Near-road sampling of PM 2 . 5 , BC , and fine particle chemical 1 components in Kathmandu Valley , Nepal

15 Semi continuous PM2.5 and BC concentrations, and 24-hour integrated PM2.5 filter samples were 16 collected near roadways in the Kathmandu Valley, Nepal. Instruments were carried by a group of 17 volunteer traffic police officers in the vicinity of six major roadway intersections in the Kathmandu 18 Valley across two sampling periods in 2014. Daily PM2.5 filter samples were analyzed for water 19 soluble inorganic ions, elemental carbon (EC) and organic carbon (OC), and 24 elements. Mean 20 PM2.5 and BC concentrations were 124.76 μg m -3 and 16.74 μgC m during the drier spring 21 sampling period, and 45.92 μg m and 13.46 μgC m during monsoonal sampling. Despite the 22 lower monsoonal PM2.5 concentrations, BC and several elements were not significantly lower 23


Introduction
Particulate matter (PM) is a worldwide air pollution burden but is often most onerous in the developing nations (Han and Naeher, 2006).One such example is the Kathmandu Valley in Nepal, where degraded air quality is a major environmental and societal issue.The valley has gone through transformative social and economic changes over last 2 decades.With its current population of about 3.5 million, the Kathmandu Valley is growing at a rate of 4 % per year (CBS, 2014).The Kathmandu Valley has the highest population density (2800 persons km −2 ) in the nation (CBS, 2014), and concomitant with population growth, the number of vehicles in the Bagmati zone, one of the 14 administrative zones where the Kathmandu Valley is located, has increased by almost ninefold over 2 decades, with a total of 922 900 vehicles in 2014-15 (Department of Transportation Management, 2015).Motorcycles and passenger vehicles (cars, jeeps, and vans) are the main vehicle types, amounting to 92 % of the total registered vehicles in the Bagmati zone by the year 2014-15.Out of total gasoline and diesel consumption across Nepal, about half of the gasoline and one-fifth of the diesel is consumed by the population of the Kathmandu Valley alone (Malla, 2014).
Continued operation of older vehicles and poor road conditions are another cause of aggravated air quality problems in the valley.Shrestha et al. (2013) estimated that vehicle emissions from similar engines driving under the poor con-Published by Copernicus Publications on behalf of the European Geosciences Union.
ditions in the Kathmandu Valley were higher compared to other cities in developing countries in Asia.Traffic-related PM is especially important because it has been implicated in reducing lung function, increasing respiratory diseases, cardiac arrhythmias, asthma, and changes in heart rate variability (Pope and Dockery, 2006;Gauderman et al., 2007;Zanobetti et al., 2010;Shakya et al., 2016).Several studies have reported the health effects associated with specific PM components (Ghio and Devlin, 2001;Janssen et al., 2013;Wu et al., 2013), emphasizing the importance of chemical composition information as essential for assessing the health effects of PM and for understanding its sources.
In addition to emissions from vehicles and resuspended road dust due to vehicles, emissions from more than 100 brick kilns, the widespread use of small captive power generators during scheduled power cuts, and burning household and/or municipal wastes are other major sources of air pollution in the valley (ICIMOD, 2007;Shakya et al., 2010;The World Bank, 2014;Kim et al., 2015).The Kathmandu Valley has been facing rapidly deteriorating air quality on a metropolitan scale resulting from rapid urbanization and modernization, high population growth, the increasing number of vehicles, and increasing fuel consumption throughout the region.
Limited information exists for PM 2.5 (particles smaller than 2.5 µm in diameter) pollution (Aryal et al., 2009;Gurung and Bell, 2012;Stone et al., 2010) in this location, and even less information is available for PM chemical composition in the valley (Shakya et al., 2010;Chen et al., 2015).As developing nations continue to grow economically and societally, they are likely to be faced with critical decisions on energy consumption, transportation infrastructure, and public health protection measures.In order to make the best informed decision on efficient and effective regulation, it is important to understand aerosol climatology by understanding spatiotemporal patterns, and to provide foundational emissions inventory data to support advanced emissions models.This is particularly true with complex meteorology, where there are significant seasonal monsoon circulation and precipitation that can substantially alter regional air quality, but its effect at the local level may be uncertain.
In this study, we investigated the variability in aerosol concentrations (PM 2.5 and black carbon, BC) and PM 2.5 chemical composition to which people are exposed in the vicinity of six major traffic intersections of the Kathmandu Valley in two different seasons (spring dry season and monsoon wet season of the year).To our knowledge, this is the first comprehensive study of wide spatial and temporal variation in PM 2.5 pollution, notably PM 2.5 chemical composition, in the Kathmandu Valley, and it builds upon related work (Shakya et al., 2016;Kiros et al., 2016) on anthropogenic air pollution exposure and health effects in this community.

Methods
Aerosol sampling was conducted on roadsides at six locations in the Kathmandu Valley during two distinct sampling periods: Phase 1 in the spring (dry) season (16 February-4 April 2014) and Phase 2 in the monsoon (wet) season (20 July-22 August 2014).The six sites were selected to observe spatial distributions of air quality across the central urban core of the Kathmandu Valley.With this aim, we selected six locations: Kalanki, Balaju, Chabahil, Koteshwor, Thapathali, and Jawalakhel (Fig. 1), and each location was sampled for 5-6 days.At each location, up to six adult volunteers, who were employed as traffic police officers, carried a small bag containing battery-operated sampling equipment.A small GPS was included to collect geolocation data every 15 s.Sampler inlet tubes were attached to the chest of the traffic volunteers and were connected to a scattering nephelometer (pDR-1500, Thermo Scientific, USA) and a microaethalometer (AE51, AethLabs, San Francisco, USA).Volunteers carried the equipment both during work hours (typically during the day) and throughout the overnight hours.Overnight sleeping quarters were located on the ground floors of buildings adjacent to these major intersections at all sites except Jawalakhel, where the sleeping quarters were located about 50 m from the main intersection.Samples were collected continuously by these officers from the beginning of their work week (Sunday mornings) through the end of their week (Friday afternoons).With this arrangement, up to six instrument packages were in use at any given time, with traffic officers stationed on or near roadways across their neighborhood during the day, and all officers returning to a common sleeping barracks at night.At the conclusion of a weeklong sampling deployment, the equipment was collected, quality assured, and relocated to a new neighborhood location.Four sites, Kalanki, Balaju, Chabahil, and Koteshwor, are located on Kathmandu's busy Ring Road, while the remaining two sites, Thapathali and Jawalakhel, are located in an urban area inside the Ring Road.
Because of the high concentrations of PM, weekly cleaning of inlet cyclones was performed with deionized water.The pDR nephelometer instruments were zeroed each week using a HEPA filter, and volumetric flow rates were checked, and adjusted if necessary, with a traceable flowmeter.Filter tabs for the micro-aethalometer were always changed at least once every day, or more frequently depending on the warning given by the micro-aethalometer.Final concentrations from filter measurements for various chemical constituents were blank corrected by subtracting corresponding field blank values.
The pDR is calibrated against Arizona Test Dust (ISO 12103-1, Powder Technology, Inc, USA) by the manufacturer and operates at an accuracy of 5 %.The instrument operated at 1.52 L min −1 on a 5 min time base.Similarly, the micro-aethalometers operated at a 5 min time base, but at 50 cc min −1 because of the high BC loading anticipated in this measurement location.The micro-aethalometer was calibrated by the manufacturer before deployment in the field.Because of the high PM levels observed in this study, techniques such as optimized noise-reduction averaging were deemed to be unnecessary in order to obtain high-quality data.
2.1 PM 2.5 and BC sampling PM 2.5 and BC concentrations were measured in real time, recording data every 5 min using a portable scattering nephelometer (pDR-1500, Thermo Inc., US) and a microaethalometer (AE51, AethLabs, US), respectively.Both instruments were fitted with individual PM 2.5 cyclone heads to sample only particles less than 2.5 µm, and no denuding devices were employed.The pDR-1500 measures relative humidity and makes a calibrated correction for relative humidity to compute PM 2.5 concentrations.
Daily (24 h) PM 2.5 filter samples were collected on 37 mm filters (polytetrafluoroethylene (PTFE) filters or prebaked quartz fiber filters, QFFs) by the pDR instrument at a flow rate of 1.52 L min −1 .Filters were changed every morning.The PTFE filter and QFF were alternated every other day across all six sampler sets (i.e., all six samplers collected QFF on 1 day and PTFE the next, and so on).Quartz fiber filters were baked at 850 • C for 4 h prior to use for sampling in the Kathmandu Valley.After sampling, the filters were shipped to our laboratory for further chemical composition analyses.

PM 2.5 chemical speciation
An X-ray fluorescence (XRF) spectrometer (QUANT'X, Thermo Inc., US) was used to analyze elements on the particulate matter samples collected on a PTFE filter.Five scans were made for each filter with an X-ray tube at 10-50 keV following EPA compendium methods for inorganic metal speciation (US EPA, 1999).Thin film element standards were used for calibrating elements.
A 1.24 cm 2 punch was taken from the QFF for elemental and organic carbon analysis using a Sunset Laboratory OC-EC analyzer (Sunset Labs, US) following NIOSH-based thermo-optical methods (Birch and Cary, 1996).The instrument was validated with both external sucrose standards and an internal methane standard following the manufacturer's recommendations.

Results and discussion
3.1 Seasonal and diurnal variability in PM 2.5 and BC concentrations Hourly average PM 2.5 concentrations in the vicinity of six major road intersections in the Kathmandu Valley were observed to vary in the range of nearly 0-800 µg m −3 (Figs. 2 and 3) with the 5 min maxima reaching above 1000 µg m −3 during the spring sampling period.As expected, PM 2.5 levels decreased during the monsoon season, with only one hourly average value exceeding 100 µg m −3 .The median hourly average PM 2.5 concentration in spring (101.2 µg m −3 ) was nearly 3 times higher than that in the monsoon (36.3 µg m −3 ).
It is interesting to note that BC levels were not much different during the two seasons (Fig. 2).PM 2.5 concentration showed strong diurnal variability, with two distinct peaks occurring during the mornings and evenings (Fig. 3), which correspond to rush hour traffic.Such peaks occurred during the measurements in both seasons, though unsurprisingly, they were attenuated during the monsoonal sampling.BC also exhibited similar diurnal variability, which suggests that vehicle emissions are an important PM source in the valley.The BC concentration spikes during these rush hours were more pronounced, with larger peak concentrations at rush hours compared to other hours.Past studies (Panday and Prinn, 2009;Aryal et al., 2009;Sharma   et al., 2012) have shown that morning and evening peaks for PM 10 , PM 2.5 , BC, and carbon monoxide (CO) have been observed in the Kathmandu Valley.Morning peaks during the spring were qualitatively larger than evening peaks in spring compared to the monsoon for both PM 2.5 and BC.Such differences are likely due to strong nocturnal inversion layers and stagnant conditions during spring in the Kathmandu Valley (Panday and Prinn, 2009).This latter study was based on measurement sites that were a significant distance away from busy roads, and morning and evening peaks were found even on days without a regular rush hour.
A source of PM thought to be important in Kathmandu are brick kilns, which are important economic engines that manufacture bricks for construction.These kilns are operated only in the winter and spring and often use coal for a fuel source.With one exception, the selected sampling locations were more than 10 km away from active brick kilns.The Kalanki location, however, is located approximately 5 km to the east from a small cluster of kilns and may have been affected by this emission source.
In this current study, the highest PM 2.5 concentrations occurred between 06:00 and 10:00 NST during spring and returned to pre-06:00 concentrations after 11:00 (Fig. 3a).However, during the summer monsoonal sampling (Fig. 3b), PM 2.5 concentrations quickly rose after 06:00 and these elevated concentrations persisted well into the evening.BC diurnal patterns were qualitatively similar to PM 2.5 during the monsoon and spring sampling campaigns, with a rapid increase in concentration between 06:00 and 10:00, a falling concentration during midday, and a second peak in concentration during the afternoon rush hour period.Diurnal variations for BC were similar to those for PM 2.5 (Fig. 3c and  d).
The 24 h PM 2.5 average was calculated from the hourly average of 5 min measurements.These data exceeded the WHO guidelines for 24 h mean concentration (25 µg m −3 ) for all of 32 days of the sampling period during spring, and for 20 days out of 23 days during the monsoon.The 24 h ambient PM 2.5 standard set by the Nepalese government is 40 µg m −3 .All days during spring and 13 days out of 23 days in summer exceeded the national 24 h ambient PM 2.5 standard of Nepal.The 24 h PM 2.5 mean was 124.8 ± 55.9 and 45.1 ± 16.4 µg m −3 during spring and the monsoon, respec-tively.It should be noted, however, that these data were based on samples collected by mobile traffic personnel who lived and worked near busy roadways and they reflect a composite of both on-road, near-roadway, and sometimes indoor samples; the measurements were not made from the fixed monitors typically used in regulatory monitoring and do not necessarily reflect typical urban conditions subjected to regulatory action.Early morning to late evening measurements were collected outdoors while six traffic personnel worked at one of six different locations within a distance of about 2 km of their work station.Nighttime measurements were based on indoor measurements in the traffic officers' dormitory, which was located within a few hundred meters of their on-street work location.Spikes in concentration during the daytime could be affected by their specific work location, for example, such as whether the traffic personnel were working on busy intersections or on less traveled roadside locations with lighter traffic.
Spatial variability of PM 2.5 levels was similar during both seasons.Though the monitoring at the six sites was performed during 6 different weeks, we can compare the overall variation among the sites for the same season.Balaju had the highest PM 2.5 concentrations (198.4 µg m −3 ) compared to other sites (94.3-120.6 µg m −3 ) during spring (Fig. 2).During the summer monsoon, PM 2.5 ranged from 25.6 to 57.9 µg m −3 , with the highest and lowest concentrations at Balaju and Thapathali, respectively.This was consistent with observed results from Balaju, a neighborhood adjacent to a large bus terminal where the highest PM 2.5 concentrations were observed.Thapathali and Jawalakhel, with the lowest PM 2.5 levels, are located inside the city, and have paved roads with minimal road dust compared to other sites.
Overall, there was 57-74 % reduction in PM 2.5 concentrations (based on mean concentration) at four sites during the summer monsoon season compared to spring.During the monsoon, there was greater reduction at Jawalakhel (73 %) and Thapathali (72 %) compared to Balaju (61 %) and Chabahil (55 %).The monsoon season was characterized by higher ambient temperature and more frequent rain events (∼ 80 % of total annual precipitation occurs during June-September period) leading to both less energy use (such as absence of brick production, significantly fewer captive power generator sets and less burning of trace).
The average temperature and relative humidity (Davis automatic weather station) at a monitoring station at Bode, Bhaktapur, in the Kathmandu Valley were 14.8 • C and 73.2 %, respectively, during the dry season and 23.6 • C and 88.0 %, respectively, during the monsoon (rainy) season.The total precipitation during the dry and monsoon seasons was 50.47 and 266.6 mm, respectively.
The use of air conditioners in the Kathmandu Valley is not common; however, electric home heaters and heaters that use biomass briquettes, kerosene, and liquefied petroleum gas are common in the cooler winter.Scheduled power outages are also reduced in the summer monsoon, leading to a lower usage of diesel generators in the summer monsoon.
Daytime concentrations were computed as the average from 06:00 to 20:00 and nighttime concentrations were averaged as 20:00 to 06:00 from hourly averages of 5 min measurements.PM 2.5 concentrations were higher in the spring than in the monsoon (Fig. 4a and b).However, such differences between two seasons were much larger for daytime compared to nighttime.Daytime PM 2.5 concentrations exceeded nighttime by ∼ 1.5 times during spring, while these were ∼ 3 times higher during the monsoon.Balaju had the largest daytime and nighttime PM 2.5 levels in both seasons.Many long-route night buses operate from the Balaju (Gonggabu) bus terminal, and this may partially explain these results.Balaju, Chabahil, Koteshwor, and Kalanki, sites that are located along the Ring Road, had larger spring daytime PM 2.5 levels compared to the two sites located inside the Ring Road, Thapathali and Jawalakhel.Heavy-duty trucks, vehicles failing emission tests, and trucks carrying construction material are not allowed to enter inside the Ring Road during the daytime.Diesel trucks, pickups, and jeeps are thought to cause more pollution than diesel cars and vans in the Kathmandu Valley (Ale and Nagarkoti, 2003), and heavyduty vehicles are expected to cause more road dust suspension than lighter duty vehicles (Charron and Harrison, 2005;Garg et al., 2000).This, along with poorer road conditions around the four sites on the Ring Road, might be the reason for higher PM concentrations observed at these four sites on the Ring Road compared to the other two sites during the daytime.
Daytime and nighttime BC levels in the two seasons followed a spatial variation similar to PM 2.5 (Fig. 4c and d).The seasonal difference in BC concentration was much smaller during the daytime compared to nighttime.Traffic conditions (except diesel trucks) across the two seasons are not expected to be very different, which is indicated by a much smaller difference in BC concentration during the two seasons despite the large seasonal difference in PM 2.5 .The number of diesel trucks on roads are decreased during the summer monsoon because of the reduction in brick production and less demand for other construction materials.Such trucks could comprise the significant number (more than half) of vehicle movement in certain regions in the valley (JICA, 2012).While there is somewhat enhanced BC in the spring, this may be due to the stronger nighttime inversion and increased combustion and energy demand during spring.Increased BC concentration may also be partially explained by reliance on diesel generators, which are frequently used during the winter and spring dry months in the Kathmandu Valley to meet electrical power requirements by hospitals, hotels, industries, banks, and commercial operations (The World Bank, 2014) since electricity is often in short supply throughout the region in winter and spring.Such persistent emission of BC not only poses a serious threat to residents in the Kathmandu Valley, but BC is also an important short-lived climate forcing agent.BC also contributes to atmospheric brown cloud formation and it affects regional climate and glacier melting in the Himalayan region (Ramanathan and Carmichael, 2008).
3.2 Association of BC and PM 2.5 PM 2.5 and BC were correlated with each other (Fig. 5) during both the spring (r = 0.65, p<0.001) and monsoon seasons (r = 0.70, p<0.001), suggesting that a large fraction of PM 2.5 in the valley is co-emitted with BC directly from primary emission sources.While BC concentrations did not differ much during the two seasons, PM 2.5 concentrations were substantially lower during the monsoon.The ratio of PM 2.5 to BC, based on a simple linear regression relationship, was much lower during the monsoon (slope = 1.37) compared to spring (slope = 4.24).

Carbonaceous aerosol
Carbonaceous aerosol constitutes the major fraction (64 %) of PM 2.5 concentration during spring (Table 1).EC and OC are moderately correlated with each other in spring (r = 0.37, p<0.001) (Fig. S1 in the Supplement).The 24 h average PM 2.5 EC measurements from filter analysis and 24 h mean BC (from 5 m average measurement) measured with a micro-aethalometer indicate good agreement with each other (Fig. S2).The EC concentrations were larger than BC concentrations during spring and likely suggest the possibility of overestimation of EC in some of the samples.Previous studies have showed some deviations in BC and EC measurements (Allen et al., 1999;Kim et al., 2013).Lower BC values compared to EC values have also been recorded in other studies (Salako et al., 2012), and these results are likely attributed to analytical measurement differences and measurement uncertainty.However, there are some limitations in the comparison in this study.The comparison between EC and BC concentrations is based on measurements from the six sets of instruments by 36 traffic personnel around six sites.Examining the comparisons of our measurements show that about half of measurements had a BC / EC ratio between 0.70 and 1.19.
Both OC and EC concentrations were the highest in Balaju during spring, showing this site to be more polluted compared to the other five sites in the Kathmandu Valley.QFF samples collected during the monsoonal period were contaminated irretrievably and are, unfortunately, not included in the discussion.

Water-soluble ions
Water-soluble inorganic ions exhibited high spatial and temporal variability across six sites in the Kathmandu Valley (Fig. 6).The average concentrations of 11 water-soluble ions from all sites were 23.6 and 9.1 µg m −3 during spring and the monsoon, respectively.Among ions, SO 2− 4 , NH + 4 , Ca 2+ , and K + were the major ions dominating the PM 2.5 chemical composition (by mass) during both seasons (Table 1).Though Ca 2+ concentrations were similar in both seasons, SO 2− 4 and NH + 4 concentrations were reduced by about a factor of 5 during the monsoon.NO − 3 was reduced by a factor of 2 in the monsoonal period.High concentrations of SO 2− 4 observed in the Kathmandu Valley during spring may have been derived from the increased emission of precursor gas (i.e., SO 2 ) during spring from activities such as operation of brick kilns (which use coal with a high sulfur content), diesel generators, and diesel trucks.There was a slight increase in Ca 2+ concentration in the monsoon than in spring (lacking statistical significance, however).This suggests that dust contributions are either the same or slightly enhanced in the monsoon compared to spring, a surprising finding.Ca 2+ concentrations have also been used to indicate Asian dust in several studies (Choi et al., 2001;Shen et al., 2008).Again, monsoon PM 2.5 concentrations were lower by about a factor of 3 compared to spring.Thus, the similar loading of Ca 2+ suggests an impressive persistent dust burden, even during the monsoon.Road conditions were worse around Balaju and Chabahil compared to other sites, and it is likely that resuspension of road dust was an important emission component in these areas.Thus, it is not surprising that Ca 2+ levels were the highest at Balaju and Chabahil during both the spring and monsoon seasons.
Other water-soluble ions, such as Mg 2+ and K + concentrations, decreased in the monsoon compared to spring.The concentration of K + exceeded 1 µg m −3 at all sites except Thapathali during the spring.The K + concentration ranged from 1.2 to 1.8 µg m −3 during the spring, and 0.1 to 1.0 µg m −3 during the monsoon.During the monsoon, K + concentration decreased by greater than a factor of 5 (< 0.5 µg m −3 ) at all sites except Balaju.This suggests an increased contribution from biomass burning sources during www.atmos-chem-phys.net/17/6503/2017/spring as K + is a good marker for biomass burning emissions (Andreae, 1983;Duan et al., 2004) and is likely to be related to more open burning practices and refuse burning taking place during winter and spring months.A high concentration of phosphate (PO 2− 4 ) (comparable to Mg 2+ and Na + ) also suggests the increased contribution of residential burning to aerosol mass (Anderson et al., 2010).
During the spring, the highest SO 2− 4 concentrations (15.2 µg m −3 ) were observed at Thapathali despite this site having the lowest PM 2.5 concentrations.Such SO 2− 4 spikes might be related to the increased operation of diesel-powered power generators at nearby hospitals, commercial showrooms, and many commercial businesses located in the Thapathali area.Daily power outages occur for about 12 h per day in spring, and this leads to increased use of small-scale diesel-powered generators at commercial and tourism facilities.Thapathali is a popular business district in Kathmandu.
With two exceptions (nitrite and calcium), mean concentrations of water-soluble inorganic ions were higher during spring compared to the monsoon (Table 1).The largest difference was for SO 2− 4 and NH + 4 , when the spring concentrations were higher by a factor of ∼ 5 compared to the monsoon.The NO − 3 concentrations were more than 2 times higher in spring than in the monsoon.PM 2.5 in general was higher in spring than the monsoon due to additional emission sources and atmospheric stagnation during spring.The lower levels of the water-soluble ions during the monsoon could mainly be due to wet removal of particles containing these ions.Increased levels of NO − 3 , SO 2− 4 , and NH + 4 in spring compared to the monsoon might also be contributed by the low temperature and increased relative humidity favoring ammonium nitrate or ammonium sulfate formation and their partition into the particulate phase.Not surprisingly, NH + 4 was strongly correlated with SO 2− 4 (r = 0.65 in spring and r = 0.90 during monsoon).NO − 2 and Ca 2+ were about 20 and 10 % higher, respectively, during the monsoon than during spring.

Elemental composition
Crustal elements such as Al, Si, Ca, and Fe were observed at higher concentrations (2-6 µg m −3 ) during both seasons (Table 1).Among the analyzed elements, Si mass was the highest, which reinforces the large contribution of soil or sand and crustal material to PM 2.5 mass concentrations near roadways in the Kathmandu Valley.Silica in PM also comes from cement used in construction work, road surface dust, and tire wear (Kreider et al., 2010).Crustal elements such as Al, Si, Ca, Fe, and Mg were 2-3 times higher at Balaju and Chabahil compared to the other four sites during both spring and the monsoon.These elements were found to have higher concentrations at Koteshwor during the monsoon.These elements are likely to have originated from road dust due to the poor road conditions at both Balaju and Chabahil and road construction occurring in the Koteshwor-Satdobato segment of the Ring Road during the sampling period.Additionally, some evidence suggests that Fe, Mg, and Ca can also be emitted from diesel vehicles (Sharma et al., 2005).Tracer elements (Ba, Cu, and Zn) contributed 0.8 and 1.2 µg m −3 in the Kathmandu Valley during the spring and monsoon seasons, respectively.They are likely associated with trafficrelated emissions.Zn may be attributed to several trafficrelated sources, such as tire wear, brake dust, automobile exhaust, and metallic barriers (Lough et al., 2005;Kreider et al., 2010).Cu and Ba were found to be among the major elements in abrasion dusts from brake pads manufactured in Japan (Iijima et al., 2007).Fe and titanium (Ti) were also commonly found in brake dust samples (Thorpe and Harrison, 2008).
Crustal elements were well correlated with each other, and this relationship was stronger in spring than in the monsoon sampling (Fig. 7).Elements related with traffic (such as Cu and Zn) were also positively correlated with each other.One exception to this was Ba, which was correlated with Cu only during the monsoon.Total concentrations of eight heavy metals, Cr, Mn, cobalt (Co), Cu, Zn, arsenic (As), mercury (Hg), and lead (Pb), ranged from 0.2 to 1.1 µg m −3 at six sites during both sampling periods.Jawalakhel had the highest heavy metal concentration during spring and the second highest during monsoon.This site may be influenced by emissions from small metal industries located 2-3 km from Jawalakhel.

Elemental enhancement during monsoon
Despite lower PM 2.5 concentrations in the monsoon compared to spring, most of the elements were enhanced in PM 2.5 in the monsoonal sampling (Fig. 8).An enhancement ratio (ratio of PM 2.5 -normalized elemental concentration during monsoon to spring) is plotted for 11 major elements observed in this study.
The enhancement ratio is the largest for Ba, suggesting the consistent and large contribution of traffic-related PM 2.5 sources (Lough et al., 2005;Iijima et al., 2007) in the valley.The enhancement ratios were within the range of 2-4 for Si, Ca, Fe, K, Al, Mg, and Na.High enhancement ratios for the elements suggest that emissions of these elements did not concomitantly decrease in the monsoon compared to spring even though total PM 2.5 clearly decreased during the monsoonal sampling.This finding is important for source apportionment activities, where monsoonal effects on emissions profiles might be assumed to be proportionately reduced due to washout.
Among these elements, S is the only element that was not enhanced (ratio < 1) in the monsoon relative to spring.Nonenhancement of sulfur is also on par with the decrease in secondary inorganic ion concentrations in the monsoon.Coal combustion is not likely the major sulfur emission source in the valley in the monsoon period, and if sulfur aerosol is formed from secondary sources, monsoonal loss of such sulfur is likely.The major sulfur source probably arises from diesel fuel used in the Kathmandu Valley throughout the year.An additional contribution comes from coal combustion used in brick factories during winter and spring months.
When the elements measured by XRF spectroscopy and the water-soluble ions of the same elements were compared for such enhancements, it indicated consistent en-hancement ratios between two techniques, except for Cl and Mg (Fig. S3).For chlorine, water-soluble chloride ions were more enhanced in the monsoon compared to total chlorine, while water-soluble magnesium ions were less enhanced in the monsoon compared to total magnesium.

Enrichment factor
Because of the relative importance of crustal material to aerosol loading in the Kathmandu Valley, we use an enrichment factor technique to assess potential aerosol sources.The enrichment factor (EF) is computed by using (Taylor and McLennan, 1995 where E x is the concentration of individual elements shown in Fig. S4, (E x /E Al ) aerosol is the ratio of the elemental concentration to the Al concentration in aerosol, and (E x /E Al ) crust is the ratio of the elemental concentration to the Al concentration in the upper continental crust.In this exercise, Al is used as the reference element in crustal particles.
Al was strongly correlated (r>0.8) with the crustal elements, Si, Fe, Ca, and Mg during both seasons.The crustal ratios used for the computation of the enrichment factor are based on chemical composition of a generic upper continental crust (Taylor and McLenna, 1995).Lower EF values (close to 1) suggest natural crustal origin, while the higher EF values (greater than 10) suggest anthropogenic origin (Duan et al., 2006;Cong et al., 2010).
Mg has an EF less than 1 and is not shown in the illustration (Fig. S4).Si also has an EF less than 1, but only during the monsoon.Si also had the lowest EF ratio among all the elements during spring.This suggests that Mg and Si have mostly crustal origin.During the monsoon, Na, K, Ca, Mn, and Fe all have an EF less than 10.These elements are also likely to be associated with crustal and/or dust sources during the monsoon.
The elements Ca, Mn, and Na have EFs less than 10 or around 10 during spring, suggesting them to be of mostly crustal origin.Other trace elements have much larger EFs, suggesting emissions from anthropogenic sources.Potassium has an EF < 10 during the monsoon but an EF > 10 during spring, suggesting additional anthropogenic sources of K during spring.This is consistent with the open burning practices such as refuse burning, wood combustion, forest fires, and agro-residue burning taking place during the spring months (March-May) in the region.Elements such as Cu, Zn, and nickel (Ni) have EFs larger than 100, suggesting the strong contribution of anthropogenic sources.Zn and Cu were strongly correlated (r>0.85) in both seasons, suggesting the origin of particles from similar sources or common sources, such as traffic-related sources like tire wear and braking (Kreider et al., 2010).This confirms that traffic emissions remain one of the most important PM 2.5 sources in the Kathmandu Valley, and this is consistent with the work by Sharma et al. (2000), who suggested that urban air in Kathmandu is heavily influenced by local anthropogenic sources such as automobile exhaust or fossil-fuel-related emissions.

Source apportionment by chemical components
Though more than 300 filter samples were speciated in this study, there were only 12-18 samples collected from each site in 1 week in each season.Among these 12-18 samples from each location, we analyzed elemental and water-soluble ionic species; a second set of filters was collected on a QFF, which was used for analyzing carbonaceous species.Thus, the number of samples was inadequate to perform a robust source apportionment using models such as the positive matrix factorization or a similar technique for each location.Instead of such advanced models, we used a cruder method as follows to estimate the contributions of tracer species to PM 2.5 at each of our six sampling locations in both seasons.
By pooling six sets of measurements taken during the same week and season at each of the sites, we grouped chemical components of PM 2.5 into five potential source and chemical components: the crustal component, trace elements, secondary inorganic aerosols, biomass burning tracers, black carbon, and others (Fig. 9).The crustal component was simply computed as the sum of five crustal elements: Mg, Al, Si, Ca, and Fe.Trace elements were a sum of 15 trace elements: P, Sc, Ti, V, Cr, Mn, Co, Ni, Cu, Zn, cadmium (Cd), tin (Sn), Ba, As, and Pb.The sum of three water-soluble ions, sulfate, nitrate, and ammonium, were used to indicate secondary aerosols.Cl and K are used as biomass burning tracers."Others" accounts for the sum of the remaining watersoluble ions, organic matter, and remaining trace elements.Among the known (analyzed) chemical components, BC, the crustal component or dust, and secondary aerosols were the major components of PM 2.5 .The percent contributions of BC, crustal, and trace element components increased in the monsoon compared to spring, while secondary aerosol contributions decreased in the monsoon.
During spring, OC contributed about 34 to 52 % to PM 2.5 mass on the days when PM 2.5 samples were collected on QFFs.Carbonaceous aerosol (OC and EC) is the main component (∼ 64 %) of PM 2.5 during spring in the Kathmandu Valley.In a previous study during spring at an urban location in the Kathmandu Valley, carbonaceous aerosol contributed 63 % to the total speciated aerosol (carbonaceous and ionic aerosols) (Shakya et al., 2010).
It is interesting to note that BC contribution was larger and crustal contributions were smaller at the two sites located inside inner cities, Jawalakhel and Thapathali, compared to the other four sites (except Koteshwor, where no samples were collected in spring) during the monsoon.The sum of the five crustal element concentrations were highest at Chabahil (spring: 16.9 µg m −3 , monsoon: 12.7 µg m −3 ) and Balaju (spring: 17.7 µg m −3 , monsoon: 10.5 µg m −3 ), and the lowest at Thapathali (spring: 4.4 µg m −3 , monsoon: 1.8 µg m −3 ) and Jawalakhel (spring: 9.6 µg m −3 , monsoon: 5.8 µg m −3 ).BC concentrations exceed crustal elemental concentrations by a factor of 1.3-1.8 at Balaju, Chabahil, Jawalakhel, Kalanki, and Koteshwor.This suggests that dust and traffic emissions are the most important PM emission sources in the valley.One exception is Kalanki, where the BC / dust ratio was 3.2, and Thapathali with a BC / dust ratio of 2.8 and 6.2 during spring and the monsoon, respectively.Concentrations of biomass burning tracers were the largest at Balaju and Chabahil during both spring (4.1-4.6 µg m −3 ) and the monsoon (1.1-1.5 µg m −3 ).Chabahil is located near Pashupati, which is the main cremation site for the community.This process is biomass burning intensive and occurs year round, which may partially explain these results.

Spatial variability in exposure within the sites
The coefficient of divergence (COD) was computed between four locations for PM 2.5 , BC, and several chemical species (Fig. 10) to explore the differences in concentrations between each of these specific locations due to the mobile nature of the sampling by six traffic volunteers at the same time.COD is expressed as (Wilson et al., 2005) COD where S ij and S j k are the concentrations (of PM 2.5 or other parameters) for sampling day i for individual traffic personnel working at the j and k locations; p represents the number of observations.The higher COD values indicate heterogeneity and lower COD values indicate homogeneity (Wilson et al., 2005).The COD values lower than 2 are considered as homogeneous (Thornburg et al., 2009).Two locations (i.e., Kalanki and Koteshwor) did not have enough data points for both seasons, and they were excluded from the analysis.The COD is utilized to quantify the heterogeneity of the parameters measured from the same set of instruments carried around by six traffic police officers in the vicinity of each site.The six traffic police officers were within about 2 km from each other, and all the measurements were taken concurrently for a period of 6 days.Overall, Jawalakhel and Thapathali were the most heterogeneous for chemical species, suggesting the presence of more diverse sources in their vicinity.Balaju was the least heterogeneous for most of the chemical species.BC and PM 2.5 were the most homogeneous (COD < 0.2) among the four sites, while chemical components were the most heterogeneous.BC had the least spatial variation among all sites except Thapathali.This might be due to BC loadings from additional significant sources such as diesel generators at Thapathali.
Tracers of dust particles and secondary inorganic ions had the highest intra-site variation.A large fraction of these samples were in the immediate vicinity of roads with heavy to moderate traffic and thus the traffic emissions were the most prevalent source.Despite the similarity of measurements (within short distance and similar local settings) in Kathmandu Valley, chemical species were still variable within such short distances among these six sets of measurements.The COD values show that concentrations could vary even among the similar sites within a short distance.The differences in contributions of local sources, transient emission events, and measurement error (Pinto et al., 2004) might explain the high spatial variability observed in the Kathmandu Valley.

Conclusions
This study documented distinct seasonal (dry season versus wet season) and diel variations in PM 2.5 and BC levels in the Kathmandu Valley.The variability in PM 2.5 and BC was greater for nighttime levels than daytime levels, suggesting that local PM 2.5 emissions were not much reduced during the monsoon.Meteorological factors such as higher temperature and wind speeds, change in wind directions, increased number and frequency of rainfalls, absence of certain types of emissions (such as trash burning and brick production), and a weaker nocturnal boundary layer may have played the main roles in lower concentrations of PM 2.5 in the Kathmandu Valley during the monsoon.BC was only marginally reduced by monsoonal sampling conditions.This indicates that there is a prevalence of BC sources near roads, most likely traffic emissions, in the Kathmandu Valley, which constantly pollutes the valley's air with BC throughout the year.
Organic carbon was an abundant and ubiquitous aerosol component at all six locations in the Kathmandu Valley.OC aerosol contributed the largest fraction of PM 2.5 , followed by EC, SO 2− 4 , NH + 4 , and crustal elements.Not surprisingly, all samples from the six sites were heavily influenced by trafficrelated emissions and dust, and vehicle emissions were found to be a major source of PM 2.5 chemical components in these locations.High concentrations of SO 2− 4 in the Kathmandu Valley point to the influence of diesel and coal combustion in the valley's air pollution, particularly during the winter and spring months with fairly similar emission and meteorological characteristics.Despite the close proximity and similarity of the sites (close to busy traffic with a similar upwind regional emissions source), PM 2.5 chemical species were found to be spatially variable across specific chemical species, but less variable for bulk measurements of PM 2.5 and BC.
For air quality management purposes in the Kathmandu Valley, this study suggests that traffic-related emissions and soil-dust-construction materials are the main sources of PM 2.5 near roadside locations.Chemical component data also suggest that biomass burning, secondary ions, and dust contribute to PM 2.5 during the drier spring, with a diminished effect in the rainier summer.In contrast, dust contribution is much enhanced in PM 2.5 during the monsoon period, while water-soluble ion concentrations were reduced in the same period as they are efficiently washed out by rain.During the monsoon, frequent rains on unpaved roads may facilitate the resuspension of dust particles.
More broadly, this unique dataset highlights a divergence in concentrations that were thought to be affected downward by large-scale meteorological effects.While PM 2.5 was substantially reduced during the monsoon, there was no such decrease in BC levels; similar findings were observed for a number of trace elemental components in ambient aerosol.These data have important relevance for promulgating optimized air quality control measures, protection of human health, and assessment of climate forcing effects from localized emissions.As a result, there remains much to understand in how highly polluted communities in the developing world can affect local and regional air quality.
Data availability.Raw and other data from this work can be obtained by e-mailing the corresponding author.
Competing interests.The authors declare that they have no conflict of interest.

Figure 2 .
Figure 2. Time series of BC concentrations (upper panel) and PM 2.5 concentrations (lower panel) observed at six near-road locations in the Kathmandu Valley during measurement periods in the spring and monsoon seasons of 2014.The thin solid line in each panel represents hourly average concentrations, while the thick line represents 24 h average concentrations.

Figure 3 .
Figure 3.Diurnal variations in PM 2.5 and BC concentrations observed at six near-road sites in the Kathmandu Valley for entire sampling periods in the spring season (a, c) and monsoon season (b, d).The lower end and upper end of each box represent the 25th and 75th percentiles, respectively; whiskers represent 1.5 × interquartile range; the horizontal black line in the middle of each box is the median for each hour of the day.

Figure 4 .
Figure 4.The mean concentrations of PM 2.5 (a, b) and BC (c, d) at nighttime (20:00-06:00 NST, Nepal standard time) and daytime (06:00-20:00 NST) over the sampling periods in the spring season and monsoon season of 2014 observed at six near-roadside locations in the Kathmandu Valley.Error bars are the standard deviation of the measurements.

Figure 5 .
Figure 5. Correlation between PM 2.5 concentrations and BC concentrations observed at six near-road sites in the Kathmandu Valley over the sampling periods in the (a) spring season and (b) monsoon season in 2014.

Figure 6 .
Figure 6.The average ambient concentrations of ions and elements in PM 2.5 over the two sampling periods in the spring season and monsoon season at six locations in the Kathmandu Valley.

Figure 7 .
Figure 7.The matrix of correlation coefficients between elements, ions, BC, and PM 2.5 concentrations observed at six sites in the Kathmandu Valley in the spring season (above diagonal of white boxes) and monsoon season (below diagonal of white boxes).

Figure 8 .
Figure 8. Enhancement ratio of elements in the monsoon season compared to the spring season in the Kathmandu Valley.

Figure 9 .
Figure 9.Chemical composition of ambient PM 2.5 (crustal elements (Mg, Al, Si, Ca, and Fe), heavy metals (Cr, Mn, Co, Cu, Zn, As, Hg, and Pb), secondary ions (NH + 4 , SO 2− 4 , and NO − 3 ), biomass burning tracers (K and Cl), and others (including organic material)) observed at six sites in the Kathmandu Valley for two sampling periods in the spring season (upper panel) and monsoon season (lower panel) in 2014.

Figure 10 .
Figure 10.Mean coefficient of divergence (COD) for 24 h average concentrations of PM 2.5 , BC, and PM 2.5 chemical constituents at four sites during 2014.

Table 1 .
Summary of PM 2.5 chemical composition (µg m −3 ) in the Kathmandu Valley during the spring season (Phase 1) and monsoon season (Phase 2).