Multi-model ensemble projections of future extreme temperature change using a statistical downscaling method in eastern Australia

Projections of changes in temperature extremes are critical to assess the potential impacts of climate change on agricultural and ecological systems. Statistical downscaling can be used to efficiently downscale output from a large number of general circulation models (GCMs) to a fine temporal and spatial scale, which now provides the opportunity for future projections of extreme temperature events. This paper presents an analysis of extreme temperature in data downscaled from ensembles of 13 selected GCMs, out of 28 GCMs, contributing to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC-AR5) under two Representative Concentration Pathways (RCP4.5 and RCP8.5) in eastern Australia. The statistical downscaling procedure begins with spatial interpolation of the monthly gridded data to specific locations of interest using an inverse distance-weighted method, followed by a bias correction towards historical observed climate. Daily climate data for each location are then generated by a modified version of the WGEN stochastic weather generator. The extremes of temperature are described by eleven indices, namely, the annual maximum daily Tmax (TXx), the annual maximum daily Tmin (TNx), the annual minimum daily Tmax (TXn), the annual minimum daily Tmin (TNn), the number of hot days (HD) and frost days (FD), warm days (TX90p) and nights (TN90p), cold days (TX10p) and nights (TN10p) and extreme temperature range (ETR). The results show that downscaled data from most of the GCMs reproduced the correct sign of recent trends in all the extreme temperature indices (except TN10p) although there was much more variation between the individual model runs. An independence weighted mean method was used to calculate uncertainty estimates, which verified that multi-model ensemble projections produced a good consensus compared to the observations in magnitude of the trend in TXx, TN90p, HD, TNn, ETR for the period 1961-2000 when averaged across eastern Australia. In the 21st century the frost days, cold days and nights decrease while more frequent warm days and nights and hot days are projected in the New South Wales (NSW) wheat belt. The changes in temperature extremes under RCP8.5 are more pronounced than that under RCP4.5. Greater warming occurs in the east and northeast of the NSW wheat belt by the end of the 21st century and increases the risk of exposure to hot days around wheat flowering date, which results in farmers needing to reconsider wheat varieties suited to maintain yield. This analysis provides a first overview of projected changes in climate extremes from the ensemble of 13 CMIP5 models with statistical downscaling data in eastern Australia, and supplies important information to mitigate the adverse effects of climate extremes on NSW wheat belt and improve the regional strategy for agricultural systems.