Browsing by Author "Edae, Belete Diriba"

Now showing 1 - 1 of 1
  • No Thumbnail Available
    Item
    Flood magnitude estimation: a comparison of frequency analysis and rainfall-runoff modelling approaches
    (University of Dar es Salaam, 1994) Edae, Belete Diriba
    The objective of this study has been to investigate the performances of two methods in determining the magnitude flood quantile corresponding to any return period. The methods are Regional Flood Frequency Analysis, (RFFA), an rainfall-runoff transformation model, specifically The Linear Perturbation Model. The former is discussed here in the context of fitting selected distributions to an at-site sample data using the method of probability weighted moments. Annual maximus discharge data was collected for six gauging stations in the Indian ocean drainage basin and were prepared for use. Four distributions, namely EV1, GEV, 4-Parameter Wakebyan 5-Parameter Wakeby, were fitted to the data. Acceptable performance is observed from the fitting. Further, regiona analysis was performed on regionally averaged standardise probability weighted moment values. Parameters of the four distributions were determined and using these, flood magnitudes of different return period were computed for all stations finally probability plots were prepared. Observation of the plots showed that EV1 fitted best to the observed discharge. Thus this distribution is selected to determine flood magnitude in the region. In the second part of the work the Linear Perturbation Model was applied to rainfall and discharge data for all sub-catchments considered. The model is based on the assumption that departures of rainfall and discharge from seasonal means are linearly related. Seasonal mean values were smoothed and pulse response ordinates were obtained by using ordinary Least squares. After allowing the model to operate in updated mode only one sub-catchment showed results with good performance. Analysis was made for the catchment which performed well in applying the model. Using rainfall as an input discharge was generated from the model and to the annual maximum series of this EV1 distribution was fitted to obtain flood quantiles. Standard errors of the flood quartiles of different return periods obtained by flood frequency analysis and rainfallrunoffmodelling approaches were determined. From these results, it is seen that the standard errors obtained for the former case were twice as large as those obtained for the later case.