RESUMO
This paper presents a novel framework that links imprecision (through a fuzzy approach) and stochastic uncertainty (through a Bayesian approach) in estimating flood probabilities from historical flood information and systematic flood discharge data. The method exploits the linguistic characteristics of historical source material to construct membership functions, which may be wider or narrower, depending on the vagueness of the statements. The membership functions are either included in the prior distribution or the likelihood function to obtain a fuzzy version of the flood frequency curve. The viability of the approach is demonstrated by three case studies that differ in terms of their hydromorphological conditions (from an Alpine river with bedrock profile to a flat lowland river with extensive flood plains) and historical source material (including narratives, town and county meeting protocols, flood marks and damage accounts). The case studies are presented in order of increasing fuzziness (the Rhine at Basel, Switzerland; the Werra at Meiningen, Germany; and the Tisza at Szeged, Hungary). Incorporating imprecise historical information is found to reduce the range between the 5% and 95% Bayesian credibility bounds of the 100 year floods by 45% and 61% for the Rhine and Werra case studies, respectively. The strengths and limitations of the framework are discussed relative to alternative (non-fuzzy) methods. The fuzzy Bayesian inference framework provides a flexible methodology that fits the imprecise nature of linguistic information on historical floods as available in historical written documentation.
RESUMO
A Bayesian approach in a possibilistic context, when the available data for the underlying statistical model are fuzzy, is developed. The problem of point estimation with fuzzy data is studied in the possibilistic Bayesian approach introduced. For calculating the point estimation, we introduce a method without considering a loss function, and one considering a loss function. For the point estimation with a loss function, we first define a risk function based on a possibilistic posterior distribution, and then the unknown parameter is estimated based on such a risk function. Briefly, the present work extended the previous works in two directions: First the underlying model is assumed to be probabilistic rather than possibilistic, and second is that the problem of Bayes estimation is developed based on two cases of without and with considering loss function. Then, the applicability of the proposed approach to concept learning is investigated. Particularly, a naive possibility Bayes classifier is introduced and applied to some real-world concept learning problems.