The region of influence (ROI) approach for selecting sites is examined in conjunction with a Bayesian generalised least squares (GLS) regional flood frequency regression. The GLS procedure regionalises the mean, standard deviation and skewness of the log-Pearson III distribution with simultaneous consideration of model and sampling error. The ROI approach is used to enhance the traditional fixed region GLS analysis. It starts with the 15 nearest sites to the site of interest. The regional model is calibrated to this site data (using the catchment characteristics identified in the fixed region GLS analysis) and the model error variance noted. Then the ROI is expanded to include the 20 nearest sites. This process is repeated until the region producing the smallest model error variance is identified. A case study was undertaken for 55 catchments located in eastern New South Wales. Using an approach similar to stepwise regression, the best model for the mean was found to use catchment area and 50-year, 12-hour rainfall intensity as explanatory variables, whereas the models for the standard deviation and the skewness only had a constant term. One-at-a-time cross validation was performed for the 55 sites. Of significance it was found there were no outlier sites. Moreover, the ROI GLS approach produced more accurate and consistent results than a fixed region GLS model, highlighting the superior ability of the ROI approach to deal with heterogeneity.
32nd Hydrology and Water Resources Symposium (H2009). H2009: Proceedings of H2009, the 32nd Hydrology and Water Resources Symposium (Newcastle, N.S.W. 30 November - 3 December, 2009) p. 603-615