Flood hazard in the Mekong Delta – a probabilistic, bivariate, and non-stationary analysis with a short-termed future perspective
N. V. Dung1,3, B. Merz1, A. Bárdossy2, and H. Apel11GFZ German Research Centre for Geosciences, Section 5.4 Hydrology, Potsdam, Germany 2Institute of Hydraulic Engineering, University of Stuttgart, Stuttgart 70569, Germany 3Southern Institute of Water Resources Research SIWRR, Ho Chi Minh City, Vietnam
Received: 01 Feb 2013 – Accepted for review: 06 Feb 2013 – Discussion started: 20 Feb 2013
Abstract. In this paper we present a novel approach for flood hazard analysis of the whole Mekong Delta with a particular focus on the Vietnamese part. Based on previous studies identifying the flood regime in the Mekong delta as non-stationary (Delgado et al., 2010), we develop a non-stationary approach for flood hazard analysis. Moreover, the approach is also bi-variate, as the flood severity in the Mekong Delta is determined by both maximum discharge and flood volume, which determines the flood duration. Probabilities of occurrences of peak discharge and flood volume are estimated by a copula. The flood discharges and volumes are used to derive synthetic hydrographs, which in turn constitute the upper boundary condition for a large-scale hydrodynamic model covering the whole Mekong Delta. The hydrodynamic model transforms the hydrographs into hazard maps. In addition, we extrapolate the observed trends in flood peak and volume and their associated non-stationary extreme value distributions to the year 2030 in order to give a flood hazard estimate for the near future. The uncertainty of extreme flood events in terms of different possible combinations of peak discharge and flood volume given by the copula is considered. Also, the uncertainty in flood hydrograph shape is combined with parameter uncertainty of the hydrodynamic model in a Monte Carlo framework yielding uncertainty estimates in terms of quantile flood maps. The proposed methodology sets the frame for the development of probabilistic flood hazard maps for the entire Mekong Delta. The combination of bi-variate, non-stationary extreme value statistics with large-scale flood inundation modeling and uncertainty quantification is novel in itself. Moreover, it is in particular novel for the Mekong Delta: a region where not even a standard hazard analysis based on a univariate, stationary extreme value statistic exists.
Dung, N. V., Merz, B., Bárdossy, A., and Apel, H.: Flood hazard in the Mekong Delta – a probabilistic, bivariate, and non-stationary analysis with a short-termed future perspective, Nat. Hazards Earth Syst. Sci. Discuss., 1, 275-322, doi:10.5194/nhessd-1-275-2013, 2013.