Rapid attribution analysis of the July 2017 heat across Australia
Australia experienced its highest average daily maximum temperatures for July on record in 2017. Below I outline the methods used to investigate the role of climate change in this event.
I first checked which of the 16 models (Taylor et al. 2012; Table 1) with the required simulations for the analysis have a July average maximum temperature anomaly distribution that is compatible with the observed distribution (following the method of Lewis and Karoly 2013; King et al. 2015). Maximum temperature anomalies were extracted from the Bureau of Meteorology’s observational data (using a 1961-1990 baseline) and compared with historical model simulations using the same baseline.
Table 1. Models and simulations used in this analysis. In bold are model simulations used for calculating natural baseline and the 1.5°C and 2°C warmer worlds. Other historical simulations were only used for model evaluation.
I first checked which of the 16 models (Taylor et al. 2012; Table 1) with the required simulations for the analysis have a July average maximum temperature anomaly distribution that is compatible with the observed distribution (following the method of Lewis and Karoly 2013; King et al. 2015). Maximum temperature anomalies were extracted from the Bureau of Meteorology’s observational data (using a 1961-1990 baseline) and compared with historical model simulations using the same baseline.
Table 1. Models and simulations used in this analysis. In bold are model simulations used for calculating natural baseline and the 1.5°C and 2°C warmer worlds. Other historical simulations were only used for model evaluation.
Twelve models passed the evaluation test for Australia and were used in the attribution analysis.
The change in the likelihood of hot Julys (above the previous record set in 1975) was computed between a natural model ensemble, based on historicalNat simulations, and an all-forcings current-world ensemble, based on RCP8.5 for 2006-2026. The Risk Ratio for hot Julys in the current world relative to the natural world was calculated using all available model simulations and on 10000 bootstrapped sub-ensembles (50% of paired complete simulations) so the sampling uncertainty could be quantified. The reported result is the conservative 10th percentile value of the change in likelihood.
The likelihood of hot Julys, similar to the event we have seen in 2017, is estimated for two future scenarios. Model ensembles representing global warming of 1.5°C and 2°C above a pre-industrial baseline were extracted from the model projections (following King et al. (2017)). These ensembles were used to examine the likelihood of high July temperatures at the policy-relevant Paris global warming targets. The 90% confidence intervals shown are derived from bootstrapped subsamples of the model simulations.
The change in the likelihood of hot Julys (above the previous record set in 1975) was computed between a natural model ensemble, based on historicalNat simulations, and an all-forcings current-world ensemble, based on RCP8.5 for 2006-2026. The Risk Ratio for hot Julys in the current world relative to the natural world was calculated using all available model simulations and on 10000 bootstrapped sub-ensembles (50% of paired complete simulations) so the sampling uncertainty could be quantified. The reported result is the conservative 10th percentile value of the change in likelihood.
The likelihood of hot Julys, similar to the event we have seen in 2017, is estimated for two future scenarios. Model ensembles representing global warming of 1.5°C and 2°C above a pre-industrial baseline were extracted from the model projections (following King et al. (2017)). These ensembles were used to examine the likelihood of high July temperatures at the policy-relevant Paris global warming targets. The 90% confidence intervals shown are derived from bootstrapped subsamples of the model simulations.
