Data: 17/09/2014 à 19/09/2014
Local: São Paulo - Brazil
Application of Bfs to Flood Probabilistic Forecasting for Wangjiaba Cross-Section of Huaihe River,China (PAP014841)
Flood forecasting and early warning systems
Because of the complexity of hydrological processes and the limitations of knowledge of Human being about the nature world, there inevitably exists uncertainties in input, model structure and parameter of the procedures of real-time flood forecasting, which in turn results in uncertainties for the final forecasting outcomes. Conventional approaches of real-time flood forecasting provide deterministic or unique value, which in essence excludes considering the inevitably uncertainties in the proceesses of flood forecasting, consequently it is unable to quantity the forecasting uncertainties, and can not evaluate the possible risk of decision-making which is on the basis of flood forecasting. Therefore, quantitatively estimating the forecasting uncertainties and developing flood probabilistic forecasting(FPF), supplying with abundant information for flood control and decision-making, is of important both in theory and application. In this study, the FPF for Wangjiaba corss-section of Huaihe River was achieved by means of Bayesian Forecasting System (BFS). The sources of floods at Wangjiaba were generalized as floods from the upper reaches, namely Xixian, Huangchuan and Bantai cross-sections, as well as floods from the intermediate reach between the upper cross-sections and Wangjiaba cross-section. The operational rainfall-runoff relations of these cross-sections and the intermediate reach were used to provide the initial predictions. Then the methodology of the hydrologic uncertainty processor (HUP) with in BFS was adopted, i.e., using prior distribution to describe the natural uncertainty of the hydrological process, and the likelihood function to describe uncertainties of model structure and parameter; normalizing the initial forecasting data and observations of discharge by Meta-Gaussian model; estimating the posterior distribution of predictand by Bayesian theorem, so that obtaining the probabilistic forecasting for flood events. The procedure was applied to flood events from yr 2000 to yr 2012, It indicated that the approach identifies model and parameter uncertainties in the procedure of flood forecasting, and is easy to use for flood probabilistic foresting.