Bushfire smoke plume composition and toxicological assessment from the 2019–2020 Australian Black Summer

Many of the population centres in southeast Australia were swathed in bushfire smoke during the 2019–2020 austral summer. Bushfires burning during what is now known as the Black Summer was historically large and severe, and the fire season historically long. The chemical composition in the gas and aerosol phase of aged plumes measured near Wollongong, NSW in early 2020 is reported in this work. Enhancement ratios to carbon monoxide are presented for thirteen species (acetaldehyde, acetone, acetonitrile, black carbon aerosol, benzene, methane, methacrolein + methyl vinyl ketone, methyl ethyl ketone, methanol, ammonium ion PM1 fraction, nitrate ion PM1 fraction, organic PM1 fraction and PM2.5). Observed plume composition is comparable to that measured in fresh smoke from Australian fires reported in the literature. Enhancements of biogenic volatile organic compounds such as isoprene (smoke-effected period mean 1 ppb, maximum 6 ppb) were observed along with elevated concentrations of particulate variables. Enhancement ratios reported here can be used in plume modelling of landscape-scale fires and assist in concentration estimates of infrequently measured atmospheric pollutants. The relative toxicological contribution of species present in the plumes was determined for plume exposure at the measurement site and for concentrated plumes at a population centre case study. Similar results were apparent at both locations. Contributions to the toxicological loading were dominated by respirable particles (~ 52–63% total contribution), formaldehyde (~ 30–39% total contribution) and acrolein. This is a reminder to consider the toxicological contributions in the gas phase when considering health impacts of population exposure to bushfire smoke.


Introduction
Biomass burning is the largest contributor to fine carbonaceous aerosol globally and the second largest emission source of atmospheric trace gases (Akagi et al. 2011). A significant portion of total biomass burning emissions can be attributed to emissions from wildfires, which provide a climatically relevant source of black carbon aerosol and greenhouse gases to the global atmosphere  and contribute to enhancements of air quality pollutants in densely populated areas (Carvalho et al. 2011). Australia is a region of global significance in the biomass burning field as fires occur frequently (van der Werf et al. 2006) and Australian emissions (expressed as TgC year −1 ) are estimated to contribute 6-8% to the global biomass burning total (Kasischke and Penner 2004;van der Werf et al. 2006).
Biomass burning is estimated to contribute 53 ± 14% to Australia's annual black carbon aerosol emissions (Qi and Wang 2019). Savanna fires are the dominant contributor (83%) to biomass burning carbon dioxide (CO 2 ) emissions in Australia (Shi et al. 2015). Forest fires along the east coast of the Australian continent also contribute to biomass burning in Australia, responsible for 13% of biomass burning CO 2 emissions annually (Shi et al. 2015). These fires are more episodic in nature, with area burnt varying significantly from year to year (Australian Government Department of Industry Science Energy and Resources 2020). The variability of fire weather is influenced by a number of inter-annual climatological processes, most notably the El Niño Southern Oscillation (ENSO) (Dowdy 2018;Harris and Lucas 2019).
The atmospheric impact of forest fire smoke along the east coast of Australia is reasonably well documented.
Smoke from large fires in 2013 in the Blue Mountains region of New South Wales (NSW) was demonstrated to have had detrimental impacts on air quality as far away as Brisbane, Queensland as well as in closer population centres such as Sydney and Wollongong, NSW. . Emission factors (EFs) for CO 2 , carbon monoxide (CO), nitrous oxide (N 2 O) and methane (CH 4 ) were calculated for these fires using in situ Fourier transform infrared spectrometer (FTIR) measurements from Wollongong, NSW (Rea et al. 2016). Fresh smoke from the firegrounds of a series of prescribed burns throughout NSW and Victoria from 2010-2015 were measured by open path FTIR (Paton-Walsh et al. 2014), Selected Ion Flow Tube Mass Spectrometer (SIFT-MS) and grab samples . Emission factors were calculated for greenhouse gases and a large range of volatile organic compounds (VOCs) from these fires Paton-Walsh et al. 2014), and a dependence on Modified Combustion Efficiency (MCE: CO 2 /(CO + CO 2 )) was observed for some species . Filter measurements of particulate matter less than 2.5-µm diameter (PM 2.5 ) from the fireground of four prescribed burns in temperate eucalypt forest in Victoria, southern Australia have been used to produce emission factors for this air quality pollutant (Reisen et al. 2018). Measurements of trace species emitted by forest fire in south-eastern Australia have also been taken from satellite (Young and Paton-Walsh 2011;Paton-Walsh et al. 2008) and aircraft platforms (Hurst et al. 1996).
The negative health implications of exposure to smoke components are well known, e.g. (Lelieveld et al. 2015;Naeher et al. 2007;Pope III et al. 2002), specifically in the Australian context (Walter et al. 2020). Positive relationships between population exposure to forest fire smoke plumes and health metrics including adult asthma and chronic obstructive pulmonary disease hospital admissions (Morgan et al. 2010), non-accidental mortality (Johnston et al. 2011) and emergency ambulance dispatches (Salimi et al. 2017) have been observed in eastern Australian cities. Each of these studies uses particulate matter with diameter less than 10-µm (PM 10 ) measurements to define plume exposure. The majority of the negative health effect of exposure to bushfire smoke is attributable to the fine particulate pollution present in smoke (Johnston et al. 2012). Nevertheless, a recent study estimated the contribution of gas phase toxics to smoke toxicity and demonstrated that considering exposure to a range of air toxics is critical to determining health impacts of bushfire smoke in Australia (MacSween et al. 2020). Historically, particulate and CO pollution experienced in eastern Australia's population centres is attributable to smoke events Paton-Walsh et al. 2019;Keywood et al. 2015), with air quality outside of these events normally relatively good. However, the episodic influence of smoke is becoming more frequent, and more sustained air quality impacts have been observed recently (Di Virgilio et al. 2021;Ryan et al. 2021).
The 2019-2020 Australian bushfire season-known colloquially as the 'Black Summer'-was amongst the most significant in Australia's recent history, driven by historical dryness across the landscape coupled with meteorological conditions promoting wildfire (Nolan et al. 2020). It is estimated that 7.4 million hectares of temperate forest was burnt from October 2019 to February 2020 (Australian Government Department of Industry Science Energy and Resources 2020). Fires were widespread, stretching from near the NSW-Queensland border to the Gippsland region of Victoria, a distance of over 1000 km. The fires were featured prominently in global news reports, e.g. (BBC News 2020; Cave 2020). Smoke caused exceedances of air quality monitoring criteria pollutants on an unprecedented scale at sites across Australia's east coast from Brisbane, Queensland (Queensland Government 2020) to Melbourne, Victoria (Environment Protection Authority Victoria 2020). During the fire season (October 2019-February 2020), PM 2.5 concentrations exceeding the 95th percentile of historical daily mean values were observed by at least one monitoring station in Australia's eastern states on 94% of days (Borchers Arriagada et al. 2020). A total of 417 (95% CI, 153-680) excess deaths have been attributed to exposure to smoke from the 2019-2020 bushfire season along with thousands of hospital admissions (Borchers Arriagada et al. 2020). These statistics demonstrate the 2019-2020 Black Summer bushfires in the temperate forests of southeast Australia were of unique magnitude, duration and severity.
Fire seasons similar to the 2019-2020 austral summer may become more frequent in the future. The frequency of fire weather in southeast Australia has increased in the past four decades likely due to the influence of anthropogenic climate change (Harris and Lucas 2019). A trend is also observed in the onset of fire weather, with dangerous conditions occurring earlier in the year (Dowdy 2018) lengthening the so-called 'fire season'. Forest fire frequency and intensity in the Australian southeast are predicted to increase into the future in a changing climate (Keywood et al. 2013). It is therefore critical to understand the chemical composition of smoke plumes impacting population centres and assess whether current exposure limits and public health messaging are appropriate.
The Characterizing Organics and Aerosol Loading over Australia (COALA-2020) campaign was a comprehensive atmospheric measurement campaign which occurred from January to March 2020 at Cataract Scout Park, close to Wollongong, NSW (Paton-Walsh et al. 2022). Whilst the primary objective of the campaign was to explore the relationship between biogenic VOC emissions and organic aerosol, the first 3 weeks of measurements were significantly impacted by bushfire smoke. The measured plumes were complex and contained smoke of various ages generated by multiple fires burning across the southeast Australian landscape (Mouat et al. 2021b).
This paper presents detailed measurements of bushfire smoke from the 2019-2020 fire season in the aerosol and gas phases. This dataset is the most comprehensive measurement of bushfire smoke from the historical 2019/2020 Australian bushfire season. This paper seeks to answer two scientific questions using these measurements: 1) Firstly, what was the chemical composition of aged smoke plumes from the Black Summer fires, such as those that impacted major population centres, and how does the composition compare to previously observed plumes in the region? 2) Secondly, what is the relative toxicological impact of each chemical species present in these plumes?
In this paper, we present enhancement ratios to the commonly measured atmospheric pollutant carbon monoxide (CO) for several trace gas and aerosol-phase chemical species. These enhancement ratios can be used to estimate the pollution levels of relevant pollutants throughout the air quality monitoring networks in eastern Australia wherever CO is measured and used in epidemiological studies quantifying the health impacts of the 2019/2020 bushfire season. Assessment of the relative toxicological contribution of chemical species observed in smoke plumes that impacted Australia's major cities is also made using current exposure limit concentrations. The methodology and results of composition determination will be presented first, followed by the methodology and results of toxicological impact determination and finally the application of this preceding work to a case study in Canberra, Australia.

Determining smoke composition
Cataract Scout Park (34°14′44″S, 150°49′26″ E), at which the COALA campaign took place, is located 20 km northnorthwest of Wollongong on the east coast of NSW, Australia. The site is marked on the map presented in Fig. 1. The areas surrounding the site are heavily forested. Eucalypt species dominate the canopy. There are numerous longwall (underground) coal mine heads close to the measurement site. The Appin Colliery is approximately 1.5 km to the north-northeast, and the West Cliff Colliery approximately 2.5 km to the north. This site was selected for the COALA campaign as the primary objective of the campaign was to learn more about biogenic VOC emissions and secondary organic aerosol (SOA) formation from southeast Australian eucalypt forest.
The COALA campaign was conducted from January 17th till March 23rd, 2020, covering the second half of the Australian summer and the start of Autumn. Measurements presented in this study span from the beginning of the campaign until February 5th, a period of 20 days. This period will hence be referred to as the 'smoky period'. A widespread significant rain event occurred in NSW beginning on February 7th, "cleaning" the local atmosphere of smoke and extinguishing most fires in the state (Bureau of Meteorology 2020). All dates and times are reported in local (Australian Eastern Daylight) time.

Instrumentation
Detailed descriptions of all instruments used in COALA are provided in an overview publication (Paton-Walsh et al. in prep.). This publication focuses on particle-phase and VOC measurements of bushfire smoke. Instruments used heavily in the present analysis are described below. All measurements were averaged to 1-min time resolution for analysis.
A Spectronus Fourier Transform Infrared Spectrometer (FTIR) (Griffith et al. 2012) was used to make continuous, high-precision measurements of CO 2 , CO, CH 4 , N 2 O, ∂ 13 C in CO 2 and ∂ 18 O in CO 2 ). This instrument measures each trace gas simultaneously at 2-min resolution. It functioned near-continuously for the measurement period. Ambient air was drawn through approximately 7 m of ¼″ Dekabon tubing from an inlet 4.7 m above ground level, before being dried by Nafion dryer and magnesium perchlorate before passing into the instrument. The instrument was calibrated before and after the campaign, and a reference cylinder was measured each day to account for any drift in the calibration. The instrument ran for the majority of the campaign, excepting during a power failure that occurred on February 7 th .
A TSI 3776 Ultrafine Condensation Particle Counter (UCPC) was used to measure condensation nuclei number concentration greater than 3 nm (CN 3 ) (TSI Incorporated, Shoreview, MI, USA) (Keywood et al. 2020). The instrument was operated at a sample flow rate of 300 mL min −1 . Sample air was drawn from a common aerosol bypass inlet.

3
The inlet was located 5 m above ground level. Measurements were recorded at 1-Hz temporal resolution.
A Scanning Mobility Particle Sizer (SMPS) was used during COALA to measure aerosol size distribution between 14 and 670-nm mobility diameter . Full scans of this size range were recorded every 5 min. The system consisted of an X-ray aerosol neutralizer and 3071 Long Electrostatic Classifier (TSI Incorporated, Shoreview, MI, USA) coupled to a 3772 CPC (TSI Incorporated, Shoreview, MI, USA). Sample was drawn from the same inlet as used by the UCPC.
The 5012 Multi-Angle Absorption Photometer (MAAP; Thermo-Fisher Scientific, Waltham, MA, USA) is used to measure the concentration of black carbon (BC) aerosol . Sample was drawn down a ½″ inlet at 1000 L min −1 . Measurements of black carbon aerosol concentration were recorded at 1-Hz resolution. The inlet was located 4.8 m above ground level, within 1 m of the aerosol inlet, and equipped with a PM 10 inlet cap.
An air quality monitoring station (AQMS) owned and operated by the NSW Government Department of Planning, Industry and Environment was installed at Cataract Scout Park for the COALA campaign (Gunashanhar et al. 2021). This station included measurements of temperature, wind speed, wind direction, PM 10 , PM 2.5 , ozone (O 3 ), sulphur dioxide (SO 2 ), oxides of nitrogen (NO x ), CO and visibility. NSW air quality stations are accredited by the National Association of Testing Authorities and use Standards Australia methods for ambient air quality measurements where possible (Department of Planning Industry and the Environment 2020). Further details of specific measurement techniques are available on the Department's website (Department of Planning Industry and the Environment 2020). Inlet heights for the station were as follows (metres above ground Fig. 1 A satellite image of southeast Australia from January 30th, 2020, sourced from the NASA Worldview online viewer. Major cities of Sydney, Melbourne and Wollongong are marked, as is the COALA-2020 measurement site at Cataract Scout Park. Fires and thermal anomalies detected by VIIRS are plotted in red. The widespread smoke plume over SE Australia is evident. The inset shows the area surrounding the measurement site in more detail level): temperature: 4.5 m, wind: 5.6 m, PM 10 : 4.9 m, PM 2.5 : 4.9 m, gases: 4.7 m.
A Proton Transfer Reaction Time-of-Flight Mass Spectrometer (PTR-ToF-MS) (Ionicon 4000; Ionicon Analytik Ges.m.b.H., Eduard-Bodem-Gasse 3, 6020 Innsbruck, Austria) was used to retrieve concentrations of VOCs during COALA (Mouat et al. 2021a). Sample air was drawn from an inlet on a 10-m mast, down a 20-m PTFE inlet line using a bypass flow of 1.5-3 L min −1 . Calibration occurred in situ using standard cylinders and is reported with an uncertainty of ± 20%. Further details of instrumental parameters and calibration processing can be found in Mouat et al. (2021b). Deployment of this instrument was delayed due to airfreight cancellations following a volcanic eruption near Manila in the Philippines, and measurements began on February 2 nd . As a result of this delay, there are only 2 days of VOC measurements when significant smoke impacted the measurement site.
Measurements of particulate matter with diameter smaller than 1 µm (PM 1 ) aerosol composition were taken using a Time-of-Flight Aerosol Chemical Speciation Monitor (ACSM; Aerodyne Research Inc., Billerica, MA, USA). Mass concentrations of Organics (Org), sulphate (SO 4 2− ), nitrate (NO 3 − ), ammonium (NH 4 + ) and chloride (Cl − ) in the aerosol fraction 40-1000-nm vacuum aerodynamic diameter range, referred to as PM 1 , are reported. Measurements were taken at 10-min resolution and down sampled to 1-min resolution using cubic spline interpolation. Sample was drawn from the aerosol inlet common to the CPC and SMPS and dried using a Nafion dryer to < 40% RH before sampling.
Further instrumental and sampling information are available in the metadata associated with the published data archive for each measurement set. Please see the Data Availability Statement for details.

Determination of enhancement ratios and emission factors
An enhancement ratio describes the way the concentration of a selected pollutant scales with the concentration of another reference pollutant during a period in which both pollutants are enhanced above the local background concentrations. Enhancement ratios provide a useful way of characterizing the composition of a plume without having to deploy specialised and expensive instrumentation, e.g. (Lonsdale et al. 2020;Decker et al. 2021). In the present analysis, the trace gas CO was selected as the reference pollutant. This is a very common choice, e.g. (Rea et al. 2016;Keywood et al. 2015) as CO concentrations are often measured at air quality monitoring stations globally. Although the largest Australian source is secondary chemistry production (Fisher et al. 2017), CO has a large biomass burning source in Australia. Enhancement ratios were calculated by fitting a linear regression to a correlation plot between the mole fraction of the pollutant of interest x and the mole fraction of CO. The enhancement ratio of x to CO can be taken as the gradient of the line of best fit. Although an orthogonal regression as used in previous studies , weighting the regression for uncertainties in x and y may be most appropriate for this kind of analysis, estimates of precision for each individual species (as required for orthogonal regression) were not readily available in the present study. The high precision of the CO measurement and the location of this measurement of the x-axis also increase the suitability of simple linear regression for this study as this regression minimises residuals in y. Uncertainties quoted are total uncertainties. The type of regression used has been shown to have limited influence on the value of the slope of the line of best fit (Wu and Yu 2018) where strong correlations are observed. As enhancement ratios were determined only for periods of strong correlation (R 2 > 0.7), in this analysis a simple linear regression was used. Further to this, the extremely tight correlations between CO and other atmospheric species as presented in Fig. 2 and very precise measurements captured by the FTIR (Griffith et al. 2012) mean that the expected real variability in enhancement ratios is much greater than the measurement uncertainty for this analysis. This method of determining enhancement ratios is only reliable when correlation between the species of interest and the reference species is high, as this implies limited influence from other atmospheric sources on the species in question. Previous studies have applied this method of enhancement ratio calculation only in cases where R 2 > 0.5 (Sinha et al. 2003;Desservettaz et al. 2017). Given the complexity of the smoke plumes observed during the present study, a stricter limit of R 2 > 0.7 has been applied. An advantage of calculating enhancement ratios from the slope of correlation plots is that no assumptions with respect to the site background must be made , which is important at the Cataract site as many proximal sources of measured species were present which made determining an atmospheric background concentration difficult. Following the exclusion of relationships without strong X:CO correlation, some events were excluded manually after examination of the correlation plots.
Enhancement ratios were calculated event-wise. Events were identified as periods of strong correlation between known smoke tracers (e.g. CO, black carbon aerosol, acetonitrile) using a series of iterative plots. Correlation plots showing strong relationships between CO and other atmospheric species are presented for the example event Smoke I in Fig. 2. Similar plots for other events are presented in the supplement as Figure S1. Details of the timing of each smoke event, along with summary tables of enhancement ratio values, are presented in supplementary Tables S1 and S2 respectively. Sixteen events were identified and assigned alphabetical names. A 'dilute smoke' characterization was assigned to periods between identified events. This category is distinct from the true background air of the region as CO, and black carbon aerosol remained present at concentrations significantly above measurements taken following the heavy rain which occurred February 7-9 th . Identified events along with the 'dilute smoke' periods (grey) are plotted in Fig. 3. The PTR-ToF-MS began sampling on February 2 nd , capturing events Smoke I and later.
The final enhancement ratio reported for each species is the mean slope for all events for which the relationship between x and CO was sufficiently strong to meet the criteria for inclusion. The standard deviation, range and number of events included in this calculation are also presented. These values are reported in Table 1.
Emission factors for some species were also calculated. Emission factors for CO, CO 2 and CH 4 were determined for the only period where correlations between all three  /m3)   200  300  400  500  600  200  300  400  500  600  200  300  400  500  600   200  300  400  500  600  200  300  400  500  600  200  300  400  500  600   200  300  400  500  600  200  300  400  500  600  200  300  400  500  600   200  300  400  500  600  200  300  400  500  600  200  300  400  500   CO2:CH4 = 0.90), overnight on February 5 th . The dilute nature of the sampled plumes, coupled with the surrounding biogenic sink for CO 2 , led to a decoupling of the expected relationship between CO 2 and CO in sampled smoke. The approach taken by Mouat et al. (Mouat et al. 2021b) was used to calculate emission factors for CO, CO 2 and CH 4 , where the enhancement ratio of the focus species is ratioed to the sum of the enhancement ratios of CO, CO 2 and CH 4 . This result is then multiplied by a fuel carbon content and the ratio of the focus species molecular mass to the molecular mass of carbon. Calculated emission factors for CO (99 g kg −1 fuel), CO 2 (1665 g kg −1 fuel) and CH 4 (4 g kg −1 fuel) are comparable to mean values reported by Paton-Walsh et al. (2014) for a series of hazard reduction burns in similar forest types. This adds confidence to the use of a single period for the calculation of emission factors for these species. Emission factors for black carbon aerosol, PM 2.5 and selected PM 1 fractions were determined by multiplying observed enhancement ratios to CO by the emission factor to CO determined from the measurements. To compensate for the lack of variability in CO emission factors calculated in this study, the relative error for the CO emission factor is taken from Paton-Walsh et al. (2014) and added in quadrature to the variance in focus species enhancement ratio to provide uncertainty in reported emission factors as presented in Table 2. Emission factors for long lived VOC species determined from PTR-ToF-MS measurements taken during COALA-2020 are published in Mouat et al. (2021b).

Meteorology and general atmospheric composition during the smoky period
Wind directions observed throughout the smoke-influenced period of COALA-2020 were variable. Winds from all directions were observed with some periods of sustained easterly and southerly airflow. The influence of anthropogenic emissions from nearby sources such as the cities of Sydney and Wollongong cannot be discounted. However, fires can be assumed to be the dominant source of most pollutants during this time, given the widespread, persistent nature of the smoke plume covering the measurement site for much of this period (e.g. Fig. 1). Local wind speeds were generally low, with a mean wind speed of 1.52 m s −1 observed. The average daily maximum ambient temperature was above 29 °C. The average overnight low was 19 °C. Three days with a maximum temperature > 35 °C were observed, with the maximum recorded temperature 44 °C. Limited precipitation was observed during the smoky period.
Considering the highly vegetated setting of the measurement site, it is unsurprising that a diurnal cycle in CO 2 was observed. The magnitude of this cycle is approximately 10 ppm, with minimum concentrations occurring in the afternoon, coincident with the highest values of photosynthetically active radiation and hence photosynthetic drawdown (Wayne 1991). Very high CH 4 concentrations (up to 36 ppm) were observed during the campaign. The extremely high concentrations observed and the very short (< 6-h) duration of the events suggest they are from local point sources. The significant long-wall coal mining in the area under and surrounding the measurement site has caused significant subsidence and cracking impacts (Kay et al. 2006;NSW Department of Planning 2008). Observed very high CH 4 concentrations may have originated from nearby mine heads or from outgassing from waterways or cracks in local geology. Enhanced CH 4 concentrations observed in nearby Wollongong have previously been linked to mining emissions (Buchholz et al. 2016).
Significant concentrations of isoprene (up to 6 ppb, mean 1 ppb), isoprene oxidation products and monoterpenes were observed during the campaign. Eucalypt forests of southeast Australia are known to be large sources of these biogenic VOCs (Benjamin et al. 1996;Kesselmeier and Staudt 1999), and emissions are known to be proportional to local ambient temperature (Emmerson et al. 2020;Ramirez-Gamboa et al. 2021). Biogenic VOCs such as those observed during COALA can act as precursors to secondary organic aerosols (SOA) (Hu et al. 2013;McFiggans et al. 2019). This was a potentially significant particle source during the smoky period of the COALA campaign. However, biogenic SOA can be separated from particles attributable to wildfire smoke by their differing size characteristics. Recently formed SOA are expected to be ultrafine aerosol, with aerodynamic diameters 10 nm and below, e.g. (Yu et al. 2014;Roldin et al. 2015). This size range was captured by the CPC during COALA. Occasions of very high CN 3 concentrations, possibly indicating recent particle nucleation, were observed during the smoky period of COALA. Smoke particles mostly occupy larger particle fractions (Sillanpää et al. 2005;Bouya and Box 2011) and were therefore detectable in PM 2.5 and black carbon records.
Relatively low concentrations of the anthropogenic VOCs benzene, toluene and xylene were observed in the smoky period. The mean concentration of both these species was < 0.5 ppb during the smoky period with no significant enhancements observed. This implies limited influence on the measurement site during the smoky period from nearby roads ). Significant concentrations of ozone were observed, closely correlated with temperature. The 4 days on which the highest ozone concentrations were observed were the 4 days on which temperatures were observed above 35 °C. The maximum   recorded concentration was 106 ppb, observed on the hottest day when the temperature reached 44 °C. The presence of biogenic and biomass burning VOCs in the local atmosphere (Granier et al. 2000), combined with high temperatures, provided conditions favourable for ozone formation (Utembe et al. 2018). The mean ozone concentration during the smoky concentration was 33 ppb. NO x concentrations were low with a mean concentration of 3.6 ppb. The two largest enhancements in NO x in which concentrations > 20 ppb were observed occurred during periods of strong westerly winds at night. Some simultaneous enhancement is visible in SO 2 . A possible source for these enhancements is a national highway (Pettit et al. 2021), the Hume Motorway, which is located 15 km west of the site. Carbon monoxide concentrations throughout the smoky period were elevated significantly above the expected background for the measurement site. For example, the clean air mean CO concentration at nearby Wollongong was 63 ppb between 2011 and 2014 (Buchholz et al. 2016). The mean CO concentration during the smoky period of 181 ppb is greater than twice the mean for the COALA campaign after the smoky period (78 ppb). The scale and extent of bushfires burning prior to and during the smoky period gave an obvious dominant source for elevated CO during this period. Other pollutants closely associated with wildfire smoke (e.g. acrolein, black carbon aerosol, PM 2.5 and organic aerosol fraction) were also enhanced during the smoke period. Enhancements vary in magnitude as plumes containing varying concentrations of pollutants passed over the measurement site. These enhancements are discussed further in the following analysis.

Source of biomass burning influence
The Black Summer fires burning during COALA-2020 were numerous and occurred over a very large geographical area. Extensive fires were burning along the southern coast of NSW, in eastern Victoria and in the alpine regions during the campaign. The campaign began months after the start of the fire season. By mid-January thousands of hectares of forest had already burned (Australian Government Department of Agriculture 2021), creating a significant atmospheric smoke burden in the measurement region. These dilute plumes had continued influence on the local atmosphere for the entire summer, influenced by local meteorology such as sea breeze recirculation (Ryan et al. 2021). The numerous fire sources and complex meteorology of the region made it impossible to link any observed smoke plumes directly to a specific fire emission source. Back trajectories produced using the HYSPLIT model were used to assist in this process. Trajectory information for most plumes was difficult to interpret, as trajectories often passed near multiple burning areas in the hours and days before air mass measurement and did not capture well the local meteorology which may have influenced smoke recirculation over the measurement site.
We can deduce, however, that smoke sampled during COALA-2020 had aged significantly before measurement. There is evidence for this in both the particle and gas phases. In the particle phase, ageing of the aerosols present can be approximated using observations from the ToF-ACSM, where the fraction of organic PM 1 with the mass-to-charge ratio of 44 (f44) serves as a marker for photochemically aged organic aerosols (Aiken et al. 2008;Ng et al. 2010). During this period of abundant smoke, f44 consistently ranged between 0.20 and 0.30, presented in Fig. 4, with nearly 60% of data points being over 0.25. In contrast, observations of f44 in other studies of biomass burning-affected environments show that f44 rarely exceeded 0.25 (Carbone et al. 2013;Bougiatioti et al. 2014;Zhang et al. 2015;Milic et al. 2017). The high concentrations of the aged aerosol marker f44 observed suggest dominant contributions of aged smoke to the particle phase during the campaign.
Fresh smoke VOC tracers were also present in low concentrations in the present measurements. Using the same measurements discussed here, Mouat et al. (2021b) observe oxidative products of the short-lived smoke tracer furan at concentrations indicating smoke ageing. The authors also estimate smoke transport time as ~ 8 h for the overnight period on February 3, the youngest plume sampled with PTR-ToF-MS, using HYSPLIT trajectories. The locations of active fires during the campaign lead us to suggest smoke sampled during the discussed measurements is at least 8 h old.

Comparison of enhancement ratios to literature values
Enhancement ratios to CO were determined for 13 measured variables using the methodology outlined in Sect. 2.2, as reported in Table 1. Enhancement ratios determined for measured particle-phase variables will be presented first, followed by enhancement ratios determined for gas-phase species. A plot of all enhancement ratio to CO values for each species is presented in Fig. 5. Emission factors for select particle species are presented in Table 2.

Particle-phase variables
Periods of strong correlation were observed between CO and several particle-phase variables, notably black carbon aerosol, organic fraction of PM 1, and PM 2.5 . However, not all aerosol variables displayed periods of good correlation to CO: Aitken and accumulation mode aerosol (as measured by SMPS), CN 3 and PM 10 were not observed to have a strong relationship with CO during any of the selected events. A lack of correlation to CO for periods dominated by biomass burning influence suggests either the influence of a separate source on the measured airmass, or extensive chemistry occurring since emission changing the concentration of the variable in the plume. Elevated concentrations of Aitken mode (10-100-nm diameter) particles, accumulation mode (100-1000-nm diameter) particles and CN 3 are often associated with secondary aerosol formation ) as they are aerosol count measurements and therefore are heavily influenced by small, freshly nucleated particles. The measurement site, as described in Sect. 2, is surrounded by native SE Australian forest dominated by Eucalypt species. This forest type is known to be a significant source of biogenic VOC species Emmerson et al. 2018) which are associated with the formation of secondary organic aerosol (Cope et al. 2014;Dominick et al. 2019). Therefore, it is suggested that particle formation originating from biogenic precursors may disrupt the relationship between biomass burning tracers such as CO and measurements of ultrafine aerosol variables. The condensation of biogenic VOCs onto existing smoke particles may also have influenced observed ultrafine particle variables during COALA-2020. Another potential disruption in the relationship between biomass burning tracers such as CO and ultrafine aerosol relates to plume ageing. It is wellknown that the size and composition of aerosols emitted from wildfire change as the plume ages, e.g. (Hodshire et al. 2019;May et al. 2015;Sillanpää et al. 2005). Although the organic aerosol production has been found to be negligible in photochemically aged plumes (May et al. 2015;Zhou et al. 2017), the opposing processes of VOC condensation into the aerosol phase driven by atmospheric oxidation and VOC evaporation as the plume dilutes lead to complex changes in aerosol size distribution and composition (Garofalo et al. 2019). It is suggested these processes, along with the influence of locally emitted biogenic VOCs, possible biogenic SOA formation or condensation of these species onto existing smoke particles disrupted the relationship between CO and ultrafine aerosol variables during COALA. PM 10 is not an ultrafine aerosol measure and did not display strong relationship to CO. A possible confounding source of PM 10 is present during COALA-2020 is lofted mineral dust. Widespread landscape dryness (Nolan et al. 2020) and high wind speeds (Deb et al. 2020) present in the region during the fire period provided conditions conducive to the formation and transport of dust plumes from inland regions of southeast Australia.
The concentration of black carbon aerosol was strongly correlated to CO concentration for thirteen selected events. In most cases, the correlation was very strong (Fig. 2, supplementary Table S1). Calculated enhancement ratios to CO ranged between 2.1e − 03 and 6.3e − 03 μgm −3 ppb −1 CO . The enhancement ratio to CO from the Kilmore East fire, a large 2009 fire in central Victoria which consumed largely similar fuel to the Black Summer fires, was calculated as 4.5e − 3 µgm −3 ppb −1 from the emission factor published by Surawski et al. (2016). The mean enhancement ratio from the present values is similar, 4.0 ± 1.5e − 03 μgm −3 ppb, and the value derived from literature agrees within one standard deviation. The similarity of these values suggests that the ageing process occurring in the plumes measured during COALA-2020 did not significantly change the concentration of black carbon aerosol. Atmospheric ageing of smoke from fires in the USA has been observed to have limited impact on black carbon concentration up to 5 h after emission (May et al. 2015). The plumes sampled during COALA-2020 are likely older than this. However, the relative chemical inactivity of this species and atmospheric lifetime of 4-12 days under typical oxidative conditions (Cape et al. 2012) provides an explanation for an aged smoke enhancement ratio value similar to the emission ratio calculated from a southeast Australian forest fire. PM 2.5 was well-correlated with CO for seven of the identified smoke periods. The calculated enhancement ratio to CO ranged from 4.4e − 02 to 1.2 − e01 µgm −3 ppb, with the mean reported value 8.6 ± e − 02 μgm −3 ppb. This value is a factor of three higher than that reported from measurements of aged, diluted smoke in Sydney . Interestingly, however, the mean PM 2.5 enhancement ratio to CO for more than 300 North American wildfire measurements is three times greater than the value observed during COALA-2020 (Prichard et al. 2020 Fig. 3. The red diamond marks the mean of all values. The small black diamond marks a literature value as cited in Table 1 ratio to CO of approximately 2.6e − 01 μgm −3 ppb −1 can be calculated from emission factors reported for PM 2.5 and CO for both flaming and smouldering fires from prescribed burn measurements in southeast Australian eucalypt forests (Reisen et al. 2018). This suggests relative concentrations of PM 2.5 decreased between emission and sampling in the plumes measured during COALA-2020. As discussed above, there are multiple complex physical, chemical and dilution factors that change the physical and chemical characteristics of smoke plume aerosol populations after emission. The variability in reported enhancement ratios of PM 2.5 to CO demonstrates this complexity. We suggest that coagulation to larger particle sizes and removal via dry deposition contribute to the lower relative concentration of PM 2.5 observed during COALA-2020 compared to freshly emitted wildfire plumes. Further, plumes originating from the uncontrolled high-intensity wildfire as occurred during the Black Summer may have significant contributions from flaming combustion which is associated with lower particulate emission (Reisen et al. 2018).
Organic aerosol was the dominant fraction in speciated PM 1 measured by ACSM during COALA-2020. Organic aerosol was well-correlated with CO for all selected smoke events (correlation coefficients between 0.86 and 0.99). Of fractions reported here, organic aerosol contributed ~ 90% to the PM 1 loading. This value is similar to those observed in measurements from multiple smoke plumes of varying ages in the western USA (Garofalo et al. 2019). This factor is expected to be independent of smoke age as the oxidation of primary organic aerosol is balanced by the formation of SOA, giving a net maintenance of total organic aerosol loading (Enders et al. 2021). The nitrate (NO 3 − ) aerosol fraction was also frequently well-correlated to CO. Ammonium (NH 4 + ) aerosol fractions were occasionally well-correlated, with sulphate (SO 4 − ) and chloride (Cl − ) aerosol not wellcorrelated. All fractions displayed similar levels of variability to other pollutants, with the enhancement ratio to CO range of similar magnitude to the mean value (that is, the range is approximately 100% of the reported value). It is possible that sulphate and chloride aerosol may have been influenced by local marine emissions with a source in the Tasman Sea, lying 20 km to the east of the measurement site as marine aerosol is a source of both of these ions (Xiao et al. 2018). It is also possible that all three of these fractions may display large influence from nearby anthropogenic sources, particularly those involving fossil fuel combustion as this process emits sulphate (Lighty et al. 2000) and chloride (Luo et al. 2019).
Emission factors calculated in the current study (Table 2) are also compared to published values. The emission factor for black carbon aerosol, reported as 0.34 g kg −1 fuel is similar to the adjusted value used by Surawski et al. (2016) for a landscape-scale fire in temperate Australian forests, 0.55 g kg −1 . The emission factor for PM 2.5 calculated in the present study as 7.7 g kg −1 is lower than the globally reported value for temperate forests, 13 g kg −1 fuel (Akagi et al. 2011), and a factor of two lower than the factor calculated from flaming combustion during controlled burns in temperate Australian forests (Reisen et al. 2018). Speciated PM 1 emission factors have been calculated from fresh smoke measurements from flaming and smouldering combustion during Australian prescribed burns (Reisen et al. 2018). Ammonium and nitrate PM 1 fraction emission factors calculated during COALA were higher than observed in fresh smoke. This may indicate different combustion processes during wildfires or ammonium and oxidative nitrate particle formation during smoke transport. The organic PM 1 fraction emission factor was higher during COALA than from a Northern Australian savannah fire counterpart value ; however, the differing fuel and fire types make this comparison difficult to draw conclusions from.

Gas-phase variables
Ambient concentrations of a range of VOC species were determined from PTR-ToF-MS measurements. Details of the instrumental techniques and spectrum analysis used in retrieval are presented in Mouat et al. (2021b). Many VOC species were well-correlated to CO during selected smoke events. For example, acetaldehyde, benzene, methane, methyl ethyl ketone and methanol demonstrated strong relationships with CO for six or more of a possible eight events. As with particle phase variables, the enhancement ratios to CO determined from the COALA-2020 measurements are similar to those found in the literature. Acetaldehyde was determined to have a mean enhancement ratio to CO of 3.2e − 03 ± 1.3e − 03 ppb ppb −1 from seven smoke events. The range of measurements is of similar magnitude to the measurement. This value is smaller than the enhancement ratio to CO calculated from fresh smoke emissions during a series of laboratory burns of southeast Australian fuels by Guérette et al. (2018) of 7.0e − 03 ppb ppb −1 . The short atmospheric lifetime of acetaldehyde may contribute to this observation; however, formation of this species through oxidation of other VOCs present in the plume would also be expected. Acetone and benzene enhancement ratios to CO calculated from COALA-2020 were similar to those determined by Guérette et al. (2018) with similar relative ranges. These values are reported in Table 1. Robust relationships were determined periodically also between CO and the m/z 71 (reported as the sum of methacrolein and methyl vinyl ketone). The enhancement ratio for the sum of methacrolein and methyl vinyl ketone to CO observed during COALA-2020 is five times less than the value reported by Guérette et al. (2018). The hours-long atmospheric lifetime for these species may partially explain this difference between fresh and aged smoke; however, more continuous measurements of ageing plumes are required to properly understand plume chemistry. No field measurements containing an enhancement ratio for methyl ethyl ketone to CO was found in the literature; however, a value from laboratory burns of Australian fuels produced a value similar to that observed in the present study (Guérette 2016). The reported enhancement ratio to CO for methane is 7.9 ± 3.4e − 02 ppb ppb −1 from eight smoke events. This value is consistent with that calculated by Paton-Walsh et al. (2014) from open path FTIR measurements on the firegrounds of hazard reduction burns in SE Australian temperate forests. The same is true when comparing the enhancement ratio to CO determined in this study for methanol to that calculated from the results of Paton-Walsh et al. (2014). It is interesting to note that the enhancement ratio to CO values calculated from aged smoke during COALA-2020 are for many species close to those reported in the literature from fresh smoke measured on the fire ground. This suggests that dilution and chemistry effects are not significantly altering the relative amounts of these VOCs in the plume during transport from the emission site to the measurement site.
The only reported species for which the enhancement ratios to CO are significantly higher to those found in the literature is acetonitrile. Acetonitrile was calculated to have an enhancement ratio to CO of 5.0e − 04 ppb ppb −1 using the results of Paton-Walsh et al. (2014). During COALA-2020, the mean repor ted enhancement ratio to CO is approximately twice this value at 9.6 ± 2.5e − 04 ppb ppb −1 . This may suggest a higher emission factor of this species from the Black Summer fires, or the formation of acetonitrile in the plume during transport to the measurement site. Higher enhancement ratios of acetonitrile have been observed outside the Australian context, with a review of field measurements reporting a mean value of 2.0e − 03, twice as large as the COALA value (Huangfu et al. 2021).
The determination of these enhancement ratios reveals that plumes measured at Cataract during the Black Summer bushfires had chemical composition similar to fresh plumes previously observed in southeast Australian temperate forest fires. This is an interesting result considering that sampled smoke had been significantly aged before reaching the measurement site. This finding provides an indication that enhancement ratios determined from firegrounds in temperate Australia may be appropriate for use in estimating population exposure at distant receptor sites in some cases. Determining the composition of these plumes is crucial in assessing the health impacts of exposure to this smoke. The relative toxicological contribution of different atmospheric species as measured during the Black Summer to the health burden of smoke exposure will now be investigated.

Determination of relative toxicological impact
Measurements from COALA-2020 can be related to exposure limit concentrations in order to estimate the potential toxicological contribution for each measured variable. Chemical species included in this analysis are those shown to have a > 1% contribution to the toxicological impact of wildfire smoke exposure in Australia (MacSween et al. 2020) and in the western USA (O'Dell et al. 2020). The enhancement ratios to CO used from sources other than the current COALA-2020 study are presented in Table 3. The sources for these enhancement ratios (MacSween et al. 2020;Guérette et al. 2018) were selected as they are recent values measured in smoke from fires burning in similar forests to those consumed during the Black Summer fires.
Measured concentrations were compared to the Safe Work Australia (SWA) time weighted average exposure limits (Safe Work Australia 2019) for atmospheric pollutants. These limits provide a mean concentration that should not be exceeded over the course of an 8-h working day. No specific limit concentration exists for smoke particles of indeterminate composition, though limits for other types of respirable particles (e.g. wood dust, coal dust, graphite dust) are published between 1 and 5 mg m −3 . A value of 3 mg m −3 is used here for respirable smoke particles given this is the limit specified for other carbon-dominated dusts like coal and graphite dust (Reisen and Brown 2009). This value has also been used in other publications in Australia and overseas (e.g. (Reisen and Brown 2009;MacSween et al. 2020;Miranda et al. 2010)). National Environmental Protection (Ambient Air Quality) Measure (NEPM) (National Environmental Council 1998) limit concentrations are also discussed. Measured concentrations of atmospheric variables were temporally averaged to the resolution specified by the exposure limit for comparison. Limit concentrations included in this analysis are presented in Table 4. A series of mean concentrations were calculated from the smoke-influenced measurements of COALA-2020. An overall mean concentration was calculated for smoke events with sufficient duration for comparison to an 8-h limit concentration. Smoke H (duration 7 h) and the immediately consecutive events Smoke I-L (combined duration 9 h) were selected, as were the mean concentrations determined for the entire smoke-influenced period. For species listed in Table 3, which were not directly measured, the cited enhancement ratio to CO was used to calculate an assumed concentration. Mean concentrations were divided by exposure limit concentrations listed in Table 4 to obtain a percentage of exposure for each species. Following the methodology MacSween et al. (2020), atmospheric pollutants were grouped into toxicological classes with similar physiological effects from the MIXIE web tool (Vyskocil et al. 2007). A tabulation of these classes, the regions of the body they impact and the atmospheric pollutants that fall into each class is available in MacSween et al. (2020). Finally, the percentage exposure for each species in each toxicological class was summed together to provide a cumulative exposure estimate for each class. The physiological effect of each class is summarised in Table 5. Vyskocil et al. (2007) provide a more thorough description of these classes.

Relative contribution of atmospheric variables to toxicological load
The exposure limits selected for comparison in this study are time-weighted average workplace exposure limits set by SWA (Safe Work Australia 2019), the national government agency responsible for work health and safety. These are limits that must not be exceeded in the workplace over an 8-h working day, assuming a standard 5-day week. This limit is therefore not designed to be a measure for continuous exposure, nor the limit of exposure to air pollutants for vulnerable members of the population such as elderly people or those with respiratory illness. The concentration limits set by SWA are higher than those reported in the national measure for ambient air quality-the NEPM (National Environmental Council 1998). For example, the 8-h CO NEPM standard is 9 ppm, compared to a SWA limit of 30 ppm averaged over the time period. The SWA limit for respirable particles (classified as PM 3 ) is 1-5 mg m −3 , compared to the NEPM daily value of 0.025 mg m −3 . The NEPM ambient exposure limits for SO 2 and NO 2 are also much lower than their SWA equivalents. The NEPM framework is more suitable for contextualizing the health risk of exposure to air pollutants in the ambient setting but does not contain limits for VOC species. This must be kept in mind when assessing long-term population exposure to wildfire plumes such as those produced during the Black Summer. This gives rise to near-continuous exposure to pollutants for large populations, including vulnerable populations, for a period far extending the 8-h TWA averaging time. Results from smoke events H and I-L are presented in Fig. 6 and discussed, along with the mean results for all events.
The breakdown of species contribution to the cumulative toxicological load of exposure to Black Summer plumes is consistent between events as presented in Fig. 6. This suggests that any differences in composition observed within the plumes sampled during COALA-2020 have limited effect on the toxicological impact of exposure. The percentage of the SWA limit observed in the dilute smoke sampled at Cataract is relatively low even during smoke events (~ 2%) whilst averaged over the entire smoky period of measurements the  Disruption of oxygen transport 9 Antithyroid effect 11 Liver disorder 13 Kidney disorder 19 Central nervous system disorder 27 Skin disorder 30 Embryonic or foetal disorder 32 Cancer concentration limits are closer to ~ 1% of the SWA limit. It must be noted here that these percentage-of-limit exposure values are calculated from workplace exposure limits. The measurement site was also distant from smoke sources, and therefore plumes impacting the site were significantly diluted compared to plumes impacting population centres closer to active fires. It must also be noted that observations from COALA-2020 were taken weeks after the peak fire activity. The cumulative week-to-month long exposure and exposure of vulnerable populations to plume from the Black Summer had serious health impacts (Ryan et al. 2021;Walter et al. 2020) despite low reported exposure percentages from the COALA-2020 measurement site. The relative contribution of particulate and gaseous species is important in the cumulative toxicological burden of exposure to plumes measured during COALA-2020. Respirable particles as PM 3 (measured as PM 2.5 during this study) contribute significantly to the cumulative exposure burden for each class they are included in: Classes 1, 2, 3, 27 and 32. In each of these classes for all analysed periods, the contribution from respirable particles is more than 50%. Formaldehyde also contributes significantly to each of these classes (30-39% for all analysed periods) Acrolein also contributes to Classes 1, 2, 3, 27 and 32 with contributions between 15 and 24%. Carbon monoxide has the dominant (> 95%) contribution to Class 4. Overall, respirable particles contributed 52-53% to the total exposure loading during Smoke H and during the entire smoky period. During smoke events I-L, the particulate contribution was 63%.
These results are very similar to those found in Mac-Sween et al. (2020) who used enhancement ratios derived from fresh smoke across a number of fires to find respiratory particulates, formaldehyde, acrolein and CO were the major contributors to cumulative exposure estimates. Formaldehyde and acrolein were also found to dominate the acute health risk hazard index calculated by O'Dell et al. (2020), who also reported significant contribution from benzene from VOC measurements in western US wildfire plumes. Interestingly, O'Dell et al. report the acute risk from acrolein in smoke decreases as smoke ages. The similar contribution of acrolein to the exposure contributions in Fig. 6 provides another piece of evidence that mostly aged plumes measured during COALA did not vary significantly (i.e. by days) in age. It must also be noted here that the enhancement ratio Other to CO used to calculate formaldehyde concentrations was sourced from measurements of fresh smoke. Formaldehyde is known to be a product of smoke plume chemistry in southeastern Australian bushfire plumes (Young and Paton-Walsh 2011), and as such concentrations of this compound can increase after emission. The real concentrations, and hence contribution of formaldehyde to the percentage cumulative exposure, could be higher than reported.
Smoke measured during COALA-2020 was aged and diluted. Although the maximum measured CO concentration (> 1 ppm) during the campaign is many times greater than the background expected for the site, much higher pollutant concentrations associated with smoke plumes form the Black Summer were observed in other regions of Australia (Health Protection Service 2021). For example, CO concentrations greater than 20 ppm and PM 2.5 concentrations greater than 1000 µg m −3 in early January (Health Protection Service 2021) in Canberra, 200 km from the measurement site. The relatively homogeneous chemical composition of distinct smoke plumes observed during COALA-2020 allows the estimation of plume composition in sites closer to active fires impacted more heavily by Black Summer plumes. To assess the relative toxicological impact of atmospheric species on exposure to smoke plumes from the Black Summer, a case study site impacted heavily by undiluted plumes has been performed.

Application to a case study site: Canberra, Australia
A case study of population exposure in an urban centre heavily impacted by undiluted plumes from the Black Sumer fires was also completed. The city of Canberra ( Fig. 1) was selected as the smoke impacts were particularly severe during the Black Summer. Canberra is an Australian city located approximately 200 km southwest of Cataract (Fig. 1). The air quality measurements from each of Canberra's three AQMSs were downloaded from the public data repository (Health Protection Service 2021). The AQMS at Monash was selected for analysis, as this suburb has the largest population of the three AQMS sites in Canberra. The PM 2.5 concentrations were assessed for other sites (Civic and Florey) and found to be of similar magnitude to those observed at Monash. The CO record at Florey was also similar to that recorded at Monash. There is no Civic CO record for the analysed period. It is therefore assumed that the Monash concentrations are regionally representative of Canberra's air quality for the measured period, especially considering the spatially large pollutant source and proximity of the three AQMSs.

Toxicological contributions in Canberra
Figure 7 plots hourly CO and PM2.5 measurements from an air quality monitoring station (AQMS) in Monash, Canberra during the period affected by smoke from the Black Summer fires. The 8-h NEPM standard for CO was exceeded for 2 days, January 01 and January 02. The maximum 8-h mean CO concentration recorded at the Monash AQMS was 15.9 ppm. This is 176% of the NEPM limit concentration for this pollutant. The maximum 24-h mean PM 2.5 concentration observed at Monash was 968 µg m −3 . This is 38 times the NEPM daily limit concentration of 25 µg m −3 . As evident in Fig. 7, the hourly mean PM 2.5 concentration was greater than the NEPM daily limit for the majority of the smoke-effected period from early December 2019 and early February 2020. Forty-nine of 64 days between December 07, 2019, and February 08, 2020, had daily mean PM 2.5 concentrations greater than 25 µg m −3 . The above statistics demonstrate that air quality was poor in Canberra during the Black Summer, and especially bad on January 01, 2020, and the following week.
Considering this, the health effects of exposure to this smoke was calculated for three selected periods: the median 8-h rolling mean CO and PM 2.5 concentrations from the entire smoke affected period (December 08-February 08), the median 8-h rolling mean CO and PM 2.5 concentrations from the worst-affected week (December 31-January 06) and the maximum 8-h rolling mean CO and PM 2.5 concentrations (January 01). Rolling mean 8-h CO concentrations were calculated for each of these periods, and a bootstrap median 8-h mean CO concentration retrieved. From these median 8-h concentrations, cumulative exposure estimates and the relative toxicological contributions of each of the species listed in Table 2 were calculated using the method described above. The results of this analysis are presented in Fig. 8. A limited number of species dominate the toxicological contributions. Respirable particles (as PM2.5) are again the dominant contributor to toxicological classes 1,2,3, 27 and 32. For the entire smoke-affected period and for the worst week of air quality particulate contributions in classes 1, 2, 3, 27 and 32 were 54-71%. During the worst day of air quality, particulate contributions were lower (39-52%), suggesting higher relative contributions of CO. Formaldehyde and acrolein also significantly contribute to classes 1, 2 and 3, with formaldehyde also prominent in classes 27 and 30, similar to the results from Cataract. The overall particulate contribution was 51% when considering the entire smoke period, 56% when considering the worst week of air quality and 36% when considering the worst day of air quality. The identity and magnitude of chemical species contribution are similar when comparing calculated concentrations in Canberra to (largely) measured concentrations at Cataract. This similarity suggests the enhancement ratios determined from Other measurements and selected from literature are appropriate for use in this context. Workplace exposure limits are designed for workers who may be exposed for eight hours of a 24-h cycle. However, concentrations of CO, PM 2.5 and other atmospheric pollutants were elevated consistently on a timescale of weeks to months in many Australian cities during the Black Summer. Daily workplace toxicological loads were multiplied by three for the worst day of air quality observed in Canberra. This is an attempt to scale an exposure limit for 8 h out of a 24-h cycle to continuous 24-h exposure. When this scaling is applied, five toxicological classes exceed 200% of the adjusted exposure limit (Classes 1, 2, 3, 27, 32). One other class (4) exceeds 100% of the adjusted exposure limit. A similar experiment was performed on the weekly values calculated above. Weekly values were scaled by a factor of 40/ (24*7), representing the proportion of a 40-h working week specified in the SWA exposure limits of the total hours in a week. This scaling results in exposure limit exceedances for two classes (Classes 1 and 2), and exposures > 80% of the limit for three further classes (3, 27, 32).

Implications
Enhancement ratios to CO from aged smoke measured during the historic Black Summer Australian bushfires have been determined in this analysis. For many gas-phase and particle species, the values determined from the present experiment are not significantly different to those found in literature calculated from fresh smoke measurements in similar forest types. The determination of these values from aged smoke is significant as it provides a starting point for future detailed epidemiological study of the health effects of smoke exposure in receptor regions distant from the fire source. This type of exposure is common in the Australian context. With the frequency and magnitude of fire events in southeast Australia predicted to increase in the future (Keywood et al. 2013), along with the population of the Australian eastern seaboard, it is critical to have an understanding of the composition of aged smoke plumes from landscape scale fire events. The publication of enhancement ratios for harmful atmospheric species will allow for the modelling of population exposure to diverse chemical species at receptor sites using a CO tracer. However, it must be noted that not all species of concern were measured during COALA-2020. Of particular note is formaldehyde, as it contributes significantly to the exposure health burden (MacSween et al. 2020;O'Dell et al. 2020) and has a known source in the Australian bushfire smoke. Furthermore, fine particle variables were not observed to correlate strongly with CO during the present measurements. Therefore, more research is required to better understand the physics and chemistry that drives variability in these species, which are of particular concern from a health perspective given their ability to penetrate deeply into the respiratory system. It must also be noted that spatio-temporal heterogeneity of smoke plume composition is being assumed in the reporting of these values. It is understood that combustion type, fuel type and meteorological conditions all influence the emissions from wildfire, and therefore this assumption is not ideal. The variability in calculated enhancement ratio to CO values presented in Fig. 4 illustrates this. A recent study has also demonstrated that despite the landscape-scale pollutant source, spatio-temporal patterns in pollutant concentrations have been observed in the Sydney region. This reinforces the need to use locally measured CO tracer values for calculation of exposure risk when possible (Di Virgilio et al. 2021).
The calculated toxicological contributions of select atmospheric species also appears similar to analyses performed on fresh smoke (MacSween et al. 2020). The importance of considering the additive effects of species impacting the same part of the body is becoming increasingly recognised. The relatively low percentage of Safe Work Australia exposure limits reached during measurements of COALA-2020 highlights the inadequacy of using 'workplace' exposure limits for assessing ambient exposure. An ambient exposure limit for chemical species beyond those covered in the NEPM is recommended. The United States Environmental Protection agency provides a relevant document for assessing the risk of ambient exposure to identified Hazardous Air Pollutants (US EPA 2020).
The relative toxicological contribution of measured chemical species from COALA-2020 also provide some necessary insight. The importance of particulate species to the health burden of smoke exposure is highlighted, with respirable particles contributing > 50% to cumulative exposure values for most averaging periods in Canberra and Cataract. However, the contribution of infrequently measured gases such as formaldehyde and acrolein cannot be discounted. As these gases are not included in the NEPM or widely measured, their contribution to the toxicological load from smoke plumes may be being neglected. It should also be noted that the air quality impacts of the Black Summer fires demonstrate that traditional public health measures to lessen exposure risk must be adapted to account for events on the scale of the Black Summer. Near-continuous PM 2.5 concentrations observed in Canberra > 25 µgm −3 over an 8-week period (Fig. 6) invalidates the traditional advice for at-risk populations to limit time spent outside during smoke-affected periods. The contribution of gaseous species to the toxicological loading in plumes measured during the Black Summer also highlights the need to develop health advice beyond the use of personal face masks. Masks can be effective at particulate removal but have little impact on the amounts of gaseous pollutants inhaled.