Author(s): Elena Kirilova Bojilova
Linked Author(s):
Keywords:
Abstract: The best available methods for the generation of multi-seasonal, multi-site streamflow sequences are perhaps disaggregation models. Disaggregation modelling is a process by which time series are generated dependent on a time series already available. Typically, the independent series has been previously generated, by any stochastic model desired. For any disaggregation model the basic requirement is to preserve the statistical properties for both short and long time intervals (annual, monthly. .. ). The two basic forms of disaggregation modelling are temporal and spatial. The focus of this study was application of disaggregation models in the temporal domain. The original extended model of Mejia and Rouselle (1976), is rejected because of its inconsistent model structure, which results in a parameter estimation problem (Lane, 1982; Stedinger and Vogel, 1984). The Mejia and Rouselle model failed to preserve one of the important characteristics of the disaggregation models which is the additivity. A technique was developed by Lin (1990) to derive the corrected parameter estimation equations associated with the use of the full Mejia and Rouselle model. Lin found that corrected parameter estimation equations are conditional on its key series (input). The new approach can preserve exactly the important moments of interest and the additivity. In this research the original extended model of Mejia and Rouselle and the corrected Lin model, (into two forms-one and two-stage disaggregation), were applied to the streamflow data of the chosen rivers in Bulgaria. The new approach succeeded in the preservation of the additivity as well as the moments. Applying the Lin model in one and two-stage disaggregation results in consistent parameter estimates. Furthermore, in the study both single and multi-site disaggregation were applied. The multi-site approach of the Lin model results in decreasing the number of parameters but numerical problems then appear with increasing of the number of sites. The advantage of the multi-site approach is the preservation of some additional correlations.
Year: 1999