In the 2016 release of Australian Rainfall and Runoff (ARR), variability of flood influencing factors (such as losses and temporal pattern) is now explicitly considered through application of Monte Carlo and ensemble approaches. However, hydrologic models are commonly an input for 2D hydraulic modelling, where computational constraints prevent a large number of floods from being modelled. Practitioners are therefore often required to select ‘representative floods’ for use in hydraulic modelling. When making this selection there is an implicit assumption that the probability of rainfall directly corresponds to the probability of flood levels—that is, the transformation is ‘probability neutral’. This assumption is often untested. The ‘representative flood’approach was benchmarked against probability neutral estimates of flood depths constructed through Monte Carlo application of a 2D hydraulic model. Hydrographs that were found to be probability neutral with regards to peak outflows from the hydrologic model did not necessarily result in probability neutral estimates of flood depths. This demonstrates the need for running a greater selection of events to avoid generating biassed hydraulic model results.