HR: 09:00h
AN: H41A-05 INVITED     [Abstracts]
TI: Bayesian Combination of Regional and Local Information Using Some Common Distributions in Hydrology
AU: * SEIDOU, O
EM: ousman_seidou@ete.inrs.ca
AF: INRS-ETE, 490 rue de la Couronne, Quebec, QC G1K9A9, Canada
AU: OUARDA, T B
EM: taha_ouarada@ete.inrs.ca
AF: INRS-ETE, 490 rue de la Couronne, Quebec, QC G1K9A9, Canada
AB: The main challenge in flood frequency analysis is to find relevant and sufficient information to fit a local distribution with an acceptable precision to the variable of interest. This precision impacts the cost and reliability of hydraulic structures as well as the safety of downstream communities. If the site of interest has been monitored for a sufficiently long period (more than 30-40 years), at-site frequency analysis can be used to estimate flood quantiles with a fair precision. Otherwise, regional estimation may be used to mitigate the lack of data, but local information is then ignored. The authors propose a Bayesian method in this paper that uses both sources of information for even more precise quantile estimation. The proposed method uses the classical log- linear regression as regional model and assumes that the local flood peaks are GEV, Gamma, Weibull, Log- Normal or exponentially distributed. The method works even with a single local observation besides relaxing the hypothesis of normality of the quantiles probability distribution that is used in the empirical Bayes approach. A thorough performance assessment was made with the GEV distribution and it was shown that a) when the regional model is unbiased, the proposed method gives better estimation of the GEV quantiles and parameters than the local, regional and empirical Bayes estimators; b) even when the regional model displays a severe relative bias when estimating the quantiles, the proposed method still gives the best estimation of the GEV shape parameter and outperforms the other approaches on higher quantiles provided that the relative bias is the same for all quantiles; c) the gain in performance with the new approach is considerable for sites with very short records. Theoretical developments and some preliminary results are presented for the other distributions. Keywords: regionalization, probability distribution, linear regression, empirical Bayesian method, flood frequency analysis, combination.
DE: 1817 Extreme events
DE: 1821 Floods
DE: 1874 Ungaged basins
SC: Hydrology [H]
MN: 2007 Joint Assembly