic tree (Mean, median, mode. Does it matter how to separate epistemic and aleatory? Not for mean hazard, but…)

8. 6 Problems and possible strategies
8. 6. 1 model redundancies 
8. 6. 2 Makeing trees understandable. Trees with a backbone and how to generate them. Example Chile model.


0.0. XXXXXXXXX

XX



Pages:
.............. 8.1. XXXXXXXXX
.............. 0.0.0. XXXXX

Frank Scherbaum (2015), Fundamental concepts of Probabilistic Seismic Hazard Analysis, Hazard Classroom Contribution No. 001

Comment