Atmospheric environment volume 78 issue 2013 doi 10 1016%2fj atmosenv 2012 05 006 hai, cao dung; kim oanh, nguyen thi effects of local, regional meteorology and emission sources on mass and compositions of

Effects of local, regional meteorology and emission sources on mass and compositions of particulate matter in Hanoi Cao Dung Hai 1 , Nguyen Thi Kim Oanh * Environmental Engineering and Management, School of Environment, Resource and Development, Asian Institute of Technology, Thailand highlights < PM 2.5 and PM 10e2.5 were investigated during polluted winter period in Hanoi. < Variation in PM mass and composition were analyzed with meteorology and emission. < Reconstructed mass and PMF model were used for source identification and apportionm ent. < Contribution of local sources and long range transport to PM 2.5 were analyzed. article info Article history: Received 2 October 2011 Received in revised form 26 April 2012 Accepted 3 May 2012 Keywords: Fine particulate matter EC/OC Diurnal variation Meteorology Source apportionment Hanoi abstract Intensive monitoring for PM mass and composition was conducted during December 2006eFebruary 2007 at a mixed site in Hanoi, Vietnam. Fluctuations in levels of 24 h PM 2.5 (26e143 m gm À3 ) and PM 10 (37e165 m gm À3 ) were examined in relation to weather conditions and regional synoptic patterns. High 24 h PM levels were observed at low wind speeds when a stagnating ridge governed over Northern Vietnam. Diurnal variations of PM mass and composition, analyzed using 4 h-samples, reflected the influence of local emissions with peaks of PM mass and EC/OC observed during morning and evening rush hours. The 24 h EC and OC levels in PM 2.5 ,1.5e4.9 m gm À3 and 10e39 m gm À3 , respectively, constituted about 90% of the total EC and OC in PM 10 . Ionic species (SO 2À 4 ,NO À 3 ,NH þ 4 ,Cl À and K þ ) were the major composition of PM 2.5 , whereas elements (Ca, Si, Al and Fe) were the major components in the coarse fraction (PM 10e2.5 ). The reconstructed mass, explained 78 Æ 11% of PM 2.5 and 61 Æ 11% o f PM 10e2.5 , suggested significant contributions of secondary PM (OM, SO 2À 4 ,NO À 3 and NH þ 4 ) and combustion (biomass, diesel, etc.) to PM 2.5 . PMF model revealed 7 source factors of PM 2.5 : secondary mixed PM (40%), diesel traffic (10%), residential/commercial cooking (16%), secondary sulfate rich (16%), aged seasalt mixed (11%), industry/incinerator (6%), and construction/soil (1%). Contributions from local emission sources and the potential long-range transport were analyzed using the PM compositions and the diurnal variations in relation to local source activities, location of local sources, winds and air mass HYSPLIT trajectories. Major part of PM 2.5 mass appeared to link to local emission origins. Additional measurement data are required to characterize the weekendeweekday and inter-seasonal patterns of PM. Ó 2012 Elsevier Ltd. All rights reserved. 1. Introduction Atmospheric particulate matter (PM) is the most notable air pollution problem in developing Asian cities (Gupta et al., 2006; Tsai and Chen, 2006; Hopke et al., 2008; Kim Oanh et al., 2006). PM gets increasing attention because of the adverse effects on human health, the atmosphere and climate (Bond et al., 2004). The extent of these effects depends on PM size and compositions. Fine parti- cles measured as PM 2.5 (particles with the aerodynamic diameter 2.5 m m) can penetrate the lungs deeper than large particles, hence are more damaging (Pope et al., 2009). Systematic PM records, however, remain limited in Asian developing countries, especially for PM 2.5 . Until recently, available national monitoring networks focused more on the total suspended PM (TSP) but at present PM 10 (particles with the aerodynamic diameter 10 m m) is also monitored routinely in many countries (HEI, 2010). Available data show high levels of PM in Asian cities that often exceeded the respective national ambient air quality standards (NAAQS). In particular, PM 2.5 and PM 10 normally exceed * Corresponding author. E-mail address: kimoanh@ait.ac.th (N.T. Kim Oanh). 1 Current address: Petrovietnam Overseas Exploration Production Operating Company, Hanoi, Vietnam. Contents lists available at SciVerse ScienceDirect Atmospheric Environment journal homepage: www.elsevier.com/locate/atmosenv 1352-2310/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved. doi:10.1016/j.atmosenv.2012.05.006 Atmospheric Environment 78 (2013) 105e112 the 24 h WHO air quality guidelines by a factor of two or more (Hopke et al., 2008; Kim Oanh et al., 2006). In Vietnam, in the past only TSP was measured and over a short sampling period (e.g. 1 h) but recently PM 10 mass is continuously monitored in some urban areas by a network of automatic stations. Data on PM 2.5 have been made available through research projects (Hien et al., 2001, 2002; Kim Oanh et al., 2006) and commonly show high levels in large cities. In Hanoi, the capital city of Vietnam, higher PM levels are observed during the dry season which can be explained by high emission strength and limited atmospheric dispersion (Kim Oanh et al., 2006). Some data on PM 2.5 mass and composition have been reported earlier (Hien et al., 2004; Kim Oanh et al., 2006) but mainly based on 24 h PM sampling. This paper presents the results of more time-resolved PM 10 and PM 2.5 monitoring, with 4 h averaging time (6 samples per day), conducted at a mixed site in Hanoi during winter. In addition to the water soluble ions and element compositions that were partly reported in early studies, this paper also provides the EC and OC data. The chemical composition of size-segregated atmospheric particles was used to quantitatively assess contributions from sources using the reconstructed mass approach and the Positive Matrix Factorization (PMF) receptor model (Paatero and Tapper, 1994). 2. Monitoring program design 2.1. Study area The monitoring was conducted during the period of December 20 06eFeburary 2007 in the Hanoi capital city of Vietnam. Then the Hanoi Metropolitan Region (HMR) comprised of 7 inner districts and 5 suburban districts with a total area of around 921 km 2 and the population of 3.3 million (GSO, 2006). Since 2008, the HMR has been expanded to cover a total area of 3348 km 2 and the population of over 6.7 million (GSO, 2009). Hanoi is located in Red River delta, about 100 km from the East Vietnam Sea, and has a tropical climate with 4 seasons. During winter and early spring, the area is under the influence of northeast monsoon while during summer the influence of southeast monsoon is pronounced. The sampling site (20.983  N; 105.784  E) was located in the Thuong Dinh industrial zone, Hanoi city (Fig. S1, supplementary information). The PM samplers and meteorology equipment were placed on the rooftop of a building (15 m above the ground) of the Hanoi University of Science (HUS) and about 100 m from the heavily travelled Nguyen Trai road (Fig. S1, SI). This road, 60 m wide with 6 auto-vehicle and 2 non-auto vehicle lanes, is the main southwestward transport route from the Hanoi city. The Thuong Dinh industrial zone is the largest among those located on the right bank of the Red River in HMR. There are also several large univer- sities in the area which contribute to the high traffic flow in this street (Truc and Kim Oanh, 2007). Main air pollution emission sources in this area include road traffic, residential cooking, industrial activities and construction activities. Surrounding agri- cultural areas also contribute emission from agroresidue field burning to the site but this is a seasonal emission source, e.g. intensive rice straw field burning is observed around June and November each year. In peri-urban areas around the site, solid waste open burning is also intensive when weather is dry. 2.2. Sampling and analytical methods One Andersen dichotomous sampler (dichot) for simultaneous fine (PM 2.5 ) and coarse (PM 10e2.5 ) fractions, two MiniVol Samplers (minivol) for PM 2.5 and PM 10 , one Sibata 30 L sampler for PM 10 , and one portable weather station (GroWeather Davis) were deployed. The meteorology conditions (temperature, wind speed and direction) were recorded every 30 min during sampling. Before sampling all PM samplers were calibrated to obtain the recom- mended flow rates (15.03 L min À1 for fine and 1.67 L min À1 for coarse fractions by dichot; 5 L min À1 for Minivols, and 30 L min À1 for Sibata 30 L). The entire sampling period was divided into two sub-periods: the routine 24 h sampling, 23 December 2006e7 January 2007, when 15 pairs of 24 h PM 10 and PM 2.5 samples were collected and the intensive sampling, 12 Januarye11 February 2007, when 92 pairs of 4 h PM samples (6 samples per day) were collected. Detail on the sampling equipment and on filters used is given in Table S1, SI. The QA/QC of the sampling and analysis were following the same procedures described in Kim Oanh et al. (2009). The filter conditioning and pre-weighing (using a microbalance) were done at the Asian Institute of Technology (AIT). Quartz filters were pre- fired at 550  C for about 6 h to remove any carbonaceous/organic pollutants. After sampling, each quartz filter was put in a Petri dish and kept in a separate airtight bag. The samples were refrigerated at HUS and were transported (in an ice box) to the AIT laboratory in Thailand for subsequent analyses. Immediately after collected, each mixed cellulose esters filter was cut into 2 equal parts. One part was kept in a tube containing Milli-Q water and refrigerated while stored in HUS to minimize the loss of volatile ions. At AIT the tube content was extracted and analyzed for water soluble ions (Na þ , NH þ 4 ,K þ ,Mg 2þ and Ca 2þ ;Cl À ,NO À 3 and SO 2À 4 ) by Ion Chromatog- raphy (IC). The other part was refrigerated in HUS and upon arriving at AIT was analyzed for elements (20 species) using the Inductively Coupled Plasma-Optical Emission Spectroscopy (ICP-OES). The quartz filters were first used for mass determination at AITand then sent to the University of Illinois at Urbana Champaign for EC/OC analysis by a Sunset Analyzer (thermal-optical transmittance method, the NIOSH 5040 protocol). Blank reference filters were weighed before and after real samples on each weighing day to ensure that the filter mass was unchanged over time under the weighing conditions (US EPA, 1998). At least one trip blank was made for a composition group (EC/OC, ions and elements) in each sampling sub-period. The blank filters (brought to the site but were not exposed during a sampling event) were stored and transported together with the samples. On average, the ion levels in the (mixed cellulose) blanks were below 1.5 m g per filter for K þ ,Mg 2þ and Na þ , and in the range of 3e4 m g per filter for Cl À ,SO 2À 4 ,NO À 3 ,NH þ 4 and Ca 2þ . The element content in the mixed cellulose blanks was below 0.03 m g per filter for most of the analyzed species, except for Fe, Si and Mg that was 0.2e0.5 m g per filter and Ca of 3 m g per filter. OC content in the quartz blanks was 1.4e3.7 m g per cm 2 while EC blanks were close to zero. The composition results were all blank corrected. Note that a denuder was not used hence a certain loss of easily volatile nitrate particles was possible. However, the loss during sampling was expected to be minimized due to the moderate ambient air temperatures during the period, 13e20  C, coupled with the low sampling flow rates and relatively short sampling period (4 h). 3. Results and discussion 3.1. Daily PM levels in relation to meteorology The daily (24 h) mass concentrations of PM 2.5 and PM 10 during the sampling period were widely fluctuating (Fig. 1). The 24 h PM 2.5 levels were averaged at 76 Æ 32 m gm À3 , ranged from 26 to 143 m gm À3 , i.e. all measurements were above 24 h WHO guideline of 25 m gm À3 . The 24 h PM 10 levels during the period were 98 Æ 35 m gm À3 (37e165 m gm À3 ) which exceeded the 24 h Vietnam NAAQS of 1 50 m gm À3 in 2 days and exceeding the WHO guideline (50 m gm À3 ) in 27 days (87% of the measurements). The monitoring C.D. Hai, N.T. Kim Oanh / Atmospheric Environment 78 (2013) 105e112106 period (DecembereFebruary) includes the months with the highest PM levels in Hanoi (Hien et al., 2002). The range obtained by Hien et al. (2002) for DecembereFebruary in 1998e1999 was compa- rable with our results for fine PM (PM 2.2 w25e150 m gm À3 ) but higher for PM 10 (w50e350 m gm À3 ). Our results also show that PM 2.5 constituted the major fraction of PM 10 (PM 2.5 /PM 10 w 0.76 Æ 0.08) which emphasizes the impor- tance of the combustion emission and/or secondary particles. The regression line between PM 2.5 and PM 10 (PM 2.5 ¼ 0.9 PM 10 À 15) had R 2 ¼ 0.92 indicating a strong correlation which further suggests that the change in PM 10 was mostly driven by the change in PM 2.5 . Consistently high PM levels during the first 9 sampling days (Fig. 1) were observed with the highest levels during the New Year holiday period (31st December and 1st January). In principle, less activities in the industrial zone and universities (hence less traffic emission) were expected during the holidays but only on Sunday, 31st December a reduction in PM was observed. Examining local and regional meteorology during the sampling period should help to better understand the daily PM variations. The weather during the monitoring period was generally dry with some drizzle and foggy days typically observed during winter in Hanoi. The wind speeds (30-min averaged) measured at the site were light, ranging from calm to about 2 m s À1 . Only a few obser- vations recorded speeds of about 3 m s À1 with a maximum of 4.8 m s À1 seen on 7 February 2007. Daily PM mass during the period appears to inversely proportional to the local wind speed (Fig. 1). Obviously, lower wind speeds resulted in poor dispersion of air pollution during the New Year holidays that could offset the impacts of reduced source activities on the PM levels. In the following week (2e7 January 2007), when wind speeds increased, lower PM levels were observed. Daily windroses were examined and a summary on wind directions together with major features of regional synoptic meteorology on the sampling days are given in Table S2, SI. On the regional scale, during winter Northern Vietnam is frequently influenced by a high pressure ridge extending from anticyclones centered in Siberia and China. When such a ridge reaches Northern Vietnam the sea level pressure increases, temperature and relative humidity drop, while wind with more northerly directions and increased speeds are observed. This principal relationship between the weather conditions and the synoptic patterns was also observed during the sampling days. From 25th December 2006 to 7 th January 2007 the prevalent wind directions were gaining more Northerly components and eventu- ally to predominant Northerly during 27 th e29 th . The pressure was the highest during 29e31 December 2006 whereas the surface temperature dropped to below 20  C. The synoptic chart (0:00 UTC) shows a strong ridge over Northern Vietnam (Fig. S2,a, SI). The influence of this stagnating ridge (typically with subsidence and radiative inversions) over Northern Vietnam resulted in stagnant air and high PM levels in the initial monitoring days before 2 January 2007. After that, the ridge was weakening and a low pressure system developed over SEA (Fig. S2,b). The wind speed in Hanoi increased (daily average of 1.5 m s À1 ) and remained high until 7th January due to a relatively large pressure gradient observed over the area that induced a better dispersion hence lowered PM levels. During 12the15th January, Northern Vietnam was again under the influence of a stagnating ridge (Fig. S2,c) and the PM levels were high. The levels declined from 16th January onward when SEA was under influence of a low pressure system and wind speed increased (Fig. S2,d). 3.2. Diurnal variations of PM and major composition components The 4 h sampling of PM 10 and PM 2.5 was conducted, starting from 6:00 am each day to yield 6 samples per day during the intensive sampling, 12 th e20 th January and 05 th e11 th February. Fig. 2 presents the average diurnal variations of PM 10 and PM 2.5 Fig. 2. Diurnal variations of PM mass, major components in PM 2.5 and wind speed based on 4 h-sampling in JanuaryeFebruary 2007 (25  C, 1 atm). Fig. 1. 24 h PM mass concentrations (25  C, 1 atm) and daily average wind speed during the monitoring period. C.D. Hai, N.T. Kim Oanh / Atmospheric Environment 78 (2013) 105e112 107 mass and major ionic components in PM 2.5 . High PM levels observed in the morning rush hours (6:00e10:00) and evening rush hours (18:00e22:00) suggested higher contributions of primary particles especially the road traffic. Note that diurnal variations were clearly seen for PM 2.5 while for the coarse fraction (PM 10e2.5 is the gap between PM 10 and PM 2.5 in Fig. 2) there was only a small reduction in the early morning samples (2:00e 06:00). This is because these two fractions were originated from different sources/processes. The coarse particles are normally emitted directly from industrial and construction activities, and/or wind- blown dust while fine particles are either directly emitted from combustion sources (primary particles) or formed in the atmo- sphere (secondary particles) (Chow, 1995). The diurnal variations of EC and OC also followed the variations of PM 2.5 mass with peaks in the morning and evening rush hours (Fig. 2). EC is directly released from incomplete combustion (Bond et al., 2004), mainly in fine PM, and can be used as a marker of this source. OC includes both primary and secondary particles but they were not segregated in our results presented in Fig. 2. Sulfate and nitrate, major secondary inorganic PM (fine fraction), did not show a clear diurnal variation as compared to PM mass and EC/OC (Fig. 2). Chloride (Cl À ), in fact, was following the PM mass diurnal pattern but the peaks were quite small. Daily patterns of primary particles were expected to reflect daily variations in the local emission. Truc and Kim Oanh (2007) reported a very high traffic density in the same road with the daily flow on weekdays of 145,000 vehicles (from 7:00 to 19:00) that had high peaks (mainly motorcycles and cars) during morning and evening rush hours. The fleet contained 96% motorcycles, 3% gasoline powered vehicles (taxi, cars, etc.) and about 1% diesel powered vehicles. Two major types of diesel powered vehicles, truck and bus, were the principal source of EC at the monitoring site. The bus hourly density remained quite stable (about 120e180 buses) during the daytime period of 5:00e19:00, declined afterward (20e30 buses) and virtually no buses were observed after 22:30. The truck hourly density was w100 vehicles during the daytime (8:00 e17:00) and reduced to about 30 trucks at night but then with more heavy duty vehicles. In addition, it was observed that cooking activities for meals at homes and in small restaurants in the surrounding areas were also coincidentally more intensive during the morning and evening rush hour periods. The daily patterns of PM mass and EC/OC thus appear to follow the daily patterns of these two major sources in the study area. Other important factors determining the PM diurnal variations are related to the meteorology (mixing height, wind, etc.). Average wind speed was the highest (1.8 m s À1 , Fig. 2) around noon (10:00e14:00), which to some extent may lower PM levels at this sampling time (in addition to reduced emission strength). However, the second highest wind speed (1.6 m s À1 ) was observed during the morning rush hours (6:00e10:00) when maximum PM and EC/OC were observed suggesting that the enhanced dispersion by the wind could not offset the increased emission strength. In principle, the vertical mixing also changes during a day, better around noon and poorer late at night, especially for winter when radiative inversion frequently exists in Hanoi (Hien et al., 2002). Thus, low levels of PM and EC/OC in the afternoon (14:00e18:00) may be attributed to both the enhanced mixing and reduced emission strength (a shorter averaging period is desirable as the evening rush hours in principle start at around 17:00). Note that PM and EC/OC levels in the samples collected at night and early morning (22:00e 02:00 and 2:00e6:00) were still lowest in a day which confirmed the predominant influence of diurnal variations of the local emission. Sample-wise PM mass and EC/OC variations (Fig. S3 and S4, SI) showed two highest EC/OC peaks, both in the Friday evening (18:00e22:00), January 12 (EC ¼ 7.3 m gm À3 ,OC¼ 56 m gm À3 ) and February 9 (EC ¼ 8.8 m gm À3 ,OC¼ 44 m gm À3 ). To examine further the weekdayeweekend fluctuation, the diurnal variations for different days of a week were analyzed (Table S3, SI). The- highest 4 h average EC level (18:00e22:00) was on Friday (5.9 Æ 3.7 m gm À3 ), followed by Sunday (4.5 Æ 2.5 m gm À3 ) and the lowest were on other working days (MondayeThursday), 2.4 Æ 0.5 m gm À3 . OC levels largely followed the EC patterns. High traffic density on Friday and Sunday evenings, related to personal trips from and to Hanoi for the weekend activities, may partly explain these high PM levels. Overall, the diurnal patterns of EC and OC on different days of a week were similar to the PM 2.5 pattern but generally different from the PM 10e2.5 pattern. The fact that fine particles contain EC and OC emitted from combustion and that majority of EC and OC in our samples were found in PM 2.5 explains the similarity in their diurnal patterns. However, the number of monitoring days was still small (3 Saturdays, 2 Sundays, 3 Fridays, and 8 other working days) to properly characterize the weekdayeweekend variations. Additional data are required for the purpose, including the time-resolved air pollution, meteorology and source activities (i.e. traffic counts, cooking pattern) during different days of a week. 3.3. Daily PM composition 3.3.1. EC/OC The 24 h average of EC and OC levels in PM 2.5 are presented in Fig. 3 which show the same trend as PM presented in Fig. 1.Notethat EC/OC data were available for all 92 PM 2.5 samples as compared to only 30 PM 10e2.5 samples, accordingly EC/OC were estimated for only 30 PM 10 samples. In PM 2.5 , the average EC levels during the sampling period were 2.7 Æ 1.5 m gm À3 (range of 24 h levels: 1.5e4.9 m gm À3 ) and OC were 18.3 Æ 11.9 (10e39 m gm À3 ). In PM 10 the EC levels were 3.0 Æ 1.7 m gm À3 (1.5e5.8 m gm À3 ) and OC levels were 21 .5 Æ 12.3 m gm À3 (11e43 m gm À3 ). Majority of EC (about 89 Æ 4%) 0 10 20 30 40 50 12 Jan (Fri) 13 Jan (Sat) 14 Jan (Sun) 15 Jan (Mon) 16 Jan (Tue) 17 Jan (Wed) 18 Jan (Thur) 19 Jan (Fri) 20 Jan (Sat) 04 Feb (Sun) 05 Feb (Mon) 06 Feb (Tue) 07 Feb (Wed) 08 Feb (Thur) 09 Feb (Fri) 10 Feb (Sat) 11 Feb (Sun) EC/OC, µg m -3 OC EC Fig. 3. Daily average EC and OC concentration in PM 2.5 calculated from 4 h data for intensive sampling period (whiskers represent one standard deviation). C.D. Hai, N.T. Kim Oanh / Atmospheric Environment 78 (2013) 105e112108 and OC (90 Æ 4%) was associated with PM 2.5 . Overall, EC constituted only a small fraction of PM 2.5 mass (3.4 Æ 1.1%) while OC share was significant (24.4 Æ 7.3%). PM 10e2.5 contained only a low level of EC (0.5 Æ 0.3 m gm À3 ) and OC (2.2 Æ 1.3 m gm À3 ). The ratio between EC and total carbon (TC ¼ EC þ OC) in PM 2.5 also varied during the monitoring period with lower EC/TC ratios (0.11e0.12) observed during 12 th e16 th January and the highest ratios (w0.15e0.2) during 7 th e9 th February (Fig. S5, SI). In principle, the diesel exhaust would have a high ratio (EC/TC w0.7, Kim Oanh et al., 2010) while biomass open burning would have a lower ratio (w0.15 for rice straw field burning, Kim Oanh et al., 2011). The EC and OC absolute levels as well as EC/TC ratios appeared to change with the main wind directions despite of the overall low wind speeds. When northerly winds (from residential and commercial areas to the site, see Fig. S1, SI) were observed, e.g. 12e16 January (Table S2, SI), higher EC and OC levels were obtained but lower EC/TC ratios. The cooking in homes and restaurants and other related activities in the residential area located just to the north of the site may be the influencing factor. Winds were southerly and southeasterly during the rest monitoring period that brought in more fresh traffic emission from the road to the site hence during this period higher EC/TC ratios were obtained although the levels of EC and OC were generally lower. 3.3.2. Ionic species The results of 8 water soluble anions are summarized in Fig. 4 for the fine (PM 2.5 ) and coarse fraction (PM 10e2.5 ). The highest levels were found for SO 2À 4 ,NO À 3 , and NH þ 4 , predominantly in the fine fractions, indicating significant contributions of secondary inor- ganic particles. Local sources of the gas precursors of these particles may include the vehicle exhausts (SO 2 ,NO x ), residential cooking in the study area with smoky coal briquettes (SO 2 ), as well agriculture activities (NO x ,NH 3 ). Cl À and K þ were also found at higher levels in the fine fraction as compared to the coarse fraction that may be linked to biomass burning. Na þ and Cl À may link to seasalt including fresh table salt from cooking. Hanoi is located about 100 km west to the Bac Bo (Tonkin) Gulf and the S-SE winds may bring in the aged seasalt particles through the regional transport. High levels of Ca 2þ observed in the coarse fraction probably link to construction activ- ities around the site as well as soil/road dust. The ion balance was constructed for the QA/QC purpose using the sum of anions and sum of cations in the molar concentration unit (equivalents). For PM 2.5 , the regression line between anions and cations had a slope of 0.97 (more anions or slightly acidic) and R 2 ¼ 0.79 while that for PM 10 has a slope was 1.1 (slightly alkaline) and R 2 ¼ 0.62 (Fig. S5, SI). 3.3.3. Element composition The results of the element analysis are summarized in Fig. 5. Major elements in PM 10e2.5 included Ca (highest, reaching above 2 m gm À3 ), Si, Fe and Al which indicate the contribution from soil/road dust and construction activities. PM 2.5 contained noticeable levels of most detected elements but mostly below 0.5 m gm À3 in the high level group and below 0.2 m gm À3 in the low level group (Fig. 5) which indicate a combustion origin. Lead (Pb) was quite significant (w0.17 m gm À3 ) and was mainly found in PM 2.5 , also suggesting a combustion origin. The Pb level was well below the Vietnam 24 h NAAQS (1.5 m gm À3 ) and that was expected following the successful phase out of leaded gasoline in the country in 2001 (ESMAP, 2002). 3.4. Source apportionment of PM 3.4.1. Reconstructed PM mass The reconstructed mass was done using 8 mass groups following the method presented previously (Kim Oanh et al., 20 06) to provide information on major contributing sources. These included (K-biomass ¼ KÀ0.6 Fe), crustal (oxides of crustal ¼ 1.16 Â (1.9 Al þ 2.15 Si þ 1.41 Ca þ 1.67 Ti þ 2.09 Fe)), organic matter (OM ¼ 1.4 OC), soot (EC), seasalt (2.54 Â Na þ ), NH þ 4 , NO À 3 and SO 2À 4 for secondary organic particles, and trace metals (remaining analyzed elements) as presented in Table 1. The calcu- lated mass in principle should be lower than the measured mass because not all the components were analyzed. However, due to uncertainties related to sampling and analytical methods, a few PM 2.5 samples (7 samples) had more than 100% mass explained hence they were excluded from the statistics presented in Table 1. For the coarse fraction, all samples had mass explained below 100%. On average, the portion of measured mass explained by the 8 groups was 78 Æ 11% for PM 2.5 and 61 Æ 11% for PM 10e2.5 . The K-biomass was used to identify the presence of biomass burning smoke in the PM and it was not the absolute contribution of the biomass burning. It was excluded from the sum of the mass explained to avoid double counting because K was already included in the “trace elements” group. The K-biomass group was more significant in PM 2.5 than PM 10e2.5 as most of PM emitted from biomass burning is expected to be in the fine fraction (Kim Oanh et al., 2011). The crustal group was predominantly found in the coarse frac- tion indicating a significant contribution from soil/road dust and construction activities. The seasalt group, estimated based on the content of Na þ and Cl À hence indicating “fresh” seasalt, was small in both fractions. The trace elements (remaining analyzed elements) group was generally found at low levels and more in the fine PM (than the coarse PM fraction) and may be linked to a combustion origin. 3.4.2. Source apportionment of PM 2.5 by PMF The PMF receptor model (Paatero and Tapper, 1994), used in this study, assumes non-negativity of the factors, both loadings (source contributions) and scores (source profiles). PMF2, applicable for two-dimensional arrays, was applied for the PM 2.5 source appor- tionment using the composition data of 92 PM 2.5 samples taken during the intensive sampling sub-period that had complete EC, OC, ions and element composition data. Due to the lack of EC/OC data in a large number of coarse PM samples the PMF analysis was not done for this fraction. The measurement uncertainty of each species was estimated following the method presented in Kim Oanh et al. (2009). PMF was run with F peak ranged from À0.6 to þ0.6 and F peak ¼ 0.5 produced seven source factors with the most explainable source profiles. The contributions of these source factors to PM 2.5 (Fig. 6) listed in the reducing order included: 1) secondary mixed (local) PM (31 Æ 15 m gm À3 ), 2) residential/commercial combustion (12 Æ 10 m gm À3 ), 3) secondary sulfate rich (LRT) (12 Æ 11 m gm À3 ), 4) aged seasalt mixed (9 Æ 7 m gm À3 ), 5) traffic (diesel) (8 Æ 5 m gm À3 ), 6) industry/incinerator (4 Æ 5 m gm À3 ), and 7) Fig. 4. Water soluble ions concentrations in fine and coarse PM (whiskers represent one standard deviation). C.D. Hai, N.T. Kim Oanh / Atmospheric Environment 78 (2013) 105e112 109 construction and soil/road dust (1 Æ 1 m gm À3 ). The scatter plot between the scaled calculated mass (x) and measured mass (y) yields a regression equation y ¼ 0.97x þ 2.7 with R 2 ¼ 0.64. The profiles of the identified source factors are presented in Fig. S7, SI and the temporal variations in the contributions are in Fig. S8, SI. The residential/commercial had more pronounced contributions during January 12e16, when the local wind was weak and had the directions of N-NW, and lower contributions during the rest of the monitoring period when the wind directions changed to S-SE. A populated residential area was located to the north of the site (Fig. S1, SI), food stall/restaurants and commercial activities were densely distributed along the road and other small streets inside the residential area. Commercial cooking normally lasted from 4:00 am to late evening hours with the highest activ- ities between 18:00e22:00 while residential cooking was normally more intensive in the evening. Various fuel types were used for cooking in the area including LPG and wood fuel at home, and commonly smoky coal briquettes in food stalls/restaurants. The coal briquettes used for cooking were made of rejected coal parti- cles which would produce high sulfur and heavy metal emission. Sawdust or peat mixed in the briquettes, and wood fuel would contribute high levels of K þ shown in the source profile. Low combustion temperature of the residential/commercial cooking as well as solid/yard waste burning (observed in dry weather) is ex- pected to emit higher OC than EC. Accordingly, the residential/ commercial combustion had a source profile with specific markers including high OC and EC (with the ratio EC/OC < 1.0), relative abundance of NH þ 4 , sulfate, nitrate, Na þ ,K þ and Cl À , and some elements (Si, Al, Pb, Cr, Mn). Note that Na þ and Cl À present in this source profile may directly link to table salt used in cooking that may also explain the diurnal Cl À variation pattern discussed (Fig. 2). The traffic (diesel) exhaust profile is characterized by high EC with the ratio EC/OC > 1.0 (Fig. S7, SI) which is remarkably different from the residential/commercial source discussed above. A rela- tively high abundance of V, Zn, Si and Mg 2þ in this source might be originated from the fuel additives (Kim Oanh et al., 2010). This source was quite stable during the monitoring period but was more pronounced during February 9e11 when the wind (SE, <1ms À1 ) was blowing from the road to the site (Fig. S1, SI). Diesel powered buses and trucks running in this street were probably the most important sources of fresh PM 2.5 with high EC to the site. A more clear diurnal contribution pattern was also observed for this source (Fig. S8, SI) with low emission in the late night and early morning hours (22:00e 6:00) when the public buses stopped operation. Note that at night heavy trucks were observed traveling in this road and other side streets in the surrounding area hence the truck contribution may be not strongly directional. The secondary mixed PM (local) was identified by a profile with abundance of sulfate, nitrate, ammonium and high OC but with low EC (Fig. S7, SI). Overall, its contribution was quite stable during the sampling period, but showed some dependence on wind speeds (not on wind directions). This source factor had slightly lower contributions when higher wind speeds were observed, e.g. January 17e18 and vice versa (Fig. S8 and Table S2, SI). The HYSPLIT (Draxler and Hess, 1998) 5-day backward trajectories were constructed, starting at 0:00 UTC and 1000 m above the sampling site for each sampling day (Fig. S9, SI), and showed that the variations in this source factor contribution were not strongly associated with the air mass types. In the dry weather observed during the sampling period a long atmospheric life of PM was expected that enhanced the mixing and uniform distribution of the secondary PM in the study area. This explained why the contributions of this source factor Fig. 6. Percentage of average source contributions to PM 2.5 at Thuong Dinh, Hanoi (based on 4 h sampling, January 12eFebruary 20, 2007). 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.0 0.5 1.0 1.5 2.0 2.5 3.0 Fe Zn Si Al Ca M g Ni Pb Sr Ti V Be Cd Co Cr Cu Li Mn Mo Tl Low conc. g rou p ( µg m -3 ) High conc. group (µg m -3 ) Fine fraction Coarse fraction Low conc. species High conc. species Fig. 5. Element compositions of fine fraction and coarse PM (whiskers represent one standard deviation). Table 1 Summary of reconstructed mass for fine and coarse fraction based on 4 h PM sampling, m gm À3 . Parameters PM 2.5 (n ¼ 85) PM 10e2.5 (n ¼ 85/30) b Max Min Av Æ std Max Min Av Æ std PM mass 162 32 78 Æ 33 66 18 29 Æ 10 K-biomass a 3.2 0.0 0.9 Æ 0.8 2.2 0.0 0.1 Æ 0.3 Crustal 5.4 0.38 2.2 Æ 1.1 13 0.6 4.8 Æ 2.5 OM 54 6.2 21 Æ 8.5 5.5 1.5 2.9 Æ 1.1 Soot 4.8 0.79 2.2 Æ 0.9 1.6 0.1 0.5 Æ 0.5 NH þ 4 19 0.02 7.9 Æ 4.8 2.8 0.02 0.9 Æ 0.7 NO À 3 17 0.02 6.9 Æ 3.8 6.4 0.02 1.7 Æ 1.8 SO 2À 4 39 7.2 17 Æ 8.1 4.8 0.02 1.2 Æ 1.3 Seasalt 2.8 0.04 1.2 Æ 0.8 8.1 0.04 1.8 Æ 1.8 Trace metals 3.1 0.31 0.8 Æ 0.5 1.0 0.05 0.3 Æ 0.2 Mass explained, % 99 39 78 Æ 11 93 36 61 Æ 16 a K-Biomass is not included in the calculated mass to avoid double counting. b Only 30 coarse PM samples had EC/OC data. C.D. Hai, N.T. Kim Oanh / Atmospheric Environment 78 (2013) 105e112110 were more depending on local wind speeds than on air mass types (with characteristic wind directions). To distinguish this source factor from the secondary sulfate appeared later it is named the secondary PM (local) although the precursors of this secondary PM may have both local and regional/LRT emission origins. The “ industry/incinerator” contribution was identified by higher levels of several elements (Mg 2þ , Fe, Pb, Zn, Al, Cr and Mn) in the source profile. This source factor had a relatively small contribution (6%) and high temporal variations (Fig. S7, SI). A few peaks were observed at late evening and early morning hours and were in fact higher on Saturdays and Sundays. Several industrial activities were located on the opposite side across the road, but in principle should not have a high contribution at late hours and on weekend. It was noted that there was a hazardous hospital waste incinerator (stack height of 20 m) located about 10 km to the NW of the monitoring site that may explain higher contributions from this source during 12e16 January (NW-N wind predominant). In addition, there is a waste recycling village (Trieu Khuc, Fig. S1, SI) located about 2 km to the SW of the site which often had waste burning in the late evening hours. Further investigation is still required to explain the detailed source activities of this factor. The aged seasalt (mixed) was identified by higher levels of Na þ , Mg 2þ , as well as nitrate and sulfate in the profile. It contributed about 11% of PM 2.5 mass at the site. Examining the HYSPLIT 5-day backward trajectories during the period, confirmed that the aged seasalt (mixed) contribution was higher during the days when a long marine pathway (15e16, 18 January and 5e7 February) was observed (Fig. S9, SI). The coastal line of the East Vietnam Sea is about 100 km away from the site. During the transport to the site, the marine air masses likely have Cl À depleted (Lee et al., 1999), i.e. replaced by nitrate and sulfate. Likewise, various secondary and primary air pollutants were also picked up and brought to the site hence this aged seasalt mixed was not the same as the (fresh) seasalt calculated by reconstructed mass (Table 1). Overall, this source factor was likely associated with the regional/LRT air pollution that had a marine pathway. The secondary sulfate rich PM (LRT), with a distinguished high abundance of sulfate, contributed about 16% to PM 2.5 mass. As compared to the secondary mixed PM (local) this source factor had a low level of nitrate but more abundance of K þ ,Mg 2þ and NH þ 4 . The temporal variation (Fig. S8, SI) showed higher contributions when relatively strong winds were recorded at the site, i.e. during 17e19 January and 11 February. The backward trajectories showed that this source factor had higher contributions when the air masses arriving at the site with a longer continental pathway, i.e. moving along the coastal line of Southern China on 13th January and 17th January, or with a continental pathway over Northern Vietnam and Southern China, or arriving from the west neighboring countries/ territories (9e11 February). This source factor was likely linked to the LRT of sulfate-rich PM that was originated from continental territories on the pathway of air masses before arriving to the site. The contribution from soil/construction activities to the site was identified by a source profile with a high abundance of Ca, Si and Al (Fig. S7, SI). This source contributed only 1% of PM 2.5 mass at the site. As expected, there was no wind direction dependent and a few peaks (Fig. S8, SI) may be linked to drier weather conditions when more intensive construction activities and soil/road dust released. 3.4.3. Discussion: contributions from local emission sources and LRT For the purpose of the air quality management it is always of interest to estimate the contribution from the local sources and the LRT air pollution but this is a challenging task, especially when only monitoring data are available. This is because PM locally emitted or formed in a study area is mixed with those of regional and LRT origin. The precursors are also both locally and regionally emitted. At the monitoring site, the most likely part of PM having association with the regional and LRT origins included the secondary sulfate rich (LRT) and aged seasalt mixed, that collectively contributed about 27% of PM 2.5 mass. This was estimated using the source profiles, knowledge of local sources and their variations, the vari- ations in the source contributions in relation to local meteorology (wind) and pathways of air masses. Nevertheless, it is important to note that not the entire contributions of these two source factors (27%) could be attributed to the LRT because the presence of PM of local origins in the mixture was anticipated. Likewise, the secondary mixed PM (local), contributing about 40% PM 2.5 mass, may also have a part associated with the LRT origin. However, intensive emission from local sources, for example traffic and cooking, may contribute significantly to air pollution, both primary PM and gaseous precursors for secondary PM, in Hanoi. As motorcycles have high daily driving activities and scattered over the city (Phuong, 2009), their contributions would also be widely scattered. Lack of a source profile of the gasoline powered vehicles prevents from estimation of this source contribution. Preliminary results of the emission inventories for Hanoi, using the Interna- tional Vehicle Emissions modeling, show that in 2008 the total motorcycle fleet in Hanoi contributed 2.4 kt PM 10 (Phuong, 2009) which was about 15% of the PM 10 emission estimated from the diesel bus fleet alone in 2010 (Trang, 2011). Thus, the gasoline vehicles may be more important in the contribution of the gaseous precursors for secondary PM formation than directly to the primary PM in the city. An important local source that is observed to dras- tically increase around the city is the rice straw field burning. However, the burning is mainly in June and November after the harvest of two rice crops, respectively, which was not covered by our monitoring. Hien et al. (2004) conducted the PMF modeling for fine and coarse PM samples collected during 1999e2001 in Hanoi for 3 types of air mass backward trajectories. Majority of the backward trajectories in our study period (Fig. S9a & b, SI) can be roughly classified into Type 2 (Northeasterly) of Hien et al. (2004). The authors identified 6 contributing source factors to fine PM for this air mass type (LRT, local burning, soil dust, local secondary aerosol, marine aerosol and Cl-depleted aerosol). These are largely equiva- lent to our identified source factors except for the traffic (diesel) exhaust emission that was not identified in their study. With the available EC/OC data and element composition for the PMF input our study can identify the contribution from traffic (diesel) emis- sion, about 10% of PM 2.5 mass. Also our study revealed that the LRT was likely to contribute a smaller part of PM 2.5 pollution than the local emission. The difference between these two studies was large and may be partly explained by the rapid urbanization and motorization in Hanoi during the 7e8 year gap. Note that the average PM 2.5 data in our study period was 76 Æ 32 m gm À3 that was about 2 times of the measured levels for Type 2 (43 m gm À3 )inHien et al. (2004). 4. Conclusions High PM levels were observed at a mixed site in Hanoi during the dry winter period with the fine fraction (PM 2.5 ) contributed the majority (76%) to PM 10 . The daily PM 2.5 levels were found to closely link to local wind speeds and directions (relative to local emission sources) and regional synoptic meteorology suggesting that both local sources and LRT contributed to the high PM levels at the site. Diurnal variations in PM 2.5 , EC, OC and EC/TC were similar and can be largely explained by the fluctuations in local source activities. The coarse fraction (PM 10-2.5 ) was less fluctuated during a day. Reconstructed mass served as useful tool in identifying potential C.D. Hai, N.T. Kim Oanh / Atmospheric Environment 78 (2013) 105e112 111 source factors for subsequent PMF analysis, although for a site located 100 km away from the sea like this, the estimated (fresh) seasalt group may not necessarily relate to the sea spray. PMF identified seven source factors for PM 2.5 that can be explained using the knowledge on local emission sources and their variations, wind and HYSPLIT air mass backward trajectories. The major source factors included the secondary mixed PM (largely with local origins), local emission (primary PM from cooking, traffic and industry/incinerator) and LRT (secondary sulfate rich and aged seasalt mixed). For PM 10e2.5 , the reconstructed mass suggested that soil/road dust and construction activities were the main sources. Overall, the LRT was likely to contribute a smaller part of PM 2.5 mass (about 27% have an association with LRT origins) measured at the site while the majority appears to originate from local emis- sions. The secondary PM (local) contributed a large part of PM mass (about 40% of PM 2.5 ) hence to improve PM air quality in the city the control efforts should also focus on the emission of gaseous precursors. Cleaner fuel-cookstove systems in residential/ commercial cooking, cleaner diesel vehicles, as well as emission reduction for the gasoline vehicle fleet (including the fleet density reduction) should be addressed. Further studies should be con- ducted for other periods of the year especially during the rice straw burning months of June and November to reveal the contribution from this source. Acknowledgment We would like to acknowledge the AIRPET project (http://www. serd.ait.ac.th/airpet) sponsored by Sida and the ARCP2007-07CMY project sponsored by the Asia-Pacific Network for Global Change Research for the partial funding support provided to cover the consumables for the data collection. Dr. Hoang Xuan Co at Hanoi University of Science and Dr. Nghiem Trung Dung at Hanoi University of Technology and their team are specially thanked for their great assistance extended during the sampling period in Hanoi. Colleagues at UIUC are highly appreciated for the generous support for EC/OC analysis. Appendix A. Supplementary information Supplementary information related to this article can be found online at doi:10.1016/j.atmosenv.2012.05.006. References Bond, T.C., Streets, D.G., Yarber, K.F., Nelson, S.M., Woo, J H., Klimont, Z., 2004. A technology-based global inventory of black and organic carbon emissions from combustion. Journal of Geophysical Research 109, 1e43. Chow, J.C., 1995. Measurements methods to determine compliance with ambient air quality standards for suspended particles. Journal of Air and Waste Manage- ment Association 45, 320e382. Draxler, R.R., Hess, G.D., 1998. An overview of the HYSPLIT-4 modelling system for trajectories, dispersion and deposition. Australian Meteorological Magazine 47, 295e308. ESMAP, 2002. An overnight success: Vietnam’s switch to unleaded gasoline. Available at: http://www.esmap.org/esmap/node/872. GSO, 2006. Population and population density by province in 2006. Available on: http://www.gso.gov.vn/. GSO, 2009. Population and population density by province 2009. Available on: http://www.gso.gov.vn/. Gupta, P., Christopher, S.A., Wang, J., Gehrig, R., Lee, Y., Kumar, N., 2006. Satellite remote sensing of particulate matter and air quality assessment over global cities. Atmospheric Environment 40 (30), 5880e5892. HEI, 2010. Outdoor Air Pollution and Health in the Developing Countries of Asia: a Comprehensive Review. Health Effects Institute, Massachusetts, USA. Hien, P.D., Bac, V.T., Tham, H.C., Nhan, D.D., Vinh, L.D., 2002. Influence of meteo- rological conditions on PM2.5 and PM2.5e10 concentrations during the monsoon season in Hanoi, Vietnam. Atmospheric Environment 36, 3473e3484. Hien, P.D., Bac, V.T., Thinh, N.T.H., 2004. PMF receptor modeling for fine and coarse PM10 in air masses governing monsoon conditions in Hanoi, northern Vietnam. Atmospheric Environment 38 (2), 189e201. Hien, P.D., Binh, N.T., Truong, Y., Ngo, N.T., Sieu, L.N., 2001. Comparative receptor modelling study of TSP, PM 2 and PM 2À10 in Ho Chi Minh City. Atmospheric Environment 35 (15), 2669e2678. Hopke, P., Cohen, D., Begum, B., Biswas, S., Ni, B., Pandit, G., Santoso, M., Chung, Y., Davy, P., Markwitz, A., Waheed, S., Siddique, N., Santos, F., Pabroa, P., Seneviratne, M., Wimolwattanapun, W., Bunprapob, S., Bac, V.T., Hien, P.D., Markowicz, A., 2008. Urban air quality in the Asian region. Science of the Total Environment 404, 1103e1112. Kim Oanh, N.T., Pongkiatkul, P., Upadhyay, N., Hopke, P., 2009. Design a monitoring program of ambient particulate matter for source apportionment study by receptor modeling. Atmospheric Environment 43, 3334e3344. Kim Oanh, N.T., Thiansathit, W., Bond, T.C., Subramanian, R., Ekbordin, W., Ittipol, P., 2010. Compositional characterization of PM 2.5 emitted from in-use diesel vehicles in Asia. Atmospheric Environment 44 (1), 15e22. Kim Oanh, N.T., Thuy, L.B., Tipayarom, D., Manadhar, D.R., Pongkiatkul, P., Simpson, C.D., Liu, S.L.J., 2011. Characterization of particulate matter emission from open burning of rice straw. Atmospheric Environment 45 (2), 493e502. Kim Oanh, N.T., Upadhyay, N., Zhuang, Y.H., Hao, Z.P., Murthy, D.V.S., Lestari, P., Villarine, J.T., Chengchua, K., Co, H.X., Dung, N.T., Lindgren, E.S., 2006. Particulate air pollution in six Asian cities: spatial and temporal distributions, and asso- ciated sources. Atmospheric Environment 40, 3367e3380. Lee, E., Chan, C.K., Paatero, P., 1999. Application of positive matrix factorization in source apportionment of particulate pollutants in Hong Kong. Atmospheric Environment 33, 3201e3212. Paatero, P., Tapper, U., 1994. Positive matrix factorization: a non-negative factor model with optimal utilization of error estimates of data values. Environmetrics 5, 111e126. Pope III, C.A., Ezzati, M., Dockery, D.W., 2009. Fine particulate air pollution and life expectancy in the United States. The New England Journal of Medicine 360, 376e386. Phuong, M.T.T., 2009. Emission Inventory for Motorcycles in Hanoi, Vietnam Using the International Vehicle Emissions Model. Master thesis, Asian Institute of Technology, Pathumthani, Thailand. Tsai, Y.I., Chen, C.L., 2006. Atmospheric aerosol composition and source appor- tionments to aerosol in southern Taiwan. Atmospheric Environment 40, 4751e4763. Trang, T.T., 2011. Emission Inventory of Passenger Transport Fleet in Hanoi to Assess Air Quality and Climate Co-benefits Associated With Various Technology Scenarios. Master thesis, Asian Institute of Technology, Pathumthani, Thailand. Truc, V.T.Q., Kim Oanh, N.T., 2007. Roadside BTEX and other gaseous air pollutants in relation to emission sources. Atmospheric Environment 41, 7685e7697. US EPA, 1998. Quality Assurance Guidance Document. Method Compendium: PM 2.5 Mass Weighting Laboratory Standard Operating Procedure for the Performance Evaluation Program. C.D. Hai, N.T. Kim Oanh / Atmospheric Environment 78 (2013) 105e112112 . Effects of local, regional meteorology and emission sources on mass and compositions of particulate matter in Hanoi Cao Dung Hai 1 , Nguyen Thi Kim Oanh * Environmental Engineering and Management,. 5 m gm À3 ), and 7) Fig. 4. Water soluble ions concentrations in fine and coarse PM (whiskers represent one standard deviation). C.D. Hai, N.T. Kim Oanh / Atmospheric Environment 78 (2013) 105 e112 109 construction. h PM mass concentrations (25  C, 1 atm) and daily average wind speed during the monitoring period. C.D. Hai, N.T. Kim Oanh / Atmospheric Environment 78 (2013) 105 e112 107 mass and major ionic