1. Introduction
An exceptionally severe heatwave event struck southeastern Australia during the two weeks of 27 January–8 February 2009 (National Climate Centre 2009). A widespread region in the southeast Australian states of South Australia, Victoria, and Tasmania experienced record-breaking hot conditions, with anomalous daily maximum temperature reaching 12°–15°C above the climatological (1971–2000) normal across Victoria and southern South Australia (National Climate Centre 2009). The extremely warm conditions peaked on 7 February 2009 when many weather stations in Victoria set their highest temperature records that still stand to date. The intense and prolonged heatwave conditions led to major health concerns and infrastructure failure (Bettio et al. 2019) and was followed by wildfire events across Victoria, commonly referred to as the Black Saturday bushfires, which caused numerous fatalities and significant economic loss (Victorian Bushfires Royal Commission 2010). Considering the severity of the Black Saturday heatwave event and its devastating impacts, a comprehensive understanding of the underlying physical processes and a skillful forecast for such an extreme event, at least a week in advance, would be of substantial value to help communities and emergency services prepare in advance (e.g., Perkins-Kirkpatrick et al. 2016).
Multiple climate drivers on different time scales are shown to be associated with southeast Australian heatwaves, which often last for 1–2 days but sometimes can even persist for 10 days or longer (King and Reeder 2021). These longer-lived heatwaves in southeast Australia are associated with the anomalous northerly flow, with high pressure systems to the east (often over the Tasman Sea) is the most common synoptic feature (Boschat et al. 2015). On monthly to seasonal time scales, climate drivers such as positive Indian Ocean dipole (IOD; Cai et al. 2009), La Niña (Parker et al. 2014; Perkins et al. 2015), and transient circulation anomalies (Pezza et al. 2012) can act to either precondition the landscape through drying, or directly influence through changes in the location, intensity, and movement of synoptic systems (McKay et al. 2022). Marshall et al. (2014) showed that southeast Australian heatwaves tend to be enhanced by the negative phase of the Southern Annular Mode (SAM), persistent anticyclones over the Tasman Sea, and phases 2–3 of the Madden–Julian oscillation.
The summer of 2008/09 followed below-normal rainfall over the southeast Australian region, associated with a positive IOD (Cai et al. 2009). The dry conditions in winter to spring 2008 were themselves part of a much longer dry spell in the region (commonly referred to as the Millennium Drought), which commenced in the mid- to late-1990s (e.g., Cai et al. 2014). The drought conditions resulted in low soil moisture over southeast Australia, which likely amplified the intensity of the heatwave (Kala et al. 2015; Perkins et al. 2015). For the dynamics of the event, influences from both the tropics and extratropics were important. Parker et al. (2013) indicated that upper tropospheric anticyclonic circulation associated with tropical cyclone outflow reinforced the heatwave event. This anticyclonic circulation and associated surface warming were supported by tropospheric impacts from Antarctic warming at the upper levels and quasi-stationary anticyclones over the Southern Ocean (Fiddes et al. 2016).
In 2008, the empirical seasonal forecast scheme (Drosdowsky and Chambers 2001) of the Australian Bureau of Meteorology (hereafter referred to as the Bureau) suggested a high probability of unusually warm conditions likely in the summer of 2008/09 over southeastern Australia (http://www.bom.gov.au/climate/ahead/archive/temperature/20081126T.shtml), and the dynamical weather forecast also reasonably captured the heatwave conditions. Despite the existing skill to predict extreme temperatures using numerical weather prediction models, the delivery of a specific heatwave warning system was not available at the time of the Black Saturday event and it only started in 2014 (Bettio et al. 2019) to support the Australian community, partly motivated by the impact of the extreme Black Saturday heatwave event. The national heatwave service along with Emergency Management Australia has shown success in reducing morbidity and mortality due to heatwave events by providing extreme heatwave forecasts, such as the January 2014 heatwave event over southern Australia (Nitschke et al. 2016; Bettio et al. 2019). With the advancement of the weather and subseasonal forecasting capability, there is a growing demand to improve heatwave prediction skills and reduce the uncertainty in heatwave forecasts.
The other important issue with advancing prediction capability and risk management associated with extreme heat events is understanding the role of climate change in them. Recent climate attribution studies (e.g., Lewis and King 2015; Hope et al. 2016) suggested that anthropogenic climate change significantly impacts extreme warming events in Australia and globally. The frequency, duration, and intensity of extreme heatwave events across Australia have substantially increased over the last few decades (e.g., Perkins et al. 2012; King et al. 2017; Bureau of Meteorology and CSIRO 2020). Therefore, a comprehensive heatwave forecast assessment and near-real-time climate attribution to the atmospheric greenhouse gas (GHG) increase would be of considerable value in advancing prediction capability and risk management.
As an improved understanding of an extreme event and its skillful forecasts can reduce the impacts of deadly weather–climate extremes like the Black Saturday heatwave (Hope et al. 2022), we explore the heatwave event using a state-of-the-art multiweek to seasonal climate forecasting system of the Bureau, the Australian Community Climate and Earth-System Simulator-Seasonal prediction system, version 1 (ACCESS-S1; Hudson et al. 2017), and assess its prediction skill against available observations. Having established the prediction skill of the event, we explore its climate drivers including the role of climate change in amplifying the extremely warm conditions. To estimate the role of the GHG increase over the twentieth century in the Black Saturday heatwave event, we adopt an initialized climate attribution method (Wang et al. 2021). The details of the subseasonal forecast system, attribution method, and observed dataset are described in section 2, results are discussed in sections 3 and 4, and the primary conclusions of this study are provided in section 5.
2. Prediction system, attribution experiment, and observed dataset
ACCESS-S1 is based on the Met Office Unified Model Global Seasonal forecast system version 5, using the Global Coupled model configuration 2 (GC2; MacLachlan et al. 2015). It was the operational subseasonal to seasonal (S2S) forecasting system at the Bureau from 2018 to 2021 and has been used to support the attribution research. The horizontal resolution of the atmospheric model is ∼60 km in the midlatitudes, and it has 85 vertical levels with a well-resolved stratosphere (35 levels above 18 km; Walters et al. 2017). The ocean model, based on Nucleus for European Modeling of the Ocean (NEMO ORCA25; Madec 2008; Megann et al. 2014) has a 25-km horizontal resolution and 70 vertical levels (∼1-m resolution in the top 10 m). For the hindcast dataset analyzed here, the atmospheric initial conditions are derived by interpolating the ERA-Interim (Dee et al. 2011) directly onto the model grid. The oceanic initial conditions are obtained from the Forecast Ocean Assimilation Model run at the Met Office (Blockley et al. 2014), which uses three-dimensional variational assimilation of available temperature, salinity, and sea level observations. To represent the initial condition uncertainty, an additional 10-member ensemble is produced by adding or subtracting scaled perturbations for zonal and meridional winds (u, υ), air temperature (T), specific humidity (q), and surface pressure from the ERA-Interim dataset to the unperturbed (central member) atmospheric initial conditions (Lim et al. 2016; Hudson et al. 2017). In total, 11-member (an unperturbed central plus 10 perturbed) ensemble forecasts are available from the 1st, 9th, 17th, and 25th of each month during 1990–2012 with atmospheric CO2 concentration set to the observed values in every hindcast year.
To understand the impact of the increased atmospheric CO2 concentration since the 1900s on the Black Saturday heatwave event, we also reran the initialized ensemble forecast with a modified ocean–atmosphere mean state from a counterfactual world with reduced CO2 concentration in the atmosphere (297 ppm, equivalent to ∼1905 CO2 level). Our null hypothesis is that the increase of atmospheric CO2 does not make any difference, and in testing this hypothesis, internally driven variations in the current and “Low CO2” forecasts are assumed to be identical (e.g., Hope et al. 2016). Based on this assumption, the initial conditions for the “Low CO2 world” are derived by subtracting the mean climatology difference between the present period (1990–2020) and the past representative period (1861–1950) from the original initial conditions, following the method discussed in Wang et al. (2021). For this study, we obtain the ocean and atmospheric mean states for the two epochs from five-member ensemble long simulations of HadGEM3-GC2 (Williams et al. 2015). The counterfactual ocean initial conditions are then derived by subtracting three-dimensional ocean temperature and salinity differences of the current and past periods from the original ocean initial conditions, and the corresponding atmospheric initial conditions are obtained by subtracting T and q differences from the actual atmospheric T and q initial conditions (see Fig. S1 in the online supplemental material for the difference in ocean–atmospheric conditions between two epochs). The other variables (e.g., wind and relative humidity) rapidly respond to these changes in the first few days of the integration (not shown). We apply the same perturbations to the unperturbed “Low CO2” atmospheric initial conditions as in the current central member and generate an 11-member ensemble. Our 11-member ensemble “Low CO2” forecast experiment is initialized on 17 January 2009 with a modified ocean-atmospheric state and reduced atmospheric CO2 concentration. The systematic errors of lead-time-dependent forecasts are corrected by subtracting forecast climatology from the 2009 forecasts as a function of forecast initialization dates and lead times, and the resultant forecast anomalies are verified against the observed anomalies. The significance of the difference between the current and “Low CO2” forecasts is calculated using a two-tailed paired t test.
The near-surface maximum temperature forecast is evaluated against the observed daily maximum temperature (Tmax) dataset from the Australian Water Availability Project (AWAP; Jones et al. 2009). These data are available across the Australian landmass back to 1911 with a horizontal resolution of 5 km × 5 km. We also use global monthly Extended Reconstructed Sea Surface Temperature (ERSST; Huang et al. 2017), version 5, at 2° longitude × 2° latitude resolution. The daily mean sea level pressure (MSLP) and 850-hPa zonal and meridional wind analyses are taken from ERA5 (Hersbach et al. 2020), which is available globally on a 0.25° × 0.25° grid from January 1959 to the present.
3. Observed characteristics of the Black Saturday heatwave event
a. Synopsis of the heatwave event
We begin our assessment of the Black Saturday heatwave by examining the observed temperature and associated large-scale circulation anomalies (i.e., long-term climatology and seasonal cycle removed) during the event. Anomalously high temperatures are evident over most of southern Australia and Tasmania in the observed Tmax anomalies during the period of 27 January–8 February 2009 (Fig. 1a). As reported by the National Climate Centre (2009), the averaged Tmax anomalies in southeastern Australia for the 2 weeks were well above the long-term (1981–2010) mean, and they were extremely high over southeastern Australia where the heatwave event peaked. This heatwave event was associated with high pressure anomalies over the Tasman Sea and the southern Indian Ocean (∼60°S, east of 60°E), and low pressure anomalies over southern Australia (Fig. 1b). Both the high pressure anomalies developed about a week before the heatwave event and were quasi-stationary for about 3 weeks (Fiddes et al. 2016). Between these two quasi-stationary highs, a slow-moving intense low pressure system evolved, which was shown to be initiated by the wave amplification of an equivalent barotropic Rossby wave packet (e.g., Engel et al. 2013).
(a) AWAP Tmax anomalies (°C) during the Black Saturday heatwave event (27 Jan–8 Feb 2009) and (b) associated MSLP (contours; hPa) and 850-hPa wind (vectors) anomalies from ERA5 during the heatwave event and SST anomalies (shaded; °C) during January 2009 from ERSST dataset. The MSLP contours are drawn at 2-hPa interval and positive (negative) contours are displayed in green (pink) shades, with darker colors representing higher values; the zero contours are shown in gray.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0833.1
(a) AWAP Tmax anomalies (°C) during the Black Saturday heatwave event (27 Jan–8 Feb 2009) and (b) associated MSLP (contours; hPa) and 850-hPa wind (vectors) anomalies from ERA5 during the heatwave event and SST anomalies (shaded; °C) during January 2009 from ERSST dataset. The MSLP contours are drawn at 2-hPa interval and positive (negative) contours are displayed in green (pink) shades, with darker colors representing higher values; the zero contours are shown in gray.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0833.1
(a) AWAP Tmax anomalies (°C) during the Black Saturday heatwave event (27 Jan–8 Feb 2009) and (b) associated MSLP (contours; hPa) and 850-hPa wind (vectors) anomalies from ERA5 during the heatwave event and SST anomalies (shaded; °C) during January 2009 from ERSST dataset. The MSLP contours are drawn at 2-hPa interval and positive (negative) contours are displayed in green (pink) shades, with darker colors representing higher values; the zero contours are shown in gray.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0833.1
The middle (500 hPa)- and upper (200 hPa)-tropospheric geopotential heights (z500 and z200, respectively) during the heatwave event show that a Rossby wave train arced across the tropical to high latitudes over the Australian domain with a persistent high pressure anomaly over the Tasman Sea (Fig. S2). The Tasman Sea high–induced anomalous subsidence and associated adiabatic warming over southeastern Australia, which promoted near-surface horizontal advection of the tropical warm air to southeast Australia.
The Rossby wave propagation is likely to be associated with the enhanced convection over northern Australia and the western Pacific (Hoskins and Karoly 1981) due to the moderate La Niña conditions in the tropical Pacific (Fig. 1b). The Oceanic Niño Index (ONI) value of −0.8°C indicates that the cooling over the Niño-3.4 region (5°S–5°N, 120°–170°W) was around the La Niña threshold (the Bureau considers −0.8°C Niño-3.4 temperature as the La Niña threshold). Another possible driving factor could be the Ningaloo Niño, that is, the positive sea surface temperature (SST) anomalies along the western Australia coast (Fig. 1b, Feng et al. 2013). Marshall et al. (2015) suggested that Ningaloo Niño can influence warming over southern Victoria. However, the weak Ningaloo Niño in January 2009 was unlikely to have contributed much to the severe heatwave event.
b. The historical context of the heatwave event and its linkage to the tropical SST
To explore the general characteristics of Australian summer Tmax variability in comparison to the Black Saturday heatwave event, we conduct the empirical orthogonal function (EOF) analysis (North et al. 1982) on the observed 14-day running-averaged daily Tmax anomalies during austral summer [December–February (DJF)] of 1979–2020 without including the 2008/09 austral summer. To eliminate high-frequency weather scale noise, a 14-day running mean is applied to the daily summertime Tmax anomalies. Figures 2a–c show the three leading eigenvectors, which are well separated from their neighboring modes [confirmed using Eqs. (24)–(26) of North et al. (1982)], and they account for more than 70% of the S2S Tmax variability in the austral summer. The leading mode (EOF 1) explains about 40% of the variance and shows a continent-wide positive loading with the maxima over the western part of the country, while the second mode (EOF 2) explains 16.5% of the variance with an east–west spatial dipole structure. EOF 3, which explains 14.2% of the variance, closely resembles the Black Saturday heatwave warming pattern shown in Fig. 1a, having a north–south dipole structure.
(a)–(c) Three leading EOF patterns and (d)–(f) their corresponding PCs of fortnightly (14-day) average AWAP Tmax anomalies. The EOF analysis is based on 1979–2020 (except 2008/09) DJF Tmax anomalies. The PCs are obtained by projecting the Tmax anomalies onto the EOFs. The variance (in %) explained by each EOF and trends in PC time series during the 1971–2020 period are mentioned at the top right of the corresponding panel.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0833.1
(a)–(c) Three leading EOF patterns and (d)–(f) their corresponding PCs of fortnightly (14-day) average AWAP Tmax anomalies. The EOF analysis is based on 1979–2020 (except 2008/09) DJF Tmax anomalies. The PCs are obtained by projecting the Tmax anomalies onto the EOFs. The variance (in %) explained by each EOF and trends in PC time series during the 1971–2020 period are mentioned at the top right of the corresponding panel.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0833.1
(a)–(c) Three leading EOF patterns and (d)–(f) their corresponding PCs of fortnightly (14-day) average AWAP Tmax anomalies. The EOF analysis is based on 1979–2020 (except 2008/09) DJF Tmax anomalies. The PCs are obtained by projecting the Tmax anomalies onto the EOFs. The variance (in %) explained by each EOF and trends in PC time series during the 1971–2020 period are mentioned at the top right of the corresponding panel.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0833.1
The corresponding principal component (PC) time series are derived by projecting the observed fortnightly (14-day) averaged Tmax anomalies during the 1911–2020 DJF period onto the observed EOF eigenvectors, and the PC time series are shown along with their trends during the 1971–2020 DJF in Figs. 2d–f. The leading two PCs show moderate amplitudes in 2009, and they have linear upward trends of 0.16° and 0.18°C decade−1, respectively (both are significant at the 1% level as assessed by a two-tailed Student’s t test with 25 degrees of freedom) in the recent 50 years (1971–2020). In contrast, the PC3 time series shows some multidecadal variation, but it has no obvious linear trend. The PC3 shows its second largest amplitude (∼3σ) in 2009, confirming that the Black Saturday heatwave event was an extreme event largely associated with this mode of Tmax variability.
We build a multiple regression model by using the three observed PC time series from 27 January to 8 February 1990–2012 (except 2009) as the predictors and reconstructed the 2009 Tmax anomalies by plugging the observed PC values during the heatwave event (Fig. 3a) into the model. The reconstructed Tmax anomalies during the heatwave event (Fig. 3b) match well with the observation. While the extreme amplitude of the PC3 appears to have determined the observed meridional dipole Tmax anomalies over the mainland of Australia, the contribution of PC1 to the extraordinary warming in southern Australia (30°–37°S, 125°–155°E) is similar to PC3 (33% and 36%, respectively), and the contribution of PC2 is found to be less but still substantial (19%, Fig. 3c).
(a) Normalized observed PCs from 27 Jan to 8 Feb 2009, (b) reconstruction of Tmax anomalies (°C) during the Black Saturday heatwave event using 1990–2012 (except 2009) observed PC1–3 as the predictors, and (c) contributions of the individual PCs to the observed Tmax anomalies (in %) over southern Australia (30°–37°S, 125°–155°E).
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0833.1
(a) Normalized observed PCs from 27 Jan to 8 Feb 2009, (b) reconstruction of Tmax anomalies (°C) during the Black Saturday heatwave event using 1990–2012 (except 2009) observed PC1–3 as the predictors, and (c) contributions of the individual PCs to the observed Tmax anomalies (in %) over southern Australia (30°–37°S, 125°–155°E).
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0833.1
(a) Normalized observed PCs from 27 Jan to 8 Feb 2009, (b) reconstruction of Tmax anomalies (°C) during the Black Saturday heatwave event using 1990–2012 (except 2009) observed PC1–3 as the predictors, and (c) contributions of the individual PCs to the observed Tmax anomalies (in %) over southern Australia (30°–37°S, 125°–155°E).
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0833.1
To see if any of these leading modes of S2S Tmax variability are associated with the oceanic conditions, which could provide predictability beyond the weather time scale, we regress the observed January SST anomalies to the January PCs (except 2009 January) and scale the resultant regression coefficients by the magnitude of the corresponding PCs during January 2009. It is interesting to note in Fig. 4 that the reconstructed SST by PC3 well matches the observed January 2009 SST anomalies in the tropical Indo-Pacific region with a cooling in the central-eastern Pacific and warming in the western Pacific, the South Pacific convergence zone region, and along the Western Australia coast. The pattern correlation between observed and reconstructed SST anomalies in the tropical Indo-Pacific region (20°–20°N, 30°E–60°W) is 0.7, suggesting a likely association between the heatwave event and the La Niña condition in the tropical Pacific as in Fig. 1b. Previous studies (Parker et al. 2014; Perkins et al. 2015; Hope et al. 2022) also showed a role of La Niña in the heatwave event over southeastern Australia. Compared to PC3, the other two PCs are unrelated to the SST anomalies except for a weak association between PC1 and the warming (cooling) in the Tasman Sea (off the southwest Australian coast), which were not prominent during the heatwave event.
Observed January SST anomalies (°C) regressed onto observed (a) PC1, (b) PC2, and (c) PC3. All the regression coefficients are scaled by corresponding PC amplitude during January 2009.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0833.1
Observed January SST anomalies (°C) regressed onto observed (a) PC1, (b) PC2, and (c) PC3. All the regression coefficients are scaled by corresponding PC amplitude during January 2009.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0833.1
Observed January SST anomalies (°C) regressed onto observed (a) PC1, (b) PC2, and (c) PC3. All the regression coefficients are scaled by corresponding PC amplitude during January 2009.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0833.1
4. Prediction of the Black Saturday heatwave event
We first assess 14-day-mean Tmax prediction skill during January–February of the 23-yr (1990–2012) hindcast period in ACCESS-S1 with four start times (the 1st, 9th, 17th, and 25th of January and February; Figs. 5a–d). The prediction skill at 8-day intervals is measured by the correlation between the predicted and the observed 14-day-averaged Tmax anomalies as a function of forecast lead time. The Tmax prediction skill is found to be generally higher over the western part of the country and, to a lesser degree, over southeastern Australia at the shortest lead time. As expected, the Tmax prediction skill reduces with the longer lead times, although a moderate skill (i.e., the statistically significant correlation at a 10% level, which is 0.36 with 23 samples) is noted up to 16-day lead time.
(a)–(d) The 2-week mean Tmax prediction skill (correlation between observed and predicted 14-day-averaged Tmax anomalies; higher correlation value represents better prediction skill) at 8-day initialization intervals during January–February 1990–2012 in ACCESS-S1. (e)–(h) ACCESS-S1 Tmax (°C; shaded) and MSLP (hPa; contours; contour interval: 2 hPa) forecast anomalies for 27 Jan–8 Feb 2009 from available January forecasts start dates. (i) Observed and predicted normalized Tmax anomalies over southeastern Australia (30°–37°S, 140°–150°E) for the 27 Jan–8 Feb period in 1990–2012.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0833.1
(a)–(d) The 2-week mean Tmax prediction skill (correlation between observed and predicted 14-day-averaged Tmax anomalies; higher correlation value represents better prediction skill) at 8-day initialization intervals during January–February 1990–2012 in ACCESS-S1. (e)–(h) ACCESS-S1 Tmax (°C; shaded) and MSLP (hPa; contours; contour interval: 2 hPa) forecast anomalies for 27 Jan–8 Feb 2009 from available January forecasts start dates. (i) Observed and predicted normalized Tmax anomalies over southeastern Australia (30°–37°S, 140°–150°E) for the 27 Jan–8 Feb period in 1990–2012.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0833.1
(a)–(d) The 2-week mean Tmax prediction skill (correlation between observed and predicted 14-day-averaged Tmax anomalies; higher correlation value represents better prediction skill) at 8-day initialization intervals during January–February 1990–2012 in ACCESS-S1. (e)–(h) ACCESS-S1 Tmax (°C; shaded) and MSLP (hPa; contours; contour interval: 2 hPa) forecast anomalies for 27 Jan–8 Feb 2009 from available January forecasts start dates. (i) Observed and predicted normalized Tmax anomalies over southeastern Australia (30°–37°S, 140°–150°E) for the 27 Jan–8 Feb period in 1990–2012.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0833.1
To examine how well the Black Saturday heatwave event is predicted in ACCESS-S1, we examine the 27 January–8 February 2009 Tmax and MSLP forecast anomalies in Figs. 5e–h for the initialized forecasts from 25, 17, 9, and 1 January of that year (i.e., at 2-, 10-, 18-, and 26-day lead times). The plots are designed to be directly comparable to the observed Tmax anomalies in Fig. 1a. It should be noted that the forecast anomalies are derived by removing the 23-yr climatological seasonal cycle of the same start dates and lead times. The Tmax anomalies during the heatwave period are predictable at the lead times of up to 10 days (i.e., forecast initialized on 17 January) as demonstrated by high pattern correlation of 0.81 and 0.78 at 2-day and 10-day lead times, respectively (Figs. 6c,d), but the predicted Tmax amplitude is substantially weaker than the observed, even at the shortest lead time. The underestimation of the observed anomalies in the ensemble-mean forecast is largely due to the model’s systematic bias in producing Tmax variability over southeastern Australia. ACCESS-S1 forecast skills for the 2-week mean DJF Tmax converge to the climatology at longer lead times, suggesting a lack of predictive skill at a long forecast lead over this region (Fig. S3). The same bias likely causes the underestimation of the 2009 heatwave event.
(a) Normalized PCs during the Black Saturday heatwave event (27 Jan–8 Feb 2009) in observation and ACCESS-S1 forecast with start dates of 25, 17, 9 and 1 Jan of that year, respectively. The vertical whiskers indicate the upper and lower bounds of the ensemble PC forecast. (b)–(d) Observed and ACCESS-S1 predicted Tmax anomalies from 25 to 17 Jan start dates during the heatwave event. (e)–(g) Reconstruction of observed and ACCESS-S1 predicted Tmax (°C; shaded) anomalies with 25 and 17 Jan start dates from 27 Jan to 8 Feb 2009 using a combination of 3 PCs during 1990–2012 (except 2008/09). The pattern correlation between observation and forecast in (b)–(d), observed/predicted Tmax anomalies, and their corresponding reconstruction in (e)–(g) are shown at the bottom-left corner of each panel.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0833.1
(a) Normalized PCs during the Black Saturday heatwave event (27 Jan–8 Feb 2009) in observation and ACCESS-S1 forecast with start dates of 25, 17, 9 and 1 Jan of that year, respectively. The vertical whiskers indicate the upper and lower bounds of the ensemble PC forecast. (b)–(d) Observed and ACCESS-S1 predicted Tmax anomalies from 25 to 17 Jan start dates during the heatwave event. (e)–(g) Reconstruction of observed and ACCESS-S1 predicted Tmax (°C; shaded) anomalies with 25 and 17 Jan start dates from 27 Jan to 8 Feb 2009 using a combination of 3 PCs during 1990–2012 (except 2008/09). The pattern correlation between observation and forecast in (b)–(d), observed/predicted Tmax anomalies, and their corresponding reconstruction in (e)–(g) are shown at the bottom-left corner of each panel.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0833.1
(a) Normalized PCs during the Black Saturday heatwave event (27 Jan–8 Feb 2009) in observation and ACCESS-S1 forecast with start dates of 25, 17, 9 and 1 Jan of that year, respectively. The vertical whiskers indicate the upper and lower bounds of the ensemble PC forecast. (b)–(d) Observed and ACCESS-S1 predicted Tmax anomalies from 25 to 17 Jan start dates during the heatwave event. (e)–(g) Reconstruction of observed and ACCESS-S1 predicted Tmax (°C; shaded) anomalies with 25 and 17 Jan start dates from 27 Jan to 8 Feb 2009 using a combination of 3 PCs during 1990–2012 (except 2008/09). The pattern correlation between observation and forecast in (b)–(d), observed/predicted Tmax anomalies, and their corresponding reconstruction in (e)–(g) are shown at the bottom-left corner of each panel.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0833.1
The extremity of the Tmax forecast during the 27 January–8 February period is highlighted by comparing the 2009 predictions with all other forecasts in the 23-yr hindcast set. Despite the underestimation of the observed warm anomalies in the forecast, the predicted positive Tmax anomalies over southeastern Australia from the 25 and 17 January 2009 start times are the largest in the 23-yr hindcasts (Fig. 5i), evident in the realistic amplitude of the normalized forecasts. Both observed and predicted Tmax are about three standard deviations (i.e., above the 95th-percentile confidence level for the index). It indicates that the extremity of the heatwave event is skillfully predicted by ACCESS-S1 as in the previous-generation Predictive Ocean Atmosphere Model for Australia (POAMA; Hudson et al. 2011).
In the forecast, the weaker warming over southeastern Australia is likely related to underestimated MSLP anomalies over the Tasman Sea and the southern Australian region. Both the Tmax and MSLP anomalies are further underestimated in longer lead forecasts, and the model shows no skill to predict the heatwave condition in the forecast initialized on 9 and 1 January 2009. Consistent with previous studies (e.g., Hudson et al. 2011), the prediction of individual heatwave events (days to multiweek) in ACCESS-S1 is dependent on reliable forecasting of the weather and synoptic events, for which predictability is limited to a week or two. It implies that the atmospheric internal processes predominantly drove the Black Saturday heatwave event, although the climate drivers including the preceding drought and the La Niña condition in the Pacific had a role in the event.
In the previous subsection, we showed that the leading three PCs of Australian summer Tmax positively contributed to the Black Saturday heatwave event and especially PC3, which is historically related to the tropical Pacific SST variations, exhibited an extraordinarily positive amplitude during the heatwave event. To examine the ability of ACCESS-S1 in capturing these relative contributions of the leading three PCs for the forecast of this event, we compare the observed and predicted PCs from 27 January to 8 February 2009. The ensemble-mean forecast PCs are obtained by projecting the predicted Tmax anomalies onto the observed EOF patterns shown in Figs. 2a–c. The predicted PC amplitudes in the forecasts initialized on all of the four different start dates of January 2009 are compared to observations during the heatwave event (Fig. 6a). The greater amplitude of the observed PC3 relative to the other two PCs during the heatwave event is well captured in the forecasts from the beginning of the month, although the predicted PC3 amplitude significantly decreases in the longer-lead forecasts. The ensemble PC forecast shows both above- and below-normal temperatures and the spread generally increases with longer lead times. However, the ensemble predictions at the short lead times are found to be skewed toward the above-normal temperatures and this skewness is more prominent for PC3 predictions. In contrast to PC3 prediction, ACCESS-S1 does not skillfully predict the contributions of the leading two modes, which appears to limit the model’s capability to predict the heatwave event beyond 10 days.
The correlation between observed and predicted PCs during the available hindcast years (1990–2012) are shown in Table 1 (and their time series are provided in displayed in Fig. S4). Consistent with the results in Fig. 5, ACCESS-S1 shows no skill to predict the Tmax PCs during the 13 days (27 January–8 February) at longer lead times with early January forecast start dates. Only a reasonably high skill (correlation > 0.5) is noted at about a 10-day lead for the forecast initialized on 17 January. Notably, PC3 is predictable at the longer lead time beyond 10 days, which is consistent with predicted PC3 for 2009 (Fig. 6a). The greatest prediction skill for PC3 achieved with a maximum correlation of 0.81 at 2-day lead time from 25 January and it is likely to have positively contributed to the skillful prediction of the spatial details of the heatwave event.
ACCESS-S1 skill to predict the three leading PCs (correlation between observed and predicted PCs) from 27 Jan to 8 Feb for all available January start dates (1990–2012).
The prediction of the leading three modes of Tmax and the 2009 austral summer heatwave event is further demonstrated by reconstructing the forecast Tmax anomalies of the heatwave event with a multiple linear regression model using three PC values in the forecast as the predictors. The synthesized Tmax anomalies are compared with the observed (Fig. 6b) and forecast anomalies (Figs. 6c,d), and their pattern correlations with the predicted Tmax anomaly patterns are mentioned at the bottom-left corner of each panel in the bottom row (Figs. 6e–g). The reconstructed Tmax anomalies appear to closely match with the 17 and 25 January forecasts of Tmax anomalies with a pattern correlation of 0.89 and 0.76, respectively. This confirms that the model’s capability to predict the leading three modes of Tmax variability results in the skillful forecast of the heatwave event at ∼10-day lead time.
We also examine the association of individual PCs with SST in the forecasts initialized on 25 and 17 January (Fig. S5). ACCESS-S1 largely captures the observed association between the SST anomalies and PC3. This association apparently explains the relatively long-lead prediction skills of PC3 up to ∼20 days.
5. Role of atmospheric CO2 increase on Black Saturday heatwave event
Noting the prediction skills of ACCESS-S1 for the austral summer Tmax variability and the Tmax anomalies for the 2009 heatwave event, this event provides an excellent opportunity for an initialized attribution experiment. An attribution forecast experiment is conducted on the Black Saturday heatwave event with the “Low CO2 world” configuration as described in section 2. We initialize this forecast experiment on 17 January 2009, and compare the results with the current (with high atmospheric CO2 levels) forecast initialized on the same start date. Based on our earlier assessment, a forecast initialized on a closer start date to the event generally provides a more skillful prediction, but we start the forecast experiment about 10 days before the event to bring the atmosphere and the ocean into a better balance under the counterfactual cold climate and allowing for the fast response of the atmosphere to stabilize the change in CO2 concentration. The Tmax and MSLP during the Black Saturday heatwave event in observation, current, and “Low CO2” forecasts are presented in Fig. S6. The observed warming and associated MSLP patterns are largely captured in current and “Low CO2” forecasts, although the predicted Tmax anomalies are weaker than observed. The areal mean temperature over southeastern Australia (30°–37°S, 140°–150°E; the box in Fig. 7a) is found to be 33.2°C in the current forecast and 30.6°C in the “Low CO2” forecast; that is, the current forecast over the region is about 2.6°C warmer than the counterfactual world.
Estimated contribution of atmospheric CO2 increase to the Black Saturday heatwave, (a) predicted Tmax (shaded; °C) and MSLP (contours; 1-hPa contour interval) difference from 27 Jan to 8 Feb 2009 between current and the “Low CO2” forecasts. The significant Tmax difference at the 5% level is stippled based on a paired t test. (b) Box-and-whisker plot showing the statistical difference for the impact of atmospheric CO2 increase on warming over southeastern Australia [30°–37°S, 140°–150°E; region marked with a rectangle in (a)] using the 11-member paired forecast ensemble. The bottom and top of the box represent the 25th and 75th percentiles, respectively, the horizontal line inside the box represents the mean deterministic forecast difference relative to the “Low CO2” forecast, and the whiskers show the nonoutlier range of the difference between current and the “Low CO2” ensemble forecasts. (c) Probability density functions (PDFs) of 27 Jan–8 Feb 2009–averaged 11-member predicted Tmax over southeastern Australia (30°–37°S, 140°–150°E) from current and the “Low CO2” forecasts. The median value of the temperature distributions is shown in vertical dashed lines.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0833.1
Estimated contribution of atmospheric CO2 increase to the Black Saturday heatwave, (a) predicted Tmax (shaded; °C) and MSLP (contours; 1-hPa contour interval) difference from 27 Jan to 8 Feb 2009 between current and the “Low CO2” forecasts. The significant Tmax difference at the 5% level is stippled based on a paired t test. (b) Box-and-whisker plot showing the statistical difference for the impact of atmospheric CO2 increase on warming over southeastern Australia [30°–37°S, 140°–150°E; region marked with a rectangle in (a)] using the 11-member paired forecast ensemble. The bottom and top of the box represent the 25th and 75th percentiles, respectively, the horizontal line inside the box represents the mean deterministic forecast difference relative to the “Low CO2” forecast, and the whiskers show the nonoutlier range of the difference between current and the “Low CO2” ensemble forecasts. (c) Probability density functions (PDFs) of 27 Jan–8 Feb 2009–averaged 11-member predicted Tmax over southeastern Australia (30°–37°S, 140°–150°E) from current and the “Low CO2” forecasts. The median value of the temperature distributions is shown in vertical dashed lines.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0833.1
Estimated contribution of atmospheric CO2 increase to the Black Saturday heatwave, (a) predicted Tmax (shaded; °C) and MSLP (contours; 1-hPa contour interval) difference from 27 Jan to 8 Feb 2009 between current and the “Low CO2” forecasts. The significant Tmax difference at the 5% level is stippled based on a paired t test. (b) Box-and-whisker plot showing the statistical difference for the impact of atmospheric CO2 increase on warming over southeastern Australia [30°–37°S, 140°–150°E; region marked with a rectangle in (a)] using the 11-member paired forecast ensemble. The bottom and top of the box represent the 25th and 75th percentiles, respectively, the horizontal line inside the box represents the mean deterministic forecast difference relative to the “Low CO2” forecast, and the whiskers show the nonoutlier range of the difference between current and the “Low CO2” ensemble forecasts. (c) Probability density functions (PDFs) of 27 Jan–8 Feb 2009–averaged 11-member predicted Tmax over southeastern Australia (30°–37°S, 140°–150°E) from current and the “Low CO2” forecasts. The median value of the temperature distributions is shown in vertical dashed lines.
Citation: Journal of Climate 36, 19; 10.1175/JCLI-D-22-0833.1
Figure 7a shows the Tmax difference between the current forecast and the “Low CO2” forecast experiment for the Black Saturday heatwave event. The higher temperature is apparent in the current forecast over eastern Australia with the Tmax difference up to 4°C in some parts of southeastern Australia. The Tmax difference is significant at the 5% level as assessed by a two-tailed paired 11-sample Student’s t test for each forecast set. With an east–west dipole Tmax difference pattern over the continent, this atmospheric CO2-driven temperature structure has some resemblance to the Tmax EOF2 (Fig. 2b). It is interesting to note that observed EOF2 time series (PC2) has a statistically significant linear warming trend (Fig. 2e). The MSLP difference between the current and “Low CO2” forecasts indicates that the low pressure anomalies over southern Australia are strengthened and extended to the Great Australian Bight in the current forecast, promoting advection of the warm continental air to southern Australia. As observed, the predicted hottest period was between 27 January and 8 February (10–22 lead days) in the 17 January forecast. Most ensemble members verified the timing of the heatwave event very well, suggesting the onset of the event was quite predictable (Fig. S7).
The role of atmospheric CO2 increase on the Black Saturday heatwave event is also estimated using 11 paired ensemble member forecasts. Generally, the current forecast ensemble members predict warmer conditions relative to “Low CO2” ensemble predictions (Fig. S7). To quantify this difference, we consider the same area over southeastern Australia (30°–37°S, 140°–150°E) and examine the various statistics of the Tmax difference between current and “Low CO2” predictions over the region (Fig. 7b). Out of 11 paired ensemble members, only one paired member shows lower temperature over southeastern Australia in the current forecast, while the remaining 10 paired members predict higher temperature due to the CO2 increase in the current forecast. To account for the uncertainty in the result, we calculate the lower, upper quartiles, and the ensemble-mean values of the Tmax difference between current and “Low CO2” forecasts using all the ensemble members [
To compare the occurrence of high temperatures over southeastern Australia in the current and “Low CO2” ensemble predictions, we calculate the probability density functions (PDFs) of predicted Tmax over southeastern Australia (Fig. 7c). For each 11-member forecast over southern Australia (sample size: 11 ensemble members × 13 latitudes × 12 longitudes = 1716), 27 January–8 February 2009 averaged Tmax values are clustered into 50 equal-width bins between 25° and 45°C. A clear warm shift is evident in the current forecast with a mean difference of about 2.65°C, which aligns with the Tmax difference noted in Fig. 7a. This result suggests that atmospheric CO2 increase positively contributed to the extremely warm conditions during the Black Saturday heatwave event.
6. Conclusions
We have examined southeastern Australia’s heatwave event in 2009 using observation, reanalysis datasets, and the Bureau’s state-of-the-art ocean–atmosphere coupled S2S prediction system, ACCESS-S1. Consistent with previous studies (e.g., Fiddes et al. 2016), our result suggests that the intense positive temperature anomalies were associated with the quasi-stationary Tasman high and low pressure anomalies over southern Australia, which favored horizontal warm-air advection from the lower latitudes to southeastern Australia. In the mid-to-upper troposphere, the eastward- and poleward-propagating Rossby wave train resulted in persistent anticyclonic circulation anomalies over the Tasman Sea and southeastern Australia (Parker et al. 2013).
To better understand the spatial characteristics of this heatwave event and its sources of predictability, we apply an EOF analysis to the 2-week running-averaged austral summer Tmax anomalies over the 1979–2020 period and show that the Black Saturday heatwave event was the second strongest episode in PC3 since 1911, which has a north–south dipole temperature anomaly pattern. Interestingly, the positive loading of PC3 is moderately related to La Niña and the associated SST anomalies in the tropical Indo-Pacific region, which appears to be a potential source of extended predictability of this type of temperature extremes beyond the weather regime. The other two leading modes, PC1 and PC2, contribute to about half of observed Tmax during the heatwave event but they are not predictable at longer lead times, which could be either due to their intrinsic lack of predictability or due to the model deficiency.
ACCESS-S1 can predict the Black Saturday heatwave event with some useful skills about 10 days in advance, though the amplitude of the predicted Tmax anomalies is significantly weaker relative to observations. The dampening of Tmax anomalies in the ensemble mean is likely due to systematic prediction bias in Tmax variability over southeastern Australia. Despite the weaker than observed temperature anomalies in the 2009 forecast, the ensemble-mean departure is the largest predicted for any period in the model hindcast period (1990–2012). Our analysis reflects the predominance of atmospheric internal processes on the Black Saturday heatwave event, which are underestimated in the forecast, especially at the longer lead times. The Tmax forecasts over the hindcast period are poor beyond the 10-day lead time, highlighting a limitation of ACCESS-S1 in predicting the Tmax variability as well as the extreme heatwave event at longer lead times.
By altering the atmospheric CO2 levels and the initial conditions to reflect an estimate of the “Low CO2” world, we show a clear role of accumulated global warming and increased CO2 concentration in the predicted temperature anomalies. Using ACCESS-S1 at the 10-day lead time, the ensemble-mean “Low CO2” forecast experiment predicts up to 4°C cooling over southeastern Australia compared to the reference (current) forecast, with a domain-averaged temperature reduction of 2.6°C. While there is abundant variability in the “Low CO2” forecast, only 1 out of 11 ensemble members is shown to be warmer than the forecast from the contemporary climate. However, the reduction in multiweek Tmax prediction skill with lead times does present some challenges to the interpretation. For example, while the ensemble-mean difference is positive at ∼2 weeks’ lead time, the predicted Tmax anomaly during the heatwave event is about one-quarter of the observed. The match is better when the lead time is shorter, but this is at the cost of using a model with less time to adjust to the reduced atmospheric CO2 levels and the modified initial conditions.
Further investigation is required to understand the occurrence of the Black Saturday heatwave event in the counterfactual “Low CO2” world and its frequency change in the current climate from the counterfactual world. As it was shown in previous studies (e.g., Hope et al. 2016) that an extreme heatwave event in the current climate was also an extreme heatwave event in the counterfactual low GHG conditions, the Black Saturday heatwave event is likely to be an extreme event in the “Low CO2” world as well. But the exact estimation requires the climatology of the “Low CO2” world, which will be particularly valuable for a comprehensive assessment of extreme heatwave events. We have attributed the Black Saturday heatwave forecast to the anthropogenic changes to the background mean state by altering atmospheric CO2 concentration and its impact on the ocean–atmosphere system, but changes in soil moisture, and other radiative forcing components such as aerosols have not been considered in this study. As atmospheric CO2 is the dominant source of anthropogenic emissions (Friedlingstein et al. 2019) causing the global temperature rise (IPCC 2021), it is reasonable to attribute the increase in warming to the recent enhancement of the atmospheric CO2 concentration. However, there is still scope to improve the attribution experiments by taking into account other GHGs and aerosols, and this will be attempted in a future study. Another caveat is that our estimation of background mean state response to anthropogenic climate change depends on the HadGEM3-GC2 simulations. The quantitative impact of atmospheric CO2 increase on the heatwave event assessment may be sensitive to the background mean state difference and lead-time choice. These uncertainties can be reduced by considering the background mean state from other models and with forecasts from multiple start dates. Caution is also warranted in interpreting 2.6°C warming due to enhanced atmospheric CO2 level, which does not guarantee its impact is limited to only 2.6°C as its influence can be complex, nonlinear, and spatially diverse.
Despite the uncertainties, this study highlights that global warming due to GHG increase positively contributed to the extreme Black Saturday heatwave event. The S2S forecast system includes observed climate change signals into the initial conditions and the updated values of GHGs. Given that the attribution system uses the same operational S2S forecast system, it could be further developed to run alongside the currently operational ACCESS-S2 (Wedd et al. 2022) and help the Bureau’s community and emergency services by providing some quantitative estimate of GHG-driven long-term climate change impact on a particular extreme weather or climate event. This will add value to the effort to earn the trust of the public and the end users of climate information and improve our understanding of the climate change impact on extreme events. Improvements in the delivery of the service can be gained through further conversations with stakeholders, and how they utilize S2S forecasts in their decision-making (e.g., VanBuskirk et al. 2021), their understanding of uncertainty estimates (Marimo et al. 2015), and the drivers of decisions for different sectors (Dilling and Lemos 2011; Klemm and McPherson 2017; Vogel et al. 2021). The skillful forecasts could have provided appropriate warning of the heatwave event and sufficient preparation time if ACCESS-S1 had been available in 2009, along with the latest warning systems (Bettio et al. 2019). We have growing confidence that such extreme heatwave events will now be reasonably well predicted and the increasing impact of climate change on them will be better quantified and communicated (Fischer et al. 2021).
Acknowledgments.
This work was funded by the National Environmental Science Program (NESP) of the Australian Government. S Abhik acknowledges partial support from the Australian Research Council Discovery Project (DP220101468) and E-P Lim is partly supported by the Department of Energy, Environment and Climate Action through the Victorian Water and Climate Initiative-phase 2. We thank the Bureau’s Seasonal Prediction team for making the hindcast data available. Discussions with Fraser Lott (Met Office) and Guomin Wang (Bureau) greatly helped us to develop the climate attribution suite. We thank Tim Cowan, Andrew Marshall, and Harry H. Hendon for their insightful comments during the internal review of the manuscript, and Suzanne Rosier for her suggestions on the significance testing. We are indebted to Dàithì Stone and two anonymous reviewers for their constructive comments on the submitted manuscript. This research was undertaken with the assistance of resources and services from the National Computational Infrastructure (NCI), supported by the Australian Government.
Data availability statement.
The daily AWAP maximum temperature dataset is available to registered NCI users. The daily-mean ERA5 dataset is derived from the hourly ERA5 dataset, which was downloaded from the Copernicus climate data store: https://cds.climate.copernicus.eu/#!/search?text=ERA5&type=dataset. The monthly ERSST dataset was obtained from https://psl.noaa.gov/data/gridded/data.noaa.ersst.v5.html. The ACCESS-S1 hindcast datasets are available from the corresponding authors upon reasonable request.
REFERENCES
Bettio, L., J. R. Nairn, S. C. McGibbony, P. Hope, A. Tupper, and R. J. B. Fawcett, 2019: A heatwave forecast service for Australia. Proc. Roy. Soc. Victoria, 131, 53–59, https://doi.org/10.1071/RS19006.
Blockley, E. W., and Coauthors, 2014: Recent development of the met office operational ocean forecasting system: An overview and assessment of the new global FOAM forecasts. Geosci. Model Dev., 7, 2613–2638, https://doi.org/10.5194/gmd-7-2613-2014.
Boschat, G., A. Pezza, I. Simmonds, S. Perkins, T. Cowan, and A. Purich, 2015: Large scale and sub-regional connections in the lead up to summer heat wave and extreme rainfall events in eastern Australia. Climate Dyn., 44, 1823–1840, https://doi.org/10.1007/s00382-014-2214-5.
Bureau of Meteorology and CSIRO, 2020: State of the climate. BoM and CSIRO, 24 pp., https://apo.org.au/sites/default/files/resource-files/2020-11/apo-nid309418.pdf.
Cai, W., T. Cowan, and M. Raupach, 2009: Positive Indian Ocean dipole events precondition southeast Australia bushfires. Geophys. Res. Lett., 36, L19710, https://doi.org/10.1029/2009GL039902.
Cai, W., A. Purich, T. Cowan, P. van Rensch, and E. Weller, 2014: Did climate change–induced rainfall trends contribute to the Australian millennium drought? J. Climate, 27, 3145–3168, https://doi.org/10.1175/JCLI-D-13-00322.1.
Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553–597, https://doi.org/10.1002/qj.828.
Dilling, L., and M. C. Lemos, 2011: Creating usable science: Opportunities and constraints for climate knowledge use and their implications for science policy. Global Environ. Change, 21, 680–689, https://doi.org/10.1016/j.gloenvcha.2010.11.006.
Drosdowsky, W., and L. E. Chambers, 2001: Near-global sea surface temperature anomalies as predictors of Australian seasonal rainfall. J. Climate, 14, 1677–1687, https://doi.org/10.1175/1520-0442(2001)014<1677:NACNGS>2.0.CO;2.
Engel, C. B., T. P. Lane, M. J. Reeder, and M. Rezny, 2013: The meteorology of Black Saturday. Quart. J. Roy. Meteor. Soc., 139, 585–599, https://doi.org/10.1002/qj.1986.
Feng, M., M. J. McPhaden, S.-P. Xie, and J. Hafner, 2013: La Niña forces unprecedented Leeuwin current warming in 2011. Sci. Rep., 3, 1277, https://doi.org/10.1038/srep01277.
Fiddes, S. L., A. B. Pezza, and J. Renwick, 2016: Significant extra-tropical anomalies in the lead up to the Black Saturday fires. Int. J. Climatol., 36, 1011–1018, https://doi.org/10.1002/joc.4387.
Fischer, E. M., S. Sippel, and R. Knutti, 2021: Increasing probability of record-shattering climate extremes. Nat. Climate Change, 11, 689–695, https://doi.org/10.1038/s41558-021-01092-9.
Friedlingstein, P., and Coauthors, 2019: Global carbon budget 2019. Earth Syst. Sci. Data, 11, 1783–1838, https://doi.org/10.5194/essd-11-1783-2019.
Hersbach, H., and Coauthors, 2020: The ERA5 global reanalysis. Quart. J. Roy. Meteor. Soc., 146, 1999–2049, https://doi.org/10.1002/qj.3803.
Hope, P., G. Wang, E.-P. Lim, H. H. Hendon, and J. M. Arblaster, 2016: What caused the record-breaking heat across Australia in October 2015? Bull. Amer. Meteor. Soc., 97 (12), S122–S126, https://doi.org/10.1175/BAMS-D-16-0141.1.
Hope, P., and Coauthors, 2022: Subseasonal to seasonal climate forecasts provide the backbone of a near-real-time event explainer service. Bull. Amer. Meteor. Soc., 103 (3), S7–S13, https://doi.org/10.1175/BAMS-D-21-0253.1.
Hoskins, B. J., and D. J. Karoly, 1981: The steady linear response of a spherical atmosphere to thermal and orographic forcing. J. Atmos. Sci., 38, 1179–1196, https://doi.org/10.1175/1520-0469(1981)038<1179:TSLROA>2.0.CO;2.
Huang, B., and Coauthors, 2017: Extended reconstructed sea surface temperature, version 5 (ERSSTv5): Upgrades, validations, and intercomparisons. J. Climate, 30, 8179–8205, https://doi.org/10.1175/JCLI-D-16-0836.1.
Hudson, D., A. G. Marshall, and O. Alves, 2011: Intraseasonal forecasting of the 2009 summer and winter Australian heat waves using POAMA. Wea. Forecasting, 26, 257–279, https://doi.org/10.1175/WAF-D-10-05041.
Hudson, D., and Coauthors, 2017: ACCESS-S1: The new bureau of meteorology multi-week to seasonal prediction system. J. South. Hemisphere Earth Syst. Sci., 67, 132–159, https://doi.org/10.1071/ES17009.
IPCC, 2021: Summary for policymakers. Climate Change 2021: The Physical Science Basis, V. Masson-Delmotte et al., Eds., Cambridge University Press, 3–32, https://doi.org/10.1017/9781009157896.001.
Jones, D. A., W. Wang, and R. Fawcett, 2009: High-quality spatial climate data-sets for Australia. Aust. Meteor. Oceanogr. J., 58, 233–248, https://doi.org/10.22499/2.5804.003.
Kala, J., J. P. Evans, and A. J. Pitman, 2015: Influence of antecedent soil moisture conditions on the synoptic meteorology of the Black Saturday bushfire event in southeast Australia. Quart. J. Roy. Meteor. Soc., 141, 3118–3129, https://doi.org/10.1002/qj.2596.
King, A. D., D. J. Karoly, and B. J. Henley, 2017: Australian climate extremes at 1.5°C and 2°C of global warming. Nat. Climate Change, 7, 412–416, https://doi.org/10.1038/nclimate3296.
King, M. J., and M. J. Reeder, 2021: Extreme heat events from an object viewpoint with application to south-east Australia. Int. J. Climatol., 41, 2693–2709, https://doi.org/10.1002/joc.6984.
Klemm, T., and R. A. McPherson, 2017: The development of seasonal climate forecasting for agricultural producers. Agric. For. Meteor., 232, 384–399, https://doi.org/10.1016/j.agrformet.2016.09.005.
Lewis, S. C., and A. D. King, 2015: Dramatically increased rate of observed hot record breaking in recent Australian temperatures. Geophys. Res. Lett., 42, 7776–7784, https://doi.org/10.1002/2015GL065793.
Lim, E.-P., H. H. Hendon, D. Hudson, M. Zhao, L. Shi, O. Alves, and G. Young, 2016: Evaluation of the ACCESS-S1 hindcasts for prediction of Victorian seasonal rainfall. Australian Bureau of Meteorology Research Rep. 019, 51 pp., http://www.bom.gov.au/research/publications/researchreports/BRR-019.pdf.
MacLachlan, C., and Coauthors, 2015: Global seasonal forecast system version 5 (GloSea5): A high-resolution seasonal forecast system. Quart. J. Roy. Meteor. Soc., 141, 1072–1084, https://doi.org/10.1002/qj.2396.
Madec, G., and Coauthors, 2008: NEMO ocean engine: Note du pole de l’Institut Pierre-Simon Laplace Tech. Rep. 27, 396 pp.
Marimo, P., T. R. Kaplan, K. Mylne, and M. Sharpe, 2015: Communication of uncertainty in temperature forecasts. Wea. Forecasting, 30, 5–22, https://doi.org/10.1175/WAF-D-14-00016.1.
Marshall, A. G., D. Hudson, M. C. Wheeler, O. Alves, H. H. Hendon, M. J. Pook, and J. S. Risbey, 2014: Intra-seasonal drivers of extreme heat over Australia in observations and POAMA-2. Climate Dyn., 43, 1915–1937, https://doi.org/10.1007/s00382-013-2016-1.
Marshall, A. G., H. H. Hendon, M. Feng, and A. Schiller, 2015: Initiation and amplification of the Ningaloo Niño. Climate Dyn., 45, 2367–2385, https://doi.org/10.1007/s00382-015-2477-5.
McKay, R. C., J. M. Arblaster, and P. Hope, 2022: Tropical influence on heat-generating atmospheric circulation over Australia strengthens through spring. Wea. Climate Dyn., 3, 413–428, https://doi.org/10.5194/wcd-3-413-2022.
Megann, A., and Coauthors, 2014: GO5.0: The joint NERC–Met office NEMO global ocean model for use in coupled and forced applications. Geosci. Model Dev., 7, 1069–1092, https://doi.org/10.5194/gmd-7-1069-2014.
National Climate Centre, 2009: The Exceptional January–February 2009 Heatwave in South-Eastern Australia. Bureau of Meteorology Special Climate Statement 17, 11 pp., https://www.bom.gov.au/climate/current/statements/scs17d.pdf.
Nitschke, M., G. Tucker, A. Hansen, S. Williams, Y. Zhang, and P. Bi, 2016: Evaluation of a heat warning system in Adelaide, South Australia, using case-series analysis. BMJ Open, 6, e012125, https://doi.org/10.1136/bmjopen-2016-012125.
North, G. R., T. L. Bell, R. F. Cahalan, and F. J. Moeng, 1982: Sampling errors in the estimation of empirical orthogonal functions. Mon. Wea. Rev., 110, 699–706, https://doi.org/10.1175/1520-0493(1982)110<0699:SEITEO>2.0.CO;2.
Parker, T. J., G. J. Berry, and M. J. Reeder, 2013: The influence of tropical cyclones on heat waves in Southeastern Australia. Geophys. Res. Lett., 40, 6264–6270, https://doi.org/10.1002/2013GL058257.
Parker, T. J., G. J. Berry, M. J. Reeder, and N. Nicholls, 2014: Modes of climate variability and heat waves in Victoria, Southeastern Australia. Geophys. Res. Lett., 41, 6926–6934, https://doi.org/10.1002/2014GL061736.
Perkins, S. E., L. V. Alexander, and J. R. Nairn, 2012: Increasing frequency, intensity and duration of observed global heatwaves and warm spells. Geophys. Res. Lett., 39, L20714, https://doi.org/10.1029/2012GL053361.
Perkins, S. E., D. Argüeso, and C. J. White, 2015: Relationships between climate variability, soil moisture, and Australian heatwaves. J. Geophys. Res. Atmos., 120, 8144–8164, https://doi.org/10.1002/2015JD023592.
Perkins-Kirkpatrick, S. E., and Coauthors, 2016: Natural hazards in Australia: Heatwaves. Climatic Change, 139, 101–114, https://doi.org/10.1007/s10584-016-1650-0.
Pezza, A. B., P. van Rensch, and W. Cai, 2012: Severe heat waves in southern Australia: Synoptic climatology and large scale connections. Climate Dyn., 38, 209–224, https://doi.org/10.1007/s00382-011-1016-2.
VanBuskirk, O., P. Ćwik, R. A. McPherson, H. Lazrus, E. Martin, C. Kuster, and E. Mullens, 2021: Listening to stakeholders: Initiating research on subseasonal-to-seasonal heavy precipitation events in the contiguous United States by first understanding what stakeholders need. Bull. Amer. Meteor. Soc., 102, E1972–E1986, https://doi.org/10.1175/BAMS-D-20-0313.1.
Victorian Bushfires Royal Commission, 2010: 2009 Victorian Bushfires Royal Commission: Final Report Summary. Government Printer for the State of Victoria Tech. Rep. 332, 42 pp., https://royalcommission.vic.gov.au/finaldocuments/summary/PF/VBRC_Summary_PF.pdf.
Vogel, E., J. Lerat, R. Pipunic, A. J. Frost, C. Donnelly, M. Griffiths, D. Hudson, and S. Loh, 2021: Seasonal ensemble forecasts for soil moisture, evapotranspiration and runoff across Australia. J. Hydrol., 601, 126620, https://doi.org/10.1016/j.jhydrol.2021.126620.
Walters, D., and Coauthors, 2017: The Met Office unified model global atmosphere 6.0/6.1 and JULES global land 6.0/6.1 configurations. Geosci. Model Dev., 10, 1487–1520, https://doi.org/10.5194/gmd-10-1487-2017.
Wang, G., P. Hope, E.-P. Lim, H. H. Hendon, and J. M. Arblaster, 2021: An initialized attribution method for extreme events on subseasonal to seasonal time scales. J. Climate, 34, 1453–1465, https://doi.org/10.1175/JCLI-D-19-1021.1.
Wedd, R., and Coauthors, 2022: ACCESS-S2: The upgraded Bureau of Meteorology multi-week to seasonal prediction system. J. South. Hemisphere Earth Syst. Sci, 72, 218–242, https://doi.org/10.1071/ES22026.
Williams, K. D., and Coauthors, 2015: The Met Office Global Coupled Model 2.0 (GC2) configuration. Geosci. Model Dev., 8, 1509–1524, https://doi.org/10.5194/gmd-8-1509-2015.
