Author(s): Fiona Johnson
Linked Author(s):
Keywords: Climate change; Design rainfalls; Areal Reduction Factor; General Circulation Models; Model Evaluation
Abstract: Areal Reduction Factors (ARFs) are a common method of considering the heterogeneity of storm events over large watersheds. The ARFs assumes that the peak rainfall intensity will not occur uniformly due to the scale disparity between the storm events and the watershed. The rainfall intensities are thus reduced depending on the size of the watersheds and the meteorological conditions of the area. The concept of ARFs has been extended to analyze the outputs of General Circulation Model (GCM) simulations of extreme rainfall events. It is generally considered that GCM rainfall predictions are areally averaged over the grid cell. Therefore most GCM evaluation studies use gridded rainfall data to provide a like-with-like comparison. However for the purposes of engineering design, the actual magnitude of future extreme rainfall events at a point scale is required since most hydrological engineering methods are based on point rainfall data. In this study, the scaling of point to grid rainfall for extreme rainfall events is considered by firstly comparing point to gridded observations over the historical period. The scaling relationships for a range of grid sizes based on observations were compared to the GCM derived rainfall intensities for the historical and future climates. The differences in the scaling relationships across Australia are explored. Point estimates of extreme rainfalls are estimated using regionalized Annual Maximum Series (AMS) data. The Generalized Extreme Value (GEV) distribution was used to fit the AMS data with L-moments. For the gridded observed data and GCM data, L-moments and the GEV distribution were used to fit the AMS data extracted for grid areas ranging from 100 km2 to 40000 km2. The scaling factors are also compared to previously derived Areal Reduction Factors across Australia. It was found that the raw GCM simulations estimates of extreme rainfalls are not consistent with the existing ARF relationships; on the other hand gridded historical rainfall across a range of watershed sizes is consistent with existing ARF relationships. Bias correction of the daily rainfall GCM simulations improves the agreement of the extreme rainfalls with the expected relationships although the results vary across Australia. The results of this study will be useful for GCM model evaluation research as well as providing practical input into estimates of extreme rainfalls under climate change.
Year: 2013