Measurement report

Atmos. Chem. Phys., 22, 11557–11577, 2022

https://doi.org/10.5194/acp-22-11557-2022
© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

Measurement report: The 10-year trend of PM2.5 major

components and source tracers from 2008 to 2017 in
an urban site of Hong Kong, China
Wing Sze Chow1 , Kezheng Liao1 , X. H. Hilda Huang2 , Ka Fung Leung2 , Alexis K. H. Lau2 , and
Jian Zhen Yu1,2
1 Department of Chemistry, The Hong Kong University of Science and Technology,
Clear Water Bay, Kowloon, Hong Kong, China
2 Division of Environment & Sustainability, The Hong Kong University of Science and Technology,

Clear Water Bay, Kowloon, Hong Kong, China

Correspondence: Jian Zhen Yu ([email protected])

Received: 7 February 2022 – Discussion started: 7 March 2022

Revised: 7 August 2022 – Accepted: 14 August 2022 – Published: 8 September 2022

Abstract. Fine particulate matter (PM2.5 ) remains a major air pollutant of significant public health concern in
urban areas. Long-term monitoring data of PM2.5 chemical composition and source-specific tracers provide es-
sential information for the identification of major sources as well as evaluation and planning of control measures.
In this study, we present and analyze a 10-year dataset of PM2.5 major components and source-specific tracers
(e.g., levoglucosan, hopanes, K+ , Ni, V, Al, and Si) collected over the period of 2008–2017 in an urban site in
Hong Kong, China. The time series of pollutants were analyzed by the seasonal and trend decomposition using
the locally estimated scatter plot smoothing (LOESS) method and general least squares with the autoregressive
moving average method. Bulk PM2.5 and all its major components displayed a significant decline of varying
degrees over the decade. PM2.5 was reduced by 40 % at −1.5 µg m−3 yr−1 . PM2.5 components that are predom-
inantly influenced by local vehicular emissions showed the steepest decline, with nitrate decreasing by −66 %,
elemental carbon by −60 %, and hopanes by −75 %, confirming effective control of local vehicular emissions.
For components that are significantly impacted by regional transport and secondary formation, they had a notably
lower percentage reduction, with sulfate declining by −40 % and organic carbon by −23 %, reflecting complex-
ity in their region-wide contributing sources and formation chemistry. Levoglucosan and K+ , two tracers for
biomass burning, differed in their reduction extent, with K+ at −60 % and levoglucosan at −47 %, indicating
they likely track different biomass burning types. Dust components in PM2.5 also decreased, by −37 % for Al
and −46 % for Si. The year of 2011 was an anomaly in the overall trend in having higher concentrations of
PM2.5 and components than its adjacent years, and the long time series analysis attributed the anomaly to un-
usually lower rainfall associated with strong La Niña events. This 10-year trend analysis based on measurements
exemplifies the utility of chemical composition data in support of an evidence-based approach for control policy
formulation.

Published by Copernicus Publications on behalf of the European Geosciences Union.

11558 W. S. Chow et al.: Ten-year trend of PM2.5 components in Hong Kong

1 Introduction PM2.5 to a particular control measure. This is evident from

Fig. 1b, which shows the percentage changes vary signifi-
Air pollution controls are of both local and global impor- cantly among PM2.5 components, using the dataset to be dis-
tance. Their effectiveness needs to be periodically reviewed cussed in this work as an illustration.
for optimizing options to improve air quality and minimize The speciated analysis of PM2.5 , in particular the mea-
environmental impacts. Particulate matter with aerodynamic surement of source-specific marker species, provides valu-
diameter less than 2.5 µm, namely fine particulate matter able information for understanding the sources, formation,
(PM2.5 ), is a major air pollutant. It is a significant contrib- and evolution of the PM pollution. Long time series of such
utor to visibility reduction and climate change and has detri- chemically specific data would, on the other hand, poten-
mental effects on human health (Yang et al., 2018; Zhao et tially allow for the discernment of meaningful trends that
al., 2013; Lippmann and Chen, 2009; Ko et al., 2007; Kim are not apparent in one-time field projects as well as es-
et al., 2006; Cheung et al., 2005). Hong Kong, located in the tablish long-term knowledge about representative urban/re-
southern coastal part of China, is an important part of the gional aerosol chemistry. However, long-term measurements
Guangdong–Hong Kong–Macau Greater Bay Area (GBA), of PM2.5 species in China are very limited, more so for
which includes the Pearl River Delta (PRD) region in Guang- the source-specific markers. Our research team launched a
dong plus Hong Kong and Macao. The Hong Kong govern- filter-based PM2.5 monitoring program in mid-2007 at Tsuen
ment has been assiduous in controlling the local emission Wan (TW), an urban location, and has maintained the op-
via the Air Pollution Control Ordinance, in addition to coop- eration since then. Our monitoring program adopts a regu-
erating with the neighboring Guangdong and Macao govern- lar sampling schedule of one 24 h sample every 6 d to ensure
ments on formulating control policies to reduce air pollution temporal representativeness. High-volume samples were also
emissions in the area (HKEPD, 2021a). Ambient monitoring collected to allow for sufficient aerosol materials for anal-
of criteria air pollutants plays an important role in verifying ysis of organic source markers (e.g., hopanes for vehicular
the effectiveness of control policies. For example, from 2006 emissions and levoglucosan for biomass burning emissions).
to 2018, large reductions were documented for sulfur diox- Starting from 2011, Hong Kong Environmental Protection
ide (−81 %), nitrogen dioxide (−28 %), and PM10 (−36 %) Department (HKEPD) established a regular PM2.5 chemical
in terms of annual average concentrations (HKEPD, 2019, speciation monitoring network, with TW as one of its moni-
2020), reflecting the benefits from a series of SO2 , NO2 , and toring stations and adopting a 1-in-6 d sampling schedule as
PM reduction measures (Table S8 in the Supplement). well (e.g., Yu et al., 2012). The field sampling, laboratory
PM2.5 was introduced as a criteria pollutant in Hong Kong analyses, quality check/quality assurance, and data valida-
in 2004, while its online monitoring preceded 5 years ear- tion were conducted according to the same set of standard
lier at three sites (Tap Mun, Tung Chung, Tsuen Wan) in operating procedures which are in reference to those recom-
1999. PM2.5 mass was added as a monitoring parameter in mended by the US Environmental Protection Agency (Chow
2015 to the PRD Regional Air Quality Monitoring Network, and Watson, 1998; USEPA, 2016). The laboratory analyses
which has included 23 sites in the GBA since 2015. The of PM2.5 mass and major components (water-soluble ions,
monitoring data indicate that while substantial progress has OC, EC, and elements) were performed by our research team
been made in lowering the pollution level of PM2.5 , from except for samples from the year of 2015, which were ana-
38 µg m−3 in 1999 to 15 µg m−3 in 2020 (Fig. 1a), the cur- lyzed by the Desert Research Institute (Chow et al., 2016).
rent level still notably exceeds the most updated air quality This set of filter samples allows us to observe the long-term
guideline of an annual average of 5 µg m−3 as recommended trend of PM2.5 major components and source tracers and to
by the World Health Organization (World Health Organiza- examine variations in aerosol sources affecting Hong Kong
tion, 2021). This reality highlights the need for continued ef- urban atmosphere for the 10-year period of 2008–2017.
forts to further identify specific emission sources such that Studies of PM2.5 speciation data derived using a consistent
effective management strategies can be formulated. Different sampling and analysis methodology over a period of as long
from criteria gaseous pollutants, PM2.5 is a complex mixture as a decade and as early as 2008 are few and far between in
containing inorganic components (e.g., sulfate, nitrate, and China and elsewhere in Asia. A few multi-year studies were
ammonium), elemental carbon (EC), organic carbon (OC) documented in the literature. One study covered a rural site
(consisting of tens of thousands of individual organic com- (Wanqinsha) in the GBA in the fall and winter seasons over
pounds), and metal oxides. The accumulation of PM2.5 pol- 6 years (2007–2012) (Fu et al., 2014). In the second study,
lution could come from direct emissions from human activi- PM2.5 speciation covered six sites in Foshan, a populated
ties and biogenic sources and/or atmospheric formation pro- city in the GBA, in winter and summer seasons over 7 years
cesses. Additionally, changes in air quality could be masked (2008–2014) (Tan et al., 2016). In the third study, PM2.5 sam-
by variations in atmospheric dispersion conditions on daily, ples were collected in urban Beijing from 2011 to 2015 (Lang
seasonal, and annual bases. The multiple layers of complex- et al., 2017). Unfortunately, these long-term studies did not
ity mean that PM2.5 mass concentration alone is insufficient follow regular sampling schedules throughout the annual cy-
to identify contributing sources or to attribute a reduction in cle. Some months of the year were not sampled, biasing their

Atmos. Chem. Phys., 22, 11557–11577, 2022 https://doi.org/10.5194/acp-22-11557-2022

W. S. Chow et al.: Ten-year trend of PM2.5 components in Hong Kong 11559

Figure 1. (a) Trend of annual average PM2.5 in Hong Kong for the period of 1999–2020 and the PRD regionwide average PM2.5 during
2015–2020. The regionwide annual PM2.5 is the average of 23 monitoring stations in the PRD Regional Air Quality Monitoring Network,
including three stations located in Hong Kong. The shaded period (2008–2017) has available PM2.5 major components and select source
tracers at Tsuen Wan (TW), an urban site in Hong Kong, with the yellow dots representing gravimetrically determined PM2.5 concentrations
from the collected filter samples at TW. (b) The percentage changes of PM2.5 and its major composition at TW using 2008 as the base year.

temporal representativeness in tracking long-term trends of and located about 3.3 km north of the city’s international
PM2.5 composition. Recently, the Chinese Central Govern- shipping port (Kwai Chung and Tsing Yi Container Ter-
ment set up the National Aerosol Composition Monitoring minals). Both high-volume and mid-volume samplers were
Network in 2017 with a view to evaluating the effectiveness equipped at the station. The high-volume sampler (Ander-
of its “2+26” strategy for improving air quality (Chen et al., sen Instrument, Smyrna, GA, USA) was loaded with a pre-
2019; Dao et al., 2019). This nationwide monitoring program baked 20 cm × 25 cm quartz fiber filter and operated at a
is expected to generate high-quality PM2.5 composition data flow of 1.13 m3 min−1 . Two types of mid-volume samplers
in the long run. Yet, a long-term dataset is not available due were used in the course of 10 years. From 2008 to 2010,
to the limited operating period. a RAAS four-channel mid-volume sampler (Andersen In-
In this work, we analyzed the trends of PM2.5 and its ma- strument, Smyrna, GA, USA) was operated, and the con-
jor components and individual source marker molecules/ele- figuration of the four channels was as below: Channels 1
ments by the seasonal and trend decomposition with locally and 4 sampling at 16.7 L min−1 and loaded with one 47 mm
estimated scatter plot smoothing (LOESS), a robust method Teflon filter and one 47 mm quartz filter and Channels 2 and
for extracting trend components from concentration time se- 3 sampling at a flow rate of 7.3 L min−1 and loaded with one
ries (Cleveland et al., 1990), and cross-compared the re- 47 mm nylon filter and one 47 mm quartz filter. From 2011
sults with the non-parametric Mann–Kendall test and Sen’s to 2017, two mid-volume samplers (Partisol R&P, Model
slope. The objectives are to quantify the 10-year variations 2025, Albany, NY, USA) were operated side by side to col-
in PM2.5 chemical composition and to characterize how ma- lect one Teflon and one quartz fiber filter at 16.7 L min−1 .
jor local and regional sources impacting PM2.5 pollution in After the collections, the 47 mm Teflon filters were stored
Hong Kong have varied in the decade. The aim of this work in Petri dishes, and the 47 mm quartz filters were stored in
is to provide a well-scrutinized long-term dataset of PM2.5 the dishes lined with aluminum foil. The dishes were sealed
chemical composition and a sound analysis of source impli- with parafilm, while the 20 cm × 25 cm quartz filters were
cations for an urban location in southern China to support folded in half and stored in aluminum foil. All the filters were
studies of control measure evaluation and formulation. packed in a thermal bag for transportation to the laboratory
and refrigerated under 4 ◦ C before chemical analysis. Field
blanks were collected at a frequency of 10 % of a sampler’s
2 Data and method routine operating frequency. They were used to monitor con-
tamination throughout the process from sampling to analysis
2.1 PM2.5 chemical speciation data and for background correction. The field blank filters were
stored and analyzed following the same procedures as those
PM2.5 filter samples were collected on a 24 h basis (mid- of PM2.5 sample filters for quality assurance purposes. A to-
night to midnight) once every 6 d from 2008 to 2017 at tal of 592 sets of filter samples were collected, each set con-
Tsuen Wan (TW, 22◦ 220 18 N, 114◦ 60 52 E), an urban air qual- sisting of one 20 cm × 25 cm filter and multiple 47 mm filter
ity monitoring station (AQMS) in Hong Kong. TW is a samples for a sampling day.
station surrounded by residential and commercial buildings

https://doi.org/10.5194/acp-22-11557-2022 Atmos. Chem. Phys., 22, 11557–11577, 2022

11560 W. S. Chow et al.: Ten-year trend of PM2.5 components in Hong Kong

A suite of chemical speciation analysis was conducted on provided by the Environmental Central Facility from the
the collected sample filters (Table 1). Specifically, the 47 mm Hong Kong University of Science and Technology (HKUST
Teflon filters were used to determine the PM2.5 mass concen- ENVF, http://envf.ust.hk/dataview/gts/current/, last access:
trations by gravimetric analysis and the trace element con- 20 June 2021). The temporal variations of gaseous pollutants
centrations by energy-dispersive X-ray fluorescence spec- serve as additional data valuable for exploring the effects of
trometry (XRF). The Nylon filters from 2008–2010 and the changes in precursor gases on the secondary formed PM2.5
quartz filters from 2011–2017 were used to quantify the constituents (e.g., sulfate and/or nitrate). The decadal time
concentrations of major ions (SO4 2− , NO3 − , Cl− , NH4 + , series of temperature, relative humidity (RH), wind speed
Na+ , K+ , Ca2+ , and Mg2+ ) by ion chromatography (IC). and direction, and precipitation at the site are summarized
OC and EC were measured by a thermal–optical transmit- in Sect. S1 in the Supplement.
tance (TOT) method. Concentrations of saccharides were an-
alyzed by high-performance anion-exchange chromatogra- 2.3 Seasonal and trend decomposition with the LOESS
phy coupled with pulsed amperometric detection (HPAEC- method (STL)
PAD) (Kuang et al., 2015). Non-polar organic compounds,
including alkanes, polycyclic aromatic hydrocarbons, and For the evaluation of the overall trend of time series indepen-
hopanes, were quantified by a method coupling in-injection dently of seasonal influence, the season and trend decompo-
port thermal desorption gas chromatography with mass spec- sition was adopted to decompose a time series (Yv ) into three
trometry (TD-GC/MS) (Ho and Yu, 2004; Ho et al., 2008). components: the trend component (Tv ), the seasonal compo-
Data validation on the PM2.5 speciation was carried out nent (Sv ), and the remainder (Rv ) in an additive or a multi-
at three levels according to the publication “Guideline on plicative manner, as per Eq. (1a) and (1b) (Cleveland et al.,
Speciated Particulate Monitoring” prepared for the USEPA 1990).
by Chow and Watson (1998). Level I validation mainly
consists of flagging measurements that deviate from proce- Y v = T v + Sv + R v (1a)
dures through reviewing sampling log sheets and field quality Yv = Tv × Sv × Rv (1b)
check records and identifying invalid values. Level II valida-
tion checks the internal consistency among data from differ- The STL algorithm is performed via LOESS under two iter-
ent analyses, involving the following: (1) comparing a sum of ative loops. In comparison with other time series decompo-
measured chemical species vs. PM2.5 mass concentrations, sition techniques, such as the simplest moving average (MA)
(2) comparing total sulfur by XRF vs. sulfate by IC, (3) com- method (Molugaram and Rao, 2017), the STL method has
paring total potassium by XRF and soluble potassium by IC), more flexibility in parameter tuning, as well as higher robust-
(4) calculating anion/cation balances, and (5) examining time ness to counterpart the influences from outliers. A detailed
series data to identify and investigate outliers. Level III val- description about its algorithm is provided in Sect. S2 in the
idation is part of the data interpretation process, mainly fo- Supplement. The STL method has been readily implemented
cusing on identification of unusual values through parallel and widely tested in most programming languages such as R
consistency tests with other independent datasets. and Python. In this study, we utilized the STL function in the
Details of analytical procedures and data validation stlplus package in R for the following calculation.
are documented in our previous studies (Huang et al., Before applying the STL method, we manually inspected
2014; Chow et al., 2022) and in a series of project re- the data and removed data points exceeding the upper quar-
ports (Yu et al., 2012, 2013, 2014, 2015; Yu and Zhang, tile by 3 times the interquartile range (i.e., X75 % +3(X75 % −
2017, 2018; Chow et al., 2016), which are available X25 % )) to avoid the influence of extreme concentrations on
at http://www.epd.gov.hk/epd/english/environmentinhk/air/ the trend slope (Singh et al., 2021; Bigi and Ghermandi,
studyrpts/pm25_study.html (last access: 3 February 2022). 2014). The concentration data were found lognormally dis-
See Yu et al. (2022) for dataset access details. tributed in Q–Q plots as shown in Figs. S5 and S6 in the
Supplement. Thus, log-transformation and monthly averag-
2.2 Gaseous pollutant and meteorological parameter ing were applied to create a normally distributed time series
data with even time intervals to cope with the assumptions of the
STL model.
Criteria gaseous pollutant data (CO, SO2 , O3 , and NOx )
and meteorological parameters at TW are from the network 2.4 Generalized least squares–autoregressive moving
of air quality monitoring station operated by the HKEPD. average (GLS-ARMA) model
The criteria gaseous pollutant data are available from the
HKEPD Environmental Protection Interactive Centre (https: The trend curves from the STL method are often too irreg-
//cd.epic.epd.gov.hk/EPICDI/air/download/, last access: 17 ular to be described verbally or quantitatively. This prompts
June 2021), while the meteorological parameters can be re- us to seek a method that allows for calculation of an over-
trieved from the Atmospheric and Environmental database all changing rate for the trend component for further anal-

Atmos. Chem. Phys., 22, 11557–11577, 2022 https://doi.org/10.5194/acp-22-11557-2022

W. S. Chow et al.: Ten-year trend of PM2.5 components in Hong Kong 11561

Table 1. PM2.5 and list of components targeted for trend analysis and their measurement methods.

Species Source characteristics Measurement method

PM2.5 Gravimetry1
OC Primary emissions and secondary formation with volatile Thermal–optical analysis2
organic compounds (VOCs)
as direct precursors
EC Combustion sources
Sulfate Secondary with SO2 as direct precursor Ion chromatographic analysis of aerosol water
Nitrate Secondary with NOx as direct precursor extracts3
Ammonium Secondary, particle presence in close association with
sulfate and nitrate
K+ Biomass burning, sea salt, dust
Al, Si Soil dust Energy-dispersive X-ray fluorescence
Ni, V Residual oil combustion spectrometry3,4
Pb, Cu, Zn Coal combustion, metal industries
Hopanes Fossil fuel uses such as vehicular emission, residual Thermal desorption GC/MS5,6
oil burning
Levoglucosan Biomass burning High-performance anion-exchange chromatography
coupled with pulsed amperometric detection7
1 USEPA (2016). 2 Chow et al. (2007). 3 Huang et al. (2014). 4 Watson et al. (1999). 5 Ho and Yu (2004). 6 Ho et al. (2008). 7 Kuang et al. (2015).

ysis. When dealing with time series data with autocorrela- brid STL–GLS-ARMA method are provided in Sect. S3 in
tion (i.e., the current value (Yv ) depends on its lagged values the Supplement.
(Yv−h )), generalized least squares (GLS), instead of ordinary
least squares (OLS), is more suitable for the quantification of 2.5 Comparison with other trend analysis methods
the changing rate of the time series. In GLS, the covariance
matrix (and so the residuals) can be estimated by an autore- Additional trend analysis methods were explored to cross-
gressive moving average (ARMA(p, q)) model. Specifically, validate the results from the GLS-ARMA method. First,
the ARMA(p, q) model assumes that the current value (Xt ) Sen’s slope method (Wilcox, 2017), a non-parametric
is influenced by its p-order of lagged values (Xt−h ) and q- method, was performed on the same dataset to calculate the
order of lagged residuals (εt−i ) with the corresponding coef- changing rate. In Sen’s slope method, we first run a Mann–
ficients φh and θi as shown in Eq. (2). Kendall test to see whether the overall trend of the annual
averages is monotonic. Then the median of the slopes for all
p q
X X pairwise data points is computed and defined as Sen’s slope.
Xt = φh Xt−h + εt + θi εt−i (2)
Second, an exponential trend estimation was computed using
h=1 i=1
the compound annual growth rate (CAGR) for each pairwise
The determination of p and q in an ARMA(p, q) model combination of annually averaged values using Eq. (3).
is achieved by minimizing model selection criteria such as
1
Akaike’s Information Criterion (AIC) and Bayesian Informa- 
Xtn

tn −t0
tion Criterion (BIC). A model with a smaller AIC or BIC CAGR(%) = × 100 %, (3)
X t0
value is deemed more likely to generate the data that we
obtain, while taking both probability likelihood and model where Xtn and Xt0 are the annual averaged time series at time
simplicity into consideration. Despite the differences in as- tn and t0 , respectively. Like Sen’s slope method, the overall
sumptions, theoretically AIC and BIC give similar results in trend of a species is represented by the median value of all
most cases (Shumway and Stoffer, 2017). In our study, we the CAGR results. Testing the agreement between the linear
calculated AIC, AICc (a bias-corrected AIC), and BIC. As and nonlinear approaches helps in validating the overall trend
tabulated in Table S1 in the Supplement, these criteria indi- analysis results. Details of the CAGR results are provided in
cators show no significant difference in terms of the param- Sect. S7 in the Supplement.
eter selection outcomes. Thus, hereafter only BIC values are
reported when determining the slope of each GLS-ARMA
model analysis. The details of the methodology of this hy-

https://doi.org/10.5194/acp-22-11557-2022 Atmos. Chem. Phys., 22, 11557–11577, 2022

11562 W. S. Chow et al.: Ten-year trend of PM2.5 components in Hong Kong

3 Results and discussion of the 10-year trends for PM2.5 , its major components and
source tracers will be provided in the ensuing sections.
For a simple illustration of the 10-year change in PM2.5
3.1 PM2.5 composition chemical composition, the average chemical compositions in
the starting and the ending year of the decade are compared
PM2.5 and its major components collected over the decade in Fig. 3. On the annual average basis, the top four major
are displayed in time series of monthly averages in Fig. 2a components remain to be the same, i.e., OM, sulfate, am-
and annual averages in Fig. 2b. The time series of individual monium, and nitrate, collectively accounting for a compa-
samples are provided in Fig. S3 in the Supplement. Under rably ∼ 84 % of PM2.5 in 2008 and 2017, despite 10 years
influence of the monsoon winds, the four seasons in Hong apart. Among the four top contributors, OM has gained a
Kong are well distinguished in their meteorological charac- few percent, while nitrate has been reduced by a few per-
teristics, with summer and winter being the two longest sea- cent in proportional importance. The 10-year compositional
sons and each lasting approximately 4 months. The four sea- changes are more prominent in the seasonal averages. For
sons are approximately spring from 16 March to 14 May, winter PM2.5 , the relative importance of OM increased (up
summer from 15 May to 15 September, fall from 16 Septem- from 35 % in 2008 % to 39 % in 2017), while the relative
ber to 15 November, and winter from 16 November to 15 abundance of nitrate decreased (down from 9.1 % to 6.6 %),
March (Chin, 1986). Under the influence of the Asian mon- as well as EC (down from 6.1 % to 4.4 %). For summer
soon, the northerly prevailing winds carry dry and polluted PM2.5 , the most significant compositional changes are also
northern continental air masses to Hong Kong in winter- OM (up from 32 % to 44 %), EC (down from 17 % to 8.5 %),
time, whereas prevailing southerly and southeasterly mon- and nitrate (down from 4.3 % to 1.7 %). The proportional de-
soon winds in summertime bring largely clean marine air crease of EC was most notable, reflecting the effectiveness
masses from the South China Sea or the northwestern Pa- of local vehicular emissions control measures. “Salt”, con-
cific Ocean. As a result, PM2.5 and other pollutants show sisting of Na+ and Cl− , was a very minor part of PM2.5 , ac-
distinct winter–summer contrast in their source origins and counting for less than 4 % in all years.
in concentration levels (Yu et al., 2004). In summer, lo-
cal emissions have a dominant influence, while in winter, 3.2 Annual trend analysis
the regional/super-regional pollution significantly elevated
air pollutant levels. We thus show separate time series for Previous studies that examined the annual trend of pollu-
summer and winter seasonal average PM2.5 chemical com- tants for evaluation of pollutants reduction in Hong Kong
position in Fig. 2c and d, and a discussion of the source adopted a simple method of comparing annual average val-
trend according to seasons of summer and winter provides ues (HKEPD, 2020; Zhang et al., 2018; Lu et al., 2013; Yuan
a more direct understanding of source variations over the et al., 2013). While this approach avoids the autocorrela-
years. Spring and fall, being two short and transient seasons, tion issue – the lag value of variables (Yt−h ) influences the
display more variable and mixed influences from both local current value (Yt ) in time series – it would suffer increased
and regional/super-regional sources (Fig. S4 in the Supple- bias due to the sacrifice of the sample size for estimation.
ment). Their time series are less useful for tracking decadal In comparison, STL is a more robust method for extracting
source variations and are therefore not discussed in this pa- trend components from concentration time series (Cleveland
per. et al., 1990), with the autocorrelation issue accounted for by
As shown in Fig. 2, an overall decline trend is clearly seen GLS-ARMA (Shumway and Stoffer, 2017). The STL–GLS-
in both bulk PM2.5 and its major components over the decade ARMA method has been adopted in a few studies analyzing
of 2008–2017. Sulfate and organic matter (OM) remain to air pollutant trends (e.g., Anttila and Tuovinen, 2010; Bigi
be the top two dominant PM2.5 components throughout the and Ghermandi, 2014). It is found that STL–GLS-ARMA
decade and for both winter and summer seasons. Significant has the advantage of retaining more degrees of freedom on
monthly variations are also evident, with highest concentra- sample population and thus producing a more accurate esti-
tions in the winter months and the lowest in the summer mate than the ordinary least-squares method.
months. The highest winter month average could be more We applied STL–GLS-ARMA to the monthly average
than double the lowest summer average concentration in a concentrations of PM2.5 mass and individual species, includ-
same year, clearly indicating the significant contribution of ing major components (SO4 2− , NO3 − , NH4 + , OC, and EC)
regional/super-regional pollution to PM2.5 in Hong Kong. and source-specific molecular or elemental tracers (i.e., K+ ,
Comparing Fig. 2c and d, we see that the mass reductions in Al, Si, V, Ni, Pb, Zn, Cu, hopanes, and levoglucosan), as
the summer season over the decade are much less in compar- well as the routinely monitored criteria gaseous pollutants
ison with those seen for the winter; however, a continuous (CO, SO2 , NOx , and O3 ) (Fig. 4). Table 2 summarizes the
decline in EC is clear in the decade-long time series of the slopes obtained from the GLS-ARMA, Sen’s slope method,
summer averages, indicating success in controlling local EC and percentage change of each species over 2008–2017, to-
sources (i.e., vehicular emissions). A quantitative description gether with the annual average concentration data in 2008

Atmos. Chem. Phys., 22, 11557–11577, 2022 https://doi.org/10.5194/acp-22-11557-2022

W. S. Chow et al.: Ten-year trend of PM2.5 components in Hong Kong 11563

Figure 2. Time series of PM2.5 chemical composition from 2008 to 2017 in the form of (a) monthly averages, (b) annual average, (c) winter
seasonal averages, and (d) summer seasonal averages. In the legends, “Salts” includes Na+ and Cl− ; “Crustal” represents crustal materials,
computed to be 1.89 · Al + 2.14 · Si + 1.4 · Ca + 1.43 · Fe; and “Tracers” includes elements other than Na, Cl, S, K, Al, Si, Ca, and Fe. OM
refers to organic matter and is computed as 1.4 · OC.

and 2017. The results from both the slope-determining meth- (NH4 + , NO3 − , EC: −0.12, −0.17, −0.17 µg m−3 yr−1 ); they
ods were in good agreement for all the PM2.5 species. The accounted for a similar percentage at around 8.0 %–11 %
GLS-ARMA trend slopes are significantly less than zero at a each and a combined 31 % of the overall PM2.5 reduc-
p level of < 0.001 for all PM2.5 measurement parameters, tion. Meanwhile, other components such as biomass burning
except for V and levoglucosan, which are significant at a markers (K+ and levoglucosan), industrial and coal combus-
higher p level (0.01 and 0.05, respectively). For Sen’s slopes, tion tracers (Zn and Pb), and crustal materials (Al, Si, and
they are less than zero at a lower p level of 0.01 for most Ca) altogether explain the remaining 28 % of PM2.5 deple-
species and at p = 0.05 for levoglucosan and are not signifi- tion. Note that the tracer species only account for a minute
cant at p > 0.05 for V and hopanes. Such differences reflect amount of mass; however, they are indicative of other un-
the superiority of the GLS-ARMA method arising from re- measured PM2.5 components co-emitted with these sources.
taining more degrees of freedom on the sample population. The percentage changes during the decade are calculated
Thus, we will adopt the GLS-ARMA slopes in commenting using the annual average in 2008 and 2017 and are listed
the 10-year changing rate in later discussion. in Table 2. With the GLS-ARMA model fitted data, we can
From the results of the STL–GLS-ARMA method, a de- also calculate the percentage changes. Comparing the two
clining rate of 1.5 µg m−3 yr−1 was estimated for the PM2.5 approaches, the GLS-ARMA method yields higher percent-
mass. This decline was significantly attributed by the top age drops in K+ , NO3 − , Al, Si, Pb, and Cu than those
two major components, namely sulfate accounting for 24 % calculated using annually averaged data. This could be ex-
(Slope: −0.36 µg m−3 yr−1 ) and OM 17 % (−0.18 · 1.4 = plained by the different concentration levels fitted by the
−0.25 µg m−3 yr−1 ), respectively. Ammonium, nitrate, and GLS-ARMA model. The underestimated concentration in
EC decreased in a similar rate in mass concentration unit 2017 by GLS-ARMA results from the flattened variation in

https://doi.org/10.5194/acp-22-11557-2022 Atmos. Chem. Phys., 22, 11557–11577, 2022

W. S. Chow et al.: Ten-year trend of PM2.5 components in Hong Kong

https://doi.org/10.5194/acp-22-11557-2022
Table 2. Summary of the decadal trend slopes and percent changes from 2008 to 2017 obtained with the GLS-ARMA analysis of the deseasonalized time series and with Sen’s slope
from annual averaged time series.
Concentration Slope % Relative change (2008–2017)
Species Unit 2008 2017 ARMA model1 GLS-ARMA2 Sen’s slope2 GLS-ARMA Annual-averaged Difference3
Gaseous pollutants
CO µg m−3 621 662 ARMA (1,0) 20, (−1.5, 41) 18, (−16, 48) 28 % 6.6 % 21 %
SO2 µg m−3 28.4 10.7 ARMA (1,0) −1.2∗∗∗ , (−1.8, −0.65) −0.71∗∗ , (−1.6, −0.29) −51 % −62 % 11 %
NOx µg m−3 121 77 ARMA (1,0) −3.5∗∗∗ , (−4.5, −2.4) −2.9∗ , (−4.9, −1.5) −26 % −36 % 10 %
O3 µg m−3 30.9 41.6 ARMA (1,0) 0.94∗∗∗ , (0.41, 1.5) 0.82∗ , (0.26, 1.4) 30 % 35 % −5 %
Particle pollutants
PM2.5 µg m−3 37.5 22.4 ARMA (1,0) −1.5∗∗∗ , (−1.9, −1.1) −1.6∗∗ , (−2.3, −1.3) −40 % −40 % 0%
SO4 2− µg m−3 11.0 6.60 ARMA (0,2) −0.36∗∗ , (−0.58, −0.15) −0.43∗∗ , (−0.72, −0.19) −33 % −40 % 7%
NO3 − µg m−3 2.64 0.91 ARMA (1,0) −0.17∗∗∗ , (−0.21, −0.12) −0.23∗∗ , (−0.33, −0.073) −85 % −66 % −19 %
NH4 + µg m−3 3.99 2.41 ARMA (2,0) −0.12∗ , (−0.22, −0.027) −0.16∗∗ , (−0.31, −0.038) −32 % −40 % 8%
Na+ ng m−3 364 371 ARMA (0,1) −5.5, (−15, 3.4) −5.8, (−19, 13) −13 % 1.8 % −15 %
OC µg C m−3 8.22 6.33 ARMA (2,0) −0.18∗ , (−0.31, −0.042) −0.22∗ , (−0.43, −0.082) −23 % −23 % 0%
EC µg C m−3 2.95 1.18 ARMA (0,2) −0.17∗∗∗ , (−0.2, −0.13) −0.16∗∗∗ , (−0.26, −0.12) −56 % −60 % 4%
Al ng m−3 223 140 ARMA (1,0) −13∗∗∗ , (−17, −8) −12∗∗ , (−21, −8.5) −51 % −37 % −14 %
Si ng m−3 412 222 ARMA (1,0) −27∗∗∗ , (−35, −18) −30∗∗ , (−40, −19) −60 % −46 % −14 %
ng m−3 −0.60∗ , (−1.1, −0.14)

Atmos. Chem. Phys., 22, 11557–11577, 2022

V 23.6 15.7 ARMA (1,0) −0.53, (−1.4, 0.14) −24 % −34 % 10 %
Ni ng m−3 6.94 4.46 ARMA (1,0) −0.29∗∗∗ , (−0.42, −0.16) −0.29∗ , (−0.47, −0.11) −35 % −36 % 1%
Pb ng m−3 56.6 18.8 ARMA (1,0) −3.8∗∗∗ , (−5, −2.7) −4.7∗∗ , (−6.7, −3.5) −75 % −67 % −8 %
Zn ng m−3 185 112 ARMA (2,2) −7.0∗∗ , (−12, −2) −9.5∗ , (−16, −2.9) −39 % −40 % 1%
Cu ng m−3 20.6 11.7 ARMA (1,0) −1.1∗∗∗ , (−1.7, −0.6) −1.4∗∗ , (−2.1, −0.65) −50 % −43 % −7 %
K+ ng m−3 556 223 ARMA (1,0) −32∗∗∗ , (−41, −23) −40∗∗ , (−52, −24) −65 % −60 % −5 %
Hopanes ng m−3 0.708 0.180 ARMA (1,0) −0.052∗∗∗ , (−0.075, −0.03) −0.041, (−0.06, 0.015) −71 % −75 % 4%
Levoglucosan ng m−3 61.3 31.7 ARMA (0,1) −1.4∗ , (−2.8, −0.089) −2.6∗ , (−4.9, −1.2) −31 % −47 % 16 %
1 The optimal ARMA model parameters were selected by BIC. 2 Numbers in parentheses denote the 95 % confident interval for the slope value, while asterisks denote that the slope significantly differs from zero: ∗ p < 0.05,
∗∗ p < 0.01, ∗∗∗ p < 0.001. 3 The difference between the two methods (GLS-ARMA − annual-averaged).
11564
W. S. Chow et al.: Ten-year trend of PM2.5 components in Hong Kong 11565

Figure 3. Comparison of average PM2.5 compositions between the starting year (2008) and the ending year (2017) of the decade. In each
year, three averages are shown, corresponding to annual, winter, and summer, and the donut size is proportional to the PM2.5 concentration.

the later years (Fig. 4), hence the higher percentage changes 3.3 Trend analysis of winter and summer data
in these species. This problem was less obvious for species
with smoother declines such as SO2 , NOx , OC, EC, V, Ni, As discussed in Sect. 3.1, PM2.5 levels and dominant sources
and hopanes. Therefore, the differences in percentage change are distinctly different in winter and summer. The two sea-
between the two methods helped on identifying the different sons merit separate analysis of their 10-year trends. This is
changing characteristics along the time series. further supported by correlation and hierarchical clustering
The annual percent change rates computed using CAGR, analysis of year-by-year data, the results of which are shown
summarized in Fig. S13 and Table S5 in the Supplement, in Fig. 5 using 2008 and 2017 as examples. Figure 5 reveals
show a good agreement with those from the linear ap- that the measurement variables segregate into two clusters
proaches (i.e., GLS-ARMA and Sen’s slope). In general, the marked in black and pink linkage lines, respectively, and they
exponential approach of the CAGR method estimates a larger broadly correspond to one group of pollutants known to be
decline than the GLS-ARMA method. The maximum differ- significantly influenced by regional/super-regional sources
ence occurs with NO3 − (−10 %). The relative constant con- (e.g., OC, sulfate, nitrate, NH4 + , K+ , Pb, Zn, and Cu) and
centration levels in the later years were particularly observed a second group of species with dominant contributions from
in NO3 − , which would imply a faster reduction in an expo- local sources (i.e., NOx , EC, hopanes, Ni, and V). The re-
nential variation model and thus result in a larger discrep- gional sources have strong seasonality under the influence of
ancy. The absolute differences for all the other species are the monsoon winds. It is of interest to note that SO2 shifted
less than 5 %. For simplicity, we confine the trend discussion from the local cluster to the regional cluster over the decade,
to results from the linear approaches. reflecting the changing relative importance of local vs. re-
The 10-year percentage change in PM2.5 is −40 %. Sul- gional emission sources of SO2 over the years (see Sect. S3
fate and ammonium, with nearly identical decrease trajec- for more details).
tories due to their close chemical linkage, have percentage Figure 6 shows the 10-year variations of the seasonal av-
drops that closely match the drop of bulk PM2.5 (Fig. 1b and erage concentrations of PM2.5 and selected components for
Table 2). Other major components, however, differ in their winter and summer. Season-specific values for Sen’s slopes
percentage reductions from the reduction of bulk PM2.5 , with are listed in Table 3, expressed in both mass concentration
the reduction in nitrate (−66 %) and EC (−60 %) exceeding change rate per year and percent change rate per year. The
that of bulk PM2.5 , while OC (−23 %) falls below that of latter unit allows for a direct comparison of relative source
bulk PM2.5 . Such results reveal the effectiveness of control strength changes of local and regional sources by removing
measures in lowering EC and sulfate and the increasing im- the impact of meteorological factors (e.g., boundary layer
portance of OC in addressing PM2.5 pollution in the coming height) on ambient concentrations. Seen in Table 3, Sen’s
years. slope for bulk PM2.5 is significantly different seasonally, at
−2.0 µg m−3 yr−1 in winter vs. −0.67 µg m−3 yr−1 in sum-

https://doi.org/10.5194/acp-22-11557-2022 Atmos. Chem. Phys., 22, 11557–11577, 2022

11566 W. S. Chow et al.: Ten-year trend of PM2.5 components in Hong Kong

Figure 4. (a) Gaseous and (b) PM2.5 pollutant data over 2008–2017: monthly concentrations (yellow), trend component (blue), and the
slope line of trend determined by GLS-ARMA method (red). Note that the concentrations of hopanes and levoglucosan are in nanograms per
cubic meter (ng m−3 ), while the others are in micrograms per cubic meter (µg m−3 ).

Atmos. Chem. Phys., 22, 11557–11577, 2022 https://doi.org/10.5194/acp-22-11557-2022

W. S. Chow et al.: Ten-year trend of PM2.5 components in Hong Kong 11567

Figure 5. Correlation matrix of gaseous and particulate pollutants with hierarchical clustering results at 2008 (a) and at 2017 (b). Clusters:
pink – local sources; black – regional sources.

mer, while the percentage decline rates are comparable, at −3.5 %, −6.3 %, and −1.4 % yr−1 in percentage change rate
−3.9 % yr−1 in winter and −3.7 % yr−1 in summer. In align- for sulfate, nitrate, and ammonium, respectively (Table 3).
ment with the species segregation revealed in Fig. 5, the They are significant drivers of PM2.5 decline.
group of regional species shows significantly larger decrease While the direct precursor for sulfate is SO2 , the reduction
rates in mass concentration in winter than in summer, but the of SO2 does not necessarily translate to proportional reduc-
group of local species (EC, V, and Ni) displays a compara- tion in sulfate, as various oxidants (e.g., hydroxyl radical, hy-
ble value for Sen’s slope in both seasons. It is worth noting drogen peroxide, and ozone) participate in the oxidation for-
that summer OC does not show a discernable increase or de- mation of sulfate from SO2 , and the role of each oxidant is
crease trend over the decade, but winter OC shows a decreas- highly dynamic in both temporal and spatial scale (e.g., Xue
ing trend with a slope of −0.45 µg C m−3 yr−1 (Fig. 6). Such et al., 2019). Nevertheless, it is informative to compare the
a stark contrast indicates a significant seasonal difference in changing rates of SO2 and sulfate. As a criteria gaseous pol-
OC sources and their underlying driving factors. This also lutant, SO2 has been extensively studied, and its emission in-
implies that measures to lower the OC contribution in PM2.5 ventories for Hong Kong and Guangdong province are avail-
must consider the strong seasonality of its sources. able (HKEPD, 2021b; Li et al., 2017; Zheng et al., 2018).
Considering the diverged seasonality among major com- The SO2 emission inventory data for our study decade are
ponents and source tracers, we individually examine in the shown in Sect. S3. The top two sources for SO2 emissions in
subsequent sections the trend characteristics of the major Hong Kong are power plants and marine vessels, while the
PM2.5 constituents and important sources that have effective major SO2 sources in Guangdong are power plants and in-
tracer data. dustries (Fig. S8 in the Supplement). The emission and am-
bient concentration trends of SO2 , normalized to 2018, are
examined in Fig. S8c, showing that the yearly variation of
3.4 Secondary inorganic aerosol components ambient SO2 concentrations at TW was similar to the total
SO2 emission trend from Hong Kong and SO2 emission from
The three secondary inorganic aerosol components, namely power plants in Guangdong. Overall, the changes in ambient
sulfate, nitrate, and ammonium, are constantly prominent SO2 concentrations at TW during the 10-year period are con-
components of the PM2.5 and make up 43 %–47 % of sistent with the SO2 emissions estimated for the GBA.
PM2.5 mass over the decade. Their ambient abundances Over the decade, sulfate dropped by 40 % in annual av-
exhibit a strong seasonality, with the winter concentra- erage concentration, lagging behind the 62 % drop recorded
tions more than double the summer concentrations. Sea- for SO2 (Table 2). A close examination of the 10-year time
sonally, the wintertime levels changed by −0.50, −0.38, series of monthly concentrations of SO2 and sulfate side by
and −0.16 µg m−3 yr−1 in mass concentration change rate side (Fig. 7) shows temporally uneven reduction. The steep-
and at −3.8 %, −8.4 %, and −3.2 % yr−1 in percentage est drop in ambient SO2 occurred in 2008–2009 (from 28.4
change rate for sulfate, nitrate, and ammonium, respec- to 15.6 µg m−3 , a reduction of 45 %), corresponding to the
tively. The summertime level changed by −0.17, −0.05, and significant SO2 emission resulting from the mandated switch
−0.02 µg m−3 yr−1 in mass concentration change rate and at

https://doi.org/10.5194/acp-22-11557-2022 Atmos. Chem. Phys., 22, 11557–11577, 2022

11568 W. S. Chow et al.: Ten-year trend of PM2.5 components in Hong Kong

Table 3. Summary of the seasonal variation estimated by Sen’s slope.

Species Sen’s slope (mass concentration)1 Sen’s slope (percent change)2

Unit Winter Summer Winter Summer
Gaseous pollutants
CO µg m−3 yr−1 +8.9 +21 +1.1 % +4.4 %
SO2 µg m−3 yr−1 −1.0∗ −0.98∗∗∗ −3.9 %∗ −2.8 %∗∗∗
NOx µg m−3 yr−1 −4.4∗∗ −3.7∗∗ −3.3 %∗∗ −3.1 %∗∗
O3 µg m−3 yr−1 +1.6∗ +0.56 +4.9 %∗ +2.6 %
PM2.5 and its components
PM2.5 µg m−3 yr−1 −2.0∗ −0.67∗ −3.9 %∗ −3.7 %∗
SO4 2− µg m−3 yr−1 −0.50 −0.17 −3.8 % −3.5 %
NO3 − µg m−3 yr−1 −0.38∗∗ −0.049∗ −8.4 %∗∗ −6.3 %∗
NH4 + µg m−3 yr−1 −0.16∗ −0.019 −3.2 %∗ −1.4 %
OC3 µg C m−3 yr−1 −0.45 0.0067 −3.6 % +0.16 %
EC µg C m−3 yr−1 −0.19∗∗ −0.18∗∗ −6.2 %∗∗ −5.8 %∗∗
Al ng m−3 yr−1 −15∗∗ −5.1 −5.3 %∗∗ −4.8 %
Si ng m−3 yr−1 −35∗∗ −8.5∗ −7.0 %∗∗ −4.1 %∗
V ng m−3 yr−1 −1.2∗ −1.1∗ −7.0 %∗ −3.4 %∗
Ni ng m−3 yr−1 −0.37∗ −0.42∗ −5.9 %∗ −4.7 %∗
Pb ng m−3 yr−1 −7.0∗∗ −0.91∗ −7.7 %∗∗ −7.8 %∗
Zn ng m−3 yr−1 −18∗∗ +2.6 −6.6 %∗∗ +4.4 %
Cu ng m−3 yr−1 −1.6 −0.17 −5.8 % −2.3 %
K + ng m−3 yr−1 −62∗∗ −11 −7.5 %∗∗ −5.5 %
Hopanes ng m−3 yr−1 −0.067∗∗ −0.041∗∗ −6.4 %∗∗ −8.1 %∗∗
Levoglucosan ng m−3 yr−1 −4.7 −1.8 −3.9 % −10 %
1 Asterisks in the table denote that the slope significantly differs from zero: ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001.
2 Sen’s slopes in these two columns are obtained on normalized concentrations against those in 2008, thus providing
percentage change rates relative to 2008, with the unit of percentage per year (% yr−1 ). 3 Sen’s slope for wintertime OC is
significant at a p level of 0.11.

to ultra-low S (< 0.005 wt %) for all commercial and indus- NH4 + closely tracks that of sulfate, as confirmed in our
trial processes in 2008 (Fig. S9 in the Supplement). During dataset.
the same period, sulfate dropped by 22 % from 11.0 µg m−3 It is well established that PM2.5 nitrate is a secondarily
in 2008 to 8.6 µg m−3 in 2009. Between 2009–2014, both formed product from NOx oxidation (e.g., Griffith et al.,
ambient SO2 and sulfate dropped by the same small percent 2015). Like the formation of sulfate, the involvement of mul-
(∼ 4 %) and varied in a narrow range of 14.6–16.2 µg m−3 for tiple oxidants (e.g., hydroxy radical, O3 ) creates significant
SO2 and 8.66–9.03 µg m−3 for sulfate. Between 2015–2017, complexity so that a proportional relationship is not expected
the introduction of SO2 reduction measures targeting power between variations of NOx and nitrate (Xue et al., 2014b).
plants and shipping industry (Fig. S9) led to a decrease of Additionally, atmospheric physical conditions, such as tem-
SO2 by 22 % (from 13.8 to 10.7 µg m−3 ), while sulfate only perature and RH, also strongly influence the partitioning of
dropped by 11 % (from 7.45 to 6.60 µg m−3 ) (Fig. 7). Ev- nitrate between gas and particle phase. Comparing the reduc-
idently, the discrepant changing rates of ambient SO2 and tion rates of NOx and nitrate, we note that over the decade
sulfate confirm that sulfate reduction is generally not pro- nitrate dropped by 66 %, higher than the reduction rate of
portional to local SO2 reduction because of nonlinear for- 36 % for NOx (Table 2). While the deviation from propor-
mation chemistry of sulfate and a significant contribution to tionality reflects the nonlinear formation chemistry of nitrate,
sulfate from regional transport (Chen et al., 2021; Chow et the higher reduction rate in nitrate is seemingly counterintu-
al., 2022). itive. Unlike sulfate, which predominantly exists in the par-
The very origin of PM2.5 NH4 + , i.e., reaction of ammonia ticle phase, nitrate could be either present as nitric acid in
with sulfate aerosol, dictates its close association with sul- the gas phase or as ammonium nitrate partitioning between
fate. This relationship is expectedly confirmed in the excel- gas–particle phases. Additionally, nitrate could significantly
lent correlation of NH4 + with sulfate in all the years (Fig. 5). partition to coarse particles (PM2.5−10 ) (Xue et al., 2014a).
As NH3 is generally abundantly supplied, the variation of Thus, PM2.5 nitrate, mainly existing in the form of ammo-
nium nitrate, only represents a fraction of the total nitrate.

Atmos. Chem. Phys., 22, 11557–11577, 2022 https://doi.org/10.5194/acp-22-11557-2022

W. S. Chow et al.: Ten-year trend of PM2.5 components in Hong Kong 11569

Figure 6. The 10-year variations of seasonal average concentrations of PM2.5 and selected components for the winter (blue) and summer
(orange) from 2008 to 2017. Red lines indicate Sen’s slope, and dashed blue lines indicate the 95 % confidence intervals.

This provides possibility for a higher PM2.5 nitrate reduction ductions indicate the effectiveness of an array of control mea-
rate than its precursor NOx . Zhang et al. (2018) examined sures that have been implemented by the government since
the PM10 chemical speciation data in Hong Kong that spans 2008 (Fig. 8). Most notably, they include (1) replacing pre-
18 years (1998–2015) and found nitrate in PM10 increased Euro IV diesel commercial vehicles with higher Euro stan-
from 2002 to 2011 then decreased afterwards. Such an ob- dards vehicles since 2007, (2) implementing the Statutory
servation indirectly indicates that the significant presence of Ban against idling of motor vehicle engines in 2011, and (3)
nitrate in coarse PM could lead to divergent trends of nitrate the imposition of the emission control for petrol and LPG
in PM10 and PM2.5 . A more detailed consideration with the vehicles in 2014. It is worth noting that the vehicular traffic
aid of modeling would be needed in order to reveal the vari- local to the sampling site has increased by ∼ 20 % over the
ation extent of total nitrate and the distribution of different decade if we use the traffic flow count through Shing Mun
nitrate forms. Such an exploration requires efforts going be- Tunnel, a tunnel less than 5 km away from the site, as an in-
yond the current project, and it is therefore not pursued. dicator (Fig. S2 in the Supplement). Despite the increase of
vehicles on the road, the decrease of ambient EC and hopanes
is unambiguous, which serves as strong evidence for the ef-
3.5 Components dominated by local emissions – fectiveness of vehicular emission controls. On a separate yet
vehicular and shipping emissions relevant note, Wang et al. (2018) sampled and compared both
It has been recognized that on-road vehicles and marine ves- gaseous and particulate pollutants from fresh vehicular emis-
sels are two major local emission sources for ambient PM2.5 sions in the Shing Mun Tunnel in 2003 and 2015 and found
in Hong Kong (Guo et al., 2009; Li et al., 2012; Cheng et that OM and EC decreased by −70 % and −80 % from 2003
al., 2015; Chow et al., 2022). A steadily decreasing trend to 2015, respectively. This adds more measurement-based
was observed in the concentration levels of typical vehicular evidence for the overall decrease in PM2.5 burden from lo-
emission tracers: EC and hopanes. Over the 10-year period, cal vehicular emissions.
annual average EC and hopanes decreased by 60 % and 75 % The concentration levels of shipping emission tracers (V
in mass concentration and at a rate of −0.17 µg C m−3 yr−1 and Ni) were reduced by 34 % and 36 % by mass concentra-
and −0.052 ng m−3 yr−1 , respectively. These significant re- tion and at a rate of −0.60 and −0.29 ng m−3 yr−1 (Table 2),

https://doi.org/10.5194/acp-22-11557-2022 Atmos. Chem. Phys., 22, 11557–11577, 2022

11570 W. S. Chow et al.: Ten-year trend of PM2.5 components in Hong Kong

Figure 7. The 10-year variation of monthly SO2 and sulfate concentration. The significant SO2 emission control measures implemented in
Hong Kong are indicated in the top plot. The unit of concentration is micrograms per cubic meter (µg m−3 ).

respectively. The percent reduction of V and Ni is less than 2 %–13 %, and dust 4 %–8 % of PM2.5 (Chow et al., 2022).
that of vehicular emission tracers because their decreasing The marker chemicals for these sources are among the chem-
trends were not obvious until 2015 (Sect. S5 in the Supple- ical composition data monitored, allowing us to track the
ment) when shipping emission control policy was first in- long-term trend of these sources.
troduced in Hong Kong to reduce SO2 emission (HKEPD, Levoglucosan, an abundant primary product formed dur-
2021a). ing pyrolysis of cellulose, is a highly specific tracer of
biomass burning emissions (Simoneit et al., 1999). K+ is also
abundantly emitted from biomass burning, especially crop
3.6 Species significantly influenced by regional
residue burning. In studies without levoglucosan data, K+ is
emission sources – biomass burning, industrial/coal
frequently used as a biomass burning tracer. However, K+ is
combustion, and dust sources
a less specific tracer, due to contributions from other sources
Biomass burning, industrial/coal combustion, and dust are such as coal combustion, dust, and sea salt (e.g., Yu et al.,
well-recognized regional sources that influence PM pollu- 2018; Chow et al., 2022). The clustering analysis results
tion in Hong Kong (e.g., Zhang et al., 2018; Chow et al., show K+ and levoglucosan were moderately correlated and
2022). Our source apportionment study of PM2.5 at six sites fell into two different clusters in the same group of regional
in Hong Kong in 2015 shows that combined industrial and origin (Fig. 5). Comparing the 10-year variations of these
coal combustion accounted for 12 %–20 %, biomass burning two tracers, we found that their reduction extents differed

Atmos. Chem. Phys., 22, 11557–11577, 2022 https://doi.org/10.5194/acp-22-11557-2022

W. S. Chow et al.: Ten-year trend of PM2.5 components in Hong Kong 11571

Figure 8. The 10-year variation of monthly EC concentration. The significant vehicular emission control measures implemented in Hong
Kong are indicated in the plot.

significantly, with K+ at −60 % and levoglucosan at −47 % term change in atmospheric oxidation capacity, which would
over the decade. When examined seasonally (Fig. 6 and Ta- exert its impact through atmospheric degradation kinetics.
ble 3), they showed more distinct differences. Specifically, The three metal species, Pb, Zn and Cu, have been consis-
wintertime K+ showed a definitive decline trend at a rate of tently detected in the PM2.5 samples over the decade, provid-
−7.5 % yr−1 (p < 0.01), while the decline of wintertime lev- ing opportunities to probe their associated sources. The three
oglucosan (−3.9 % yr−1 ) could not be discerned from zero display a strong seasonal contrast, with wintertime concen-
according to the statistical test at p < 0.05 (Table 3). The tration levels more than twice those in the summer for Cu
lack of a clear declining trend of wintertime levoglucosan and Zn and 5 times for Pb. The strong seasonality is a char-
could be visually verified in Fig. 6. Further, both summer- acteristic indication for their regional/super-regional origin,
time K+ and levoglucosan did not show a clear decreasing consistent with the cluster analysis results (Fig. 5).
trend either (Fig. 6 or Table 3). The inconsistency between Cu and Zn are associated with metal processing indus-
K+ and levoglucosan could be explained if one considers tries. Over the decade, the Zn level in the winter has been
that they track different types of biomass burning. Specifi- dropping steadily, at a rate of −6.6 % yr−1 , while the winter-
cally, K+ is a better marker for emissions from burning crop time Cu dropped at a rate of −5.8 % yr−1 . On the other hand,
residues, which are typically enriched in K+ , while levoglu- their summertime change rates were indiscernible from zero
cosan, a thermal pyrolysis product of cellulose, is commonly (Fig. 6 and Table 3). Cumulatively, from 2008 to 2017, ap-
found in burning of all types of vegetative biomass including proximately 40 % reduction was realized for these two metals
hill fires. The inconsistent trends between winter and summer (Table 2). The significant reductions were likely indicators of
could also be rationalized considering their different source benefits from industrial upgrading following the promulga-
regions, i.e., the PRD region and northern China during the tion and implementation of the Guangdong “double transfer”
winter vs. South Asia in the summertime. Overall, the chem- policy (industry and labor transfer away from primary indus-
ical tracer data indicate crop residue burning has been re- tries) since 2009 (Zhong et al., 2013; Yang and Zhu, 2017).
duced over the decade, perhaps indicating some success in Pb is likely dominated by coal combustion. This source
measures such as prohibiting crop burning and crop straw deduction is derived from data collected from a different
utilization recently implemented in China (Ren et al., 2019). project, in which we deployed an online XRF spectrometer
The lack of a consistent declining trend in levoglucosan, on to monitor hourly concentrations of As, Se, and Pb in Hong
the other hand, implies that biomass burning remains largely Kong from August 2019 to February 2021. The data show
uncontrolled and will continue to be a significant PM pol- strong correlations of Pb with As and Se (R > 0.80) (Fig. S7
lution source. We also acknowledge that the 10-year trend in in the Supplement), two well-known tracers for coal combus-
an organic tracer like levoglucosan could be affected by long- tion (Tian et al., 2010), providing compelling supporting evi-
dence for coal combustion as a dominant source for Pb. Over

https://doi.org/10.5194/acp-22-11557-2022 Atmos. Chem. Phys., 22, 11557–11577, 2022

11572 W. S. Chow et al.: Ten-year trend of PM2.5 components in Hong Kong

the decade, wintertime Pb has displayed a continuous drop- wind speed at TW under each level of ENSO events were
ping trend at a rate of −7.7 % yr−1 , implying effectiveness in compared with those on the normal days (i.e., neutral event)
reducing coal combustion emissions in the PRD region and and are summarized in Figs. 9 and S14 in the Supplement.
in northern China. It is also worth noting that in the last 3 The rainfall during all El Niño events was close to that dur-
years (2015–2017) of the study decade, the reduction of the ing neutral conditions in Hong Kong, but there was a notable
three metals stalled, suggesting that more stringent actions reduction of rainfall during strong La Niña events. In terms
are needed for further reduction in the upcoming years. of wind conditions, more westerly air masses (wind direc-
Al and Si are classical marker elements for dust particles. tion > 180◦ ) were transported during very strong El Niño
They have also decreased over the decade, by −37 % for Al and moderate to strong La Niña events, while the elevated
and −46 % for Si. The two elements are highly correlated wind speed generally occurred during weak to moderate El
(R: 0.54 in 2015 and 0.87–0.97 in the other years), reflecting Niño events.
their common material sources and spatial origins. They dis- The changes in meteorological conditions were hard to
play a distinct seasonality common to the regional sources; be visualized and quantified. To better investigate the ef-
i.e., wintertime abundance is notably higher than the sum- fect of ENSO in Hong Kong, a multiple linear regression
mertime. The decreasing rates for wintertime concentrations (MLR) was established between the observed concentrations
(−5.3 % yr−1 for Al and −7.0 % yr−1 for Si) are more sig- and a list of meteorological variables including tempera-
nificant than the summertime in terms of both mass concen- ture (Temp), RH, seasonal components, and ENSO events,
tration and percentage change (Table 3). The decline became as shown in Eq. (4). For simplicity, the definitions of sea-
flat in 2016–2017 (Sect. S5), indicating that the current poli- sons here are based on calendar months, with spring corre-
cies started to be less sufficient in reducing the dust contribu- sponding to March–April, summer to May–August, fall to
tion. September–October, and winter to November–February.

Xm = β1 Yearm + δ1 Seasonm + β2 Tempm + β3 RH

3.7 El Niño–Southern Oscillation events
+ δ2 ENSO, (4)
A closer examination reveals that in the overall monotonic
trend component (blue dotted line in Fig. 4) of most PM where Xm is the monthly averaged time series, β’s are the
species (PM2.5 , SO4 2− , NO3 − , etc.), 2011 is an anomaly coefficients of parametric variables (i.e., Year, Temperature,
year showing higher concentrations than the preceding and RH), and δ’s are the coefficients of two dummy variables
the succeeding years. This resemblance in patterns across the (i.e., season and ENSO event).
various PM2.5 components implies that a macro-factor, for This MLR equation explains additional variance by
example, sporadic meteorological El Niño/La Niña events, the ENSO variables (0.63 %–11.7 %) without any multi-
might be at play in influencing the temporal variation. collinearity issue (i.e., generalized variance inflation factor
El Niño–Southern Oscillation (ENSO) events randomly < 5). The prediction from this model is reasonable, produc-
occur during the irregular changes of oceanic temperature ing a slope of ∼ 0.8 and R values of 0.58–0.86. Briefly, the
among the tropical Pacific Ocean, with El Niño events as- coefficients for the year (β1 ’s) capture the decline of species
sociated with increase in ocean temperature and La Niña and approximately match the GLS-ARMA results. Seasonal
events associated with decrease in ocean temperature. Dur- variations were successfully reflected by the coefficients for
ing the events, atmospheric pressure above the Pacific Ocean season (δ1 ’s), temperature (β2 ), or RH (β3 ) (Table S7 in the
changes and thus causes the shift of the Walker circulation Supplement). For example, positive summer coefficients in
as well as the distortion of pollutant airflow towards Hong V/Ni indicate higher in summer, negative spring coefficients
Kong (Yim et al., 2019). The El Niño effect typically leads in levoglucosan indicate higher in winter, and negative tem-
to a rise in rainfall, less northerly/northeasterly winds, and perature coefficients in NO3 − /hopanes indicate stronger gas-
higher wind speed in Hong Kong (Wang et al., 2019; Yim et particle partition or degradation (Table S7).
al., 2019), thus enhancing the dispersion of regional pollu- The random ENSO events impose different impacts on
tants. The La Niña effect is associated with opposite changes gaseous and particle pollutants. Significant enhancement
in rainfall and wind, thus impeding the dispersion of air pol- of SO2 and NOx (i.e., p value of δ2, La Niña < 0.05) was
lutants. Over the decade, there were two El Niño and three found to be owing to the La Niña effect, while no changes
La Niña events that lasted for at least 2 months. The strength were observed for O3 (Fig. S15 in the Supplement). Con-
of ENSO can be classified by the Niño 3.4 index based on the versely, significant and positive coefficients of the strong La
averaged sea surface temperature (SST) anomalies in the Pa- Niña effect were obtained for all PM2.5 pollutants except
cific Ocean region. This classification scheme results in five V and Ni. The coefficient was particularly high and posi-
broad groups (Table S6 in the Supplement), that is, neutral tive for some (SO4 2− , NO3 − , NH4 + , OC, and levoglucosan,
(0–0.49), weak (0.5–0.99), moderate (1–1.49), strong (1.5– δ2, strong La Niña from 1.6 to 61) but less for some regional
1.99), and very strong (≥ 2) (the numbers in the parentheses source species (Pb, Cu, δ2, strong La Niña from 0.0089 to 0.029)
indicate the SST anomaly). The rainfall, wind direction, and (Fig. S15). In other words, the concentration of these species

Atmos. Chem. Phys., 22, 11557–11577, 2022 https://doi.org/10.5194/acp-22-11557-2022

W. S. Chow et al.: Ten-year trend of PM2.5 components in Hong Kong 11573

Figure 9. The temporal variation in strength of ENSO events in the 2008–2017 period (a) and the changes of rainfall (b) under different
strength levels of ENSO events.

was typically high under strong La Niña events in compari- All PM2.5 components were found reduced, with the
son with neutral days. This could be explained by the signif- overall PM2.5 mass dropping at −1.5 µg m−3 yr−1 and by
icantly suppressed rainfall during the strong La Niña event a cumulative rate of 40 % (from 37.5 to 22.4 µg m−3 ).
(Fig. 9), where the highly water-soluble ions and levoglu- The individual contributors to the PM2.5 reduction are
cosan were removed to a lesser extent via wet deposition and sulfate (−0.36 µg m−3 yr−1 ), OM (−0.25 µg m−3 yr−1 ), ni-
thus maintained higher concentrations than the normal days. trate and EC (each at −0.17 µg m−3 yr−1 ), ammonium
Regardless of the significant level of coefficients, the El Niño (−0.12 µg m−3 yr−1 ), and others (−0.39 µg m−3 yr−1 ). A dis-
effects are generally opposite to the La Niña effect, imply- proportional reduction was noted between the precursor
ing that the enhancement of pollution dispersion/deposition gases SO2 (−62 %) and NOx (−36 %) and their secondary
could happen during El Niño events (Table S7). products SO4 2− (−40 %) and NO3 − (−66 %) because of
the complexity in their formation chemistry and formation
4 Conclusions process spatial scale not confined locally to Hong Kong. A
steadily declining trend in EC and hopanes was recorded,
In this study, we analyzed the 10-year (2008–2017) time se- achieving a cumulative decrease of 60 % and 75 %, respec-
ries of PM2.5 , its major components, and selected source tively, in their ambient concentrations. These reductions ver-
markers in an urban site in Hong Kong by the STL–GLS- ify the effectiveness of a series of control measures to re-
ARMA method. The dataset was obtained by following a reg- duce vehicular emissions by the Hong Kong government. In
ular 1-in-6 d sampling schedule that ensures temporal repre- comparison, the reduction of OC was more modest, at 23 %,
sentativeness and adheres to well-established chemical spe- which reflects the many more contributing sources as well as
ciation analysis protocols adopted by the USEPA. In addi- important secondary formation contribution to OC.
tion, organic molecule marker compounds (i.e., levoglucosan Two biomass burning tracers, K+ and levoglucosan, dis-
and hopanes) were also measured for this 10-year sample set. played strong seasonality in both ambient abundance and 10-
Such a long time series of PM2.5 chemical composition data year variation trend, the PRD and northern China being the
derived using a consistent sampling and analysis methodol- source region in the wintertime and South Asia being the
ogy is rare in China and elsewhere in Asia, thus providing source region in the summertime. Wintertime K+ showed a
uniquely valuable data to support studies of control measure definitive decline trend at a rate of −7.5 % yr−1 and a cumu-
evaluation and formulation for the region and offering a use- lative −60 % reduction, while the decline of wintertime lev-
ful reference for other provinces in China in evaluating emis- oglucosan was hardly discernable from zero. In the summer-
sion control policies. time, neither K+ nor levoglucosan showed a clear decreasing
trend. The two tracers track different types of biomass burn-

https://doi.org/10.5194/acp-22-11557-2022 Atmos. Chem. Phys., 22, 11557–11577, 2022

11574 W. S. Chow et al.: Ten-year trend of PM2.5 components in Hong Kong

ing, with K+ more representative of crop residue burning and Disclaimer. The content of this paper does not necessarily reflect
levoglucosan tracking the burning of cellulose. Collectively, the views and policies of the Hong Kong Special Administrative
the biomass burning tracers indicate that crop straw burning Region of the People’s Republic of China Government, nor does
has been reduced over the decade but that biomass burning mention of trade names or commercial products constitute an
remains a largely uncontrolled regional/super-regional PM2.5 endorsement or recommendation of their use.
source for Hong Kong.
Publisher’s note: Copernicus Publications remains neutral with
The 10-year data of Zn, Cu, and Pb showed a cumula- regard to jurisdictional claims in published maps and institutional
tive reduction of −40 %, −43 %, and −67 %, respectively. affiliations.
All three metals had strong seasonality, with winter concen-
trations much higher than the summertime, the metal pro-
cessing industries/coal combustion from the GBA region and Acknowledgements. We thank the Hong Kong Environmental
northern China being the source regions. Their significant re- Protection Department (HKEPD) for making part of the PM2.5
ductions in wintertime (−6.6 %, −5.8 %, and −7.7 % yr−1 compositional data available for this work.
for Zn, Cu, and Pb, respectively) suggested benefits from
measures such as industrial upgrading and coal combustion
emission reduction that were implemented over the decade. Financial support. This research has been supported by the En-
The reduction for all three metals in the last 3 years (2015– vironment and Conservation Fund (grant no. ECF99/2017).
2017) had stalled, signaling new measures are needed for
their further reduction. Dust in Hong Kong’s PM2.5 mainly
comes from a regional contribution. The dust components in Review statement. This paper was edited by Willy Maenhaut and
PM2.5 decreased, by −37 % for Al and −46 % for Si, over reviewed by two anonymous referees.
the decade, indicating success in controlling dust generation
activities in the region.
Finally, the long time series reveals that 2011 is an
anomaly year in that most PM2.5 components were elevated
References
above the adjacent years. By establishing a multiple linear re-
gression model, we show that the heightened strong La Niña
Anttila, P. and Tuovinen, J.: Trends of primary and secondary pol-
events in 2011 resulted in unusually low rainfall, which in
lutant concentrations in Finland in 1994–2007, Atmos. Envi-
turn reduced the removal via wet deposition of aerosol con- ron., 44, 30–41, https://doi.org/10.1016/j.atmosenv.2009.09.041,
stituents. In conclusion, the long-term chemical speciation 2010.
data of PM2.5 starting as early as 2008 in Hong Kong, one of Bigi, A. and Ghermandi, G.: Long-term trend and variability
the important cities in the Great Bay Area, could be useful of atmospheric PM10 concentration in the Po Valley, Atmos.
for a multitude of purposes related to understanding decadal- Chem. Phys., 14, 4895–4907, https://doi.org/10.5194/acp-14-
scale atmospheric composition change and evaluating signif- 4895-2014, 2014.
icant control policies for the region and the nation. Chen, W., Chen, Y., Huang, Y., Lu, X., Yu, J. Z., Fung, J. C. H.,
Lin, C., Yan, Y., Peng, L., Louie, P. K. K., Tam, F. C. V., Yue,
D., Lau, A. K. H., and Zhong, L.: Source apportionment of
Data availability. Measurement data used in this study fine secondary inorganic aerosol over the Pearl River Delta re-
are available in the data repository maintained by HKUST gion using a hybrid method, Atmos. Pollut. Res., 12, 101061,
(https://doi.org/10.14711/dataset/EHHRBZ, Yu et al., 2022). https://doi.org/10.1016/j.apr.2021.101061, 2021.
Chen, Z., Chen, D., Wen, W., Zhuang, Y., Kwan, M.-P., Chen, B.,
Zhao, B., Yang, L., Gao, B., Li, R., and Xu, B.: Evaluating the
“2+26” regional strategy for air quality improvement during two
Supplement. The supplement related to this article is available
air pollution alerts in Beijing: variations in PM2.5 concentrations,
online at: https://doi.org/10.5194/acp-22-11557-2022-supplement.
source apportionment, and the relative contribution of local emis-
sion and regional transport, Atmos. Chem. Phys., 19, 6879–6891,
https://doi.org/10.5194/acp-19-6879-2019, 2019.
Author contributions. WSC and JZY formulated the overall de- Cheng, Y., Lee, S., Gu, Z., Ho, K., Zhang, Y., Huang, Y.,
sign of the study. WSC, KFL, and XHHH carried out the chemi- Chow, J. C., Watson, J. G., Cao, J., and Zhang, R.: PM2.5
cal analyses for tracers and key major components and data valida- and PM10−2.5 chemical composition and source apportion-
tion. WSC analyzed the data with contributions from KL, AKHL, ment near a Hong Kong roadway, Particuology, 18, 96–104,
and JZY. WSC and JZY prepared the manuscript with contributions https://doi.org/10.1016/j.partic.2013.10.003, 2015.
from all co-authors. Cheung, H., Wang, T., Baumann, K., and Guo, H.: In-
fluence of regional pollution outflow on the concentra-
tions of fine particulate matter and visibility in the coastal
Competing interests. The contact author has declared that none area of southern China, Atmos. Environ., 39, 6463–6474,
of the authors has any competing interests. https://doi.org/10.1016/j.atmosenv.2005.07.033, 2005.

Atmos. Chem. Phys., 22, 11557–11577, 2022 https://doi.org/10.5194/acp-22-11557-2022

W. S. Chow et al.: Ten-year trend of PM2.5 components in Hong Kong 11575

Chin, P. C.: Climate and weather, in: A Geography of Hong Kong, HKEPD (Hong Kong Environmental Protection Department):
edited by: Chiu, T. N. and So, C. L., Oxford University Press, A concise guide to the air pollution control ordinance,
New York, 69–85, ISBN 0195840631, 1986. https://www.epd.gov.hk/epd/english/environmentinhk/air/
Chow, J. C. and Watson, J. G.: Guideline on Speciated Particu- guide_ref/guide_apco.html#introduction (last access: 26 June
late Monitoring, prepared for the United States Environmental 2022), 2021a.
Protection Agency, http://www3.epa.gov/ttnamti1/files/ambient/ HKEPD (Hong Kong Environmental Protection Department):
pm25/spec/drispec.pdf (last access: 30 January 2022), 1998. Data & Statistics – Hong Kong Air Pollutant Emission In-
Chow, J. C., Watson J. G., Chen, L. W., Chang, M. C., Robinson, ventory, https://www.epd.gov.hk/epd/english/environmentinhk/
N. F., Trimble, D., and Kohl S.: The IMPROVE_A temperature air/data/emission_inve.html (last access: 22 June 2022), 2021b.
protocol for thermal/optical carbon analysis: maintaining con- Ho, S. S. H. and Yu, J. Z.: In-injection port thermal desorption
sistency with a long-term database, J. Air Waste Manage., 57, and subsequent gas chromatography–mass spectrometric anal-
1014–1023, https://doi.org/10.3155/1047-3289.57.9.1014, 2007. ysis of polycyclic aromatic hydrocarbons and n-alkanes in at-
Chow, J. C, Watson, J. G., Cropper, P. M., Wang, X. mospheric aerosol samples, J. Chromatogr. A, 1059, 121–129,
L., and Kohl, S. D.: Measurements and validation for https://doi.org/10.1016/j.chroma.2004.10.013, 2004.
the twelve-months particulate matter study in Hong Kong, Ho, S. S. H., Yu, J. Z., Chow, J. C., Zielinska, B., Watson, J. G.,
2015, https://www.epd.gov.hk/epd/english/environmentinhk/air/ Sit, E. H. L., and Schauer, J. J.: Evaluation of an in-injection
studyrpts/pm25_study.html (last access: 3 February 2022), 2016. port thermal desorption-gas chromatography/mass spectrome-
Chow, W. S., Huang, X. H. H., Leung, K. F., Huang, L., try method for analysis of non-polar organic compounds in
Wu, X., and Yu, J. Z.: Molecular and elemental marker- ambient aerosol samples, J. Chromatogr. A, 1200, 217–227,
based source apportionment of fine particulate matter at six https://doi.org/10.1016/j.chroma.2008.05.056, 2008.
sites in Hong Kong, China, Sci. Total Environ., 813, 152652, Huang, X. H. H., Bian, Q., Ng, W. M., Louie, P. K. K.,
https://doi.org/10.1016/j.scitotenv.2021.152652, 2022. and Yu, J. Z.: Characterization of PM2.5 major components
Cleveland, R. B., Cleveland, W. S., and Terpenning, I.: STL: A and source investigation in suburban Hong Kong: A one
seasonal-trend decomposition procedure based on loess, J. Off. year monitoring study, Aerosol Air Qual. Res., 14, 237–250,
Stat., 6, 3, 1990. https://doi.org/10.4209/aaqr.2013.01.0020, 2014.
Dao, X., Lin, Y. C., Cao, F., Di, S. Y., Hong, Y., Xing, G., Li, J., Fu, Kim, Y. J., Kim, K. W., Kim, S. D., Lee, B. K., and Han,
P., and Zhang, Y. L.: Introduction to the national aerosol chemical J. S.: Fine particulate matter characteristics and its im-
composition monitoring network of China: Objectives, current pact on visibility impairment at two urban sites in Ko-
status, and outlook, B. Am. Meteorol. Soc., 10, ES337–ES351, rea: Seoul and Incheon, Atmos. Environ., 40, 593–605,
https://doi.org/10.1175/BAMS-D-18-0325.1, 2019. https://doi.org/10.1016/j.atmosenv.2005.11.076, 2006.
Fu, X., Wang, X. M., Guo, H., Cheung, K., Ding, X., Zhao, X., Ko, F. W. S., Tam, W., Wong, T. W., Lai, C. K. W., Wong,
He, Q., Gao, B., Zhang, Z., Liu, T., and Zhang, Y.: Trends of G. W. K., Leung, T., Ng, S. S. S., and Hui, D. S. C.: Ef-
ambient fine particles and major chemical components in the fects of air pollution on asthma hospitalization rates in different
Pearl River Delta region: observation at a regional background age groups in Hong Kong, Clin. Exp. Allergy, 37, 1312–1319,
site in fall and winter, Sci. Total Environ., 497–498, 274–281, https://doi.org/10.1111/j.1365-2222.2007.02791.x, 2007.
https://doi.org/10.1016/j.scitotenv.2014.08.008, 2014. Kuang, B. Y., Lin, P., Huang, X. H. H., and Yu, J. Z.: Sources
Griffith, S. M., Huang, X. H. H., Louie, P. K. K., and Yu, J. Z.: Char- of humic-like substances in the Pearl River Delta, China: pos-
acterizing the thermodynamic and chemical composition factors itive matrix factorization analysis of PM2.5 major components
controlling PM2.5 nitrate: Insights gained from two years of on- and source markers, Atmos. Chem. Phys., 15, 1995–2008,
line measurements in Hong Kong, Atmos. Environ., 122, 864– https://doi.org/10.5194/acp-15-1995-2015, 2015.
875, https://doi.org/10.1016/j.atmosenv.2015.02.009, 2015. Lang, J., Zhang, Y., Zhou, Y., Cheng, S., Chen, D., Guo, X., Chen,
Guo, H., Ding, A. J., So, K. L., Ayoko, G., Li, Y. S., S., Li, X., Xing, X., and Wang, H.: Trends of PM2.5 and chemical
and Hung, W. T.: Receptor modeling of source apportion- composition in Beijing, 2000–2015, Aerosol Air Qual. Res., 17,
ment of Hong Kong aerosols and the implication of urban 412–425, https://doi.org/10.4209/aaqr.2016.07.0307, 2017.
and regional contribution, Atmos. Environ., 43, 1159–1169, Li, M., Liu, H., Geng, G., Hong, C., Liu, F., Song, Y., Tong, D.,
https://doi.org/10.1016/j.atmosenv.2008.04.046, 2009. Zheng, B., Cui, H., Man, H., Zhang, Q., and He, K.: Anthro-
HKEPD (Hong Kong Environmental Protection Depart- pogenic emission inventories in China: a review, Natl. Sci. Rev.,
ment): 2017 Hong Kong emission inventory report, 4, 834–866, https://doi.org/10.1093/nsr/nwx150, 2017.
https://www.epd.gov.hk/epd/sites/default/files/epd/data/2017_ Li, Y. J., Yeung, J. W. T., Leung, T. P. I., Lau, A. P. S., and
Emission_Inventory_Report_Eng.pdf (last access: 9 January Chan, C. K.: Characterization of organic particles from in-
2022), 2019. cense burning using an aerodyne high-resolution time-of-flight
HKEPD (Hong Kong Environmental Protection, Department): aerosol mass spectrometer, Aerosol Sci. Tech., 46, 654–665,
Guangdong-Hong Kong-Macao Pearl River Delta regional air https://doi.org/10.1080/02786826.2011.653017, 2012.
quality monitoring network – A report of monitoring results Lippmann, M. and Chen, L.: Health effects of concentrated ambient
in 2019, Environmental Protection Department the Govern- air particulate matter (CAPs) and its components, Crit. Rev. Tox-
ment of the Hong Kong Special Administrative Region, 1–38, icol., 39, 865–913, https://doi.org/10.3109/10408440903300080,
https://www.epd.gov.hk/epd/sites/default/files/epd/english/ 2009.
resources_pub/publications/files/PRD_2019_report_en.pdf (last Lu, Q., Zheng, J., Ye, S., Shen, X., Yuan, Z., and Yin, S.: Emission
access: 21 July 2021), 2020. trends and source characteristics of SO2 , NOx , PM10 and VOCs

https://doi.org/10.5194/acp-22-11557-2022 Atmos. Chem. Phys., 22, 11557–11577, 2022

11576 W. S. Chow et al.: Ten-year trend of PM2.5 components in Hong Kong

in the Pearl River Delta region from 2000 to 2009, Atmos. Envi- tion, https://apps.who.int/iris/handle/10665/345329 (last access:
ron., 76, 11–20, https://doi.org/10.1016/j.atmosenv.2012.10.062, 30 January 2022), 2021.
2013. Xue, J., Yuan, Z., Lau, A. K. H., and Yu, J. Z.: Insights into
Molugaram, K. and Rao, G. S.: Chap. 12 – Analysis of time series, factors affecting nitrate in PM2.5 in a polluted high NOx
in: Statistical Techniques for Transportation Engineering, edited environment through hourly observations and size distribu-
by: Molugaram, K. and Rao, G. S., Butterworth-Heinemann, tion measurements, J. Geophys. Res.-Atmos., 119, 4888–4902,
463–489, https://doi.org/10.1016/B978-0-12-811555-8.00012- https://doi.org/10.1002/2013JD021108, 2014a.
X, 2017. Xue, J., Yuan, Z., Yu, J. Z., and Lau, A. K. H.: An observation-based
Ren, J., Yu, P., and Xu, X: Straw utilization in China model for secondary inorganic aerosols, Aerosol Air Qual. Res.,
– status and recommendations, Sustainability, 11, 1762, 14, 862–878, https://doi.org/10.4209/aaqr.2013.06.0188, 2014b.
https://doi.org/10.3390/su11061762, 2019. Xue, J., Yu, X., Yuan, Z., Griffith, S. M., Lau, A. K. H., Seinfeld,
Shumway, R. H. and Stoffer, D. S.: Time series analysis and J. H., and Yu, J. Z.: Efficient control of atmospheric sulfate pro-
its applications: With R examples, Springer, New York, ISBN duction based on three formation regimes, Nat. Geosci., 12, 977–
9781441978646, 2017. 982, https://doi.org/10.1038/s41561-019-0485-5, 2019.
Simoneit, B. R., Schauer, J. J., Nolte, C. G., Oros, D. R., Elias, V. Yang, J. and Zhu, S.: Effect of industrial transfer policy on the evo-
O., Fraser, M. P., Rogge, W. F., and Cass, G. R.: Levoglucosan, a lution of regional product structure: Based on the “Double Trans-
tracer for cellulose in biomass burning and atmospheric particles, fer” policy in Guangdong Province, Trop. Geogr., 37, 452–461,
Atmos. Environ., 33, 173–182, https://doi.org/10.1016/S1352- https://doi.org/10.13284/j.cnki.rddl.002971, 2017.
2310(98)00145-9, 1999. Yang, Y., Tang, R., Qiu, H., Lai, P., Wong, P., Thach, T.,
Singh, V., Singh, S., and Biswal, A.: Exceedances Allen, R., Brauer, M., Tian, L., and Barratt, B.: Long
and trends of particulate matter (PM2.5 ) in five In- term exposure to air pollution and mortality in an el-
dian megacities, Sci. Total Environ., 750, 141461, derly cohort in Hong Kong, Environ. Int., 117, 99–106,
https://doi.org/10.1016/j.scitotenv.2020.141461, 2021. https://doi.org/10.1016/j.envint.2018.04.034, 2018.
Tan, J., Duan, J., Ma, Y., He, K., Cheng, Y., Deng, S. X., Huang, Yim, S. H. L., Hou, X., Guo, J., and Yang, Y.: Con-
Y. L., and Si-Tu, S. P.: Long-term trends of chemical char- tribution of local emissions and transboundary air pol-
acteristics and sources of fine particle in Foshan City, Pearl lution to air quality in Hong Kong during El niño-
River Delta: 2008–2014, Sci. Total Environ., 565, 519–528, Southern Oscillation and heatwaves, Atmos. Res., 218, 50–58,
https://doi.org/10.1016/j.scitotenv.2016.05.059, 2016. https://doi.org/10.1016/j.atmosres.2018.10.021, 2019.
Tian, H. Z., Wang, Y., Xue, Z. G., Cheng, K., Qu, Y. P., Chai, Yu, J., Yan, C., Liu, Y., Li, X., Zhou, T., and Zheng, M.: Potassium:
F. H., and Hao, J. M.: Trend and characteristics of atmo- A Tracer for Biomass Burning in Beijing?, Aerosol Air Qual.
spheric emissions of Hg, As, and Se from coal combustion Res., 18, 2447–2459, https://doi.org/10.4209/aaqr.2017.11.0536,
in China, 1980–2007, Atmos. Chem. Phys., 10, 11905–11919, 2018.
https://doi.org/10.5194/acp-10-11905-2010, 2010. Yu, J. Z. and Zhang, T.: Measurements and validation for
USEPA (United State Environmental Protection Agency): Qual- the twelve-months particulate matter study in Hong Kong,
ity assurance guidance document 2.12 – monitoring PM2.5 in 2016, https://www.epd.gov.hk/epd/english/environmentinhk/air/
ambient air using designated reference or class I equivalent studyrpts/pm25_study.html (last access: 5 February 2022), 2017.
methods, https://www3.epa.gov/ttnamti1/files/ambient/pm25/qa/ Yu, J. Z. and Zhang, T.: Measurements and validation for
m212.pdf (last access: 6 February 2022), 2016. the twelve-months particulate matter study in Hong Kong,
Wang, X., Ho, K., Chow, J. C., Kohl, S. D., Chan, C. S., Cui, L., 2017, https://www.epd.gov.hk/epd/english/environmentinhk/air/
Lee, S. F., Chen, L. A., Ho, S. S. H., Cheng, Y., and Watson, studyrpts/pm25_study.html (last access: 5 February 2022), 2018.
J. G.: Hong Kong vehicle emission changes from 2003 to 2015 Yu, J. Z., Tung, J. W. T., Wu, A. W. M., Lau, A. K.
in the Shing Mun tunnel, Aerosol Sci. Tech., 52, 1085–1098, H., Louie, P. K. K., and Fung, J. C. H.: Abundance
https://doi.org/10.1080/02786826.2018.1456650, 2018. and seasonal characteristics of elemental and organic car-
Wang, X., Zhong, S., Bian, X., and Yu, L.: Impact of bon in Hong Kong PM10 , Atmos. Environ., 38, 1511–1521,
2015–2016 El niño and 2017–2018 la niña on PM2.5 https://doi.org/10.1016/j.atmosenv.2003.11.035, 2004.
concentrations across China, Atmos. Environ., 208, 61–73, Yu, J. Z., Huang, X. H. H., and Ng, W. M.: Measure-
https://doi.org/10.1016/j.atmosenv.2019.03.035, 2019. ments and validation for the twelve-months particulate matter
Watson, J. G., Chow, J. C., and Frazier, C. A.: X-ray fluorescence study in Hong Kong, 2011, https://www.epd.gov.hk/epd/english/
analysis of ambient air samples, in: Elemental Analysis of Air- environmentinhk/air/studyrpts/pm25_study.html (last access: 5
borne Particles, Vol. 1, edited by: Landsberger, S. and Creatch- February 2022), 2012.
man, M., Gordon and Breach Science, Amsterdam, the Nether- Yu, J. Z., Huang, X. H. H., and Ng, W. M.: Measure-
lands, 67–96, ISBN 9789056996277, 1999. ments and validation for the twelve-months particulate matter
Wilcox, R.: Chap. 10 – Robust regression, in: Introduction to Robust study in Hong Kong, 2012, https://www.epd.gov.hk/epd/english/
Estimation and Hypothesis Testing, 4th edn., edited by: Wilcox, environmentinhk/air/studyrpts/pm25_study.html (last access: 5
R., Academic Press, 517–583, https://doi.org/10.1016/B978-0- February 2022), 2013.
12-804733-0.00010-X, 2017. Yu, J. Z., Huang, X. H. H, Zhang, T., and Ng, W.
World Health Organization: WHO global air quality guidelines: M.: Measurements and validation for the twelve-
particulate matter (PM2.5 and PM10 ), ozone, nitrogen dioxide, months particulate matter study in Hong Kong, 2013,
sulfur dioxide and carbon monoxide, World Health Organiza-

Atmos. Chem. Phys., 22, 11557–11577, 2022 https://doi.org/10.5194/acp-22-11557-2022

W. S. Chow et al.: Ten-year trend of PM2.5 components in Hong Kong 11577

https://www.epd.gov.hk/epd/english/environmentinhk/air (last Zhao, M., Zhang, Y., Ma, W., Fu, Q., Yang, X., Li, C., Zhou, B., Yu,
access: 5 February 2022), 2014. Q., and Chen, L.: Characteristics and ship traffic source identifi-
Yu, J. Z., Huang, X. H. H., Zhang, T., and Ng, W. M.: Measure- cation of air pollutants in China’s largest port, Atmos. Environ.,
ments and validation for the twelve-months particulate matter 64, 277–286, https://doi.org/10.1016/j.atmosenv.2012.10.007,
study in Hong Kong, 2014, https://www.epd.gov.hk/epd/english/ 2013.
environmentinhk/air/studyrpts/pm25_study.html (last access: 5 Zheng, B., Tong, D., Li, M., Liu, F., Hong, C., Geng, G., Li, H., Li,
February 2022), 2015. X., Peng, L., Qi, J., Yan, L., Zhang, Y., Zhao, H., Zheng, Y., He,
Yu, J. Z., Chow, W. S., and Huang, X. H. H.: Ten-year K., and Zhang, Q.: Trends in China’s anthropogenic emissions
(2008–2017) PM2.5 major components and select tracers at since 2010 as the consequence of clean air actions, Atmos. Chem.
Tsuen Wan, Hong Kong, DataSpace@HKUST [data set], Phys., 18, 14095–14111, https://doi.org/10.5194/acp-18-14095-
https://doi.org/10.14711/dataset/EHHRBZ, 2022. 2018, 2018.
Yuan, Z., Yadav, V., Turner, J. R., Louie, P. K. K., and Lau, A. Zhong, L. J., Louie, P. K. K., Zheng, J. Y., Yuan, Z. B.,
K. H.: Long-term trends of ambient particulate matter emission Yue, D. L., Ho, J. W. K., and Lau, A. K. H.: Science–
source contributions and the accountability of control strategies policy interplay: air quality management in the Pearl River
in Hong Kong over 1998–2008, Atmos. Environ., 76, 21–31, Delta region and Hong Kong, Atmos. Environ., 76, 3–10,
https://doi.org/10.1016/j.atmosenv.2012.09.026, 2013. https://doi.org/10.1016/j.atmosenv.2013.03.012, 2013.
Zhang, X., Yuan, Z., Li, W., Lau, A. K. H., Yu, J. Z., Fung, J. C.
H., Zheng, J., and Yu, A. L. C.: Eighteen-year trends of local
and non-local impacts to ambient PM10 in Hong Kong based
on chemical speciation and source apportionment, Atmos. Res.,
214, 1–9, https://doi.org/10.1016/j.atmosres.2018.07.004, 2018.

https://doi.org/10.5194/acp-22-11557-2022 Atmos. Chem. Phys., 22, 11557–11577, 2022