Regionalisation of rainfall statistics for the IFD Revision Project

The Australian Bureau of Meteorology has revised the Intensity-Frequency-Duration (IFD) design rainfall estimates for Australia. The revised IFDs cover Annual Exceedance Probabilities (AEPs) ranging from 50% to 1%. However, as even daily rainfall records are often much shorter than 100 years, there is significant uncertainty in the estimates of less frequent events. Regionalisation is an approach that overcomes this problem by assuming that information can be combined from multiple stations to give more accurate estimates of the parameters of the extreme value probability distributions. A Region of Influence (ROI) approach has been adopted for the IFD revision project. This paper describes the process used to define the ROIs and to determine the optimum region sizes. A number of approaches for the definition of the ROIs were considered including spatial proximity, using ellipses to define a preferred direction for spatial proximity, and using non-geographical site characteristics to define the closest stations. It was found that a simple measure of spatial proximity which included elevation as well as latitude and longitude gave the best results. In general it was found that predictive uncertainty in the rainfall quantiles was minimised when regions were sized to include 500 stationyears of data or 8 neighbours. This is consistent with earlier studies that have suggested an appropriate rule of thumb is a requirement for the size of a region to be approximately five times the largest quantile of interest.