A Long Term Study on Characterization and Source Apportionment of Particulate Pollution in Klang Valley , Kuala Lumpur

Samples of airborne particulate matter, PM2.5 and PM10–2.5 were collected using a Gent stacked filter sampler at an urban site, Klang Valley, Kuala Lumpur between January 2002–December 2011. The samples were analyzed for their elemental composition and black carbon content by Particle Induced X-ray Emission (PIXE) and light absorption, respectively. The annual average for PM2.5, PM10–2.5 and PM10 ranged from 21 to 35, 18 to 26 and 44 to 56 μg m, respectively. Factor analysis method and the Positive Matrix Factorisation (EPA PMF3) technique were also applied to the fine fraction data set in order to identify the possible sources of particulate matter and their contributions to the ambient particulate matter concentrations in the Klang Valley. A five factor PMF solution was found for PM2.5 particulate matter. The sources identified were; motor vehicles, industry, smoke/biomass burning, secondary sulphate and soil. It was found that the primary source of haze air particulate matter was locally generated mostly from vehicular emissions which contribute about 35% of the PM2.5 mass. The Hybrid Single Particle Lagrangian Intergrated Trajectory (HYSPLT) model was also used to explore possible long range transport of pollution. Smoke trans-boundary events were identified based on fine potassium from the data base in 2004, 2006 and 2008.


INTRODUCTION
Klang Valley (Fig. 1) is a rapidly growing urban area with the highest growth rate in Malaysia.The area comprises of Kuala Lumpur (lat 3°8′N; long 101°44′E), its suburbs, and adjoining cities and towns in the state of Selangor.The weather is hot and humid with uniform temperatures throughout the year from 25°C to 35°C and the humidity is almost the same throughout the year 70%-80% during the night-time and 50-60% during daytime.There are uniform periodic changes in the wind flow patterns namely, the southwest monsoon, northeast monsoon and inter-monsoon seasons.The southwest monsoon season is usually established in the latter half of May or early June and ends in September while the northeast monsoon season usually commences in early November and ends in March.During the two inter-monsoon seasons, the winds are generally light and variable.There are many sources that contribute to the fine and coarse particles in the area.Potential sources can originate from major highways that run throughout the Klang Valley, growth in population, unplanned and uncontrolled development of industrial premises that lead to higher emissions of organic and inorganic gases, chemicals and dust as well as noise pollution and vibration disturbance.However the sources of the pollutants and their contributions not only originate from local activities to the local factors such as open burning, construction and increasing industrialization programs, but also from the foreign activities such as forest fires and land clearing in Sumatra and Kalimantan.Emissions from these fires has caused trans-boundary haze pollution events that have affected the entire Southeast Asian region.The haze episodes in Malaysia were reported as early as the 1980s followed by a number of haze episodes that were less intense and did not receive as much public attention.The first serious haze event in the country was reported in August 1991 followed by events in 1994, 1997, 1998, 2002, 2004, 2005, 2006and 2009(Keywood, 2003, Tangang et al., 2010).These phenomenon have now become regular features in Malaysia during the dry seasons in Feb-April and June-September.
For many years the International Atomic Energy Agency (IAEA) has been supporting IAEA/RCA project on air pollution.The objective of the project was to demonstrate the applicability of nuclear and related analytical techniques (mainly NAA, XRF, PIXE and ICP-MS) in studies of pollution caused by APM (airborne particulate matter) which is now being recognized as a local, regional and global problem with a serious impact on human health, particularly for young and older people, on visibility, and on climate change.The Malaysian Nuclear Agency has been involved in the project since 1998 in which the Klang Valley region has been selected as the study area.The project focused on the measurement of fine particle (PM 2.5 ), coarse particle (PM 10-2.5 ), black carbon (BC), major and trace elements, as well as source quantification, and identification of longrange transport of fine airborne particulate matter.The study of fine particles for the samples collected during the period from 2000 to 2008 at Klang Valley identified five sources of aerosols in the Klang Valley (Rahman et al., 2009a, b).These sources included sea spray, motor vehicle, smoke, industry, soil and unknown source.This paper is reporting the results of a long term study of fine particles (PM 2.5 ) for the samples collected during the 10 year study from 2002 to 2011 at Klang Valley, Kuala Lumpur.The total fine mass, chemical composition, and contributions of different sources contributing towards the growing air pollution problem in the Klang Valley are reported.The evidence that smoke from biomass burning being a significant aerosol contributor during periods of excessive haze is also discussed.

Sampling
Air particulate samples were collected using a Gent stack sampler unit (Hopke et al., 1997) provided by the IAEA.The sampler was placed on a rooftop of a two story building at the Universiti Teknologi Malaysia, Kuala Lumpur campus (3°10′30′′N, 101°43′24′′E) approximately 10 meters in height.The campus lies between two highways to the north and south, and is surrounded by busy roads closely linked to the Kuala Lumpur city center.Samples were collected between January 2002 and December 2011.The sampler was programmed to run automatically at an air flow-rate of 15 L min -1 to collect two fractions (< 2.5 µm and 2.5-10 µm aerodynamic diameter particles) of 24 hours duration samples on 47mm diameter polycarbonate filters (with 0.4 µm and 8 µm pore sizes respectively) at a frequency of a week or fortnight.Nearly 400 filters were collected during the study period.

Mass and Black Carbon Measurement
The total mass of each sample was determined by weighing the filter using a microbalance (METTLER Model MT5).The balance was equipped with a Po-212 (alpha emitter) electrostatic charge eliminator (STATICMASTER) to eliminate the static charge accumulated on the filters before each weighing.Black carbon concentration on a filter was measured by a Smoke Stain Reflectometer (EEL model 43D).The method is based on the absorption of light, in which the amount of light absorption is proportional to the black carbon concentration on the filter.The black carbon concentration values were obtained according to the method described by other workers (Edwards et al., 1983;Cohen et al., 2000;Biswas et al., 2003).

Elemental Analysis
Elemental analysis was performed by Particle Induced Xray Emission (PIXE) (Trompetter et al., 2005) at the National Isotope Centre, GNS Science, New Zealand.X-ray spectra obtained from the PIXE measurements were analyzed with GUPIX software developed by Guelph University (Maxwell et al., 1989(Maxwell et al., , 1995)).Concentration of the following twenty elements were measured and analyzed for each sample: Al, As, Br, Ca, Cl, Cr, Cu, Fe, K, Mg, Mn, Na, Ni, P, Pb, S, Si, Ti, V and Zn.Calibration of the PIXE system was performed by irradiating suitable Micromatter thin target standards.
The receptor modeling using positive matrix factorization in the form of EPA PMF3.0 (US EPA, 2010) was then applied to the data to confirm the possible sources of air pollution that contribute to the area of Klang Valley.These EPA-PMF processes and their applications have been discussed in detail elsewhere (Reff et al., 2007;Norris et al., 2008).In order to see the contributing location of distant sources (smoke) the Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT4) back-trajectories (Draxler and Rolph, 2010) were used in this study to trace possible medium and long-range transport of pollution to the Klang Valley area.Archived REANALYSIS meteorological data were used as input.This model is accessible on-line at http://www.arl.noaa.gov/ready/hysplit4.html.The model is used to calculate the pathway of the air mass being sampled backward in time from the receptor site from various starting times throughout the sampling interval.

Particulate Matter Mass Level
Table 1 below summarises the annual average concentration of PM 2.5 and PM 10-2.5 as well as PM 10 for the study period of 2002-2011.A total of 382 sample pairs (fine and coarse) were collected from the site.The average concentration of PM 2.5 is 25 µg m -3 with a maximum of 96 µg m -3 and minimum of 2.4 µg m -3 .Estimation of PM 10 was made by addition of fine fraction (2.5 µm) to the coarse fraction (2.5-10 µm).There is no Malaysian guidelines or standard for the fine and coarse fractions, but the USEPA National Ambient Air Quality Standard (NAAQS) fine particle goals of 15 µg m -3 annual average with a maximum of 35 µg m -3 averaged over 24 hour period was adopted for reference.Fig. 2 shows the annual average concentrations of PM 2.5 and PM 10-2.5 measured from the period of study 2002 to 2011 at Kuala Lumpur, Klang Valley.It was clearly seen that during the ten years of study the annual average of PM 2.5 at Klang Valley had consistently exceeded the annual standard of USEPA National Ambient Air Quality Standards (NAAQS) of 15 µg m -3 .The PM 2.5 levels were also found to be slightly higher in 2002, 2005 and 2006.This could be due to the haze pollution reported in those years (Malaysian haze, 2005;Southeast Asian haze, 2006;Haze, 2013) The trend of the annual average levels of PM 10 concentration in the ambient air between 2002 and 2011 is shown in Fig. 3 with an average of 48 µg m -3 which is just below the Malaysian Ambient Air Quality Guidelines of 50 µg m -3 .The values for 2005 and 2006 were found to exceed the Malaysian Guidelines, which could be also due to the serious haze events reported in those years.The plot of PM 10 against PM 2.5 in Fig. 4 shows that there is a good correlation between them, indicating that the air quality problem related to air particulate matter in Klang Valley (regularly monitored through PM 10 ) was mostly dominated by the fine particle fraction.

Black Carbon and Elemental Concentration
Concentration of the following twenty elements (including      BC) were analyzed for each sample: Al, As, Br, Ca, Cl, Cr, Cu, Fe, K, Mg, Mn, Na, Ni, P, Pb, S, Si, Ti, V and Zn.Table 2 below summarizes the mass and elemental concentrations including max, median and standard deviation (SD) values of PM 2.5 as well as their percentage relative to the gravimetric mass obtained for Klang Valley in 2002-2011.Black carbon (BC) and sulfur were found to be the major elemental components of the PM 2.5 with 16% and 8.5% respectively, relative to the fine particle mass.In addition the concentration of the aerosol components was also estimated using pseudo-elements analysis as described by Malm et al. (1994) and Cohen (2000).These variables are combinations of the measured composition values that help to estimate the likely major source types.Soil with weight fraction of 8.4% was estimated by summing five elements Al, Ca, Fe, Si and Ti converted to their common oxides.The estimated sulfate (36%) was calculated from sulfur concentration assuming that sulfate was fully neutralized ammonium sulfate.Other components that represent aerosol components are sea salt and smoke with weight fractions of 5.6% and 1.4% respectively.The sea salt was calculated based on the measured sodium concentration assuming that all sodium in the aerosol is associated with sea salt as sodium chloride.Smoke was calculated from non-soil potassium.Fine potassium is an accepted indicator for smoke from biomass burning, hence to obtain a reliable smoke indicator from the measured potassium, potassium associated with soil is subtracted from the total fine potassium.Reconstructed mass (RCM) is the estimated of total fine mass calculated from the sum of all the composite variables discussed above including the black carbon (Malm et al. 1994).The average RCM calculated for the dataset was about 70% as relative to the gravimetric mass.This is considered quite good mass closure for the datasets.The missing 30% of mass could be organic matter (which was not measured in this study), nitrates and water vapour (Cohen et al., 1998).

Source Apportionment Using Positive Matrix Factorization (EPA-PMF3)
Identification and apportionment of pollutants to their sources is very important in air quality management.Positive Matrix Factorization (PMF) (Noris et al., 2008) is an advanced receptor model based on least-squares techniques that uses error estimates of the measured data to provide weights in the fitting process.The method has been developed by Paatero andTapper (1993, 1994) and has been successfully applied to a number of data sets in air pollution studies (Lee et al., 1999;Chueinta et al., 2000;Song et al., 2001;Begum et al., 2004).The method is based on solving the factor analysis problem by least squares approach using a data point weighting method which decomposes a matrix of data of dimension n rows and m columns into two matrixes, G(nxp) and F(pxm), where n is the number of samples while m is the number of species that can be written as: Briefly, a data matrix X of i by j dimensions, in which i number of samples and j chemical species can be written as: where p is the number of factors, f is species profile, g is amount of mass and e ij is the residual for each sample.The task of PMF is to minimize the object function (Q), based upon the uncertainties: where s ij are the uncertainties in x ij .The results are constrained so that all species profiles (matrix F) are non-negative and each sample has a non-negative source contribution (matrix G).Solution of Eq. ( 3) and the model are described in detail elsewhere Paatero (Paatero and Tapper, 1994;Paatero, 1997).
In this study, elemental data with m = 21 and n = 382 were used to perform the PMF analysis.In order to get the average source finger print and their contribution to the fine particle mass, 5 "bad" data were removed from the analysis.The elements As and Ni were categorized as "bad" species and excluded from the modeling as more than 50% of the data was below the minimum detection limit.The best solution was found to be the five factor solution for the elemental composition of the fine particulate matter in the Klang Valley, Kuala Lumpur area.The value of true Q calculated for p = 5 factors was 2980 which is about 86% of the expected value Q exp of 3393.The result obtained is shown in Fig. 5.
• The first factor is dominated by BC, Si, S, Na and minor quantities of metal species.These components were associated with emissions from motor vehicles which contribute about 35.3% of the fine mass.Emissions from diesel heavy trucks are the major sources of BC and S. S is part of chemical make up for diesel (as well as gasoline) fuels.Sulfuric acid is produced when sulfur combines with water vapor formed during the combustion process, and some of this corrosive compound is emitted into the atmosphere through the exhaust.Ebihara et al. (2008) reported that the city of Kuala Lumpur was among the Asia cities that were heavily affected by vehicle emissions, and a recent study on source apportionment (Rahman et al., 2011) in Kuala Lumpur, Klang Valley showed that motor vehicles including two stroke were the main source for fine particles of Klang Valley, Kuala Lumpur with contribution of > 30% of the mass concentration.
• The second factor is attributed to soil dust that contains characteristic elements of Al, Si, Mg, K and Ca (Watson and Chow, 2001).This source contributed about 3.1% to the fine mass.The presence of other elements such as Fe, V, Cr, Cu and Zn indicated the influence of road dust to the factor during transport.Fe for example is a major component of vehicles and tyre wear has been reported to contribute significantly to the Zn load in road dust (Thorpe et al., 2008).
• The third factor has high loadings of BC, S and K which represents a source from smoke/biomass burning (Alexander et al., 2001).The factor contributed about 9.3% of the fine mass.Fine K is an accepted key element for smoke from biomass burning which is released especially under high temperatures in fires mostly as potassium chloride and potassium sulfate (Khalil et al., 2003, Begum et al., 2004, Baxla et al., 2009).The activities of open burning have made smoke/biomass burning as one of the urban pollutants in the area.In addition the situation was further aggravated by trans-boundary pollution of fine particles from external biomass burning activities.• The fourth factor may be attributed to a mixture of industries as characterized by the high amounts of BC, Na, Al, S, K, Ca, Fe, Zn and Pb.The high loading of BC suggesting mixing of smoke from motor vehicles during transport.This source contributed about 47.8% to the fine mass.There are significant numbers of industries that are involved in metal works scattered around Klang Valley and the main industrial area is located about 15 to 20 km to the southern/southwestern side of the sampling site (Rahman et al., 2011).The industries include iron and steel industry, paint industry, aluminum fabrication, pharmaceutical, rubber glove production as well as food industries.
• The last factor contributed about 4.5% and has been identified as a secondary sulphate source due to the dominance of sulphur in the profile (Kim et al., 2004).
As mentioned earlier in the first factor emissions from diesel heavy trucks are the major sources of sulfur.The sampling site lies between two busy highways to the north and south.It is also surrounded by roads that are closely linked to the Kuala Lumpur city center and carrying a significant amount of traffic.Another source of sulphur is from shipping activities at the Port Klang which is located about 40 km of southwest Kuala Lumpur.The presence of potassium loading in the factor could be connected with the possible contribution of local wood smoke and possibly from other combustion related sources.Fig. 6 shows the annual trend source contribution in Klang Valley, Kuala Lumpur during 2002 to 2011.Approximately 35 % to 48 % of PM 2.5 emitted in Klang Valley, Kuala Lumpur within 2002-2011 came from motor vehicles and industry respectively.In 2002 and 2003 PM 2.5 mass was found to be dominated by motor vehicles with contributions of 49.5% and 50.0%respectively.After 2003 there was an increase of contributions from industry and has been the major source of PM 2.5 mass since 2004.Contribution of natural sources; soil and biomass burning ranged from 2.1 to 5.7% and 4.0 to 15.6% respectively, whilst contributions from secondary sulphate ranged from 2.8 to 7.3%

Trans-Boundary Pollution
Trans-boundary atmospheric pollution (the haze) in Malaysia has been an issue of increasing importance over the past few years.Haze occurs when there is sufficient smoke, dust, moisture and water vapor suspended in air to impair visibility.It is mainly caused by particulate matter  from various sources including smoke, road dust and particulate matter formed when gaseous pollutants react in the air (Haze, Begum et al., 2011).The composition of smoke depends on the nature of burning fuels as well as the conditions of combustion.Smoke trans-boundry events were identified based on fine potassium (K).Potassium is released especially under high temperatures in fires mostly as potassium chloride and potassium sulfate and has been often used as indicator of biomass burning (Khalil et al., 2003;Begum et al., 2004;Baxla et al., 2009).In order to obtain a reliable smoke indicator from the fine potassium it is necessary to subtract the fine potassium associated with soil (Cohen et al., 2010).Hence, smoke can be obtained by following equation: Wind patterns play important roles in the distribution and dispersion of pollutants in the atmosphere.During the southwest monsoon season which is usually established in the later half of May or early June and ends in September, hot, dry weather and winds blowing from the southwest help fires explode in Indonesia forests for several weeks.Pollutants emitted to the atmosphere through the forest fires can be transported hundreds and even thousands of kilometers and have the capacity to reach the Peninsular Malaysia.
In this study, smoke trans-boundary events (with evidence) were identified based on the pseudo-element K during the southwest monsoon between May-September from the data base 2004 and2008;and during transition period October, 2006.In order to exclude low local source contributions, only daily events with concentrations that are two standard deviations above the mean value for a measured species were considered .The statistics and the peak value concentration (ng m -3 ) of smoke for the selected dates during the study period are given in Table 3.Air parcel back trajectories (Draxler and Rolph, 2010) beginning at noon on these dates are presented in Figs.10-12. Eidence of all the events is identified and presented in the supporting materials, Figs.SM1-SM6.
There were two smoke events identified in 2004; Jun 21, 2004 and August 23, 2004 (Fig. 7) with concentrations of 646 ng m -3 and 721 ng m -3 respectively.Back trajectory plots at different heights: 300 m, 500 m and 1000 m above ground level (AGL) were used to calculate the four day and five day backward trajectories respectively (Fig. 10).Both events were seen to be associated with international transport of smoke from the area of Sumatra Island.The smoke signature (300 m and 500 m) traveled about 315 km (Riau) to 550 km (Jambi) in southwest direction from the source location to reach the receptor site in Kuala Lumpur.The two satellite images show the evidence of fires that were burning on the Indonesian island of Sumatra in June 2004 (June 17, 2004 andJune 18 2004).When the Aqua satellite passed over the region on those days, the Moderate Resolution Imaging Spectroradiometer (MODIS) captured this image along with numerous fire detections.Pixels in which fires were detected are marked in yellow and red (see supporting material Fig. SM1).The MODIS on NASA's Aqua satellite had also detected actively burning fires (marked in red) on August 22, 2004 on the Indonesian island of Sumatra and Borneo (Fig. SM2).Clouds and smoke were found to be swirling over the island of Borneo (right), Java (bottom), and Sumatra (left).
Fig. 11 shows the trajectory plots at different heights (300 m, 500 m, 1000 m) AGL calculated during the intermonsoon period from October 10, 2006 to October 23, 2006.The smoke signature on October 10, 2006 travels about 500 km from West Sumatra and passes through the area of Riau (315 km) to reach the receptor site.The smoke concentration at Kuala Lumpur was found to reach up to 1102 ng m -3 which is more than two times the average concentration of 477 ng m -3 .From the NASA satellite image (Fig. SM3) there were haze events in several areas in the islands of     were also calculated (Fig. 12).The trajectories show that the air parcel on July 30 came from the southeast direction and traveled about 750 km from South Sumatra (province of Palembang) to reach the receptor site in Kuala Lumpur in which the smoke concentration reach up to 1995 ng m -3 .In case of trajectory plots (300 m and 500 m) on August 4, the particles seem to have traveled from the southeast direction from the same area of South Sumatra (Palembang) then changed the direction to the southwest to reach the receptor site in Kuala Lumpur.From the NASA satellite image (see supporting material Fig. SM5) there were several fire locations detected by MODIS in the areas within the islands of Sumatra and Borneo for the week of August 4-11, 2008, as well as active fires on August 6, 2008.In addition a number of hotspots were detected in the island of Sumatra in July 2008, indicating a high intensity of forest and land fires as shown in the Sumatra hotspot distribution map (Fig. SM6).This information is available free of charge via the Internet at http://modis.gsfc.nasa.gov/.

CONCLUSIONS
The yearly 24 hour average of PM 2.5 concentrations throughout the 10 year study period was consistently about twice that of USEPA National Air Quality Standards, 15 µg m -3 .Even the PM 10 concentration that is mostly influenced by the PM 2.5 concentration was also found to be at the higher end of the Malaysian Air Quality Guidelines, 50 µg m -3 .Thus, significant increases of the air pollutants during dry spells due to open burning or forest fires caused occurrence of haze episodes on one or two occasions of a year.It was proven that the air pollutants resulted from the forest fire

Fig. 6 .
Fig. 6.Annual trend in source contributions of the PMF-modeled five factors in Klang Valley, Kuala Lumpur.

Table 3 .
Statistics and the smoke concentration for the selected dates in the studying period.