HR: 0800h
AN: H31B-0356    [Abstracts]
TI: Stochastic Simulation of Daily Rainfall for Flood Risk Assessment Using a Mixed Distribution
AU: * Hundecha, Y
EM: yeshewatesfa.hundecha@rub.de
AF: Ruhr University-Bochum, Institute of Hydrology, Water Resources Management and Environmental Engineering, Universitaetsstr. 150 (IA 01), Bochum, 44801, Germany
AU: Pahlow, M
AF: Ruhr University-Bochum, Institute of Hydrology, Water Resources Management and Environmental Engineering, Universitaetsstr. 150 (IA 01), Bochum, 44801, Germany
AB: Stochastic weather generators are often used in the construction of long time series of rainfall that can be used in conjunction with rainfall-runoff models for risk assessment in the planning of water resources and flood mitigating facilities. The basic requirements of the weather generators used for such purposes are that they be able to reproduce the statistical properties of the historical rainfall series at each site and the spatial covariance structure between sites. Although a single type of distribution has frequently been implemented to model the amount of daily precipitation with seasonally varying parameters, it can sometimes be inadequate to capture some of the statistical properties of the daily rainfall that have relevance to the purpose to which the model is applied. In the present work, we demonstrate applicability of a stochastic model for the generation of daily time series of rainfall at multiple locations in which the amount of daily rainfall is modelled by a mixture of two different probability distribution functions. A two stage modelling procedure is implemented. In the first stage, a multivariate autoregressive model is used to model the local probability of occurrence of rainfall and the amount while keeping the inter-site covariance structure using a truncated and power transformed normal distribution. In the second stage, the amount simulated using the power transformed normal distribution is further transformed so that it can be regarded as coming from a mixture of Gamma and Gumbel distribution. The annual cycles of the amount as well as the temporal and spatial correlations are incorporated using a Fourier representation. Application was made on 122 stations within the Unstrut catchment in Eastern Germany. Results show that the model can fairly well reproduces the monthly mean rainfall and the corresponding variability as well as the extreme value distribution of the annual maximum daily rainfall.
DE: 1816 Estimation and forecasting
DE: 1829 Groundwater hydrology
DE: 1840 Hydrometeorology
DE: 3265 Stochastic processes (3235, 4468, 4475, 7857)
SC: Hydrology [H]
MN: 2007 Fall Meeting