Application of Bayesian GLSR to estimate sub daily rainfall parameters for the IFD revision project

The Australian Bureau of Meteorology has recently revised the Intensity-Duration-Frequency (IFD) design rainfall estimates for Australia. Although there is much better coverage of sub-daily rainfall stations that for the previous IFD data, there are still fewer continuous rainfall stations in Australia than daily rainfall stations. Bayesian Generalised Least Squares Regression (BGLSR) has been used to estimate sub-daily rainfall parameters based on site characteristics and daily rainfall statistics to improve the spatial coverage of sub-daily data. The BGLSR was applied differently across Australia depending on the density of the spatial coverage of the sub-daily rainfall stations. For areas with good spatial coverage, a process akin to a Region of Influence approach was used where the regression equations vary in space. For parts of Australia with sparser data coverage, fixed areas of analysis were used to provide estimates of the regression equation coefficients. L-moments were used to summarise the parameters of the Annual Maxima Series (AMS) data, and thus the BGLSR provides predictions of the mean-L-coefficient of variation and L-skewness. It was found that the BGLSR provided excellent estimates of the mean of the AMS, with most important predictors being the 24 hour AMS mean rainfall and the latitude of the station. This paper describes the application of the BGLSR for the project and presents the results for one analysis area in detail.