A Novel Method for Assessing Rainfall Extremes from GCMs and RCMs using Areal Reduction Factors.

There is considerable uncertainty in future projections of rainfall extremes which affects flood risk planning. One of the reasons for this uncertainty is the significant biases in both General Circulation Model (GCM) and Regional Climate Model (RCM) simulations of these events. Generally these biases mean that future extreme rainfalls are estimated using a delta-change approach where the observed rainfalls are scaled by the percentage change in the RCM current to future estimates. One aspect that has not received much attention is how spatial patterns of extreme rainfall may change and whether these patterns are represented by GCMs and RCMs with more accuracy than the magnitude of the extreme rainfall totals.

Areal reduction factors (ARFs) are a useful tool for considering the spatial patterns of extreme rainfalls. ARFs account for the spatial heterogeneity of storm events by quantifying the required reduction in a point rainfall estimate to give an equivalent area averaged rainfall estimate. The required reduction has been found to be dependent on location, watershed area and size of the rainfall event.

This study has investigated the scaling of point to grid rainfall for extreme rainfall events as modelled by a number of GCMs and for dynamically downscaled GCM simulations over south-east Australia from the Weather Research and Forecasting (WRF) model. ARFs are proposed as a novel way of evaluating both the GCM and RCM simulations. For the GCMs, the ARFs provide guidance on the models that best represent the modulation of point rainfalls expected over watershed areas as large as individual GCM cells. The advantage of this approach is that information on the scaling of extreme value distributions with increasing spatial scales can be used to correct the bias in the GCM simulations.

For the WRF downscaled results, reanalysis driven simulations show that the representation of observed ARF relationships is quite acceptable. Future changes in ARFs can therefore be considered with confidence in the fidelity of the model. It was found for rainfall events with durations shorter than 12 hours there are some significant decreases in the ARFs suggesting changes in the mechanisms leading to extreme rainfalls. For longer rainfall event durations the changes were much smaller.