Please use this identifier to cite or link to this item: https://hdl.handle.net/11681/20762
Title: State-of-the-art for assessing earthquake hazards in the United States. Report 18, Errors in probabilistic seismic hazard analysis
Authors: Massachusetts Institute of Technology. Department of Civil Engineering.
Soils and Pavements Laboratory (U.S.)
Veneziano, Daniele.
Keywords: Earthquake engineering
Earthquake hazards
Earthquake risk
Earthquakes
Earthquake intensity
Seismic risks
Seismology
Mathematics
Mathematical models
Risk prediction
Publisher: Geotechnical Laboratory (U.S.)
Engineer Research and Development Center (U.S.)
Description: Miscellaneous Paper
Abstract: After a period of fast development, procedures for the probabilistic quantification of seismic hazard have reached a state of relative maturity. Standard methods usually differ only by secondary characteristics. In spite of this operational stability, estimates of seismic hazard that are produced by different algorithms depend on and are often very sensitive to the form of the mathematical model and the choice of some of the parameters. It is important to recognize this sensitivity in order to avoid placing excessive confidence in numerical estimates of hazard, especially in the case of rare events. Errors may originate from conscious simplification of the mathematical representation (e.g., from the use of a Poisson model of earthquake occurrences as opposed to a physically more attractive point process with spatial and temporal memory), from limited statistical information, or from faulty judgment (expert errors). The purpose of this report is to review different types and sources of errors and to assess their impact on calculated seismic hazard. In the available literature, no generally accepted procedure exists to systematically deal with uncertainty on seismic hazard and, most frequently in practice, elusive arguments of conservatism are used to avoid direct confrontation with the problem. Scientific studies are sparse and fragmentary; they typically focus on narrow aspects of the problem, use different statistical techniques, and report results in different formats. Accordingly, an effort is made here to coordinate these results and provide a unified picture. A first and fundamental problem, which in the past has received only minimal attention, is to give a definition of objective or "true" hazard against which various estimates can be evaluated. One such definition is proposed here, based on the notion of hazard as relative frequency in time. Three classes of errors are then introduced and separately analyzed: (1.) Errors due to uncertainty of model parameters, such as the geometrical configuration of the earthquake sources, the parameters of the magnitude-frequency relationship (including the upper bound), and those of the attenuation function. A recently proposed analytical procedure for the analysis of such errors is reviewed and results from sensitivity studies are presented. (2.) Errors due to the use of a wrong form of model. These errors result in biases (hence, their statistical magnitude does not depend on the amount of information available) and may be very large, especially for rare events. (3.) Finally, errors of interpretation and modeling by seismologist experts. Case studies are reviewed and mathematical models for errors of this last type are proposed.
Rights: Approved for public release; distribution is unlimited.
URI: http://hdl.handle.net/11681/20762
Appears in Collections:Documents

Files in This Item:
File Description SizeFormat 
MP-S-73-1-Report-18.pdf12.01 MBAdobe PDFThumbnail
View/Open