Premium
SAMPLE UNCERTAINTY IN FLOOD LEVEE DESIGN: BAYESIAN VERSUS NON‐BAYESIAN METHODS 1
Author(s) -
Duckstein L.,
Bogárdi I.,
Szidarovszky F.,
Davis D. R.
Publication year - 1975
Publication title -
jawra journal of the american water resources association
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.957
H-Index - 105
eISSN - 1752-1688
pISSN - 1093-474X
DOI - 10.1111/j.1752-1688.1975.tb00696.x
Subject(s) - levee , bayesian probability , flood myth , bayes' theorem , margin (machine learning) , statistics , sample (material) , sample size determination , bayes estimator , sampling (signal processing) , computer science , sampling design , mathematics , econometrics , geology , machine learning , geotechnical engineering , geography , computer vision , chemistry , archaeology , filter (signal processing) , chromatography , population , demography , sociology
Bayesian and non‐Bayesian flood levee design methods that account for the uncertainty due to limited record length are compared using a case study. The first method, Bayesian decision theory (BDT), imbeds the uncertainty in the parameters of the yearly peak stage into a loss function. The optimum design of the flood levee, called Bayes design, corresponds to the minimum expected loss, called Bayes risk. The second method, induced safety algorithm (ISA), computes a margin of safety to be added to either an existing levee or a levee designed by classical benefit‐cost analysis. The design decision is shown to fluctuate as different record lengths are considered. For short record lengths, BDT, which takes small sample bias into account, appears to yield a more conservative design than ISA. On the other hand, ISA, which is simple to implement, seems to be preferable to BDT for longer record lengths.