A comprehensive continent-wide regionalisation investigation for daily design rainfall

Abstract

Study region: Australia. Study focus: Design rainfalls, in the form of Intensity Duration Frequency curves, are the standard input for most flood studies. Methods to combine rainfall data across space are required to provide optimal estimates of design rainfalls and constrain their uncertainty. This paper robustly investigates the use of a variety of regionalization methods to provide Australia wide design rainfall estimates using 8619 high quality rainfall stations. The influence of an individual large rainfall event in March 2012 on the regionalised design rainfalls is also investigated to guide decisions on when design rainfall data should be updated following record-breaking rainfall events. New hydrological insights for the region: The optimum approach was found to be circular regions of influence with a size of 500 station-years. The regions of influence are least biased when defined using spatial proximity and including elevation in the calculating spatial proximity can help to reduce the bias in design rainfall estimates in areas of strong topographic relief. The March 2012 rainfall event increased site rainfall quantiles by 14% on average. When regionalisation is used the increases were substantially moderated and were within the uncertainty bounds of the design rainfalls. It is still not routine to provide uncertainty estimates with design rainfall data, but such information is particularly important given likely future changes in extreme rainfalls due to anthropogenic climate change. Keywords: Design rainfall, Regionalisation, Region of influence