Enhanced Rainfall Event Generation for Monte Carlo Simulation Technique of Design Flood Estimation

摘要

In recent years, there has been notable research on a more holistic approach (such as the Monte Carlo simulation technique) of design flood estimation in Australia. Rahman et al (2002a) developed a Monte Carlo simulation technique that employed commonly adopted design data and models in the simulation of derived flood frequency curves. This paper builds on the work of Rahman et al (2002a) in developing Intensity Frequency Duration (IFD) tables that are used as the basis of rainfall event generation in the Monte Carlo simulation. The current approach as outlined by Rahman et al (2002a) involves identification of 'core' rainfall events, pooling the events into duration classes or bins according to their duration. Events in each duration class are then rescaled to the characteristic duration nominated for each class. A frequency analysis assuming the exponential distribution is then applied to the rescaled intensities. Intensities for various average recurrence intervals (ARls) are then calculated for each characteristic duration. Intensities for intermediate durations are calculated using fitted second order polynomial interpolation curves. From these curves an event based IFD table can be formulated. Typically, five duration classes are chosen and four to five events per year are chosen to facilitate adequate frequency analysis. This approach was a significant advance in developing Monte Carlo simulation techniques for design flood estimation - focusing practitioners on the use of the IFD table - as used in the Design Event approach outlined in the Australian Rainfall and Runoff (AR&R) (IEAust, 1987). Two aspects of this new work required further investigation: (a) how do event based IFD tables relate to the Bureau of Meteorology (BoM) burst tables as presented in AR&R (IEAust, 1987); and (b) can we rebuild the event based IFD table using the generated rainfall events? While much further work has been done on the former, there has been insufficient work in ensuring that the generated events do indeed reflect the characteristics of the source event based IFD table. Further work in this area by the author, as presented in this paper, has resulted in a refinement in the preparation of the event based IFD table. The author believes this refinement will lead to the development of a more defined relationship between event based IFD tables and burst AR&R IFD tables. Better understanding of this relationship will obviate the need to analyse raw rainfall records required to generate the event based IFD table thus facilitating wider use of the Monte Carlo simulation technique of design flood estimation. As an illustration, the refined method is applied to a number of rainfall stations in SE Queensland. Burst and event based IFD tables are compared. The paper concludes in recommending that the refined approach be considered by practitioners.