A Clean Air Plan for Sydney: An Overview of the Special Issue on Air Quality in New South Wales

The contribution of the Soil Factor to PMmass is consistently about 10% of the total mass across the sites, while Sea Salt decreased with distance from the coast and Smoke increased with distance from the coast, (but together these two classes make-up about 25%–50% of the mass of PMat all sites) [ 141 ]. The largest contribution to PMwas from industrial sources (primary and secondary) at all except the most inland site, where Woodsmoke and industry sources made an equal contribution of 40% with most of the industry component from secondary processes (80%). At most sites, primary emissions accounted for approximately 30%, and secondary reactions accounted for approximately 70% of the industry source [ 141 ]. Studies that identify the major sources of atmospheric fine particulate matter are extremely useful in helping to prioritise efforts to reduce atmospheric concentrations by emissions control measures.

Further north of Sydney is the Hunter Valley, where domestic housing is located close to major industrial sources, and communities are concerned about the impact of coal mining and associated industrial emissions on their health. Such public concerns resulted in two studies in the Hunter Valley looking at the chemical composition of particulate matter [ 141 145 ]. Two sampling periods (2012 and 2014) at six sites in the Hunter Valley and across two size fractions (PMand PM) were input to a receptor model, to determine the source of particulate matter influencing particle composition at the sites. Fourteen factors were found to contribute to particle mass [ 141 ]. Of these, three source profiles common to all sites, size fractions, and sampling periods were Sea Salt, Industry-Aged Sea Salt, and Soil. Four source profiles were common across all sites for PM, including Secondary Sulphate, Secondary Nitrate, Mixed Industry/Vehicles, and Woodsmoke. One source profile (Other Biomass Smoke) was only identified in PMat the two sites furthest from the coast, and two source profiles (Mixed Industry/Shipping and Vehicles) were only identified in PMat the four sites closest to the coast [ 141 ].

There were two days of extreme heat during the MUMBA campaign (with temperatures exceeding 40 °C). This provided a good test case to model the impact of extreme temperatures on Oformation in Sydney [ 143 ]. Ratios of biogenic VOCs measured at the MUMBA site on these hot days were different from those measured on other days of the campaign, but further measurements are needed to understand to what degree this resulted from the different vegetation types being sampled and how much was caused by the extreme heat [ 139 ].

Wollongong is well known in New South Wales as an industrial city; nevertheless, traffic emissions were shown to be the main driver of NOconcentrations at the MUMBA site, and the air-shed was VOC-limited [ 6 ]. The daily average mass concentration of fine particles (PM) was low (6.1 µg m) during the MUMBA campaign [ 142 ]. The particle number concentration was dominated by ultrafine particles (particles with a diameter between 3 and 100 nm), with an average and median value of 7.0 × 10cmand 5.2 × 10cm, respectively. Eight particle formation and growth events were identified from the particle number size distribution range from 14 nm to 600 nm dataset [ 142 ]. Particle formation and growth events occurred in air masses that travelled from the ocean and passed through populated areas, including Sydney. Anthropogenic sulphate, the photochemical age of air masses and relative humidity potentially played a role in the particle formation and growth events. Sources of particles identified included traffic emissions, industrial activities, and the marine atmosphere [ 142 ].

The MUMBA campaign ran for eight weeks from mid-December 2012 to mid-February 2013, providing a rich dataset of atmospheric composition at the marine/urban/forest interface. An episode of clean marine air enabled the measurement of background concentrations of key species (including a number of VOCs) at these latitudes [ 139 ]. MUMBA also provided interesting observations of isoprene from nearby vegetation and other biogenic VOCs, which predominantly originated from the forested escarpment that surrounds much of Wollongong. These natural biogenic emissions play an important role in the control of air quality in Australian cities, due to the remoteness from other polluting sources and the relatively low local anthropogenic emissions [ 139 ].

An important aspect of air quality management within a city is the accuracy and reliability of the operational air quality model for the region. The establishment of the CAUL hub provided the opportunity to undertake a major air quality modelling comparison that could benchmark the available air quality models against each other and against previously established standards for performance (see Section 2.2 ). However, in order to elucidate different aspects of why one model may outperform another, it is important to have very detailed atmospheric composition data that includes a range of species that are not routinely measured as part of the regulatory air quality monitoring network. The Australian atmospheric chemistry community had undertaken three relevant measurement campaigns to gather detailed atmospheric composition data, but these data had not been through their final quality assurance procedures that were required before publication. In the early stages of CAUL, efforts were concentrated on finalising these datasets from the two Sydney Particle Study campaigns (SPS1 and SPS2) [ 32 34 ] and from the Measurements of Urban, Marine, and Biogenic Air (MUMBA) campaign [ 6 140 ] from the southern city of Wollongong. These three campaigns were used as the basis of the modelling comparison described in Section 2.2 . In addition, work has been undertaken to finalise the data gathered during two campaigns in the industrialised Hunter Valley of New South Wales [ 141 ].

Outside of the major capital cities, Australia is a large and sparsely populated nation. Most of the air quality monitoring activities are undertaken in the main cities; however, remote-sensing technologies make it possible to expand some of this coverage to the whole population. Satellite-derived estimates of the total-column aerosol optical thickness (AOT) and tropospheric NOcolumn density have been shown to be sensitive to concentrations in the boundary layer, and for both aerosols and NO, concentrations tend to be highest in the lowest layers of the atmosphere [ 136 ]. This has been successfully combined within land-use regression models to predict monthly or annually-averaged NOor PMconcentrations at the earth’s surface [ 137 138 ]. Such methods have the advantage that they are not limited by state or local government boundaries and offer relative consistency in their method.

Analysis of the smoke pollution events determined that (for the measured components) the chemical composition of smoke was very similar, whether the smoke originated from domestic wood-heaters or from hazard reduction burns. During the study period in Auburn, hazard reduction burns were a greater immediate acute threat to public health (because peak concentrations of particulate matter were highest during these events). However, domestic wood-heater events produced greater cumulative exposure during the campaign, due to the greater duration of enhanced pollution from this source, so, also presenting a public health threat. Whilst both of these pollution sources vary from year to year, this study highlighted the significance of pollution from domestic wood-heaters in Sydney as an issue of importance both for the public and for policy-makers.

The urban concentrations observed in this study imply that NHis the limiting reagent for production of NHNOaerosol, but for (NHSO, SOis the limiting reagent [ 135 ]. This finding provides further evidence to support changing the legislation to reduce the maximum permitted sulfur levels in shipping fuels and vehicle petroleum.

NH 3 :CO ratios were strongly correlated with traffic volumes on nearby roads, implying that the main source of NH 3 at the site is from traffic exhaust fumes, via the operation of catalytic converters. (The NH 3 :CO ratio will decrease if the emissions are not fresh due to the shorter atmospheric lifetime of NH 3 compared to CO.)

In conjunction with the suburban balcony case-study described above, novel atmospheric composition measurements were made along an integrated open-path of nearly 400 m between a Fourier transform infrared spectrometer and an array of mirrors, across the centre of Auburn. The spectrometer operated for approximately nine months between October 2016 and September 2017, (with a break between March and May 2017) and made routine measurements of carbon dioxide (CO 2 ), CO, methane (CH 4 ), and ammonia (NH 3 ). Concentrations of methanol (CH 3 OH), acetylene (C 2 H 2 ), ethylene (C 2 H 4 ), and formaldehyde (CH 2 O) were also measured during episodes of enhanced pollution.

A companion study investigated the comparative ability of four different indigenous tree species in NSW to remove particulate matter from the atmosphere. This study showed that evergreen trees absorb particulate matter into their leaves, whilst in deciduous trees, the particulate matter deposits onto the leaf surface and can get washed off. This means that deciduous trees’ ability to scavenge particles gets renewed after rainfall events [ 134 ].

Another suggestion from the CAUL “road-shows” was to carry out a research project looking at urban greening (and mosses in particular) to mitigate air quality impacts. Research in other cities has shown that roadside trees can either decrease or increase local concentrations of air pollutants depending upon the degree to which they hinder the dispersion of pollutants, with hedges being shown to be a particularly good choice of vegetation barrier, and green roofs also an effective air pollution abatement measure [ 24 ]. Moss proved to be even more efficient at removing particulate matter from the atmosphere than the nearby native tree species that were tested [ 133 ]. This study compared the particulate matter entrapment by roadside moss turfs with that of leaves of a common native tree in the coastal city of Wollongong, NSW, Australia. Plant samples were collected from nine sites on an urban gradient, in three urban classes based on road type: low (quiet roads in peri-urban suburbs), medium (busy suburban roads), and high (freeway-type). Chlorophyll fluorescence, a common measurement of photosynthetic efficiency, was also measured as a proxy for plant stress. By dry weight, moss trapped more than the leaf samples. In addition, greater amounts of total particulate matter were trapped by mosses at the more urbanized sites, implying a positive trend along the urban gradient. The trend in particulate matter trapped by moss was similar to the trend in average ambient PMconcentrations measured for two weeks at one site from each urban class, by the deployment of a mobile sensor. The sampled vegetation was also increasingly stressed along the urban gradient (although the exact physical or chemical causes of the stress are unknown): the photosynthetic efficiency of tree leaves declined by 2% from low to high urbanisation, while moss photosynthetic efficiency declined by 40%, indicating a steeper stress gradient for mosses. While the trees appeared to be less affected, both plant types appeared to respond to urbanisation by increasing wax deposition [ 134 ].

Our case study in Auburn showed that the New South Wales air quality monitoring network provided a good representation of pollution levels at our chosen “balcony” site. Although this result cannot be generalised to all balcony locations in western Sydney, it does demonstrate the effectiveness of the regional air quality monitoring network in this case. In contrast, average roadside PMconcentrations in the sampled areas of Randwick were found to be approximately twice those measured at nearby air quality monitoring stations. We also found very high spatial variability of PMat roadside locations, meaning that roadside air quality cannot simply be evaluated by locating an air quality monitoring station at a single roadside location. Instead, estimates of the average increase of common pollutants at roadside locations (compared to regional background values) are needed to supplement regional air quality monitoring. The heightened concentrations at intersections and near bus-stops should give additional weight to the recommendation of the broad-scale adoption of anti-idling emissions control technologies in on-road motor vehicles, and improvements in road design, such as bus lanes that move the bulk of traffic further from the curb. In future, improved estimates could be made by a network of fixed roadside sensors that operate year-round, but currently, the technology is still developing [ 132 ]. From these case studies, we conclude that the existing air quality monitoring network in New South Wales is likely to be fit for purpose, with respect to representing urban background pollutant concentrations, and that outreach programmes should be undertaken to inform the public of simple steps that can be taken to minimise their exposure.

The second case study examined roadside concentrations of PMduring an intensive three-day campaign. PMconcentration measurements were made in the vicinity of a major road (known to carry heavy traffic) in the Sydney suburb of Randwick. Observed PMconcentrations were compared to regional urban background levels, and the spatial and temporal variations were analysed [ 131 ]. This study showed a highly variable spatial distribution of PMalong the main road studied. The average PMroadside concentration recorded was 13 µgm, which was approximately twice the concentration of the nearby regulatory air quality network sites. Those people residing at, (or working for long hours outdoors at), busy roadside locations are, therefore, likely to be at enhanced risk of suffering detrimental health effects associated with air pollution. PMlevels were observed to decrease by 30% at a distance of 50 m away from main road intersections, suggesting that pedestrians and cyclists should use side-streets whenever possible. PMconcentrations were recorded to be 50% higher in the morning rush hour than the afternoon rush hour at roadside locations, implying that joggers and cyclists can reduce their PMexposure by choosing to exercise in the afternoons rather than the mornings, (although avoiding busy road locations whenever possible is advised).

In the first case-study, ambient air quality measurements were made on the roof of a two-story building in the Sydney suburb of Auburn, to evaluate conditions that might represent a typical suburban balcony site. Measurements made at the balcony site were then compared to data from three nearby regulatory air quality monitoring stations [ 130 ]. Overall, the air quality at the balcony was similar to that measured at the regulatory sites. Average Oand PMconcentrations were lowest at the Auburn balcony site; nitrogen oxides were within the range measured at the other sites, and carbon monoxide was highest at Auburn. Considering that Oand PMare the pollutants of most concern in Sydney, we concluded that the existing air quality network provides a satisfactory indication of concentrations of outdoor air quality pollutants at the selected “balcony” site at Auburn.

There is increasing public concern about air quality in Sydney. This was reflected in a common question raised at the CAUL “road-show” events, which was some variation of: “are the pollution concentrations at air quality monitoring network sites around Sydney truly representative of what I am exposed to in my everyday settings”? Undeterred by the fact that this is an unanswerable question (due to the huge variability in peoples’ daily lives), we set about attempting to provide some insights to this issue via two separate case-studies:

To finalise and publish the atmospheric composition data from a number of previous measurement campaigns so that these could be used for rigorous testing of the performance of different air quality models over New South Wales.

2.2. CAUL Air Quality Modelling Comparison and Modelling Studies

Air quality modelling is an important aspect of the management of atmospheric pollution in any community. It allows for public warnings to be issued when air quality is predicted to be poor, as well as providing insights to the causes of different pollution events and analysis of the impact of future emission scenarios. A major undertaking within CAUL was the first comparison of hourly air quality models over Sydney, using a suite of six air quality modelling systems, over the time periods of the SPS1, SPS2, and MUMBA campaigns. The comparison resulted in improvements to the implementation of models over Sydney and demonstrated that air quality modelling over the greater metropolitan regions of New South Wales can meet international standards of performance [ 146 147 ].

148,

Such modelling comparisons help cross-validate the models, test their skill at reproducing observed atmospheric composition, and identify any flaws or problems in the way that the models are set up or run [ 53 147 ]. At the end of the comparison exercise, the validated models may be used to undertake a number of different studies with added confidence in the modelled output [ 143 149 ].

The air quality model comparison used two separate meteorological models (CCAM and WRF) with a total of seven different configurations and was conducted over consistent geographical domains, grid resolutions, and time periods [ 147 ]. The modelling domains were nested so that the outer grid covered the whole of Australia at 80 km resolution, whilst the innermost (of four) grids covered the Sydney basin at 3-km resolution. Comparison of the meteorology within the models identified systematic overestimates of wind speeds that were more pronounced overnight, which is a common weather model bias [ 147 ]. The temperatures were well simulated, with the largest biases also seen overnight. The models tended to have a drier lower atmosphere than observed, implying that better representations of soil moisture and surface moisture fluxes would improve the subsequent air quality simulations [ 147 ]. The local-scale meteorological features, such as the sea breeze, which is a critical feature driving ozone formation in the Sydney Basin was reasonably well-represented in the simulations. Overall, the biases between simulations and observations were generally within the recommended benchmark values with the exception of extreme (both high and low) events, when the biases tended to be larger [ 147 ].

222Rn-based stability classification technique [

The main driver of the interaction of meteorology and air quality is the degree of atmospheric mixing that acts to dilute ground-level emissions of primary pollutants. Chambers et al., 2019 [ 5 ], show the usefulness of radon as a tool for inferring atmospheric boundary layer heights to better constrain the atmospheric mixing within air quality models. In this study, the modelling comparison results were evaluated within different atmospheric “class-types” defined over 24-h periods using aRn-based stability classification technique [ 5 ]. Calculating hourly distributions of observed and simulated quantities within each class-type helped: (i) bridge the scale gap between simulations and observations, (ii) separately represent the variability associated with spatial and temporal source heterogeneity rather than it adding to bias values, and (iii) provide an objective way to group results over whole diurnal cycles that separates uncontrollable sources of uncertainty (synoptic non-stationarity, rainfall, mesoscale motions, extreme stability, etc.) from parameterisation problems, or between-model differences [ 5 ]. Meteorological model skill varied across the diurnal cycle for all seven models, with an additional dependence on the atmospheric mixing class that varied between models. Model skill regarding air quality varied strongly as a function of mixing class and was typically worst when public exposure would have been the highest (during episodes of poor air quality). This has important implications for using contemporary models to assess potential health risks in new and rapidly evolving urban regions [ 5 ].

2.5, CO, and NOX in Sydney. The methodology used could easily be applied in other parts of the world [

The CAUL hub was particularly interested in exploring Indigenous knowledge and perspectives. The annual cycles in meteorological variables in Sydney were used to identify a set of quasi-seasons using a combination of Indigenous knowledge, statistics, and historical data from the Bureau of Meteorology in Australia [ 150 ]. This approach was particularly successful in identifying the coldest time of year, when atmospheric mixing is at its lowest, and there are peak concentrations of PM, CO, and NOin Sydney. The methodology used could easily be applied in other parts of the world [ 150 ].

3, under-predicts PM2.5 and PM10 during SPS1 and MUMBA and over-predicts PM2.5 and under-predicts PM10 during SPS2. These biases are attributed to inaccurate meteorology, precursor emissions, insufficient SO2 conversion to sulphate, inadequate dispersion at finer grid resolutions, and under-prediction in secondary organic aerosol [

Improvements to air quality forecasts may be gained by increasing the complexity of models (such as coupling to ocean models), although at the cost of greater computing resources [ 151 152 ]. The relative performance of the Weather Research and Forecasting model with chemistry (WRF/Chem), with and without coupling to the Regional Ocean Model System (ROMS) (WRF/Chem-ROMS), was shown in two paired papers [ 151 152 ]. WRF/Chem-ROMS generally performs well at 3-, 9-, and 27-km resolutions for sea-surface temperature and boundary layer meteorology, despite larger under-predictions for total precipitation due to the limitations of the cloud microphysics scheme or cumulus parameterisation [ 152 ]. The model also performs well for surface O, under-predicts PMand PMduring SPS1 and MUMBA and over-predicts PMand under-predicts PMduring SPS2. These biases are attributed to inaccurate meteorology, precursor emissions, insufficient SOconversion to sulphate, inadequate dispersion at finer grid resolutions, and under-prediction in secondary organic aerosol [ 152 ]. The use of finer grid resolutions (3- or 9-km) can generally improve the performance for most variables.

3, ±25% for CH2O, ±30% for NO2, ±35% for hydrogen peroxide, ±50% for SO2, ±60% for isoprene and terpenes, ±15% for PM2.5, and ±12% for PM10 [3 predictions at 3-km for all field campaigns, surface PM2.5 predictions at 3-km for SPS1 and MUMBA, and surface PM10 predictions at all grid resolutions for all field campaigns. The chemical boundary conditions were shown to be more important in the relatively clean Southern Hemisphere, than in the more polluted Northern Hemisphere [

In the companion paper [ 151 ], the performance of WRF/Chem and WRF/Chem-ROMS are compared for their applications in Australia. The explicit air-sea interactions in WRF/Chem-ROMS led to substantial improvements in simulated sea-surface temperature, latent heat fluxes, and sensible heat fluxes over the ocean during all three field campaigns, which led to better performance of WRF/Chem-ROMS in boundary layer meteorology [ 152 ]. The percentage differences in simulated surface concentrations between the two models were mostly in the range of ±10% for CO, OH, and O, ±25% for CHO, ±30% for NO, ±35% for hydrogen peroxide, ±50% for SO, ±60% for isoprene and terpenes, ±15% for PM, and ±12% for PM 151 ]. The satellite-constrained chemical boundary conditions reduced the model biases of surface CO, NO, and Opredictions at 3-km for all field campaigns, surface PMpredictions at 3-km for SPS1 and MUMBA, and surface PMpredictions at all grid resolutions for all field campaigns. The chemical boundary conditions were shown to be more important in the relatively clean Southern Hemisphere, than in the more polluted Northern Hemisphere [ 151 ].

3 and PM2.5. For this reason, the final paper from the comparison exercise presents the overall performance of the six air quality modelling systems in predicting O3 and PM2.5, during the SPS1, SPS2, and MUMBA campaigns [3 was evaluated against measurements at 16 air quality monitoring stations. Performance for domain-wide hourly O3 was good, with the models generally meeting benchmark criteria for normalised mean bias (<15%) and correlation (>0.5). The models also reproduced the observed O3 production regime (based on the O3/NOX indicator) at 80% or more of the air quality monitoring sites. When the model output is paired with the observations, all models tend to overestimate the lowest observed hourly O3 values and underestimate the highest observed hourly O3 values; as has been observed in other comparison exercises (e.g., [3 values above 60 ppb at specific sites was generally low (0%–67%). This probability increased to 25%–80% when testing the models for daily maximum O3 values above 60 ppb in a specific region (e.g., Sydney East, Sydney North-West). Relaxing the test further to domain-wide detection of events only marginally improved the probability of detection (28%–93%) but greatly reduced the number of false alarms, with the false alarm ratio decreasing each time the test was relaxed (False alarm ratio, domain-wide: 10%–40%; region: 32%–73%; site: 40%–100%).

The continued monitoring of air quality in Sydney has shown that the city experiences exceedances for only two of the regulated pollutants, Oand PM. For this reason, the final paper from the comparison exercise presents the overall performance of the six air quality modelling systems in predicting Oand PM, during the SPS1, SPS2, and MUMBA campaigns [ 153 ]. Model performance for Owas evaluated against measurements at 16 air quality monitoring stations. Performance for domain-wide hourly Owas good, with the models generally meeting benchmark criteria for normalised mean bias (<15%) and correlation (>0.5). The models also reproduced the observed Oproduction regime (based on the O/NOindicator) at 80% or more of the air quality monitoring sites. When the model output is paired with the observations, all models tend to overestimate the lowest observed hourly Ovalues and underestimate the highest observed hourly Ovalues; as has been observed in other comparison exercises (e.g., [ 154 ]). The probability of the models predicting daily maximum Ovalues above 60 ppb at specific sites was generally low (0%–67%). This probability increased to 25%–80% when testing the models for daily maximum Ovalues above 60 ppb in a specific region (e.g., Sydney East, Sydney North-West). Relaxing the test further to domain-wide detection of events only marginally improved the probability of detection (28%–93%) but greatly reduced the number of false alarms, with the false alarm ratio decreasing each time the test was relaxed (False alarm ratio, domain-wide: 10%–40%; region: 32%–73%; site: 40%–100%).

Performance for PM2.5 was assessed using measurements at five air quality monitoring stations during the summer campaigns (SPS1 and MUMBA) and four stations during the autumn campaign (SPS2). Domain-wide model performance for daily PM2.5 (24-h averages) was variable. Most models underestimated PM2.5 concentrations during the summer campaigns and overestimated them in autumn (SPS2). The benchmark criteria for normalised mean bias (<30%) was met by only one model for SPS2 and MUMBA. Most models met the criteria for SPS1. All models met the criteria for correlation (>0.4) during SPS2, and most did during the summer campaigns. The evaluation of the performance of the models for PM2.5 was hindered by the few monitoring sites reporting PM2.5 at the time of the campaigns.

As with many other parts of the world, key challenges remain for air quality modelling, including access to accurate emissions inventories and meteorology for the modelled region [ 6 147 ]. Australia lacks a consistent national emissions inventory. Only sporadically updated regional inventories of varying resolution, composition, and methodology are produced around some of the cities. As a contribution towards improving emissions inventories, an uncertainty analysis has been made of emissions estimates in the NSW Environment Protection Authority’s Air Emissions Inventory for 2008 for the Greater Metropolitan Region [ 20 155 ].

After the completion of the comparison study, when the models have been optimized, and their performance has been validated against observations and the ensemble of other models, it is then possible to benchmark the models against international performance standards and apply the models to address particular issues of interest.

2.5, O3, and NO2 by evaluation against air quality data from the NSW DPIE air quality monitoring network [2.5 concentrations but under-predicted peak values on the most polluted days [2.5 was generally well captured (with a factor of 2) but with some underestimation of the contribution of sea-salt, ammonia, and elemental carbon. NO2 and peak O3 values were also slightly under-predicted. The study also identified possible mechanisms for future improvements in the model, including better characterization of highly variable emissions sources, such as domestic wood-heaters, traffic, and industrial emissions [

Chang et al., 2018, assessed the ability of one of the models from the comparison study (the regional air quality model, the coupled Conformal Cubic Atmospheric Model and Chemical Transport Model (CCAM-CTM)) for the NSW Greater Metropolitan Region to predict concentrations of PM, O, and NOby evaluation against air quality data from the NSW DPIE air quality monitoring network [ 146 ]. Overall, CCAM-CTM performance was shown to be comparable to that of other regional air-shed models reported in the literature [ 146 ]. Generally, the model slightly over-predicted PMconcentrations but under-predicted peak values on the most polluted days [ 146 ]. The speciation of PMwas generally well captured (with a factor of 2) but with some underestimation of the contribution of sea-salt, ammonia, and elemental carbon. NOand peak Ovalues were also slightly under-predicted. The study also identified possible mechanisms for future improvements in the model, including better characterization of highly variable emissions sources, such as domestic wood-heaters, traffic, and industrial emissions [ 146 ].

2.5 and NOX for participants who lived in Western Sydney at the baseline of that study [2 was based on a satellite-based land use regression model, and PM2.5 exposure was based upon a Chemical Transport Model (CTM) [2.5 and NO2 at baseline, with hospitalisation for all respiratory diseases over a seven-year follow-up, was assessed. The median annual concentration of PM2.5 was slightly lower for Western Sydney residents compared with the rest of Sydney (4.1 µg m−3 vs. 4.6 µg m−3); the maximum PM2.5 concentrations were higher for residents in Western Sydney compared with other areas in Sydney (13.8 µg m−3 vs. 8.11 µg m−3) [2 was lower in Western Sydney compared with other areas in Sydney. Similar to the results for the whole of Sydney [

A study utilising the ‘45 and Up’ cohort [ 156 ] estimated the exposure to PMand NOfor participants who lived in Western Sydney at the baseline of that study [ 96 ]. Exposure assessment for NOwas based on a satellite-based land use regression model, and PMexposure was based upon a Chemical Transport Model (CTM) [ 99 ]. The associations between exposure to PMand NOat baseline, with hospitalisation for all respiratory diseases over a seven-year follow-up, was assessed. The median annual concentration of PMwas slightly lower for Western Sydney residents compared with the rest of Sydney (4.1 µg mvs. 4.6 µg m); the maximum PMconcentrations were higher for residents in Western Sydney compared with other areas in Sydney (13.8 µg mvs. 8.11 µg m) [ 99 ]. Median annual concentration of NOwas lower in Western Sydney compared with other areas in Sydney. Similar to the results for the whole of Sydney [ 99 ], no associations between exposure to air pollutants and hospitalisation for all respiratory diseases in Western Sydney were found [ 99 ].

3 formation can be both NOX or VOC-limited, with the prevalence of a NOX-limited regime over Sydney confirmed during days of elevated ozone concentrations during a 2013 heatwave using the WRF-Chem model (as used in the modelling comparison exercise described above) [3 events occurred [3 pollution in Sydney are likely to be exacerbated in future years by a warming climate, with higher temperatures increasing emissions of biogenic precursors and speeding up the chemical production of O3 in approximately equal measures. Similarly, being a NOX limited regime, the benefits of all policy actions to reduce NOX will simultaneously also reduce O3 (i.e., traffic controls and the mitigation of industrial emissions).

formation can be both NOor VOC-limited, with the prevalence of a NO-limited regime over Sydney confirmed during days of elevated ozone concentrations during a 2013 heatwave using the WRF-Chem model (as used in the modelling comparison exercise described above) [ 143 ]. However, they also highlighted the importance of biogenic VOC emissions (from Eucalyptus trees, which are prevalent in the forested regions surrounding Sydney), by showing that when all biogenic VOC emissions were removed from the model, no Oevents occurred [ 143 ]. This study also highlights that problems with Opollution in Sydney are likely to be exacerbated in future years by a warming climate, with higher temperatures increasing emissions of biogenic precursors and speeding up the chemical production of Oin approximately equal measures. Similarly, being a NOlimited regime, the benefits of all policy actions to reduce NOwill simultaneously also reduce O(i.e., traffic controls and the mitigation of industrial emissions).

3 and PM2.5 in Sydney [X or VOC limited [X emissions (except for domestic wood-smoke—which is also a high NOX emitter) [

Two further papers use the CCAM-CTM model to identify the major source contributions to Oand PMin Sydney [ 148 149 ]. The most significant contributions to ozone in NSW come from biogenic VOC emissions, which dominate over anthropogenic emissions [ 149 ]. The relative importance of different emissions varies between geographic regions of NSW, depending on the ozone formation potential of the region and whether it is NOor VOC limited [ 149 ]. Commercial and domestic sources are the largest anthropogenic contributor to ozone concentrations because they combine high VOC and low NOemissions (except for domestic wood-smoke—which is also a high NOemitter) [ 149 ]. Emissions controls on power stations will reduce ozone concentrations in North West Sydney, the Lower Hunter, and Illawarra regions of NSW, whilst traffic emissions control will be the most effective policy to reduce ozone in the South West of Sydney, which is most prone to smog and ozone exceedances [ 149 ].

2.5 concentrations from CCAM-CTM were weighted by the population density (using the 1 km resolution gridded population data from the Australian Bureau of Statistics), [2.5 burden in the NSW Greater Metropolitan Region originates from natural sources (biogenic emissions, sea salt, and wind-blown dust) and 40% from anthropogenic sources. Of the anthropogenic sources, the most significant contributions to overall population-weighted PM2.5 in the NSW Greater Metropolitan Region come from wood-heaters (31%), industry (26%), on-road motor vehicles (19%), power stations (17%), and non-road diesel and marine (6%) [

To assess the impact on residents, the modelled annual averaged PMconcentrations from CCAM-CTM were weighted by the population density (using the 1 km resolution gridded population data from the Australian Bureau of Statistics), [ 148 ]. It was found that 60% of the PMburden in the NSW Greater Metropolitan Region originates from natural sources (biogenic emissions, sea salt, and wind-blown dust) and 40% from anthropogenic sources. Of the anthropogenic sources, the most significant contributions to overall population-weighted PMin the NSW Greater Metropolitan Region come from wood-heaters (31%), industry (26%), on-road motor vehicles (19%), power stations (17%), and non-road diesel and marine (6%) [ 148 ].

These modelling studies provide evidence for policy-makers of the most important source contributors of two of the pollutants of most concern in NSW (O3 and PM2.5). Such studies provide a sound scientific basis for prioritising air quality management interventions to optimise improved public health outcomes.