Impact of Atmospheric Flow Conditions on Fine Aerosols in ...

Aerosol and Air Quality Research, 17: 1646–1659, 2017 Copyright © Taiwan Association for Aerosol Research ISSN: 1680-8584 print / 2071-1409 online doi: 10.4209/aaqr.2017.02.0083 Impact of Atmospheric Flow Conditions on Fine Aerosols in Sydney, Australia Jagoda Crawford*, David D. Cohen, Alan D. Griffiths, Scott D. Chambers, Alastair G. Williams, Eduard Stelcer Australian Nuclear Science and Technology Organisation, Locked Bag 2001 Kirrawee DC NSW 2232, Australia ABSTRACT

We apply a simple objective measure of an airshed’s degree of ventilation and determine the impact on PM2.5 observations at Lucas Heights, Sydney, Australia. We extend the analysis of previous studies, which considered total PM2.5, by: using positive matrix factorisation to split the aerosol mass by source type; and using Radon-222 measurements as an independent indicator of ventilation and mixing. For this coastal airshed we found that for 64% of the time, conditions could be classified into four categories: local recirculation (LRC; 15%), stagnation (19.5%), regional recirculation (RRC; 10.9%), or ventilation (18.6%). Mean PM2.5 concentrations under recirculation (in this study separated into; LRC and RRC) were 33% higher than under stagnation and can be double that of concentrations under ventilation. Since the combination of LRC and RRC events account for around 26% of all events, recirculation effects on PM2.5 concentrations are significant. However, we found that airshed ventilation doesn’t affect PM2.5 concentrations from all sources evenly. Considering the three main sources of total PM2.5 at this site (vehicle exhaust 26.3%, secondary sulfate 23.7% and aged industrial sulfur 20.6%), conditions leading to the highest concentrations differ. The highest vehicle exhaust concentrations occur under LRC, the highest aged-industrial-sulphur concentrations occur under RRC, and secondary sulfur had similarly high concentrations under LRC and RRC. Under LRC the concentration from vehicle exhaust can be up to a factor of 3.9 greater than under ventilation. On a seasonal basis, RRC flow is most likely to occur in summer and spring (the warmer months of the year when sea breezes are more likely), whereas LRC conditions are more likely to occur in autumn and winter. These findings support those of previous studies, indicating that re-circulation can have a significant effect on PM2.5 concentrations in coastal airsheds, and the degree of impact can vary by source type. Keywords: Aerosols; Circulation; Stagnation; Positive Matrix Factorisation; Ventilation. INTRODUCTION

With population growth and industrialisation, many

cities worldwide are experiencing increasing atmospheric particulate matter levels (PM; Liang et al., 2016). Aside from affecting the Earth’s radiative balance (e.g., Charlson et al., 1992; Huebert et al., 2003; Jiang et al., 2013), PM contributes to adverse health effects, both directly and indirectly (e.g., Dockery et al., 1993; Moloi et al., 2002; Russell and Brunekreef, 2009; Lee et al., 2016). As a result, regulatory guidelines are often set to reduce releases of primary anthropogenic pollutants (e.g., USEPA, 2007; NSW, 2011; VES, 2012). In order to determine which sources should be targeted, fingerprint analysis techniques, such as Principle Component Analysis (PCA; Jollife, 1986), Positive Matrix Factorisation (PMF; Paatero and Tapper, 1994) or * Corresponding author.

Tel.: +61 2 9717 3885, Fax: +61 2 9717 9260 E-mail address: [email protected]

Umnix (where species concentrations are apportioned by a principal components analysis using constraints to assure non-negative contributions; Henry, 2002) can be used to determine the relative contribution from the various emission source types (Crawford et al., 2013; Cohen et al., 2014).

Aside from changes in emission sources and strengths, the measured PM concentration at a site can be significantly affected by the meteorological conditions (e.g., Davis and Gay, 1993; Beaver et al., 2010; Levy et al., 2010; Pearce et al., 2011; Crawford et al., 2016a), particularly in confined air-sheds like the Sydney Basin (Crawford et al., 2016b) and Los Angeles (Cass and Shair, 1984). The meteorological processes affecting the accumulation of pollutants in the atmosphere can be classified into local-, meso- and synoptic scale (Levy et al., 2010). Local-scale primarily refers to diurnal boundary layer processes that affect vertical transport and mixing, meso-scale refers to processes such as sea/land breezes and the impact of local terrain (on a scale of few to several hundred km), and synoptic-scale refers to processes occurring on horizontal length scales of 100–1000 km. Within an airshed local concentrations of air pollution can be affected by local emissions, but also by regional transport

Crawford et al., Aerosol and Air Quality Research, 17: 1646–1659, 2017 1647

from neighbouring urban areas and, for coastal sites, can be particularly influenced by land-sea breeze recirculation (Surkova, 2013; Russo et al., 2016). Coastal recirculation has been shown to have an impact on air pollution in many parts of the world (Levy et al., 2009; and references therein). Robinsohn et al. (1992) described a specific event where about 50% of the measured SO2 concentration in the southern coastal plain of Israel was thought to have resulted from recirculation. This can particularly occur when low level inversions occur during the night, causing land breezes, and then a shift to sea breeze occurs in the late morning. Further, the timing of meso-scale circulations such as sea breezes can have a large impact on concentrations due to changes in exposure time of precursors to sunlight/UV. For example, Oh et al. (2006) observed higher O3 concentrations from late sea breezes (i.e., with an onset time later that 1200 local time) than from early sea breezes. The Sydney Basin is bounded by a mountain range to the west, smaller hills to the north and south, and open ocean to the east and recirculation is known to occur (Barros, 2001). Jiang et al. (2016) found that, due to Sydney’s subtropical coastal-basin environment, variations in air quality conditions within the airshed mainly resulted from interactions between meso- and synoptic-scale features. It therefore offers a good opportunity to study the influence of recirculation and ventilation effects on pollution dispersal in a confined coastal airshed by source type.

Allwine and Whiteman (1994) developed a simple technique for classifying these kinds of atmospheric motion based on horizontal wind speed and direction. They identified three main types of flow conditions: (a) stagnation, where wind speeds are very low or calm allowing pollutants to build up locally; (b) recirculation, where polluted air is initially carried away from the source but later returns resulting in higher subsequent pollution concentrations; and (c) ventilation, a situation where polluted air is strongly diluted or replaced by fresh air. This technique has been used in a number of studies to assess the dispersal capabilities of the atmosphere around coastal cities and industrial areas (e.g., Levy et al., 2009; Mohan and Bhati, 2012; Surkova 2013; Al-Khadouri et al., 2015; Russo et al., 2016), as well as nuclear power sites in Korea (Kim et al., 2007) and India (Nankar et al., 2009; Kumar et al., 2013). The Allwine and Whiteman (1994) approach is an objective quantitative measure which provides a straightforward method to assess the dispersal capabilities of the airshed; furthermore, the data needs are small compared to numerical modelling (Russo et al., 2016).

The general influence of local meteorology (wind speed, wind direction, temperature etc.) on source apportioned PM2.5 at Lucas Heights, Australia, was described in a previous study (Crawford et al., 2016c) to indicate the possible pollution increases in the Sydney area that may arise from future climate change. However, more detailed knowledge of the dispersive capabilities of an airshed can be important in pinpointing the conditions under which any changes of the source strengths (e.g., as from additional sources, or the application of mitigation strategies) will have the most impact. Mohan and Bhati (2012) found that NO2 and SO2

concentrations were better correlated with stagnation/ recirculation episodes than PM10. This was attributed to the fact that PM10 is a combination of primary and secondary aerosols (with reaction rates of varying time scales) and the residence time can range from a few hours (due to the larger size fraction) to a few days.

The aim of this study is to examine the influence of stagnation, recirculation and ventilation episodes in the Sydney Basin on concentrations of total PM2.5 measurements from Lucas Heights. This work further extends previous studies, on ventilation recirculation and stagnation, by identifying the impact of the flow conditions on the concentrations of separate sources of PM2.5, information that can be used in the establishment of mitigation strategies. In addition measurements of atmospheric radon (222Rn) concentrations were also available at this site. Radon is an unreactive, naturally occurring, radioactive gas commonly used as a tracer of recent (2–3 weeks) terrestrial influence on an air mass (e.g., Balkanski et al., 1992). In this study we use the Allwine and Whiteman (1994) technique to classify the atmospheric flow conditions and then we use radon concentrations to independently assess the categorisation of flow conditions between stagnation, recirculation and ventilation events (e.g., Chambers et al., 2015). STUDY SITE AND METHODS Site and Study Domain

Fine particle, meteorological, and atmospheric radon monitoring for this study was conducted at the Lucas Heights campus of the Australian Nuclear Science and Technology Organisation (ANSTO). Lucas Heights (34°03’S, 150°59’E) is located 30 km southwest of the Sydney central business district, 18 km from the coast (Fig. 1). Situated near the southern end of the Sydney Basin, land use in the vicinity is a mixture of suburban and natural vegetation (Chambers et al., 2011). Topography of the Lucas Heights region is complex, with 150 m changes in elevation within a 1 km radius of the site, primarily associated with a river valley east of the site. The measurement location is on top of a broad ridge, with significant slopes to the north and south, resulting in drainage flows that reduce the local nocturnal accumulation of pollution or radon.

Wind speed and direction are monitored every 15 minutes at 10 m above ground level with results aggregated to hourly values in post-processing.

A pronounced seasonality in wind direction exists at this site (e.g., Chambers et al., 2011; Supplementary Fig. S1). In the warmer half of the year (Oct–Mar), wind directions typically have a strong easterly component (predominantly oceanic fetch). Average directions tend to switch from south easterly at night to easterly during the day, due to the influence of sea breezes. During the cooler months (Apr–Sep) the wind direction is typically south westerly.

Annual emission of major anthropogenic sources in the Greater Sydney, Newcastle and Wollongong region are available from the New South Wales Environment Protection Agency (NSW EPA 2017a). Total emission of PM2.5 is 30,499 tonnes year–1, of which 43%, 21%, 21% and 10% is


Fig. 1. Topographic map and location of the study site, Lucas Heights. Triangles represent the locations of the coal fired power stations. The WRF domain is presented in the inset.

from industry, off-road mobile, domestic-commercial and on-road mobile sources, respectively. Total emission of NOx is 292,054 tonnes year–1, of which 60%, 30% and 8% is from industry, on-road mobile and off-road mobile sources. Total emission of SO2 is 301,863 tonnes year–1, 98% of which is from Industry. Coal combustion is the main source for electricity generation in the state of New South Wales (contributing to 70% of the electricity; LAEG, 2016). It has been shown (Cohen et al., 2012) that in the Sydney region, coal fired power stations can contribute to about 34–47% of the secondary sulfate fine aerosols. In the Greater Sydney region a large proportion of the power is supplied by eight coal fired power stations outlined in Fig. 1 and further considered in section 3.4.

Wood heaters are used in the region for domestic heating which can contribute to PM2.5 in winter (NSW EPA, 2017b) and bush fires are contributors to PM2.5 in summer (Rea et al., 2016, and references therein). The population density in the vicinity of the site with the major roads is presented in supplementary Fig. S2. Major populations are located to the north of the site (i.e., potential sources of smoke) with less population regions and natural vegetation/woodland in the south. Previous studies have found that smoke from domestic heating in winter can account for 40% of PM2.5 at Liverpool (Cohen et al., 2011) and 37% of the total PM2.5

at Richmond (Cohen et al., 2012).

Aerosol Sampling and Elemental Analysis PM2.5 has been sampled at Lucas Heights since 1998 using

an IMPROVE PM2.5 cyclone system. This system employs a 25 mm diameter Teflon filter and typically samples at a flow rate of 22 L min–1 (Cohen et al., 1996). Twenty-four-

hour (midnight-to-midnight) continuous PM2.5 samples are collected twice a week (Wednesday and Sunday). After sample collection, accelerator-based ion beam analysis (IBA) is used to determine the elemental composition of the PM2.5

samples. Source type fingerprints were determined using PMF

(Paatero and Tapper, 1994), which solves the standard bi-linear factor analysis model. In summary, the standard (bi-linear) factor analysis problem is specified as (Paatero, 2010): X = GF + E (1) or

, , , ,1

p

i j i k k j i jk

x g f e

(2)

where X is a matrix of measured elements, G and F are factor matrices to be determined, and E is a matrix of residuals. If n observations are available of m elements and assuming p contributing sources, X is an n by m matrix, i.e., xi,j represents the concentration of element j in the ith sample, G is an n by p matrix of source contributions for each sample, F is a p by m matrix of source fingerprints.

The optimisation process minimises the function, Q, while the resolved factor elements, of F and G, remain non-negative:

2,

21 1 ,

n mi j

i j i j

eQ

s

(3)


where si,,j is a specified error of the form:

, , , , ,max ,i j i j i j i j i js MDL Error x y (4)

where MDLi,j is the minimum detectable limit, Errori,j is the statistical error, and yi,j is the fitted value i.e., Y = GF.

Using this data set, source type fingerprints were determined previously (Crawford, 2016c). In particular, 836 data samples were available for this study, from which the IBA determined the concentration of 23 elements, following which using the PMF analysis, 7 source types were identified. The identified source fingerprints were vehicle exhaust (Autos; 26.3%), secondary sulfate (2ndryS; 23.7%), old sea air mixed with industrial sources (IndSaged; 20.6%), smoke from domestic heating or bush fires (Smoke; 13.7%), fresh sea salt (Sea; 10.9%), soil dust (Soil; 3.7%) and other industrial releases (1.1%). The fingerprinting method and composition of each source fingerprint is detailed in (Crawford, 2016c). Stagnation, Recirculation and Ventilation Classification

The technique for classifying atmospheric flow conditions into stagnation, recirculation or ventilation episodes was developed by Allwine and Whiteman (1994). It involves the calculation of a wind run (S), which gives the total distance that a parcel has travelled over a given transport time, irrespective of direction; and a recirculation factor (R) which gives an indication of recirculation over the same transport time (Mohan and Bhati, 2012).

Given a time series of N measurements of wind speed (Ui) and wind direction (Di) the data can be transformed into horizontal wind vectors as (Allwine and Whiteman, 1994): ui = Ui cos(Di – 180) (1) vi = Ui sin(Di – 180) (2)

where i = 1, 2, …, N. For a sample averaging interval (T) and transport time (τ), we can define the net east-west and north-south transport distances as:

i p

i jj i

X T u

(3)

i p

i jj i

Y T v

(4)

where i = 1, 2, …, N-p, p = τ/T – 1. The wind run, or total distance travelled, is calculated as:

2 2i p

i i ij i

S T u v

(5)

The resultant, straight line transport distance (which

represents the net distance a parcel will travel over the transport time, τ), is calculated as:

2 2i i iL X Y (6)

The recirculation factor is then calculated as:

1 ii

i

LR

S (7)

From the above equations we can calculate a mean daily

wind run (S) and recirculation factor (R). The variability of S and L is determined by local and large-scale factors (Surkova, 2013). If large-scale atmospheric flow dominates L is more likely to be large and therefore R small. The recirculation factor varies according to 0 ≤ R ≤ 1. When R = 0, no recirculation has occurred. When R = 1 there has been a complete recirculation and the parcel has arrived back at the origin.

The wind run, S, is used as a measure of stagnation, e.g., S = 0 equates to total stagnation. Ventilation conditions are characterised by low values of R and high values of S. To help determine atmospheric flow conditions based on calculated values of S and R, Allwine and Whiteman (1994) introduced the concept of predetermined critical transport indices (CTIs). Given predefined CTIs atmospheric flow classification can be made according to: S ≤ Sc – site is prone to stagnation (8) R ≥ Rc – site is prone to recirculation (9) S ≥ Scv and R ≤ Rcv – site is prone to ventilation (10) where Sc and Rc are the average daily CTIs for stagnation and recirculation, respectively; and Scv and Rcv are the average daily CTIs for ventilation. When τ = 24 h, Allwine and Whiteman (1994) proposed daily CTIs of Sc = 170 km, Rc = 0.4, Scv = 250 km and Rcv = 0.2. These same values have also been used in other studies (e.g., Kim et al., 2007; Nankar et al., 2009; Mohan and Bhati, 2012; Al-Khadouri et al., 2015; Russo et al., 2016) and have also been adopted in this study.

However, it should be noted that these measures are only an approximation, and will only be exact measures if the wind field at the measurement site is uniform over the region (Levy et al., 2008). It should also be noted that equation 7 has a possible singularity (i.e., when Si = 0), however in such cases Li is also 0, and in the mathematical limit Ri has a value of zero (Allwine and Whiteman, 1994). In addition not all events can be classified in one of the categories.

Further, based on Eqs. (8)–(10), we found that a number of periods were classified as both stagnation and recirculation. To address this ambiguity we introduced an additional category; local recirculation (LRC), used to refer to the instances of overlapping classification, and then we renamed the remaining cases of recirculation as regional recirculation (RRC): S ≤ Sc and R ≥ Rc – site is prone to LRC (11)


S > Sc and R ≥ Rc – site is prone to RRC (12) Back Trajectory Analysis and Mean Sea Level Pressure Maps

While for the classification in the previous section uniform conditions are considered, back trajectories take into consideration variable wind speed and direction. However, how well the back trajectories reproduce local scale air motion is dependent on the resolution of the meteoroidal data used in their calculations. Back trajectories in the current study were generated, at a starting height of 300 m above ground level, using HYSPLIT v4.0 (HYbrid Single-Particle Lagrangian Integrated Trajectory; Draxler and Rolph, 2003). The choice of the starting height was to limit the impact of local topography and to be within the boundary layer for a large proportion of the time (as it is known that the 10th–90th percentiles of the mixing layer depth are 900–3000 m; Chambers et al., 2015). The meteorological data used with the model was of 12 km horizontal resolution. This data was generated using the Weather Research and Forecasting (WRF) model version 3.5.1 with the ARW dynamical core (Skamarock et al., 2004; Skamarock and Klemp, 2008). The details of the model set up and simulations were given in the supplementary material to Crawford et al. (2016b). The domain of the WRF simulation was between longitudes 145.291° and 158.709°, and latitude –38.2885° and –27.5621° and the simulations were generated for all time from January 2007 to December 2009, with a temporal resolution of one hour. The back trajectories in this study are mainly used for the generation of density maps, to indicate the extent of reginal coverage by air masses for each of the flow classifications. At this resolution, sea breezes are expected to not be captured to their full extent, as previous numerical studies of sea breezes have used horizontal resolutions of 3 km (Comin and Acevedo, 2017) and 1 km (Arrillaga et al., 2016).

When back trajectories are presented in this work one back trajectory was determined for each hour of the day and the back trajectories were tracked for five days. Back trajectory density maps were generated, for which the horizontal position of the back trajectory (or trajectory node) was determined every 10 min (using HYSPLIT; where the velocity vectors, from the meteorological data files, are interpolated linearly both in time and space). The region was sub-dividend into grid cells of 0.1° by 0.1° dimension, and if a back trajectory node landed in the grid cell, a grid cell counter was incremented. As a different numbers of sampling days were classified in each of the four flow conditions, for comparison purposes, each grid cell value was normalised by dividing by the number of back trajectories in the group.

The time (in hours) spent by an air mass over land before arriving at the measurement site are reported in section 3.3. This was calculated by following a back trajectory from Lucas heights until it crossed the coast. This calculated time also includes those cases where the back trajectory reached the end of the domain used in the WRF simulation. Thus the reported times can be interpreted as the times spent over land in the domain of the simulation.

In this study the impact of power stations on the secondary

sulfate source fingerprint was analysed under recirculation flow conditions (as in Cohen et al., 2012). To identify whether an air mass had passed over a power station before arriving at Lucas Heights the method of Cohen et al. (2012) was used. Power stations were represented by rectangles and if a back trajectory passed over the rectangle before arriving at the measurement site, it was considered that power station contributed to fine particle measurements at Lucas Heights. The power station “rectangles” were each 0.2° by 0.2° centred at the power station locations (Fig. 1).

Composite MSLP maps were calculated using sea level pressure from the NCEP/NCAR Reanalysis (which is provided on 2.5 × 2.5 grid cell size and the data are available at 6 h time intervals; Kalnay et al., 1996). Atmospheric Radon Measurements

Radon-222 (radon) is a naturally occurring radioactive noble gas. Radon has a well-constrained terrestrial source (16–26 mBq m–2 s–1; Schery and Wasiolek, 1998; Goto et al., 2008) that exhibits little spatial and temporal variability over unsaturated, unfrozen surfaces. Radon has been used by a number of studies as a tracer of recent (2–3 weeks) terrestrial influence on an air mass for atmospheric mixing and transport studies (e.g., Balkanski et al., 1992; Zhang et al., 2008; Koffi et al., 2016). Radon has been measured at Lucas Heights from 2006 onward. A description of the radon detector at Lucas Heights has been document in Chambers et al. (2011). In this study we use radon, when available, to identify the degree of terrestrial influence on air masses. In the current analysis the 24-hour average radon measurements were used, corresponding to the PM2.5 measurement window. RESULTS AND DISCUSSION Flow Characteristics under the Classifications

Classification of flow conditions into recirculation, stagnation and ventilation states for the hourly 2001 to 2009 dataset was carried out using a transport time (τ) of 24 h. To enable direct comparison with the 24-hour integrated PM2.5 measurements, a daily average value of the flow conditions was calculated.

64% of all observations over the 9 years fell into one or other of our four flow categories: 15% (LRC), 19.5% (stagnation), 10.9% (RRC), and 18.6% (ventilation). LRC and stagnation events made up the largest proportion of classified data (34.5%), which has implications for the build-up of local sources. While only 64% of the data was classified, this value is compatible to that reported by Venegas and Masseo (1999) for five sites in Argentina. In their study, for sites closer to the coast, ventilation was more prevalent and stagnation for more inland sites. In the coastal regions of the Black Sea between 70 and 80% of the time could be classified into one of the flow categories (Surkova, 2013).

Wind roses for periods associated with each of our four flow classifications are presented in Fig. 2. Corresponding average mean sea level pressure maps are presented in Fig. 3. LRC and stagnation events are characterised by generally low wind speeds (at 10 meters above ground level; with an


Fig. 2. Wind roses for the four weather classes when all the data is used, by hour of day.

Fig. 3. Mean sea level pressure maps for days classified in each of the weather classes; (a) Stagnation, (b) LRC, (c) RRC and (d) Ventilation.


average of 1.8 m s–1, Figs. 2(a) and 2(b)) with a good representation from all sectors. They are typically associated with a high pressure system to the south-east (Figs. 3(a) and 3(b)). RRC events have a larger proportion of winds from the south to south-south-east than LRC or stagnation events, with generally larger wind speeds (2.9 m s–1 on average). RRC events are characterised by a trough over Australia, located between two high pressure systems. For ventilation events winds are more likely to be arriving from the west (characterised by the passage of a front from a westerly direction) or south-south-east (with average wind speeds of 3.2 m s–1).

On a seasonal basis, RRC flow is most likely to occur in summer and spring (the warmer months of the year when sea breezes are more likely), whereas stagnation conditions are more likely to occur in autumn and winter (Table 1). Ventilation, on the other hand has a good chance of occurring any time of the year, with a slightly lower occurrence in autumn. The largest number of LRC was recorded for autumn. Radon and Total PM2.5 Concentrations

The back trajectory density maps and the distributions of PM2.5 concentrations under each flow condition are presented in Figs. 4 and 5(a), respectively. The highest concentrations of PM2.5 occurred under LRC and RRC flow. If we use radon measurements as an indicator of the degree of recent contact with land, the LRC group contains the largest radon indicating that these air masses had the highest terrestrial influence (Fig. 5(b)), as supported by Fig. 4(b). It is interesting to note that mean radon is much higher under LRC than RRC; however PM2.5 mean values are similar for both groups. When we look at the back trajectory density maps for LRC and RRC (Fig. 4) we see that more time has been spent (overland) closer to the site under LRC (i.e., as indicated by the more orange and green in Fig. 4(b) compared to that in Fig. 4(c)), supporting the higher radon concentrations. This result is most likely due to the uniform spatial distribution of the radon source term. However, the PM2.5 sources are distributed non-uniformly across the region, and as will be shown in the next section different sources dominate under different flow classifications (and from Table 3, the sum of the averages from each source type happen to add up to similar values; 6.60 and 6.98 µg m–3, under LRC and RRC conditions, respectively).

The mean concentrations of PM2.5 (specified in µg m–3) were 5.19 (overall), 5.25 (stagnation), 6.99 (LRC), 6.97

Table 1. The fraction of occurrences of each flow classification by season, expressed as a percentage of the total number of samples classified in the particular flow type. These statistics were for the classified samples, constituting 64% of the data.

Summer Autumn Winter Spring Stag 15 32 34 19 LRC 9 38 30 23 RRC 42 13 4 41 Vent 27 20 27 26

(RRC) and 3.34 (ventilation); indicating that the mean measured concentrations under recirculation (i.e., either LRC or RRC) was double that under ventilation. The mean concentration under recirculation was 33% higher than under stagnation and 35% higher than the overall mean concentration. The difference in the means and medians of PM2.5 concentration was significant (p-value < 0.01) for all combinations of flow classification, except between LRC and RRC. Differences between these two flow conditions by PM2.5 source type are further examined in the following section. Concentrations of Separate PM2.5 Source Types

As documented previously (Crawford et al., 2016c), total PM2.5 and each of the sources exhibit seasonal variation; with higher concentration of 2ndryS, IndSaged and Sea measured in summer and higher concentration of Autos measured in winter. This is in part affected by the changes in the predominant fetch during the different seasons (section 2.1 and supplementary Fig. S1). In addition to presenting information when the whole year is considered, to reduce the impact of the seasonal variations on the analysis, here we also present results for the different source types, separated into winter and summer seasons.

The strength of the correlations between the flow classification parameters (R and S) and PM2.5 concentrations vary for the different source types (Table 2). Mohan and Bhati (2012) commented that the correlation between stagnation and recirculation was weaker when PM10 was considered as opposed to NO2 and SO2, due to the different life times and formation mechanisms of different components of PM. In their study this was particularly true for the wind run parameter. For total PM2.5, we obtain correlation values of 0.47 and –0.25 with the recirculation parameter (R) and wind run parameter (S), respectively, which compares well with the correlation coefficients of Mohan and Bhati (2012) of 0.54 and –0.23 with PM10. However, in this study, the correlation with wind run (S) improved significantly when Auto and Sea sources were considered, and the correlations with other sources were lower. The best correlation of the recirculation parameter was with Auto and 2ndryS. On annual basis, the correlations were lower than those obtained for NO2 and SO2 by Mohan and Bhati (2012) at Mumbai. However, in winter the correlation coefficients, 0.65 and 0.59, between the recirculation factor and total PM2.5 and 2ndryS, respectively, were of similar magnitude to the correlation coefficients between NO2 and the recirculation factor (of 0.62) reported by Mohan and Bhati (2012). Further, in their study the correlation coefficient between PM10 and the wind run and recirculation factor were –0.23 and 0.54, respectively, compared to –0.25 and 0.47 for PM2.5 in our study.

Differences in the correlations were seen between the seasons. For example, the correlation between wind run (S) and Smoke changed from slight positive in summer to larger negative in winter indicating that in summer, higher winds are actually bringing smoke to the site (likely from bush fires; Rea et al., 2016, and references therein). From supplementary Figs. S1 and S2, in summer higher wind


Fig. 4. Normalised back trajectory density maps of (a) Stagnation, (b) LRC, (c) RRC and (d) Ventilation. The letters represent the location of (R) Richmond, (L) Liverpool and (S) the Sydney Central Business District, (B) are locations of major bush fires during the study period, and the red circles are the locations of the power stations.

Fig. 5. Distributions of (a) PM2.5 concentrations and (b) radon, under each flow classification. The average radon measurement was used, corresponding to the PM2.5 measurement window. Boxes represent the 25th and 75th percentiles, the median is represented by the line through the box, whiskers represent the 10th and 90th percentiles and the outlier events have not been indicated.

Longitude (o)

La

titu

de

(O)

150 150.5 151 151.5 152-35

-34.5

-34

-33.5

-33

0.30.20.10.050.02

(a) Stag

B

B

B

R

L S

Longitude (o)

La

titu

de

(O)

150 150.5 151 151.5 152-35

-34.5

-34

-33.5

-33

0.30.20.10.050.02

(b) LRC

B

B

B

R

L S

Longitude (o)

La

titu

de

(O)

150 150.5 151 151.5 152-35

-34.5

-34

-33.5

-33

0.30.20.10.050.02

(c) RRC

B

B

B

R

L S

Longitude (o)

La

titu

de

(O)

150 150.5 151 151.5 152-35

-34.5

-34

-33.5

-33

0.30.20.10.050.02

(d) Vent

B

B

B

R

L S


Table 2. Correlation results (r2 values) between flow classification parameters (R and S) and the concentrations of Total PM2.5, and each of the identified source fingerprints. The correlation of entries marked with (*) were not significant at 99% confidence level (i.e., the p-value was greater than 0.01).

Correlations Total PM2.5 Auto Smoke 2ndryS Sea Soil IndSaged All data Recirculation (R) 0.47 0.4 0.24 0.45 –0.23 0.18 0.07* Wind run (S) –0.25 –0.49 –0.13 –0.29 0.43 0.02* 0.09 Summer Recirculation (R) 0.42 0.39 0.09 0.47 –0.25 0.36 0.11 Wind run (S) –0.10* –0.39 0.12 –0.31 0.45 –0.03* 0.00* Winter Recirculation (R) 0.65 0.5 0.44 0.59 –0.3 0.18 0.01* Wind run (S) –0.48 –0.51 –0.32 –0.4 0.48 –0.07* 0.05*

speeds occur from the south-east sector where more bush land is located, and from Fig. 4(a) major bush fire had occurred south-east of the site during the study period. Whereas in winter smoke is more likely from domestic wood heaters (NSW EPA, 2017b), however, the south-west sector (the sector with higher winter winds) has very low population density and as a result the emission of smoke, in winter, would be lower. In the case of soil dust the best correlations were seen with recirculation in summer.

The distributions of concentrations of each source type under each flow classification are presented in Fig. 6 and the corresponding average values presented in Table 3. The difference in the medians was significant at 95% (p-vale < 0.05) for all combinations of flow conditions other than; LRC and RRC in the case of Smoke, 2ndryS and Soil and between stagnation and RRC for Sea. When summer and winter alone were considered (plots supplied in the supplementary material; Figs. S3 and S4) similar behaviour was seen for all sources other than Smoke. The distributions of Smoke during summer and winter are presented in Fig. 7; the difference between summer and winter can be attributed to the contrasting sources; wood smoke from domestic heating in winter and bush fires in summer and also the differences in the spatial distributions of the source terms.

For Autos the highest concentrations occurred under LRC followed by stagnation. This behaviour followed closely the behaviour of radon (Fig. 5(b)), indicating similar distribution to their source term. During LRC, the air masses have spent more time over land (16 hours on average) and a high proportion of this time has been over the populated Sydney region (Fig. 4(a), where the density of roads is higher; supplementary Fig. S2). Under LRC the concentration from Auto can be up to a factor of 3.9 to that under ventilation. Under ventilation conditions (on average) the air masses have spent 3.8 hours over land before arriving at the measurement site, corresponding to less time to concentrate terrestrial sources. In addition from Fig. 4 and supplementary Fig. S2, under ventilation the air masses have travelled over the less populated region to the south.

The highest mean concentrations of 2ndryS and total PM2.5 occurred under LRC and RRC and the highest concentrations of IndSaged occurred during RRC. As seen from Fig. 4, both under LRC and RRC air masses had spent considerable time over the populated Sydney region, bringing polluted air to Lucas Heights. Under ventilation conditions the air masses tend to arrive from the less polluted

region to the south (Fig. 2(d)) in addition to dilution due to higher wind speeds (average of 3.2 m s–1, which is higher than the other classifications; Fig. 2(d)).

These results show that recirculation is important at this site, and these events occur for 25.9% of the time (i.e., when both LRC and RRC are considered). However, if one is to target the three main source types of total PM2.5 (Auto 26.3%, 2ndryS 23.7% and IndSaged 20.6%), then the conditions leading to the highest concentrations differ. The highest concentrations of Auto occur during LRC, the highest concentrations of IndSaged occurred under RRC conditions and 2ndryS had similar high concentrations under LRC and RRC. While in section 3.2, it appeared that similar PM2.5

median concentrations occurred under LRC and RRC, analysing by source types gives the decision makes more information on which source term is more important under different conditions.

Elevated pollution has been reported when polluted air masses had been swept out to sea via land breezes, followed by photochemical activity over the sea and then returned to the measurement site during the ensuing sea breezes, additionally picking up more pollution overland (e.g., Cass and Shair, 1984; Alper-Siman Tov et al., 1997; Levy et al., 2009). Concentrations of IndSaged are expected to be higher when the air mass has had higher sea influence as this source type contains sea air that has undergone reactions, accounting for the higher mean values of IndSaged during RRC conditions.

The lowest concentration of fresh sea salt was under LRC conditions, an increase was seen for stagnation, followed by RRC and finally ventilation. This corresponds well to the reduction in the number of hours that an air mass has spent over land before reaching the site, 16, 10, 12 and 5.8 hours for LRC, stagnation, RRC and ventilation, respectively (method of calculation described in section 2.4; and then calculated as an average of all samples classified into each flow type). This indicates that under ventilation, on average, the air masses have the most recent oceanic fetch.

The average Soil concentration for ventilation conditions in summer is very low due to the wind direction being dominated from the south-east over forested area and also more recent oceanic fetch. Recirculation, 2ndryS and Power Stations

Robinsohn et al. (1992) reported that recirculation, during a weak synoptic flow, contributed 50% of the measured SO2


Fig. 6. Distributions of each source types by flow classification, when all the data is used; (a) Autos, (b) Smoke, (c) 2ndryS, (d) Sea, (e) Soil and (f) IndSaged. Boxes and whiskers as in Fig. 5.

concentration (associated with a power plant), and pointed out that during such synoptic situations a switch to fuel with low-level of sulfur would be beneficial. For the Sydney region it has been shown that at Richmond, located north-

west of Lucas Heights, about 34–47% of the 2ndryS could be associated with air mass passage over one of the eight nearest power stations before arrival at the measurement site (Cohen et al., 2012).


Table 3. Mean concentrations (µg m–3) under the different flow classifications for all data as well as separately for summer and winter.

Total PM2.5 Auto Smoke 2ndryS Sea Soil IndSaged All Data LRC 6.99 2.21 1.01 2.28 0.12 0.2 0.78

Stag 5.25 1.69 0.76 1.42 0.25 0.17 0.87 RRC 6.97 1.24 0.95 2.19 0.42 0.65 1.53 Vent 3.34 0.56 0.51 0.5 0.84 0.15 0.83

Summer LRC 7.42 1.86 0.34 3.17 0.04 0.18 1.75 Stag 6.13 1.18 0.42 2.24 0.37 0.07 1.66 RRC 7.74 1.25 0.78 2.87 0.43 0.48 1.9 Vent 4.66 0.45 0.57 0.83 1.2 0.05 1.48

Winter LRC 6.76 2.69 1.43 1.71 0.08 0.2 0.36 Stag 4.52 2.15 0.83 0.84 0.13 0.16 0.35 RRC 5.71 1.52 1.16 1.39 0.4 0.28 0.94 Vent 2.1 0.6 0.37 0.29 0.54 0.12 0.42

Fig. 7. Distribution of Smoke by weather classes for (a) summer and (b) winter. Boxes and whiskers as in Fig. 5. There were small differences between the seasons for the other source types (a full set of plots is presented in the supplementary material; Figs. S3 and S4).

From Table 3 we see that the highest (average) values of 2ndryS occur under LRC and RRC conditions. We have repeated the analysis of Cohen et al. (2012), for those days classified in either LRC or RRC for 2007 to 2009 (as the WRF data was available for this period). In summary, for each day on which both PM2.5 data was available and the day was classified in either LRC or RRC, 24 back trajectories were generated (one for each hour of the day) and it was tested if the back trajectories crossed over a 0.2° by 0.2° rectangle centred at a power station. Of the total of 75 days, a power station was crossed for 57 of these days. The mean 2ndryS concentration when a power station was crossed was 2.3 µg m–3, whereas when no power station was crossed the mean concentration was 1.4 µg m–3. Of the 13 days which had a 2ndryS concentration of one standard deviation above the mean (3.69 µg m–3), 11 occurred when a power station was crossed and 2 when no power station was crossed; indicating that power stations can have an effect on the 2ndryS concentrations at Lucas Heights during recirculation conditions.

CONCLUSIONS

The classification method of Allwine and Whiteman (1994) was applied to conditions in the Sydney airshed, using data from Lucas Heights. Using this method as originally described resulted in a number of days being classified into both stagnation and recirculation, hence we introduced a new class; local recirculation (LRC) and referred to the remaining recirculation as regional recirculation (RRC). At this coastal airshed it was found that 64% of the time could be classified into one of; local recirculation (LRC; 15%), stagnation (19.5%), regional recirculation (RRC; 10.9%) and ventilation (18.6%). The fraction of time classified compares well with a study in Argentina (Venegas and Mazzeo, 1999), however in the Argentina study the fraction of recirculation was lower. On a seasonal basis, RRC flow is most likely to occur in summer and spring (the warmer months of the year when sea breezes are more likely), whereas LRC and stagnation conditions are more likely to occur in autumn and winter.

In a study by (Mohan and Bhati, 2012) the correlation


coefficient between PM10 and the wind run and recirculation factor were –0.23 and 0.54, respectively, compared to –0.25 and 0.47 for PM2.5 in our study. In this study, better correlations were seen when winter alone was considered.

The highest concentrations of total PM2.5, vehicle exhaust, secondary sulfur and smoke were seen under LRC conditions. Stagnation conditions had the next highest concentrations of vehicle exhaust, whereas RRC conditions had the second highest secondary sulfate. However, if one is to target the three main source types of total PM2.5 (Auto 26.3%, 2ndryS 23.7% and IndSaged 20.6%), then the conditions leading to the highest concentrations differ. The highest concentrations of Auto occur during LRC, the highest concentrations of IndSaged occurred under RRC conditions and 2ndryS had similar concentrations under LRC and RRC.

These results show that for coastal confined airsheds recirculation can increase the overall PM concentrations. However, to apply mitigation strategies the impact of the different PM sources is required. For example, at this site vehicle exhaust is important in winter and 2ndryS in summer, and switching to better quality vehicle fuel during LRC conditions in winter and lower sulfur content coal during recirculation conditions in summer would reduce the total PM2.5 load to the region around the site. Consideration of the flow conditions for planned back burning would also be beneficial.

ACKNOWLEDGMENTS

The authors would like to acknowledge NCRIS funding

for accelerators and Centre for Accelerator Science staff for access to their ion beam analysis facilities. We also thank Leisa Dyer and Jim Pascoe for the monitoring of meteorological data at Lucas heights. The NOAA Air Resources Laboratory (ARL) made available the HYSPLIT transport and dispersion model. We would also like to thank the Mesoscale and Microscale Meteorology (MMM) Division of NCAR for supporting WRF development. NCEP Reanalysis data provided by the NOAA/OAR/ESRL PSD Boulder, Colorado, was used in this work SUPPLEMENTARY MATERIAL

Supplementary data associated with this article can be found in the online version at http://www.aaqr.org. REFERENCES Al-Khadouri A., Al-Yahyai S. and Charabi Y. (2015).

Contribution of atmospheric processes to the degradation of air quality: Case study (Sohar Industrial Area, Oman). Arabian J. Geosci. 8: 1623–1633.

Allwine, K.J. and Whiteman, C.D. (1994). Single-station integral measures of atmospheric stagnation, recirculation and ventilation. Atmos. Environ. 28: 713–721.

Alper-Simon Tov, D., Peleg, M., Matveev, V., Mahrer, Y., Seter, I. and Lurai, M. (1997). Recirculation of polluted air masses over the East Mediterranean coast. Atmos. Environ. 31: 1441–1448.

Arrillaga, J.A., Yagüe, C., Sastre, M. and Román-Cascón, C. (2016). A characterisation of sea-breeze events in the eastern Cantabrian coast (Spain) from observational data and WRF simulations. Atmos. Res. 181: 265–280.

Balkanski, Y.J., Jacob, D.J., Arimoto, R. and Kritz, M.A. (1992). Distribution of 222Rn over the North Pacific: Implications for continental influences. J. Atmos. Chem 14: 353–374.

Barros, J.A. (2001). Sydney Basin: Air toxic emissions and health update. http://www.areco.org/SYDNEY%20BASIN.pdf.

Beaver, S., Palazoglu, A., Singh, A., Soong, S.T. and Tanrikulu, S. (2010). Identification of weather patterns impacting 24-h average fine particulate matter pollution. Atmos. Environ. 44: 1761–1771.

Cass, R.G. and Shair, F.H. (1984). Sulfate Accumulation in a Sea/Breeze/Land Breeze Circulation System. J. Geophys. Res. 89: 1429–1438.

Chambers, S., Williams, A. G., Zahorowski, W., Griffiths, A. and Crawford, J. (2011). Separating remote fetch and local mixing influences on vertical radon measurements in the lower atmosphere. Tellus B 63: 843–859.

Chambers, S.D., Williams, A.G., Crawford, J. and Griffiths, A.D. (2015). On the use of radon for quantifying the effects of atmospheric stability on urban emissions. Atmos. Chem. Phys. 15: 1175–1190.

Charlson, R.J., Schwartz, S.E., Hales, J.M., Cess, R.D., Coakley, J.A., Hansen, J.E. Jr. and Hofmann, D.J. (1992). Climate forcing by anthropogenic aerosols. Science 255: 423–430.

Cohen, D.D., Bailey, G.M. and Kondepudi, R. (1996). Elemental analysis by PIXE and other IBA techniques and their application to source fingerprinting of atmospheric fine particle pollution. Nucl. Instrum. Methods Phys. Res. B 109: 218–226.

Cohen, D.D., Stelcer, E., Garton, D. and Crawford, J. (2011). Fine particle characterisation, source apportionment and long-range dust transport into the Sydeny Basin: A long term study between 1998 and 2009. Atmos. Pollut. Res. 2: 182–189.

Cohen, D.D., Crawford, J., Stelcer, E. and Atanacio, A.J. (2012). Application of positive matrix factorization, multi-linear engine and back trajectory techniques to the quantification of coal fired power station pollution in metropolitan Sydney. Atmos. Environ. 61: 204–211.

Cohen, D.D, Stelcer, E., Atanacio, A. and Crawford, J. (2014). The application of IBA techniques to air pollution source fingerprinting and source apportionment. Nucl. Instrum. Methods Phys. Res. B 318: 113–118.

Comin, A.N. and Acevedo, O.C. (2017). Numerical simulation of sea breeze convergence over antarctic peninsula. Adv. Meteorol. 2017: 7686540.

Crawford, J., Cohen, D.D., Chambers, S., Williams, A. and Stelcer, E. (2013). Incorporation of Radon-222 as a parameter in ME-2 to improve apportionment of PM2.5

sources in the Sydney region. Atmos. Environ. 80: 378–388.

Crawford, J., Chambers, S., Cohen, D., Williams, A., Griffiths, A. and Stelcer, E. (2016a). Assessing the


impact of atmospheric stability on locally and remotely sourced aerosols at Richmond, Australia, using Radon-222. Atmos. Environ. 127: 107–117.

Crawford, J., Griffiths, A., Cohen, D.D., Jiang, N. and Stelcer, E. (2016b). Particulate pollution in the Sydney region: Source diagnostics and synoptic controls. Aerosol Air Qual. Res. 16: 1055–1066.

Crawford, J., Chambers, S., Cohen, D., Williams, A., Griffiths, A., Stelcer, E. and Dyer, L. (2016c). Impact of meteorology on fine aerosols at Lucas Heights, Australia. Atmos. Environ. 145: 135–146.

Davis, R.E. and Gay, D.A. (1993). An assessment of air quality variations in the south-western USA using an upper air synoptic climatology. Int. J. Climatol. 13: 755–781.

Dockery, D.W., Pope III., C.A., Xiping, X., Spengler, J.D., Ware, J.H., Fay, W.E., Ferris Jr., B.G. and Speiser, F.E., (1993). An association between air pollution and mortality in six US cities. N. Engl. J. Med. 329: 1753–1759.

Draxler, R.R. and Rolph, G.D. (2003). Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT), model. http://www.arl.noaa.gov/ready/hysplit4.html.

Goto, M., Moriizumi, J., Yamazawa, H., Iida, T. and Zhuo, W. (2008). Estimation of global radon exhalation rate distribution. In The Natural Radiation Environment: 8th International Symposium, Paschoa, A.S. (Ed.), American Institute of Physics, pp. 169–172.

Henry, R.C. (2002). Multivariate receptor models – current practice and future trends. Chemom. Intell. Lab. Syst. 60: 43–48.

Huebert, B.J., Bates, T., Russell, P.B., Shi, G., Kim, Y.J., Kawamura, K., Carmichael, G. and Nakajima, T. (2003). An overview of ACE-Asia: Strategies for quantifying the relationships between Asian aerosols and their climatic impacts. J. Geophys. Res. 108: 8633.

Jiang, N., Scorgie, Y., Hart, M., Riley, M.L., Crawford, J., Beggs, P.J., Edwards, G.C., Chang, L., Salter, D. and Di Virgilio, G. (2016). Visualising the relationships between synoptic circulation type and air quality in Sydney, a subtropical coastal-basin environment. Int. J. Climatol. 37: 1211–1228

Jiang, Y., Liu, X., Yang, X.Q. and Wang, M. (2013). A numerical study of the effect of different aerosol types on East Asian summer clouds and precipitation. Atmos. Environ. 70: 51–63.

Jollife I.T. (1986). Principal Component Analysis. Springer, New York.

Kalnay, E., Kanamitsu, M., Kistler, R., Collins, W., Deaven, D., Gandin, L., Iredell, M., Saha, S., White, G., Woollen, J., Zhu, Y., Leetmaa, A., Reynolds, R., Chelliah, M., Ebisuzaki, W., Higgins, W., Janowiak, J., Mo, K.C., Ropelewski, C., Wang, J., Jenne, R. and Joseph, D. (1996). The NCEP/NCAR 40-year reanalysis project. Bull. Am. Meteorol. Soc. 77: 437–471.

Kim, E.H., Suh, K.S., Hwang, W.T., Jeong, H.J., Han, M.H. and Moon, J.Y. (2007). Analysis of the site characteristics of Korean nuclear power sites from the meteorological aspects. Ann. Nucl. Energy 34: 719–723.

Koffi, E.N., Bergamaschi, P., Karstens, U., Krol, M., Segers, A., Schmidt, M., Levin, I., Vermeulen, A.T., Fisher, R.E., Kazan, V., Klein Baltink, H., Lowry, D., Manca, G., Meijer, H.A.J., Moncrieff, J., Pal, S., Ramonet, M., Scheeren, H.A. and Williams, A.G. (2016). Evaluation of the boundary layer dynamics of the TM5 model over Europe. Geosci. Model Dev. 9: 3137–3160.

Kumar, D., Kumar, A., Kumar, V., Kumar, J. and Ravi, P.M. (2013). Study of atmospheric stagnation, recirculation and ventilation potential at Narora Atomic Power Station NPP site. Environ. Monit. Assess. 185: 2887–2894.

Lee, M., Koutrakis, P., Coull, B., Kloog, I. and Schwartz, J. (2016). Acute effect of fine particulate matter on mortality in three Southeastern states from 2007-2011. J. Exposure Sci. Environ. Epidemiol. 26: 173–179.

Levy, I., Dayan, U. and Mahrer, Y. (2008). Studying coastal recirculation with a simplified analytical land-sea breeze model. . J. Geophys. Res. 113: D03104.

Levy, I., Mahrer, Y. and Dayan, U. (2009). Coastal and synoptic recirculation affecting air pollutants dispersion: A numerical study. Atmos. Environ. 43: 1991–1999.

Levy, I., Dayan, U. and Mahrer, Y. (2010). Differing atmospheric scales of motion and their impact on air pollutants. International J. Climatol. 30: 612–619.

Liang, C.S., Duan, F.K., He, K.B. and Ma, Y.L. (2016). Review on recent progress in observations, source identifications and countermeasures of PM2.5. Environ. Int. 86: 150–170.

Live Australian Electricity Generation, http://www.energymatters.com.au/energy-efficiency/australian-electricity-statistics/, Last Accesses: 10 November 2016.

Mohan, M. and Bhati, S. (2012). Wind flow conditions as an indicator to assimilative capacities of urban airsheds towards atmospheric pollution potential. J. Civil. Environ. Eng. S1: 003.

Moloi, K., Chimidza, S., Lindgren, E.S., Viksna, A. and Standzenieks, P. (2002). Black carbon, mass and elemental measurements of airborne particles in the village of Serowe, Botswana. Atmos. Environ. 36: 2447–2457.

Nankar, D.P., Patra, A.K., Dole, M.U., Venkataraman, S. and Hegde, A.G. (2009). Atmospheric stagnation, recirculation and ventilation characteristics at Kakrapar atomic power station site. Ann. Nucl. Energy 36: 475–480.

NSW (2011). Air pollution – Particulate matter (PM2.5). Office of Environment and Heritage, NSW Government. ISBN 978 1 74293 266 8.

NSW EPA (2017a). Air pollution – Where does it come from? Air emissions inventory for the Greater Metropolitan Region in NSW. http://www.epa.nsw.gov.au/resources/ air/aeibroc07335.pdf.

NSW EPA (2017b). http://www.epa.nsw.gov.au/woodsmoke/.

Oh, I.B., Kim, Y.K., Lee, H.W. and Kim, C.H. (2006). An observational and numerical study of the effects of the late sea breeze on ozone distributions in the Busan metropolitan area, Korea. Atmos. Environ. 40: 1284–1298.

Paatero, P. and Tapper, U. (1994). Positive matrix factorization: A non-negative factor model with optimal


utilization of error estimates of data values. Environmetrics 5: 111–126.

Paatero, P. (2010). User’s Guide for Positive matrix Factorization program PMF2 and PMF3, Part 1 and 2.

Pearce, J.L., Beringer, J., Nicholls, N., Hyndman, R.J. and Tapper, N.J. (2011). Quantifying the influence of local meteorology on air quality using generalized additive models. Atmos. Environ. 45: 1328–1336.

Rea, G., Paton-Walsh, C., Turquety, S., Cope, M. and Griffith, D. (2016). Impact of the New South Wales fires during October 2013 on regional air quality in eastern Australia. Atmos. Environ. 131: 150–163.

Robinsohn, J., Mahrer, Y. and Wakshal, E. (1992). The effects of mesoscale circulation on the dispersion of pollutants (SO2) in the eastern Mediterranean, southern coastal plain of Israel. Atmos. Environ. 26: 271–277.

Russell, A.G. and Brunekreef, B. (2009). A Focus on Particulate Matter and Health. Environ. Sci. Technol. 43: 4620–4625.

Russo, A., Gouveia, C., Levy, I., Dayan, U., Jerez, S., Mendes, M. and Trigo, R. (2016). Coastal recirculation potential affecting air pollutants in Portugal: The role of circulation weather types. Atmos. Environ. 135: 9–19.

Schery, S.D. and Wasiolek, M.A. (1998). Radon and Thoron in the Human Environment, Chap. Modeling Radon Flux from the Earth’s Surface, World Scientific Publishing, Singapore, pp. 207–217.

Skamarock, W.C. (2004). Evaluating mesoscale NWP

models using kinetic energy spectra. Mon. Weather Rev. 132: 3019–3032.

Skamarock, W.C. and Klemp, J.B. (2008). A time-split nonhydrostatic atmospheric model for weather research and forecasting applications. J. Comput. Phys. 22: 3465–3485.

Surkova, G. (2013). Air recirculation and ventilation in the coastal regions of the Black Sea. Cent. Eur. J. Geosci. 5: 196–207.

USEPA (2007). Guidance on the Use of Models and Other Analyses for Demonstrating Attainment of Air Quality Goals for Ozone, PM2.5, and Regional Haze. United States Environmental Protection Agency, EPA-454/B-07-002.

Venegas, L.E. and Mazzeo, N.A. (1999). Atmospheric stagnation, recirculation and ventilation potential of several sites in Argentina. Atmos. Environ. 52: 43–57.

VES (2012). Vehicle Emissions Standards, http://www.infrastructure.gov.au/roads/environment/emission/index.aspx.

Zhang, K., Wan, H., Zhang, M. and Wang, B. (2008). Evaluation of the atmospheric transport in a GCM using radon measurements: Sensitivity to cumulus convection parameterization. Atmos. Chem. Phys. 8: 2811–2832.

Received for review, February 19, 2017 Revised, April 11, 2017

Accepted, April 30, 2017

Impact of Atmospheric Flow Conditions on Fine Aerosols in ...

Documents

Transcript of Impact of Atmospheric Flow Conditions on Fine Aerosols in ...