BAMOS Vol 30 No. 2 2017 | Page 18

18 BAMOS June 2017 Article Can Regional Climate Models simulate heatwaves for New South Wales and the Australian Capital Territory? An overview of Gross et al. 2017 Mia Gross Climate Change Research Centre and ARC Centre of Excellence for Climate System Science, University of New South Wales, Sydney, Australia, [email protected]. Introduction Extreme heat events are one of the most perceivable aspects of climate change, causing substantial damage to both people and infrastructure. Heatwaves, for example, have clear links to human mortality and morbidity. In 2003, at least 25,000 people died from a heatwave across Europe (D’ippoliti et al., 2010), while the 2009 heatwave in Victoria before the Black Saturday bushfires resulted in 374 excess deaths (Victorian Government Department of Human Services, 2009). Alarmingly, many studies point to future increases in heatwave frequency, intensity and duration (e.g. Alexander and Arblaster, 2009; Cowan et al., 2014). These projections are made possible through the use of Global Climate Models (GCMs). Robust conclusions rely on the ability of GCMs to realistically simulate the past and current climate. While GCMs can reasonably simulate temperature extremes from observations on a global and continental scale (e.g. Alexander and Arblaster, 2009), they lack the finer scale detail needed for regional applications. Robust regional scale projections of how heatwaves might change are crucial, as characteristics of heat-health relationships can be dependent on location (e.g. Curriero et al., 2002). Regional Climate Models (RCMs) provide simulations at a finer scale which is more appropriate for investigating regional changes in heatwaves. Using Regional Climate Models to simulate heatwaves To help reduce uncertainties that are inherent in the models, multi-model ensembles can be used which sample a range of different physical parameterisations. Multi-model RCM simulations are provided in the New South Wales/Australian Capital Territory Regional Climate Modelling (NARCliM) project (Evans et al., 2014). NARCliM implements a technique known as dynamical downscaling, which uses RCMs to downscale coarse resolution GCMs to much finer resolutions. However, biases within the GCMs are inherited by the RCMs in the process. NARCliM provides simulations from a 12-member RCM ensemble that includes four GCMs which are downscaled using three different configurations of the Weather Research and Forecasting (WRF) RCM. The simulations are available in both raw model output, as well as bias-corrected output — ­­ a term used to describe model output that has been adjusted to reduce the inherited biases and resemble the observations more closely. In our study, we used NARCliM simulations, alongside observational data, to investigate if uncorrected and bias-corrected RCMs can represent heatwave characteristics, using measures that are relevant to the heat-health relationship. We combined an increasingly used heatwave index known as the Excess Heat Factor (EHF) (Nairn and Fawcett, 2013), which accounts for acclimatisation and heat stress in its calculation, with standard heatwave metrics related to frequency, intensity and duration (e.g. as in Perkins and Alexander, 2013). In this methodology, heatwaves are defined as events which last for at least three consecutive days. The EHF-derived indices were calculated for both uncorrected and bias-corrected NARCliM model output for the recent climate (i.e. 1990–2009), to be evaluated against observational data. Observations from 25 stations located in New South Wales and the Australian Capital Territory were selected from the Australian Climate Observations Reference Network—Surface Air Temperature (ACORN-SAT) dataset (Trewin, 2013, see Fig. 1). ACORN-SAT represents a high-quality daily temperature dataset of station-based observations, however, the RCM simulations are based on gridded output. We therefore chose to additionally include gridded data from the Australian Water Availability Project (AWAP) (Jones et al., 2009). The grid points closest to the ACORN-SAT stations were then used to allow comparison between the datasets. The results of the RCM simulations of the recent climate could then be used to infer results of bias-corrected and uncorrected model simulations of future changes in heatwaves for the period 2060–2079. The importance of bias-correction and metric selection Overall, while no individual NARCliM simulation of the EHF- derived heatwave indices perfectly resembles the observations, the ensemble of uncorrected output performs reasonably well against the observations. Simulations of stations further inland