| Summary: | Thesis (Ph.D.)., Memorial University of Newfoundland, 1998. Engineering and Applied Science Bibliography: leaves 193-201 This thesis deals with the correlation structure of annual peak flow series in detail. The thesis is in divided into four major parts. -- Part one takes a closer look at the correlation structure of annual peak flow series at two scales: scale of one that measures short-term behaviour by the lag-one autocorrelation coefficient, r(1), and scale of n, that measures long-term behaviour by Hurst's K. It is shown that there are significant correlation and dependence between Hurst's K. and r(1) for both observed data and data from Monte Carlo experiments which imply that short- and long-term behaviour cannot be treated separately as is current practice. -- Part two suggests a new approach for quantitatively describing long-term correlation that is rooted in an independent series. The results indicate that long-term correlation rooted in a short-term independent series can be quantitatively estimated, and the simultaneous occurrence of high values of Hurst's K and low values of r(1) is, in fact, not an uncommon phenomenon. A new method of testing for long-term correlation that takes the short-term correlation into account is developed. -- Part three further looks at peak flow correlation structure across scales based on the perspective of fractal geometry. A family of probability-scale-threshold curves which contain more information about the correlation structure of peak flows, are constructed and the scaling behaviour of peak flow series is explored. -- In order to take serial correlation into account in flood risk analysis, the concept of scaling plotting positions (SPP), is developed in part four. It takes scaling behaviour of peak flows into account and develops a new plotting position formula in estimation of future floods. The results of Monte Carlo simulation showed that the estimated quantiles of SPP are more efficient and robust when compared with current estimators of flood quantiles. -- The study presented in this thesis has provided a view of the correlation structure of peak flows across scales so that flood risk can be better estimated.
|
|---|